CN110926359A - A three-dimensional crack surface contour detection method for optimized two-phase scanning pavement - Google Patents
A three-dimensional crack surface contour detection method for optimized two-phase scanning pavement Download PDFInfo
- Publication number
- CN110926359A CN110926359A CN201911069593.6A CN201911069593A CN110926359A CN 110926359 A CN110926359 A CN 110926359A CN 201911069593 A CN201911069593 A CN 201911069593A CN 110926359 A CN110926359 A CN 110926359A
- Authority
- CN
- China
- Prior art keywords
- matrix
- data
- crack
- data matrix
- dimensional
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
本发明公开了一种优化双相扫描路面的三维裂缝曲面轮廓检测方法，首先对路面三维图像数据矩阵进行滤波处理，得到去燥后的三维图像数据矩阵I；将滤波后的三维数据矩阵按行排列，对矩阵中每行数据批量进行水平处理，得到数据矩阵I1，再对矩阵中每列数据批量进行垂直处理，得到数据矩阵I2，然后将两个数据矩阵I1和I2,相加，得到矩阵I3；过滤矩阵I3，获得仅包含裂缝信息I4的二进制图像；最后进行裂缝检测定性分析和ROC曲线分析。该方法以曲线拟合与平滑操作相结合，利用平滑矩阵作为滑动窗算子对裂缝数据矩阵进行滤波，同时进行曲线拟合，得到PSP曲线。可以提高获得路面曲面的精度。The invention discloses a three-dimensional crack surface contour detection method for optimizing two-phase scanning pavement. First, a three-dimensional image data matrix of the road is filtered to obtain a three-dimensional image data matrix I after de-drying; the filtered three-dimensional data matrix is divided into rows. Arrange, perform horizontal processing on each row of data in the matrix to obtain a data matrix I 1 , and then perform vertical processing on each column of data in the matrix to obtain a data matrix I 2 , and then combine the two data matrices I 1 and I 2 , phase Add to obtain matrix I 3 ; filter matrix I 3 to obtain a binary image containing only crack information I 4 ; finally, perform qualitative analysis of crack detection and ROC curve analysis. The method combines curve fitting and smoothing operation, uses smoothing matrix as a sliding window operator to filter the fracture data matrix, and performs curve fitting at the same time to obtain the PSP curve. The accuracy of obtaining the road surface surface can be improved.
Description
Technical Field
The invention belongs to the technical field of road engineering, relates to a method for detecting a road surface, and particularly relates to a method for detecting a three-dimensional crack curved surface profile of an optimized two-phase scanning road surface.
Background
With the vigorous development of the highway industry, the traditional pavement crack two-dimensional detection technology cannot meet the requirements of pavement management systems, intelligent traffic systems and increasingly perfect pavement evaluation technology development, and the pavement three-dimensional detection technology is becoming the development direction of pavement detection. In recent decades, researchers at home and abroad always pursue an efficient and accurate automatic detection and identification method for pavement cracks. With the progress of science and technology and the improvement of the requirements for detecting pavement cracks, improved crack detection algorithms are continuously appeared. In crack detection technologies at home and abroad, a CCD (charge coupled device) camera is mostly adopted to obtain a road surface image, and then the acquired two-dimensional image is subjected to subsequent processing to identify cracks. Due to the limitation of hardware conditions of a two-dimensional image acquisition system and the influence of external illumination, the acquired image has low contrast between cracks and a background, the information amount is small, a large amount of randomly distributed noise, oil stains, tree shadows, black spots and other impurities exist, the crack identification is greatly interfered, the subsequent algorithm has large calculation amount, low accuracy and poor stability, and the processing of a large amount of pavement images is not facilitated. The three-dimensional crack recognition technology is a new field, compared with a two-dimensional image crack recognition technology, the data used by the method is three-dimensional data, and the acquired data is more sufficient and accurate because the acquisition of the three-dimensional data has the remarkable advantage of being not influenced by illumination, oil stains and the like. Therefore, the accuracy and the stability of the three-dimensional crack identification are better. Currently, the three-dimensional crack recognition technology is studied as follows: huang et al propose a multi-scale fracture classification method based on fracture fundamental elements (CFE); tsai et al propose a 3D laser technology-based method for accurately evaluating asphalt pavement cracks under different illumination and low contrast conditions; peng et al propose a three-dimensional road surface image multi-target clustering automatic crack detection method based on 1mm resolution.
