CN111882664B - Multi-window accumulated difference crack extraction method - Google Patents
Multi-window accumulated difference crack extraction method Download PDFInfo
- Publication number
- CN111882664B CN111882664B CN202010652509.XA CN202010652509A CN111882664B CN 111882664 B CN111882664 B CN 111882664B CN 202010652509 A CN202010652509 A CN 202010652509A CN 111882664 B CN111882664 B CN 111882664B
- Authority
- CN
- China
- Prior art keywords
- point
- crack
- pavement
- window
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/05—Geographic models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Remote Sensing (AREA)
- Computer Graphics (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The embodiment of the invention provides a crack extraction method of multi-window accumulated difference, which comprises the following steps: acquiring three-dimensional pavement data of the target pavement after the attitude and deformation information are removed; for each point on the section in the three-dimensional pavement data, calculating the accumulated difference characteristic under each window based on the preset window number and the point elevation to obtain the point-by-point multi-window accumulated difference characteristic corresponding to the target pavement; non-supervision clustering is carried out on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a Kmeans clustering algorithm, or supervision classification based on different source samples is carried out on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a trained supervision classifier model, so that all crack objects in the target pavement are obtained. The embodiment of the invention has the characteristic of sample sharing capability, can realize large-scale three-dimensional pavement crack extraction under a small quantity of marked samples, and provides a stable and robust method for actual pavement crack detection.
Description
Technical Field
The invention relates to the technical field of line scanning three-dimensional data processing, in particular to a multi-window accumulated difference crack extraction method.
Background
With the development of line scanning three-dimensional measurement technology, more and more three-dimensional pavement data which can be acquired by a three-dimensional measurement system comprise data of different pavement backgrounds and different crack types. The existing method can analyze the cracks with clear features and obvious features in part of the pavement so as to finish crack detection. However, the existing methods are difficult to be applied to actual pavement crack detection tasks with complex background, large background difference and large crack type and characteristic difference.
On the other hand, the traditional machine learning and even the deep learning are difficult to directly scan the three-dimensional pavement data crack extraction on line to obtain better effects, one side of the linear scanning three-dimensional data has the influence of factors such as driving gesture deformation diseases, and the like, the robustness of an actual crack detection task to the method is higher, and the data suitable for different types of cracks and different pavement backgrounds are needed; on the other hand, accurate marking information is difficult to obtain from the identical data, and the traditional template matching and edge detection methods and even the deep learning methods have limited applicability to different pavement cracks of different data.
Therefore, a three-dimensional pavement crack method capable of overcoming the defects of multiple crack types, complex background and machine learning adaptation is needed, and a multi-window cumulative difference crack extraction method is proposed under the background.
Disclosure of Invention
In order to solve the above problems, an embodiment of the present invention provides a crack extraction method for multi-window cumulative difference.
In a first aspect, an embodiment of the present invention provides a method for extracting a crack by using multiple window cumulative differences, where the method includes:
acquiring three-dimensional pavement data of the target pavement after the attitude and deformation information are removed;
For each point on a section in the three-dimensional pavement data of the target pavement, calculating the accumulated difference characteristic under each window based on the preset window number and the point elevation by taking the current point as a starting point to obtain the point-by-point multi-window accumulated difference characteristic corresponding to the target pavement;
Performing unsupervised clustering on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a Kmeans clustering algorithm to obtain all crack objects in the target pavement;
or performing supervised classification based on different source samples on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to the trained supervised classifier model, and acquiring all crack objects in the target pavement.
Preferably, the acquiring the three-dimensional pavement data of the target pavement after the removal of the posture and deformation information further includes:
Modeling and characterizing the fluctuation characteristics of the cross section cracks and textures of the three-dimensional pavement data by adopting point-by-point multi-window accumulated difference characteristics based on a plurality of three-dimensional pavement data samples from which the gesture and deformation information are removed, and obtaining point-by-point multi-window accumulated difference characteristics corresponding to each sample;
if the labeling information of all samples is unknown, performing unsupervised clustering by utilizing point-by-point multi-window accumulated difference features corresponding to all samples through a Kmeans clustering algorithm to obtain trained Kmeans clustering algorithm parameters;
If the labeling information of part of the samples in all the samples is known, training the supervised classifier model by using the samples with the known labeling information, and obtaining the trained supervised classifier model.
Preferably, the calculating the cumulative difference feature under each window based on the preset window number and the point elevation to obtain the point-by-point multi-window cumulative difference feature corresponding to the target road surface specifically includes:
For each acquisition point of each cross section in the three-dimensional pavement data of the target pavement, taking the current acquisition point as a starting point, and acquiring a point-by-point multi-window accumulated difference characteristic for each cross section based on the elevation of the acquisition point of the cross section and the number of preset windows;
Specifically, the point-to-point multi-window accumulated difference characteristic is calculated by the following expression:
DN＝[d1,d2,…,di,…,dN],i∈[1,2,3,…,N]，
Wherein D N represents a point-by-point multi-window cumulative difference feature, i represents a window range, D i represents a window range threshold, e represents an elevation of an acquisition point of the cross section, e p represents an elevation corresponding to a p-th point, and N represents a preset window number.
Preferably, the performing unsupervised clustering on the point-by-point multi-window accumulated difference feature corresponding to the target pavement according to the acquired Kmeans clustering algorithm parameter, to acquire all crack objects in the target pavement, specifically includes:
Performing unsupervised clustering on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a Kmeans clustering algorithm, respectively acquiring the average depth of each clustering category according to the point set with consistent clustering labels, and taking the clustering category with the minimum average depth as a suspected crack category;
forming a suspected crack object according to the connected domain by using the obtained point set binary image corresponding to the suspected crack category;
and selecting all crack objects in the final target pavement from the suspected crack objects.
Preferably, the training the supervised classifier model by using the sample of the known labeling information, to obtain a trained supervised classifier model specifically includes:
Randomly selecting crack sample points and non-crack sample points from marking data sets of different three-dimensional pavements to serve as training samples, wherein the training samples are samples with known marking information;
Performing supervised classifier model training according to the supervised classifier model and the cross section point-by-point multi-window accumulated difference characteristics of each training sample to obtain a trained supervised classifier model;
the supervision classifier model comprises a Support Vector Machine (SVM), a K neighbor classifier (KNN) or a random forest RF classifier.
According to the multi-window accumulated difference crack extraction method provided by the embodiment of the invention, the characteristics that the fluctuation characteristics of pavement textures are stable within a certain range are utilized, the characteristics that the section part with cracks has elevation fluctuation trend are modeled, the noise and background difference are overcome, the acquired templates can be clustered among homologous data or samples among different homologous data are shared, and the threshold value setting in the crack extraction process or the dependence on homologous labeling samples is reduced. Based on the characteristics, the technical route for extracting the three-dimensional pavement data cracks by line scanning is realized, and the three-dimensional pavement non-supervision crack information extraction is realized by utilizing the proposed point-by-point multi-window accumulated difference characteristics and a typical non-supervision machine learning method, such as Kmeans, under the non-supervision condition, namely under the condition of no sample marking at all. Under the condition that a small amount of cracks are marked, the marked information and the proposed point-by-point multi-window accumulated difference features are trained to form a feature template, so that the sharing of different source samples can be realized, the requirement of a supervised machine learning method on sample marking is reduced, and the three-dimensional pavement crack extraction result is more accurately and rapidly obtained. In addition, the proposed point-by-point multi-window accumulated difference feature-based technical route has high adaptability to the existing non-supervision and supervision machine learning method, is introduced by a high-efficiency machine learning method, has the feature of sample sharing capability, can realize large-scale three-dimensional pavement crack extraction under a small number of marked samples, and provides a stable and robust method for actual pavement crack detection.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a multi-window cumulative difference crack extraction method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a line scanning three-dimensional pavement crack extraction route based on point-by-point multi-window cumulative difference features according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of three-dimensional pavement data including two exemplary cracks and a schematic diagram of corresponding crack labeling according to an embodiment of the present invention;
FIG. 4 is a schematic view of two cross-sectional slits and a portion thereof in accordance with an embodiment of the present invention;
FIG. 5 is a schematic diagram of a cross-section point-by-point multi-window cumulative difference feature calculation and a basic abstract model in an embodiment of the invention;
FIG. 6 is a graph showing an example of cross-sectional point-by-point multi-window cumulative difference feature calculation in accordance with an embodiment of the present invention;
FIG. 7 is a graph showing exemplary multi-window cumulative difference features for fracture and fracture points and texture points;
FIG. 8 is a schematic diagram of an unsupervised three-dimensional pavement crack extraction example 1 based on point-by-point multi-window cumulative difference features and kmeans clustering in an embodiment of the present invention;
FIG. 9 is a schematic diagram of an unsupervised three-dimensional pavement crack extraction example 2 based on point-by-point multi-window cumulative difference features and kmeans clustering in an embodiment of the present invention;
FIG. 10 is a schematic diagram of a crack/texture point feature template obtained by training a random forest classifier based on point-by-point multi-window accumulated difference features according to an embodiment of the present invention;
FIG. 11 is a schematic diagram of an example 1 of extraction of a three-dimensional pavement crack of different sources based on point-by-point multi-window accumulated difference features and random forests according to an embodiment of the present invention;
fig. 12 is a schematic diagram of an example 2 of three-dimensional pavement crack extraction of different sources based on point-by-point multi-window accumulated difference features and random forests according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The existing pavement crack detection and information extraction method mainly comprises the following steps: the pavement crack detection method based on the two-dimensional visual characteristics comprises the following steps: the pavement crack detection method based on the two-dimensional visual features mainly obtains pavement data through an optical camera, a video and the like, and analyzes and extracts the pavement data through the edge, gray level characteristics and pavement background differences of pavement cracks. For example, the crack seed region is extracted by using the gray level of the crack lower than the pavement background, and the shape characteristic of the crack is combined for judgment; and acquiring crack information and the like in the two-dimensional data by utilizing an edge detection operator commonly used in image processing in combination with the edge characteristics of the cracks.
The pavement crack extraction method based on the two-dimensional visual characteristics has the following defects: the pavement crack extraction method based on the two-dimensional visual characteristics mainly obtains pavement data through an optical camera, a video and the like, extracts through the gray level or shape difference between the gray level and shape characteristics of the pavement crack and other indexes of the pavement, and judges by extracting a crack seed area by utilizing the gray level of the crack lower than the pavement background and combining the shape characteristics of the crack; and acquiring crack information and the like in the two-dimensional data by utilizing an edge detection operator commonly used in image processing in combination with the edge characteristics of the cracks. The method can not overcome the influence of ambient light, shadow, road surface tire wear marks, oil stains and the like on the marking detection, and has limited applicability.
The pavement crack detection method based on the three-dimensional pavement data comprises the following steps: such methods typically use vehicle-mounted line-scanned three-dimensional data, combined with three-dimensional data accuracy (lateral resolution 1 mm), for detection using the characteristic that the elevation of a road surface crack is typically below the road surface background. For example, in the level of collecting the cross section of the elevation data, the sharp V-shaped characteristic of the crack area in the elevation section data is utilized to obtain a crack seed area, and then the shape characteristic of the crack is combined for judgment; and extracting cracks at the elevation data point cloud level by using a local threshold value or sparse representation method.
The road surface detection method based on the three-dimensional road surface data has the following defects: such methods typically use on-board three-dimensional laser scan data, combined with data accuracy (lateral resolution 1 mm) to detect pavement crack locating information. Although the method can overcome the defect that the traditional two-dimensional gray image method is easily influenced by illumination and shadow, the existing method for detecting cracks by utilizing the three-dimensional characteristics of the pavement also has the following defects: when the precision of the three-dimensional data is high enough (the transverse resolution is 1mm, and the elevation resolution is less than or equal to 0.5 mm), the three-dimensional pavement elevation data contains relatively complex pavement scene information, and not only cracks, but also pavement deformation, marking, repair and pavement textures are contained; in addition, in the high-precision data, different types of pavement diseases or indexes and the like have certain influence on crack extraction. For example, the texture fluctuation in a road surface with thicker texture is similar to the depth characteristic of the crack, and the robustness and practicality of the crack extraction method can be influenced by only utilizing the crack depth characteristic without considering the influence of the road surface construction depth.
In the cross section of line scanning three-dimensional pavement data, cracks are interfered by various factors, and the cracks are not always in clear-edge and V-shaped structures, and more, due to the interference of pavement texture background, many V-shaped structures in the data are not true crack positions. And due to the influence of factors such as driving gesture, deformation disease, pavement material abrasion, crack type, pavement background difference and the like, the cracks in the line scanning three-dimensional pavement data cross section data easily show an asymmetric V-shaped or non-V-shaped structure with unclear edges. Therefore, for line scanning three-dimensional data crack extraction, the traditional template matching and edge detection methods are difficult to obtain ideal crack detection effects.
In addition, at present, a more mature method is still fresh, so that the purpose of extracting the data cracks of the online scanning three-dimensional pavement by the machine learning method is achieved by fully utilizing different source data samples. In practical application, as more and more systems can acquire three-dimensional data, differences of pavement texture backgrounds and differences of crack types contained in the acquired data are gradually highlighted, the existing machine learning method is difficult to extract cracks of different pavement backgrounds and different crack types under the condition that samples are not marked enough, and the defect greatly limits practical application of machine learning and even deep learning in online scanning of three-dimensional pavement data.
Aiming at the problems, the embodiment of the invention firstly provides a point-by-point multi-window accumulated difference characteristic, modeling is performed by utilizing the fluctuation characteristic of pavement texture within a certain range, and the characteristic that the section part with cracks has elevation fluctuation trend, so that the acquired templates can be clustered among homologous data or samples among different homologous data to be shared while the noise and background difference are overcome, and the threshold value setting in the process of extracting the cracks or the dependence on the homologous labeling samples is reduced.