The existing three-dimensional detection technology for pavement cracks is to acquire depth information of cracks based on optical three-dimensional information. However, the image-based pavement crack detection methods that have been developed today, either based on two-dimensional images or using captured 3D images, are often less than satisfactory for crack detection results.
Because the precision of the existing three-dimensional crack detection method is relatively low and the high-precision requirement of people on crack collection is met, the research and development of a new pavement three-dimensional crack image detection method for improving the crack detection precision is one of the subjects concerned by the applicant.
Disclosure of Invention
Aiming at the defects or shortcomings existing in the process of collecting data by utilizing laser during vehicle driving, the invention aims to provide a three-dimensional crack curved surface contour detection method for optimizing a two-phase scanning road surface.
In order to realize the task, the invention adopts the following technical solution:
a three-dimensional crack curved surface contour detection method for optimizing a two-phase scanning road surface is characterized in that firstly, a road surface three-dimensional image data matrix is subjected to filtering processing to obtain a three-dimensional image data matrix I after drying is removed; arranging the filtered three-dimensional data matrix I according to rows, and horizontally processing each row of data in the three-dimensional data matrix I in batches to obtain a data matrix I1Then, each line of data in the three-dimensional data matrix I is processed vertically in batches to obtain the data matrix I2Then two data matrices I1And I2Are added to obtain a matrix I3(ii) a Filter matrix I3Obtaining information I containing only cracks4A binary image of (a); and finally, carrying out fracture detection qualitative analysis and ROC curve analysis.
According to the invention, the filter matrix I3The method comprises the following steps:
1) labeling matrix I using 8 neighborhood principle3: for matrix I3Scanning the data points from top to bottom and from left to right, searching unmarked data points, and distributing new marks for the data points; then it is determined whether 8 neighborhoods of each current center point are fracture points. If there is an unmarked crack data point in the region, the crack data point is classified into the region of the current scanning center point, and the mark of the current center point is assigned to the newly found crack data pointCrack data points, updating the center to a new crack point, scanning again and judging 8 neighborhoods of the new center until the data matrix I3All points in (a) are marked;
2) for each marked region, the zero-order distance is first calculated to obtain the area of the region, and then the value of the region is evaluated, the region smaller than the threshold value represents an isolated region which should be regarded as a noise region and is assigned a value of 0 in the filter matrix.
Further, the two data matrixes I1And I2The calculation of the addition is as follows:
calculating a data matrix I1And a data matrix I2Thereby obtaining InInitial matrix of representations I3(ii) a For those points that satisfy the following condition: present in the data matrix I1Or a data matrix I2Is absent from InIn (1), calculating a data matrix I1Or a data matrix I2And InThe shortest distance between the two matrixes is compared with a specified threshold value, and if the shortest distance does not exceed the threshold value, the result is stored in the matrix I3In (1).
The three-dimensional crack curved surface contour detection method for optimizing the two-phase scanning pavement combines curve fitting with smoothing operation, utilizes the smoothing matrix as a sliding window operator to filter a crack data matrix, and simultaneously carries out curve fitting to obtain a PSP curve. The precision of obtaining the curved surface of the road surface can be improved.
Drawings
FIG. 1 is a schematic view of an unbalanced-drive vehicle;
fig. 2 is a graph showing three lateral crack detection results, in which (a) is a 3D display of three-dimensional pavement crack data, (b) is an image obtained by a single two-phase scanning pavement crack detection method, and (c) is an image obtained by a crack detection algorithm based on psp (primary surface profile).
Fig. 3 is a graph of three longitudinal crack detection results, in which (a) is three-dimensional pavement crack data, (b) is an image obtained by a single two-phase scanning pavement crack detection method, and (c) is an image obtained by a crack detection algorithm based on psp (primary Surface profile).