Based on the point-by-point multi-window accumulated difference feature, the technical route for extracting the three-dimensional pavement data cracks is realized, and particularly under the condition of non-supervision (no sample marking at all), the three-dimensional pavement non-supervision crack information extraction is realized by utilizing the proposed point-by-point multi-window accumulated difference feature self-aggregation and a typical non-supervision machine learning method, such as Kmeans. Under the condition that a small number of cracks are marked, the marked information and the proposed point-by-point multi-window accumulated difference features are trained to form a feature template, so that the sharing of different source samples can be realized, the requirement of a supervised machine learning method on sample marking is reduced, and the three-dimensional pavement crack extraction result is more accurately and rapidly obtained.
In addition, the proposed point-by-point multi-window accumulated difference feature-based technical route has high adaptability to the existing non-supervision and supervision machine learning method, the efficient machine learning method is introduced, and features with sample sharing capability are beneficial to realizing large-scale three-dimensional pavement crack extraction under a small number of marked samples, so that a stable and robust method is provided for actual pavement crack detection.
Fig. 1 is a flowchart of a multi-window cumulative difference crack extraction method according to an embodiment of the present invention, as shown in fig. 1, where the method includes:
s1, acquiring three-dimensional pavement data of a target pavement after attitude and deformation information are removed;
firstly, initial three-dimensional pavement data of a target pavement, namely a pavement needing crack detection, is obtained, a line scanning three-dimensional measuring sensor is utilized to collect a series of section contours of the target pavement along a measuring direction, and the series of section contours of the pavement are spliced to obtain the three-dimensional pavement data.
Because of the influence of noise, the abnormal values generated by system abnormality and environment abnormality need to be corrected, the cross section profile is converted from an image space to an object space through a calibration file, and cross section data of the object space are obtained to correct the system error in the measurement system, and the method specifically comprises the following steps:
Processing partial abnormal zero-value noise points of the pavement section profile measured by the three-dimensional measuring sensor due to the interference of the measuring environment to obtain an image space section profile; and the calibration file is utilized to effectively correct systematic errors caused by sensor installation, laser linear arc degree and uneven light intensity in the road surface section profile measured by the three-dimensional measuring sensor, and perform image direction object space conversion to acquire real object space profile information of a target road surface, thereby providing good data input for subsequent crack detection and attribute information extraction thereof.
In addition, in the process of collecting the road surface, the vehicle-mounted three-dimensional system has obvious low-frequency amplitude gesture fluctuation information in the collected original three-dimensional data due to the influences of bump fluctuation of a vehicle, deformation diseases existing on the road surface and the like, and cracks are hidden in the macroscopic change information. In order to reduce the influence of data posture fluctuation on crack extraction and subsequent crack depth information extraction, it is necessary to remove posture fluctuation contained in data by adopting a related algorithm, and a method for removing the posture adopts a three-dimensional pavement data component analysis method provided in the prior art: and removing low-frequency components in the three-dimensional cross section data, and only reserving the sum of the sparse components and the vibration components as input of subsequent processing to obtain the three-dimensional pavement data of which the target pavement is removed by the attitude and deformation information.
On the premise of not losing crack information, the influence of driving fluctuation and pavement deformation on crack detection is reduced, so that crack display and elevation information extraction are facilitated.
S2, calculating the accumulated difference characteristic under each window based on the preset window number and the point elevation by taking the current point as a starting point for each point on the section in the three-dimensional pavement data of the target pavement, and obtaining the point-by-point multi-window accumulated difference characteristic corresponding to the target pavement;
because the line scanning three-dimensional pavement data cross section is interfered by various factors, the cracks are not always in clear-edge and V-shaped structures, and more, because of the interference of pavement texture background, many V-shaped structures in the data are not true crack positions. And due to the influence of factors such as driving gesture, deformation disease, pavement material abrasion, crack type, pavement background difference and the like, the cracks in the cross section data of the line scanning three-dimensional pavement data easily show an asymmetric V-shaped or non-V-shaped structure with unclear edges.
Therefore, for line scanning three-dimensional data crack extraction, traditional template matching and edge detection are difficult to obtain a more ideal crack detection effect. According to the embodiment of the invention, on the basis of three-dimensional pavement data after the attitude and deformation information are removed, the fluctuation characteristics of cross section cracks and textures of the three-dimensional pavement data are modeled and represented by adopting the cross section point-by-point multi-window accumulated difference characteristic, the processed basic unit is the cross section of the line scanning three-dimensional data, the data acquisition principle is more met, the V-shaped and non-V-shaped structures of the cracks in the cross section data in actual conditions are fully considered, the fluctuation characteristics of the pavement textures in a certain range are utilized to be stable, the characteristics of elevation fluctuation trend exist in the section part with the cracks are modeled, the noise and background difference are overcome, the acquired templates can be clustered among homologous data or the samples among different homologous data are shared, and the threshold value setting in the crack extraction process or the dependence on homologous labeling samples is reduced.
S3, performing unsupervised clustering on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a Kmeans clustering algorithm to obtain all crack objects in the target pavement;
or performing supervised classification based on different source samples on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to the trained supervised classifier model, and acquiring all crack objects in the target pavement.
Specifically, after the point-by-point multi-window accumulated difference feature corresponding to the target pavement is extracted, non-supervision clustering can be performed on the point-by-point multi-window accumulated difference feature corresponding to the target pavement by using a Kmeans clustering algorithm according to the clustering property of the point-by-point multi-window accumulated difference feature, so that all crack objects in the target pavement are extracted. And classifying the point-by-point multi-window accumulated difference features corresponding to the target pavement by using the trained supervision classifier model, and extracting all crack objects in the target pavement.
According to the multi-window accumulated difference crack extraction method provided by the embodiment of the invention, the characteristics that the fluctuation characteristics of pavement textures are stable within a certain range are utilized, the characteristics that the section part with cracks has elevation fluctuation trend are modeled, the noise and background difference are overcome, the acquired templates can be clustered among homologous data or samples among different homologous data are shared, and the threshold value setting in the crack extraction process or the dependence on homologous labeling samples is reduced. On the basis of the point-by-point multi-window accumulated difference characteristics, a technical route for extracting the data cracks of the line scanning three-dimensional pavement is realized. The method specifically comprises the step of extracting the three-dimensional pavement non-supervision crack information by utilizing the aggregation of the proposed features and a typical non-supervision machine learning method, such as Kmeans under the non-supervision condition (no sample marking at all). Under the condition that a small amount of cracks are marked, the marked information and the point-by-point multi-window accumulated difference features are trained to form a feature template, so that the sharing of different source samples can be realized, the requirement of a supervised machine learning method on sample marking is reduced, and the three-dimensional pavement crack extraction result is accurately and rapidly obtained. In addition, the proposed point-by-point multi-window accumulated difference feature-based technical route has high adaptability to the existing non-supervision and supervision machine learning method, is introduced by a high-efficiency machine learning method, has the feature of sample sharing capability, can realize large-scale three-dimensional pavement crack extraction under a small number of marked samples, and provides a stable and robust method for actual pavement crack detection.