Fig. 4 is a graph of three kinds of mesh crack detection results, in which (a) is three-dimensional road crack data, (b) is an image obtained by a single two-phase scanning road crack detection method, and (c) is an image obtained by a crack detection algorithm based on psp (primary Surface profile).
FIG. 5 is a ROC plot, wherein (a) is a transverse fracture plot, (b) is a longitudinal fracture plot, and (c) is a reticular fracture plot.
The present invention will be described in further detail with reference to the following drawings and examples.
Detailed Description
There are certain problems in using lasers to collect data during vehicle driving. A significant key issue is that the automated pavement performance data collection process may change frequently. These variations are mainly due to two reasons, namely:
(1) unbalanced driving of the vehicle (fig. 1), which will result in a road inclination in the collected data set;
(2) positive and negative noise caused by the laser effect. Specifically, positive noise refers to noise generated on a road by reflected laser light, and then captured by a 3D camera. Negative noise refers to a phenomenon of laser extinction, which often occurs when data is collected in a narrow area (Liet al 2009).
In view of these issues, variations in the data collection process should be considered in developing the crack detection algorithm. Therefore, the applicant designed a crack detection algorithm based on psp (primary Surface profile).
The algorithm can be described as the information containing two main components in the pavement crack image data matrix, namely: background information and fracture information. Typically, the values representing the background information are much higher than the areas of disruption in a certain data matrix. According to this feature, the horizontal direction (or the vertical direction) of the road surface curve can be obtained from the road surface crack image, and then these principal surface profiles are used as a reference for detecting the severity of cracks appearing in the image. After obtaining the PSP curves, calculating the difference between the relative values of the raw data and the corresponding PSP curves may provide actual road surface crack information.
The embodiment provides a three-dimensional crack curved surface contour detection method for optimizing a two-phase scanning road surface, which specifically comprises the following steps:
a first part: image information acquisition and processing
Step 1: acquiring pavement three-dimensional image data by using laser three-dimensional imaging equipment, and filtering the acquired pavement three-dimensional image data matrix to obtain a three-dimensional image data matrix I after drying is removed;
step 2: arranging the filtered three-dimensional data matrix I according to rows, and horizontally processing each row of data in the data matrix I in batches to obtain the data matrix I1；
Step 2.1: using a sliding window of size 3 x 3 on the ith row R of the data matrix IiPerforming smoothing operation (equation 1) to obtain a smoothed row data matrix R'i(formula 2); then, for the smoothed row data matrix R ″iPerforming curve fitting operation to obtain a data matrix R ″i(formula 3); then, a matrix R with the original data is creatediArray Pro of the same sizei(equation 4), the main surface profile data value is stored.
Ri＝(Zi1，Zi2，......，Zin)，i＝1，2，......，m (1)
R′i＝(Z′i1，Z′i2，......，Z′in)，i＝1，2，......，m (2)
R″i＝(Z″i1，Z″i2，......，Z″in)，i＝1，2，......，m (3)
Proi＝(proi1，proi2，......，proin)，i＝1，2，......，m (4)
Where m, n represent the rows and columns of the data matrix, respectively.
Step 2.2: the difference between the smoothed value and the fitted value is calculated using the following equation 5:
wherein, Z'ijAnd Z ″)ijRespectively representing the smoothed data matrix and the fitted data matrix. SiRepresenting the difference between the smoothed data and the fitted data in row i.
Step 2.3: the absolute difference between the smoothed value and the fitted value is calculated using the following equation (6). The significance of Δ is then evaluated using equation (7), and if the value satisfies this criterion, column J is marked and stored into the new array J.
Δ＝|Z′ij-Z″ij| (6)
Δ＞5×Si(7)
Where Δ represents the absolute difference between the smoothed data matrix and the fitted data matrix.
Step 2.4: and the column numbers in the array J are continuously divided and expanded according to the values of the column numbers, and the column numbers in the columns are sorted. The left and right ends extend to the intersection of the curve fitting result and the smoothing result, and the expanded column region is stored in the array J again.
The left and right ends of the column number should satisfy the conditions of formula (8) and formula (9), respectively:
(Z′i(a-1)-Z″i(a-1))×(Z′ia-Z″ia)≤0，a＝1，2，......，m (8)
(Z′ia-Z″ia)×(Z′i(a+1)-Z″i(a+1))≤0，a＝1，2，......，m (9)
wherein, Z'i(a-1)，Z′iaAnd Z'i(a+1)And three neighborhoods, Z ", in the smoothed data matrixi(a-1)，Z″iaAnd Z ″)i(a+1)Representing three neighborhoods in the fitted data matrix;
step 2.5: the data values are assigned to the principal surface profile matrix using equation (10). If a particular column is stored in array J, then the particular value in J should be replaced with a curve fit value, otherwise, the smoothed value is used as the main surface contour value.
Wherein, Z'ij，Z″ijThe smoothed and fitted values are indicated separately.
Step 2.6: calculation of the raw data matrix R using equation (11)iAnd a major surface profile value ProijThe standard deviation between, and is denoted as Ti：
Wherein Z isijIndicating the position of the original data value, proijRepresenting the corresponding position value in the PSP curve.
Step 2.7: r was evaluated using the following formula (12)iThe data of (1). If the result satisfies this condition, the point is marked as a crack and the data value is assigned to 1, otherwise it is considered as background and set to 0.
|Zij-proij|≥k×Ti(12)
Where K represents the parameter identified by the row, and according to the experiments performed herein, the best performance can be obtained when K selects a value of 2 to 5.
Carrying out the same operation on each data point in the data matrix I to obtain a matrix I1The horizontal processing results are shown.
And step 3: operating from the vertical direction (column direction), similarly to step 2, a further data matrix I can be obtained2And vertical information representing image data.
And 4, step 4: two data matrixes I1And I2Are added to obtain a matrix I3。
Step 4.1: calculating a data matrix I1And a data matrix I2Thereby obtaining InInitial matrix of representations I3。
Step 4.2: for those points that satisfy the following condition: exist in I1Or I2In but not storeIs characterized in thatnIn, calculate I1Or I2And InThe shortest distance between them, and comparing the shortest distance with a specified threshold. If the shortest distance does not exceed the threshold, the result is stored in matrix I3In (1).
And 5: the matrix I obtained in the step 43Carrying out filtering operation;
step 5.1: tagging I Using 8 neighborhood principles3For matrix I3The data points in (1) are scanned from top to bottom and from left to right, the data points which are not marked are searched, and new marks are distributed to the data points. Then it is determined whether 8 neighborhoods of each current center point are fracture points. If there is an unmarked fracture data point in the region, the fracture data point is classified into the region where the current scanning center point is located, and the mark of the current center point is assigned to the newly found fracture data point. And updating the center to a new crack point, and scanning again and judging 8 neighborhoods of the new center until the matrix I3All points in (a) are marked.
Step 5.2: for each marked region, the zero-order distance is first calculated, resulting in the area of the region. The value of this region is then evaluated, the region smaller than the threshold representing an isolated region which should be considered as a noise region, and is assigned a value of 0 in the filter matrix.
Step 6: finishing the treatment to obtain the product containing only the crack information I4The binary image of (2).
A second part: qualitative analysis of crack detection results and ROC curve analysis of crack detection results
(1) Qualitative analysis of crack detection results
To evaluate the performance of the proposed crack identification method, three types of pavement crack samples were selected in this example: i.e., transverse fractures, longitudinal fractures, and map fractures. Therefore, a single two-phase scanning pavement crack detection method and a crack detection algorithm based on PSP (Primary Surface Profile) are respectively adopted to carry out comparison tests on all three cracks. The processing results are shown in fig. 2, 3 and 4. Wherein fig. 2(a), 3(a) and 4(a) depict a 3D display of pavement crack data; FIG. 2(b), FIG. 3(b) and FIG. 4(b) depict crack detection results of a single two-phase scanning pavement crack detection method; fig. 2(c), 3(c) and 4(c) depict crack detection results using a psp (primary Surface profile) -based crack detection algorithm.