According to the embodiment of the invention, whether the sample has labeling information or not is determined to be learned by an unsupervised algorithm or classified by a supervised model, and on the basis of the embodiment, preferably, the three-dimensional pavement data of the target pavement after the posture and deformation information is removed is obtained, and the method further comprises the following steps:
modeling and characterizing the fluctuation characteristics of the cross section cracks and the textures of the three-dimensional pavement data by adopting the cross section point-by-point multi-window accumulated difference characteristics based on a plurality of three-dimensional pavement data samples from which the gesture and deformation information are removed, and obtaining the point-by-point multi-window accumulated difference characteristics corresponding to each sample;
after the accumulated difference characteristics of the sections point by point and multiple windows are obtained, various non-supervision and supervision machine learning can be performed on the basis of the characteristics so as to achieve the purpose of extracting crack information in data, and when the selection is actually performed, the non-supervision learning or the supervision learning can be selected according to whether the acquired samples have labeling information or not.
The three-dimensional pavement data sample after the removal of the gesture and deformation information is obtained, the sample is used as a training sample, the point-by-point multi-window accumulated difference characteristic corresponding to each sample is obtained, the obtaining process is the same as that of the point-by-point multi-window accumulated difference characteristic corresponding to the target pavement, and the description is omitted here.
If the labeling information of all samples is unknown, performing unsupervised clustering by utilizing point-by-point multi-window accumulated difference features corresponding to all samples through a Kmeans clustering algorithm to obtain trained Kmeans clustering algorithm parameters;
Specifically, under the condition of no sample labeling at all, the aggregation of the point-by-point multi-window accumulated difference characteristics and a typical non-supervision machine learning method, such as Kmeans, are utilized to realize the non-supervision crack information extraction of the three-dimensional pavement.
If the labeling information of part of the samples in all the samples is known, training the supervised classifier model by using the samples with the known labeling information, and obtaining the trained supervised classifier model.
Specifically, under the condition that labeling information of a small number of samples is known, the labeling information and the point-by-point multi-window accumulated difference features are trained to form a feature template, so that sharing of different source samples can be realized, the requirement of a supervised machine learning method on sample labeling is reduced, and a three-dimensional pavement crack extraction result is accurately and rapidly obtained.
In addition, the proposed point-by-point multi-window accumulated difference feature-based technical route has higher adaptation degree to the existing non-supervision and supervision machine learning method, is introduced by a high-efficiency machine learning method, has the feature of sample sharing capability, can realize large-scale three-dimensional pavement crack extraction under a small number of marked samples, and provides a stable and robust method for actual pavement crack detection.
On the basis of the foregoing embodiment, preferably, performing unsupervised clustering on the point-by-point multi-window accumulated difference feature corresponding to the target road surface according to the acquired Kmeans clustering algorithm parameter, to acquire all crack objects in the target road surface, including:
Performing unsupervised clustering on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a Kmeans clustering algorithm, respectively acquiring the average depth of each clustering category according to the point set with consistent clustering labels, and taking the clustering category with the minimum average depth as a suspected crack category;
forming a suspected crack object according to the connected domain by using the obtained point set binary image corresponding to the suspected crack category;
and selecting all crack objects in the final target pavement from the suspected crack objects.
Specifically, the point-by-point multi-window accumulated difference feature of the three-dimensional pavement data obtained after the gesture and deformation information are removed is combined with Kmeans clustering to perform unsupervised clustering on the point-by-point multi-window accumulated difference feature of the pavement data.
And respectively acquiring the average depth of each clustering category according to the obtained clustering categories and the point sets (namely, how many categories correspond to how many clustering tags) with consistent clustering tag pairs, and taking the category with the minimum average depth as the suspected crack category.
After the post-processing of the cracks, a more determined crack object is selected from the suspected crack objects as an extraction result, and the method specifically comprises the following steps:
The class with the smallest average depth is suspected crack class Ks, the other classes are set to 0, and the point corresponding to the Ks class is marked as 1. And forming an image object according to the connected domain by using the obtained point suspected crack point set binary image as a crack post-processing basic unit. And respectively solving an area mean value A of the suspected crack object, a minimum external rectangular length-width ratio mean value R and an area mean value SA of the adjacent object. The crack post-treatment deletes the object with the area of the object smaller than A and the aspect ratio of the minimum external moment smaller than R, and deletes the object with the area of the object smaller than A and the area of the adjacent object smaller than SA. According to the crack post-processing method, a more determined crack object is selected from the suspected crack objects to serve as an extraction result.
The method is an example of an unsupervised mode implementation of a three-dimensional pavement crack information extraction technical route based on point-by-point multi-window accumulated difference characteristics.
On the premise that no marking information is available, accurate and rapid crack information extraction is completed by utilizing the characteristic of accumulating difference characteristics and low three-dimensional pavement crack depth in a point-by-point multi-window mode, no complex threshold setting is adopted, and simple and effective line scanning three-dimensional pavement crack extraction with universality is realized.
On the basis of the foregoing embodiment, preferably, the training the supervised classifier model by using the sample of known labeling information, to obtain a trained supervised classifier model specifically includes:
Randomly selecting crack sample points and non-crack sample points from marking data sets of different three-dimensional pavements to serve as training samples, wherein the training samples are samples with known marking information;
Performing supervised classifier model training according to the supervised classifier model and the point-by-point multi-window accumulated difference characteristics of each training sample to obtain a trained supervised classifier model;
the supervision classifier model comprises a Support Vector Machine (SVM), a K neighbor classifier (KNN) or a random forest RF classifier.
According to the embodiment of the invention, the point-by-point multi-window accumulated difference characteristics obtained by three-dimensional pavement data samples after attitude deformation information removal are combined with a source A three-dimensional pavement small quantity labeling sample by using a supervision classifier training model (such as a typical random forest RF classifier, a support vector machine SVM or a K neighbor classifier and the like) on the basis of the point-by-point multi-window accumulated difference characteristics.
Wherein both the fractured sample points and the non-fractured sample points are randomly selected from the labeled dataset. The crack marking required by training is unlimited for the crack type, the road background and the acquisition system, and the three-dimensional road surface data containing the crack marking with the transverse sampling interval of 1mm can be used as a training set. The section point-by-point multi-window accumulated differential features and the labeling set, and the obtained supervised classifier training model can provide classification models for a large number of different three-dimensional pavement data.
And using the random forest training model M obtained by the data training of the A as a different source sample supervision classification for the point-by-point multi-window accumulated difference features of the three-dimensional pavement data section of the source B. Namely, for the test data B, the crack extraction of the data B can be completed through a random forest classifier by using the model M trained by the data mark A after the accumulated difference features of the cross section point-by-point multi-window are extracted without any marking and preset conditions of the data.
The method is an example of a supervision classification mode implementation of a three-dimensional pavement crack information extraction technical route based on point-by-point multi-window accumulated difference characteristics, and can be also applied to other typical supervision classifiers, such as a support vector machine and the like.
Based on the section point-by-point multi-window accumulated difference characteristics, the model trained by the data A is utilized to realize the classification of the data B with different sources, so that the sample sharing of the three-dimensional pavement crack marking data set is realized, and the limitation that the traditional supervision classifier can only realize the supervision classification of the same data on the same data sample can be overcome. The effectiveness of accumulating difference features in a point-by-point multi-window mode is demonstrated by two specific technical routes, and clustering is carried out on data without labels according to feature aggregation to obtain crack information; and marking information of different source data can be utilized to realize different source supervision classification results and accurate and rapid line scanning three-dimensional pavement crack extraction.
On the basis of the foregoing embodiment, preferably, the calculating the cumulative difference feature under each window based on the preset window number and the point elevation to obtain the point-by-point multi-window cumulative difference feature corresponding to the target road surface specifically includes:
For each acquisition point of each cross section in the three-dimensional pavement data of the target pavement, taking the current acquisition point as a starting point, and acquiring a point-by-point multi-window accumulated difference characteristic for each cross section based on the elevation of the acquisition point of the cross section and the number of preset windows;
Specifically, the point-to-point multi-window accumulated difference characteristic is calculated by the following expression:
DN＝[d1,d2,…,di,…,dN],i∈[1,2,3,…,N]，
Wherein D N represents a point-by-point multi-window cumulative difference feature, i represents a window range, D i represents a window range threshold, e represents an elevation of an acquisition point of the cross section, e p represents an elevation corresponding to a p-th point, and N represents a preset window number.
Fig. 2 is a schematic diagram of a line scanning three-dimensional pavement crack extraction route based on a point-by-point multi-window cumulative difference feature according to an embodiment of the present invention, fig. 3 is a schematic diagram of two three-dimensional pavement data including a crack and a corresponding crack labeling schematic diagram provided in an embodiment of the present invention, fig. 3 (a) is a schematic diagram of a three-dimensional pavement including a crack, fig. 3 (b) is a schematic diagram of a corresponding crack labeling in fig. 3 (a), fig. 3 (c) is a schematic diagram of another three-dimensional pavement including a crack, fig. 3 (d) is a schematic diagram of a corresponding crack labeling in fig. 3 (c), an influence of gesture deformation information in original data on crack extraction can be seen, fig. 4 is a schematic diagram of two cross-sectional cracks and a partial display thereof, fig. 4 (a) is a schematic diagram of one cross-sectional crack and a partial display thereof, and fig. 4 (b) is another cross-sectional crack and a partial display thereof, from which an actual pavement crack does not always show a simple V-shaped structure, so that a method for realizing cross-sectional information extraction by using a template matching or an edge method is typically easy to cause a large influence of background noise on crack detection. Through the characteristic of the crack in the cross section of the three-dimensional data scanned by the observation line, fig. 5 is a schematic diagram of the cross section point-by-point multi-window accumulated difference characteristic calculation and a schematic diagram of a basic abstract model in the embodiment of the invention, the crack in the cross section is abstracted into the model shown in fig. 5, and the local waveform characteristic of the crack and the texture on the elevation of the cross section is utilized to model the crack/non-crack point.
Specifically, on the basis of the three-dimensional pavement data from which the gestures are removed, taking the current point as a starting point for each point of the section, respectively taking windows of 1,2, 3, …, i, … and N, and calculating the point-by-point multi-window cumulative difference characteristics according to the mode.
N represents a window range threshold value, which can be set according to actual needs, and the value in the embodiment of the invention is 200.
The following is a preferred embodiment of the present invention, and the method specifically comprises the following steps:
(1) Data source
In the embodiment of the technical scheme of the invention, the three-dimensional data of the asphalt pavement containing the crack is taken as an example, and the method for extracting the object-oriented line scanning three-dimensional pavement crack and the attribute thereof is described.
(2) Data preprocessing
Because of the interference of the measuring environment (water stain, oil stain or foreign matter in the area to be measured on the road surface), partial abnormal noise (zero value points) may exist in the acquired data, and the abnormal noise points are replaced by the non-abnormal sampling points close to the central area of the section in the embodiment of the invention; and correcting systematic errors caused by sensor installation, laser line camber and uneven light intensity distribution in the cross section profile of the object measured by the three-dimensional measuring sensor by using the calibration file, and simultaneously converting image space data into object space data. And simultaneously splicing the pretreated series of sections along the driving direction to obtain the three-dimensional data of the asphalt pavement.
(3) Three-dimensional pavement data attitude fluctuation information removal
The low-frequency component in the three-dimensional pavement data is removed by adopting the analysis of the three-dimensional pavement data provided in the prior art, and only the sum of the sparse component and the vibration component is reserved as the input of the subsequent processing. The processing can reduce the influence of three-dimensional data gesture fluctuation acquired by the vehicle-mounted three-dimensional system on crack extraction, and is beneficial to the extraction of subsequent crack depth information. For example, two typical three-dimensional pavement data containing cracks, such as that shown in fig. 3, can be seen in combination with labeling information, and the pose deformation information in the data is detrimental to the visualization and extraction of cracks.
Fig. 6 is a diagram showing an example of calculation of cumulative difference characteristics of cross sections from point to point in a multiple window manner in the embodiment of the present invention, where fig. 6 (a) shows a line-scanned three-dimensional road cross section including a crack, fig. 6 (b) shows a cross section subjected to removal of attitude deformation, fig. 6 (c) shows cumulative difference characteristics of points of a cross section when the windows are 5, 15, 30, 50, 80, and the effect of the section level is shown in fig. 6 (a) (b). Fig. 8 (b) and 9 (b) show three-dimensional road surface data obtained by splicing the road surfaces one by one in cross section and three-dimensional road surface high-frequency data obtained by splicing the road surfaces one by one in cross section and removing the posture and fluctuation information.
(4) Section point-by-point multi-window accumulated difference feature acquisition and characterization
Based on the three-dimensional data with the attitude and fluctuation information removed, taking the current point as a starting point, respectively taking windows of 1,2,3, …, i, … and N (N is adjustable and is set as 200 by default) for each point of the section, and calculating a point-by-point multi-window accumulated difference characteristic D N according to the formula. Fig. 6 (c) illustrates the cumulative difference feature d 5、d15、d30、d50、d80 at each point of the cross section of fig. 6 (b) with the windows 5, 15, 30, 50, 80 removed. When n=200, the multi-window cumulative difference characteristic D N of the section in fig. 6 (b) is shown in fig. 7 (a). Specifically, fig. 7 shows an exemplary graph of multi-window cumulative difference characteristics of fracture and crack and texture points, fig. 7 (a) shows a multi-window cumulative difference characteristic D N (n=200) of the fracture shown in fig. 6 (a), and fig. 7 (b), 7 (c) and 7 (D) show the effect of the D N characteristic pattern visualization of the crack and texture points in the fracture.
Assuming that the cross-section length is L, the 200-dimensional D N features are found for all the first L-N points of the cross-section shown in FIG. 6 (b). And acquiring multi-window cumulative difference characteristics D N of each point for each section in the three-dimensional data according to the mode.