(2) ROC curve analysis of crack detection results
Another widely used performance evaluation index is the Receiver Operating Characteristic (ROC) curve, which can fully evaluate the performance of the method. ROC is plotted using FPR (false positive rate) as an axis value, and TPR (true positive rate) as an axis value. FPR can be calculated by equation (15), and TPR (also referred to as Recall) can be calculated by equation (14). Generally, as FPR increases, TPR also increases. The closer the TPR is to the value 1, the better the performance of the method. Specifically, the method can be considered to have perfect performance if it can obtain a final TPR value equal to 1 when the FPR becomes 1.
Fig. 5 shows ROC graphs of detection results of a transverse crack (fig. 5a), a longitudinal crack (fig. 5b) and a reticular crack (fig. 5c), which are respectively detected by the three-dimensional crack curved surface profile detection method for the optimized two-phase scanning pavement and the single two-phase scanning pavement crack detection method of the embodiment. The ROC curves show that the three-dimensional crack curved surface contour detection method for optimizing the two-phase scanning pavement provided by the embodiment can improve the detection result of each crack, and particularly, compared with the transverse crack detection, the detection result of each crack is improved obviously for longitudinal cracks and reticular cracks.
TP: predicting the crack (P) and actually predicting the correct crack (T), namely judging the correct rate of the crack;
TN: predicting as a background (N), and actually predicting as correct (T), namely judging as the correct rate of the background;
FP: predicting as a crack (P), actually predicting as an error (F), and judging the background as a crack by a false alarm rate;
FN: the prediction is background (N), actually, the prediction is wrong (F), and the report missing rate refers to the crack judgment as background.
In conclusion, the single two-phase scanning pavement crack detection method can obtain satisfactory detection accuracy, but the recall rate in the experimental result is not satisfactory. The reason for this is to ignore factors of an inclined road surface which can cause variations in the collected data set. The PSP curve is defined as a reference value during the double-phase crack detection, so that the method has stronger anti-interference performance on the change of a data set, and the overall crack detection result is improved. Therefore, the three-dimensional crack curved surface contour detection method for optimizing the two-phase scanning pavement provided by the embodiment has better possibility, and can be integrated into an automatic pavement distress detection system to facilitate the system management of the whole pavement.
Claims (3)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911069593.6A CN110926359B (en) | 2019-11-05 | 2019-11-05 | A three-dimensional crack surface contour detection method for optimized two-phase scanning pavement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911069593.6A CN110926359B (en) | 2019-11-05 | 2019-11-05 | A three-dimensional crack surface contour detection method for optimized two-phase scanning pavement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110926359A true CN110926359A (en) | 2020-03-27 |
| CN110926359B CN110926359B (en) | 2021-10-29 |
Family
ID=69852296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911069593.6A Expired - Fee Related CN110926359B (en) | 2019-11-05 | 2019-11-05 | A three-dimensional crack surface contour detection method for optimized two-phase scanning pavement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110926359B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4796998A (en) * | 1985-10-03 | 1989-01-10 | Pasco Corporation | Method for mobile survey of road surface |
| CN1563891A (en) * | 2004-04-20 | 2005-01-12 | 长安大学 | System and method for discriminating road gap |
| CN103872983A (en) * | 2014-04-04 | 2014-06-18 | 天津市鑫鼎源科技发展有限公司 | Device and method for detecting defects on surface of solar cell |
| CN104504388A (en) * | 2014-12-17 | 2015-04-08 | 长安大学 | Pavement crack identification and feature extraction algorithm and system |
| CN104574393A (en) * | 2014-12-30 | 2015-04-29 | 北京恒达锦程图像技术有限公司 | Three-dimensional pavement crack image generation system and method |
| CN105654088A (en) * | 2016-03-07 | 2016-06-08 | 西安交通大学 | Tobacco stem extracting method based on local symmetric relation |
-
2019
- 2019-11-05 CN CN201911069593.6A patent/CN110926359B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4796998A (en) * | 1985-10-03 | 1989-01-10 | Pasco Corporation | Method for mobile survey of road surface |
| CN1563891A (en) * | 2004-04-20 | 2005-01-12 | 长安大学 | System and method for discriminating road gap |
| CN103872983A (en) * | 2014-04-04 | 2014-06-18 | 天津市鑫鼎源科技发展有限公司 | Device and method for detecting defects on surface of solar cell |
| CN104504388A (en) * | 2014-12-17 | 2015-04-08 | 长安大学 | Pavement crack identification and feature extraction algorithm and system |
| CN104574393A (en) * | 2014-12-30 | 2015-04-29 | 北京恒达锦程图像技术有限公司 | Three-dimensional pavement crack image generation system and method |
| CN105654088A (en) * | 2016-03-07 | 2016-06-08 | 西安交通大学 | Tobacco stem extracting method based on local symmetric relation |
Non-Patent Citations (5)
| Title |
|---|
| 冯笑然: "Noise Robust Method for Analytically Solvable Chaotic Signal Reconstruction", 《CIRCUITS SYSTEMS AND SIGNAL PROCESSING》 * |
| 孙朝云，赵海伟，李伟，郝雪丽，呼延菊: "基于双相扫描检测的路面三维裂缝识别方法", 《中国公路学报》 * |
| 孙朝云，邹鹏，李伟，郝雪丽: "基于OpenGL路面裂缝三维重构技术研究", 《中外公路》 * |
| 宋宏勋，马建，王建锋，崔国丽: "基于双相机立体摄影测量的路面裂缝识别方法", 《中国公路学报》 * |
| 李伟，呼延菊，沙爱民，孙朝云，郝雪丽: "基于3D数据和双尺度聚类算法的路面裂缝检测", 《华南理工大学学报（自然科学版）》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110926359B (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113109368B (en) | Glass crack detection method, device, equipment and medium | |
| Crognale et al. | Damage detection with image processing: a comparative study | |
| CN104504388A (en) | Pavement crack identification and feature extraction algorithm and system | |
| JP2008158958A (en) | Road surface determination method and road surface determination device | |
| JP2008286725A (en) | Person detector and detection method | |
| Abbas et al. | Automated pavement distress detection using image processing techniques | |
| CN117710399B (en) | Crack contour extraction method in geological survey based on vision | |
| CN111524154B (en) | Image-based tunnel segment automatic segmentation method | |
| CN115294159A (en) | Method for dividing corroded area of metal fastener | |
| JPH11351862A (en) | Forward vehicle detection method and device | |
| CN117853484A (en) | Intelligent bridge damage monitoring method and system based on vision | |
| Ryu et al. | Feature-based pothole detection in two-dimensional images | |
| CN113513991A (en) | Battery pole piece burr height detection method and device | |
| CN108563994A (en) | A kind of parking position recognition methods based on image similarity degree | |
| CN119295494B (en) | Intelligent identification and maintenance method and system for potholes on road and bridge surfaces based on computer vision | |
| CN112508913A (en) | Cable section edge detection method based on image detection | |
| CN119006505B (en) | Automatic detection method and system for highway construction quality | |
| CN113066041A (en) | A pavement crack detection method based on stack sparse self-encoding deep learning | |
| CN110926359B (en) | A three-dimensional crack surface contour detection method for optimized two-phase scanning pavement | |
| CN107832732B (en) | Lane line detection method based on ternary tree traversal | |
| CN119313629A (en) | A visual inspection system and method based on automobile mold | |
| CN113947711A (en) | Dual-channel flame detection algorithm for inspection robot | |
| CN116912805B (en) | Well lid abnormity intelligent detection and identification method and system based on unmanned sweeping vehicle | |
| Kembhavi et al. | Low-cost image processing system for evaluating pavement surface distress | |
| JP2004326559A (en) | Lane mark detector |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20211029 |