The D N features of the typical fracture and texture points of the section of fig. 7 (b) were selected for visualization as shown in fig. 7 (b), 7 (c) and 7 (e). It can be seen that the window cumulative difference feature obtained by the actual data according to the method satisfies the theoretical model shown in fig. 5.
(5) Unsupervised three-dimensional pavement crack extraction based on point-by-point multi-window accumulated difference features and kmeans clustering
According to the embodiment of the invention, the point-by-point multi-window accumulated difference characteristics of the three-dimensional pavement data obtained after the attitude deformation is removed are combined with Kmeans clustering to perform unsupervised clustering on the pavement data D N characteristics, and the preset clustering number is K (K is required to be greater than 2, and the method defaults to 4). And respectively acquiring average depths of all kinds of clustering categories according to the point set with consistent clustering label pairs, taking the category with the minimum average depth corresponding to the K category labels as suspected crack category Ks, setting other categories as 0, and marking the point corresponding to the Ks category as 1. And forming an image object according to the connected domain by using the obtained point suspected crack point set binary image as a crack post-processing basic unit. And respectively solving an area mean value A of the suspected crack object, a minimum external rectangular length-width ratio mean value R and an area mean value SA of the adjacent object. The crack post-treatment deletes the object with the area of the object smaller than A and the aspect ratio of the minimum external moment smaller than R, and deletes the object with the area of the object smaller than A and the area of the adjacent object smaller than SA. According to the crack post-processing method, a more determined crack object is selected from the suspected crack objects to serve as an extraction result.
Fig. 8 is a schematic diagram of an example 1 of an unsupervised three-dimensional pavement crack extraction based on a point-by-point multi-window cumulative difference feature and kmeans cluster in the embodiment of the present invention, fig. 8 (a) is a line-scanned three-dimensional pavement data depth-to-grayscale map, fig. 8 (b) is three-dimensional pavement data after gesture and deformation information removal, fig. 8 (c) is a suspected crack obtained by feature clustering, fig. 8 (d) is a result of crack post-treatment, fig. 9 is a schematic diagram of an example 2 of an unsupervised three-dimensional pavement crack extraction based on a point-by-point multi-window cumulative difference feature and kmeans cluster in the embodiment of the present invention, fig. 9 (a) shows a line-scanned three-dimensional pavement data depth-to-grayscale map, fig. 9 (b) shows three-dimensional pavement data after gesture deformation removal, fig. 9 (c) shows suspected cracks obtained by feature clustering, fig. 9 (d) shows a result of crack post-treatment, and fig. 8, 9 exemplifies the above processes.
(6) Random forest classifier training based on point-by-point multi-window accumulated difference features
Fig. 10 is a schematic diagram of a crack/texture point feature template obtained by training a random forest classifier based on point-by-point multi-window accumulated differential features according to an embodiment of the present invention, as shown in fig. 10, according to an embodiment of the present invention, for point-by-point multi-window accumulated differential features D N obtained by three-dimensional road data after gesture deformation removal, a random forest RF classifier and a small number of labeling samples of a source a three-dimensional road are combined on the basis, and the random forest classifier training is performed, wherein both crack sample points and non-crack sample points are randomly selected from a labeling data set.
The sample size can be preset according to the requirement, 20% of the crack labeling data are randomly selected for training by the samples in the embodiment, but the ratio of crack sample points to non-crack sample points is required to be smaller than 1:1 and larger than 1:10 so as to meet the practical priori condition that the cracks in the pavement occupy smaller area, and the trained classifier is prevented from extracting too much or too little crack points. The crack marking required by training is unlimited for the type of the crack, the background of the road surface and the acquisition system, and the three-dimensional road surface data containing the crack marking crack with the transverse sampling interval of 1mm can be used as a training set. The feature D N and the labeling set can be used for obtaining a random forest classifier training model, and can be used for providing classification models for a large number of different three-dimensional pavement data.
(7) Different source three-dimensional pavement crack extraction based on point-by-point multi-window accumulated difference features and random forest
And (3) using the random forest training model M obtained by the data training in the step (6) as a different source sample supervision classification for the three-dimensional pavement data D N characteristics of the source B. Namely, for the test data B, the crack extraction of the data B can be completed by using the data A to label the trained model M after extracting the feature D N without any labeling and preset conditions of the data. And (3) carrying out post-processing on the classification result of the data B according to the step (5) due to the differences of the system and the background characteristics, so as to obtain a crack extraction result of the three-dimensional data of the data source B on the basis of a small quantity of labels of the data source A. Based on the feature D N, the model trained by the data A is utilized to realize the classification of the data B with different sources, so that the sample sharing of the three-dimensional pavement crack marking dataset is realized, and the limitation that the traditional supervision classifier can only realize the supervision classification of the same data on the same data sample can be overcome.
Fig. 11 is a schematic diagram of an example 1 of extracting a crack from a three-dimensional pavement of different sources based on a point-by-point multi-window cumulative difference feature and a random forest according to an embodiment of the present invention, where fig. 11 (a) and fig. 11 (B) are a three-dimensional diagram of data a and data after being removed by gesture deformation, fig. 11 (c) is a three-dimensional diagram of data B1 without labels, and fig. 11 (d) is a crack extraction result of data B1 obtained by using a small number of labels of data a; fig. 11 (e) is a three-dimensional view of the data B2 without labeling, and fig. 11 (f) is a crack extraction result of the data B2 obtained with a small number of labeling of the data a.
Fig. 12 is a schematic diagram of an example 2 of extracting a crack from a three-dimensional road surface based on point-by-point multi-window cumulative difference features and random forests according to an embodiment of the present invention, where fig. 12 (a) and fig. 12 (B) are a three-dimensional graph of data a and data after being removed by gesture deformation, fig. 12 (c) is a three-dimensional graph of data B1 without labels, and fig. 12 (d) is a crack extraction result of data B1 obtained by using a small number of labels of data a; fig. 12 (e) is a three-dimensional view of the data B2 without labeling, and fig. 12 (f) is a crack extraction result of the data B2 obtained with a small number of labeling of the data a.
The results of fig. 11 and 12 illustrate the above procedure, with classification results using different sources being superior to the unsupervised clustering results in step (5) from the two different sets of source random forest classification results. Both groups of experiments show the effectiveness of accumulating difference features in a point-by-point multi-window mode, and for data without labels, clustering can be carried out on the data according to the feature aggregation in the step (5) to obtain crack information; and (3) different source supervision classification results can be realized by using the labeling information of different source data according to the step (7), and higher crack extraction precision can be realized.
The technical problems solved by the invention are as follows:
(1) Data source and destination. The invention designs a point-by-point multi-window accumulated difference characteristic which can be suitable for various pavement backgrounds and various crack types by utilizing pavement data acquired based on a line scanning three-dimensional measuring sensor, and performs non-supervision/supervision three-dimensional pavement data crack extraction on the basis.
(2) And acquiring and characterizing the accumulated difference characteristics of the sections point by point and multiple windows. In the cross section of line scanning three-dimensional pavement data, cracks are interfered by various factors, and the cracks are not always in clear-edge and V-shaped structures, and more, due to the interference of pavement texture background, many V-shaped structures in the data are not true crack positions. And due to the influence of factors such as driving gesture, deformation disease, pavement material abrasion, crack type, pavement background difference and the like, the cracks in the line scanning three-dimensional pavement data cross section data easily show an asymmetric V-shaped or non-V-shaped structure with unclear edges.
Therefore, for line scanning three-dimensional data crack extraction, traditional template matching and edge detection are difficult to obtain a more ideal crack detection effect. On the basis of three-dimensional pavement data after the attitude and deformation information are removed, the three-dimensional pavement data cross-section crack and texture fluctuation characteristics are modeled and characterized by adopting the cross-section point-by-point multi-window accumulated difference characteristics.
(3) And (3) carrying out unsupervised three-dimensional pavement crack extraction based on point-by-point multi-window accumulated difference characteristics and kmeans clustering. The traditional unsupervised machine learning is difficult to directly obtain a better effect in online scanning three-dimensional pavement data crack extraction, on one hand, the online scanning three-dimensional data has the influence of factors such as driving gesture deformation diseases, and on the other hand, the traditional template matching and edge detection method has limited applicability to different pavement cracks of different data. On the basis of line scanning three-dimensional data with attitude deformation removed, the method overcomes the differences of different crack types of different pavement backgrounds by utilizing point-by-point multi-window accumulated difference features, can directly acquire crack information from the data by utilizing a widely-used Kmeans clustering method, and can acquire the line scanning three-dimensional pavement crack information more accurately and rapidly without marking information.
(4) Random forest classifier training based on point-by-point multi-window accumulated difference features. The traditional supervised learning method or the deep learning method has extremely high requirement on the homogeneity of training samples and test data, almost cannot achieve the purpose of sharing samples among different data, and can acquire a better supervised classification effect only by using labeled samples from the same data.
The method greatly limits the effect of the machine learning model in the practical application of online scanning three-dimensional pavement crack detection. According to the method, the point-by-point multi-window accumulated difference characteristics of the three-dimensional pavement data after the gesture deformation is removed are obtained, a random forest RF classifier and a small amount of marking samples of the source A three-dimensional pavement are combined on the basis, the random forest classifier training is carried out, and the crack sample points and the non-crack sample points are randomly selected from a marking data set.
The crack marking required by training is unlimited for the type of the crack, the background of the road surface and the acquisition system, and the three-dimensional road surface data containing the crack marking crack with the transverse sampling interval of 1mm can be used as a training set. The obtained random forest classifier training model can provide classification models for a large number of different sources of three-dimensional pavement data.
(5) And (3) extracting different source three-dimensional pavement cracks based on point-by-point multi-window accumulated difference features and random forests. The random forest training model M obtained through the data training of the A is used for carrying out different source sample supervision classification on the DN characteristics of the three-dimensional pavement data of the source B.
Namely, for the test data B, the crack extraction of the data B can be completed by only extracting the characteristics and then marking the trained model M by the data A without any marking and preset conditions of the data.
Based on the proposed features, the model trained by the A data is utilized to realize the B data classification of different sources, so that the sample sharing of the three-dimensional pavement crack marking data set is realized, and the limitation that the traditional supervision classifier can only realize the same data supervision classification on the same data sample can be overcome.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (3)
1. The multi-window cumulative difference crack extraction method is characterized by comprising the following steps of:
acquiring three-dimensional pavement data of the target pavement after the attitude and deformation information are removed;
For each point on a section in the three-dimensional pavement data of the target pavement, calculating the accumulated difference characteristic under each window based on the preset window number and the point elevation by taking the current point as a starting point to obtain the point-by-point multi-window accumulated difference characteristic corresponding to the target pavement;
Performing unsupervised clustering on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a Kmeans clustering algorithm to obtain all crack objects in the target pavement;
Or performing supervised classification based on different source samples on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to the trained supervised classifier model to obtain all crack objects in the target pavement;
the method for calculating the accumulated difference feature under each window based on the preset window number and the point elevation to obtain the point-by-point multi-window accumulated difference feature corresponding to the target pavement specifically comprises the following steps:
For each acquisition point of each cross section in the three-dimensional pavement data of the target pavement, taking the current acquisition point as a starting point, and acquiring a point-by-point multi-window accumulated difference characteristic for each cross section based on the elevation of the acquisition point of the cross section and the number of preset windows;
Specifically, the point-to-point multi-window accumulated difference characteristic is calculated by the following expression:
DN＝[d1,d2,…,di,…,dN],i∈[1,2,3,…,N]，
Wherein D N represents a point-by-point multi-window cumulative difference feature, i represents a window range, D i represents a window range threshold, e represents the elevation of an acquisition point of the cross section, e p represents the elevation corresponding to the p-th point, and N represents a preset window number;
the method for obtaining the three-dimensional pavement data of the target pavement after the removal of the attitude and deformation information comprises the following steps:
Modeling and characterizing the fluctuation characteristics of the cross section cracks and textures of the three-dimensional pavement data by adopting point-by-point multi-window accumulated difference characteristics based on a plurality of three-dimensional pavement data samples from which the gesture and deformation information are removed, and obtaining point-by-point multi-window accumulated difference characteristics corresponding to each sample;
if the labeling information of all samples is unknown, performing unsupervised clustering by utilizing point-by-point multi-window accumulated difference features corresponding to all samples through a Kmeans clustering algorithm to obtain trained Kmeans clustering algorithm parameters;
If the labeling information of part of the samples in all the samples is known, training the supervised classifier model by using the samples with the known labeling information, and obtaining the trained supervised classifier model.
2. The method for extracting the crack of the multi-window accumulated difference according to claim 1, wherein the non-supervised clustering is performed on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to the Kmeans clustering algorithm parameters, so as to obtain all the crack objects in the target pavement, and the method specifically comprises the following steps:
Performing unsupervised clustering on the point-by-point multi-window accumulated difference features corresponding to the target pavement according to a Kmeans clustering algorithm, respectively acquiring the average depth of each clustering category according to the point set with consistent clustering labels, and taking the clustering category with the minimum average depth as a suspected crack category;
forming a suspected crack object according to the connected domain by using the obtained point set binary image corresponding to the suspected crack category;
and selecting all crack objects in the final target pavement from the suspected crack objects.
3. The method for extracting the multi-window cumulative difference crack according to claim 1, wherein the training the supervised classifier model by using the samples of the known labeling information, and obtaining the trained supervised classifier model, specifically comprises:
Randomly selecting crack sample points and non-crack sample points from marking data sets of different three-dimensional pavements to serve as training samples, wherein the training samples are samples with known marking information;
Performing supervised classifier model training according to the supervised classifier model and the point-by-point multi-window accumulated difference characteristics of each training sample to obtain a trained supervised classifier model;
the supervision classifier model comprises a Support Vector Machine (SVM), a K neighbor classifier (KNN) or a random forest RF classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010652509.XA CN111882664B (en) | 2020-07-08 | 2020-07-08 | Multi-window accumulated difference crack extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010652509.XA CN111882664B (en) | 2020-07-08 | 2020-07-08 | Multi-window accumulated difference crack extraction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111882664A CN111882664A (en) | 2020-11-03 |
| CN111882664B true CN111882664B (en) | 2024-08-09 |
Family
ID=73150846
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010652509.XA Active CN111882664B (en) | 2020-07-08 | 2020-07-08 | Multi-window accumulated difference crack extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111882664B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112785548B (en) * | 2020-12-29 | 2023-07-18 | 中央财经大学 | Road surface crack detection method based on vehicle-mounted laser point cloud |
| CN113506257B (en) * | 2021-07-02 | 2022-09-20 | 同济大学 | Crack extraction method based on self-adaptive window matching |
| CN113658345B (en) * | 2021-08-18 | 2024-12-06 | 杭州海康威视数字技术股份有限公司 | A sample labeling method and device |
| CN115410342B (en) * | 2022-08-26 | 2023-08-11 | 安徽省地质矿产勘查局332地质队 | Landslide hazard intelligent early warning method based on real-time monitoring of crack meter |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105334219A (en) * | 2015-09-16 | 2016-02-17 | 湖南大学 | Bottleneck defect detection method adopting residual analysis and dynamic threshold segmentation |
| CN111223183A (en) * | 2019-11-14 | 2020-06-02 | 中国地质环境监测院 | Landslide terrain detection method based on deep neural network |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013020143A1 (en) * | 2011-08-04 | 2013-02-07 | University Of Southern California | Image-based crack quantification |
| JP6099479B2 (en) * | 2013-05-21 | 2017-03-22 | 大成建設株式会社 | Crack detection method |
| CN107154040B (en) * | 2017-05-08 | 2021-03-23 | 重庆邮电大学 | A method for detecting cracks in tunnel lining surface images |
| CN108229562B (en) * | 2018-01-03 | 2020-07-07 | 重庆亲禾智千科技有限公司 | Method for obtaining classification condition of concrete pavement damage |
| CN108564569B (en) * | 2018-03-23 | 2019-11-26 | 石家庄铁道大学 | A kind of distress in concrete detection method and device based on multicore classification learning |
| CN108765376B (en) * | 2018-05-03 | 2022-03-22 | 武汉光谷卓越科技股份有限公司 | Linear scanning three-dimensional pavement data component analysis method |
| CN109029381B (en) * | 2018-10-19 | 2021-04-06 | 石家庄铁道大学 | Tunnel crack detection method and system and terminal equipment |
| CN110132237A (en) * | 2019-05-05 | 2019-08-16 | 四川省地质工程勘察院 | A kind of method of urban ground deformation disaster EARLY RECOGNITION |
| CN110111322A (en) * | 2019-05-13 | 2019-08-09 | 招商局重庆交通科研设计院有限公司 | A kind of tunnel defect identifying system based on image |
| CN110378879B (en) * | 2019-06-26 | 2021-03-02 | 杭州电子科技大学 | Bridge crack detection method |
| CN110909628A (en) * | 2019-11-05 | 2020-03-24 | 长安大学 | A Natural Light Compensation Method for Detecting Cracks in Pavement with Shadows |
| CN111310558B (en) * | 2019-12-28 | 2023-11-21 | 北京工业大学 | An intelligent extraction method for pavement defects based on deep learning and image processing |
-
2020
- 2020-07-08 CN CN202010652509.XA patent/CN111882664B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105334219A (en) * | 2015-09-16 | 2016-02-17 | 湖南大学 | Bottleneck defect detection method adopting residual analysis and dynamic threshold segmentation |
| CN111223183A (en) * | 2019-11-14 | 2020-06-02 | 中国地质环境监测院 | Landslide terrain detection method based on deep neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111882664A (en) | 2020-11-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111882664B (en) | Multi-window accumulated difference crack extraction method | |
| Akagic et al. | Pavement crack detection using Otsu thresholding for image segmentation | |
| CN110264448B (en) | A method of insulator fault detection based on machine vision | |
| CN109804232B (en) | A detection method for crack development degree of asphalt pavement based on infrared thermal image analysis | |
| Ouma et al. | Pothole detection on asphalt pavements from 2D-colour pothole images using fuzzy c-means clustering and morphological reconstruction | |
| CN109101924B (en) | A method for road traffic sign recognition based on machine learning | |
| AU2016385541B2 (en) | Object surface deformation feature extraction method based on line scanning three-dimensional point Cloud | |
| Oliveira et al. | Automatic road crack detection and characterization | |
| CN109978032B (en) | Bridge crack detection method based on space pyramid cavity convolution network | |
| CN108520514B (en) | Consistency detection method of printed circuit board electronic components based on computer vision | |
| CN110288571B (en) | High-speed rail contact net insulator abnormity detection method based on image processing | |
| Mathavan et al. | Use of a self-organizing map for crack detection in highly textured pavement images | |
| CN103761743B (en) | A kind of solid wooden floor board detection method of surface flaw based on image co-registration segmentation | |
| CN110473187B (en) | Object-oriented line scanning three-dimensional pavement crack extraction method | |
| CN101718528B (en) | Digital image based rapid solving method of circle parameters | |
| CN105825169B (en) | A Pavement Crack Recognition Method Based on Road Image | |
| CN110910350B (en) | Nut loosening detection method for wind power tower cylinder | |
| CN110276747B (en) | A method for fault detection and fault rating of insulators based on image analysis | |
| Xie et al. | Fabric defect detection method combing image pyramid and direction template | |
| Yun et al. | Crack recognition and segmentation using morphological image-processing techniques for flexible pavements | |
| Bosurgi et al. | An automatic pothole detection algorithm using pavement 3D data | |
| CN109886932A (en) | Surface defect detection method of wheel speed sensor ring gear based on SVM | |
| Daniel et al. | Automatic road distress detection and analysis | |
| Ghanta et al. | Automatic road surface defect detection from grayscale images | |
| CN112308819A (en) | Pore identification method based on multi-feature gray level image of scanning electron microscope |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 430223 No.6, 4th Road, Wuda Science Park, Donghu high tech Zone, Wuhan City, Hubei Province Applicant after: Wuhan Optical Valley excellence Technology Co.,Ltd. Address before: 430223 No.6, 4th Road, Wuda Science Park, Donghu high tech Zone, Wuhan City, Hubei Province Applicant before: Wuhan Wuda excellence Technology Co.,Ltd. Address after: 430223 No.6, 4th Road, Wuda Science Park, Donghu high tech Zone, Wuhan City, Hubei Province Applicant after: Wuhan Wuda excellence Technology Co.,Ltd. Address before: 430223 No.6, 4th Road, Wuda Science Park, Donghu high tech Zone, Wuhan City, Hubei Province Applicant before: WUHAN WUDA ZOYON SCIENCE AND TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |