CN113655002B - Quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology - Google Patents

Abstract

The invention provides a quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology. The recycled aggregate transport module is used to transport the recycled aggregates to the collection area of the hyperspectral image acquisition module; the light source module is a hyperspectral image acquisition module. The acquisition area provides illumination; the hyperspectral image acquisition module is used to collect the original data of the spectral reflection intensity of recycled aggregates, and convert the original data into a pseudo-color image and output it to the deep learning module; the deep learning module selects samples by using the KS algorithm, and The extracted features are used for model training, and the trained model is used for online detection; the image processing module provides parameter input for the regression analysis module; the construction of the regression analysis module is related to the water absorption and apparent density of recycled aggregates regression model.

Description

Quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology

Technical field

The invention relates to the field of quality grade detection of recycled aggregates containing mortar on their surfaces, and in particular to a quality detection system for recycled aggregates containing mortar on their surfaces based on hyperspectral technology.

Background technique

As our country continues to invest in infrastructure, the rapidly growing amount of aggregates has led to a serious shortage of primary aggregates in some places. On the other hand, our country produces a large amount of construction waste every year, and the shortage of storage land leads to environmental pollution, and waste concrete is The main component of construction waste. During the surface strengthening and crushing process of waste concrete, mortar will inevitably remain on the surface of the recycled aggregate. The presence and content of mortar have a serious impact on the quality of the recycled aggregate. Therefore, the quality grade of the recycled aggregate is detected and the quality of the recycled aggregate is determined. Whether the quality meets the requirements for reuse and thus used in the production of concrete. Developing an efficient and high-quality recycled aggregate quality grade detection system is of great significance for accelerating the utilization rate and extent of recycled aggregates in my country.

Contents of the invention

The main technical problem to be solved by this invention is to provide a quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology, which can judge the quality level of recycled aggregates, thereby controlling the production process of recycled aggregates and solving the problem of construction waste disposal. and shortage of virgin aggregates.

In order to solve the above technical problems, the present invention provides a quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology, including: a recycled aggregate transportation module, a light source module, a hyperspectral image acquisition module, a deep learning module, and image processing Module, regression analysis module;

The recycled aggregate transport module is used to transport recycled aggregate to the acquisition area of the hyperspectral image acquisition module;

The light source module provides illumination for the acquisition area of the hyperspectral image acquisition module;

The hyperspectral image acquisition module is used to collect original data of spectral reflection intensity of recycled aggregates, and convert the original data into pseudo-color image output;

The deep learning module selects samples by using the KS algorithm, then further differentiates the hyperspectral data features of the samples through data preprocessing, then performs dimensionality reduction processing on the preprocessed sample data features, and finally uses the extracted features For model training, and use the trained model for online detection;

The different objects detected online by the trained model are output in the form of pictures. The image processing module counts the number of pixels of different objects on the output classified pictures, then counts the pixel ratio of the mortar, and calculates the area of the recycled aggregate. The size and convex hull ratio provide parameter input for the regression analysis module;

The regression analysis module constructs a regression model related to the water absorption and apparent density of recycled aggregates.

In a preferred embodiment: the recycled aggregate transportation module includes a vibrating dispersed feeding device, an encoder, and a conveyor belt; the vibrating dispersed feeding device is used to provide stable and dispersed recycled aggregate to the conveyor belt; the encoding The device is used to read the current speed of the conveyor belt; the conveyor belt device transports the recycled aggregate dispersed by the vibration dispersion device to the acquisition area of the hyperspectral image acquisition module.

In a preferred embodiment: the light source module includes two halogen lamps.

In a preferred embodiment: the hyperspectral image acquisition module is a hyperspectral camera.

In a preferred embodiment: the deep learning module includes a light sample selection module, a data preprocessing module, a feature extraction module, a model training module, and an online detection module;

The sample selection module screens the collected hyperspectral data and selects samples; the data preprocessing module further expands the features of the selected samples to make the feature differences of the samples more obvious; the feature extraction module Perform data dimensionality reduction on hyperspectral data and remove features with greater correlation; the model training module trains the extracted features to obtain a model with strong generalization ability and obtain model parameters for online detection; The online detection module uses the trained model parameters for online classification of recycled aggregates, and outputs the classification results as pictures.

In a preferred embodiment: the image processing module includes image preprocessing and feature parameter extraction;

The image preprocessing uses Gaussian filtering to clean the noise of the classified image and remove smaller particles in the image; the feature extraction performs statistics on the mortar and native aggregate pixels in the classified image, and calculates the mortar The proportion of pixels is used to calculate the area and convex hull ratio of recycled aggregates.

In a preferred embodiment: the regression analysis module includes parameter input and acquisition of a regression analysis model;

The parameter input inputs the area ratio of the mortar, the area of the recycled aggregate and the convex hull ratio of the recycled aggregate into the regression model; the acquisition regression analysis model will input the area ratio of the mortar, the area of the recycled aggregate and the convex hull ratio of the recycled aggregate. The convex hull ratio of the material, water absorption, and apparent density are subjected to multiple regression to obtain regression parameters and construct a regression equation for quality grade detection of recycled aggregates.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction will be made below to the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without exerting any creative effort.

Figure 1 is a schematic diagram of the device of the quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology provided by the present invention.

Figure 2 is an overall flow chart of the quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology provided by the present invention.

Figure 3 is a detailed flow chart of the quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology provided by the present invention.

Detailed ways

The technical solutions in 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; obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on The embodiments of the present invention and all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "inner", "outer", "top/bottom", etc. are based on the orientation shown in the drawings. or positional relationships are only for the convenience of describing the present invention and simplifying the description, but do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance.

In the description of the present invention, it should be noted that, unless otherwise clearly stated and limited, the terms "installed", "provided with", "set/connected", "connected", etc., should be understood in a broad sense, such as " "Connection" can be a wall-mounted connection, a detachable connection, or an integral connection. It can be a mechanical connection or an electrical connection. It can be a direct connection or an indirect connection through an intermediate medium. It can be internal to two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood in specific situations.

Referring to Figures 1-3, this embodiment provides a quality detection system for recycled aggregate containing mortar on the surface based on hyperspectral technology, including: a recycled aggregate transportation module 10, a light source module 20, a hyperspectral image acquisition module 30, and deep learning. Module 40, image processing module 50, regression analysis module 60;

The recycled aggregate transport module 10 is used to transport recycled aggregate to the collection area of the hyperspectral image collection module 30;

The light source module 20 provides illumination for the collection area of the hyperspectral image collection module 30;

The hyperspectral image acquisition module 30 is used to collect the original data of the spectral reflection intensity of the recycled aggregate, and convert the original data into a pseudo-color image for output;

The deep learning module 40 selects samples by using the KS algorithm, further differentiates the hyperspectral data features of the samples through data preprocessing, then performs dimensionality reduction processing on the preprocessed sample data features, and finally extracts the features Used for model training and using the trained model for online detection;

The different objects detected online by the trained model are output in the form of pictures. The image processing module 50 counts the number of pixels of different objects on the output classified pictures, then counts the pixel proportions of the mortar, and calculates the proportion of recycled aggregates. The area size and convex hull ratio provide parameter input for the regression analysis module 60;

The regression analysis module 60 constructs a regression model related to the water absorption and apparent density of the recycled aggregate.

The workflow of the recycled aggregate transportation module is shown in Figure 1. The controller controls the speed of the conveyor belt to transport the recycled aggregate placed on the conveyor belt to the data acquisition area of the hyperspectral image acquisition module, and the encoder reads the current speed of the conveyor belt.

The workflow of the light source module is shown in Figure 1. Two 50W halogen lamps are used. The halogen lamps are symmetrically distributed on both sides of the hyperspectral camera to provide a stable, uniform and ideal lighting environment for the measurement area of the hyperspectral camera. It is possible to reduce the impact of the light source on the acquired spectral data.

In this embodiment, the recycled aggregate conveying module 10 includes a vibrating dispersing feeding device 11, an encoder 12, and a conveyor belt 13; the vibrating dispersing feeding device 11 is used to provide stable and dispersed recycled aggregate to the conveying belt 13; The encoder 12 is used to read the current speed of the conveyor belt 13; the conveyor belt device 13 transports the recycled aggregate dispersed by the vibration dispersion device to the acquisition area of the hyperspectral image acquisition module 30.

The light source module 20 includes two halogen lamps 21 .

The hyperspectral image acquisition module 30 is a hyperspectral camera 31 .

The deep learning module 40 includes a light sample selection module 41, a data preprocessing module 42, a feature extraction module 43, a model training module 44, and an online detection module 45;

The sample selection module 41 screens the collected hyperspectral data and selects samples; the data preprocessing module 42 further expands the characteristics of the selected samples to make the characteristics of the samples more obvious; the characteristics The extraction module 43 performs data dimensionality reduction on hyperspectral data and removes features with greater correlation; the model training module 44 trains the extracted features to obtain a model with strong generalization ability and obtain model parameters. For online detection; the online detection module 45 uses the trained model parameters for online classification of recycled aggregates, and outputs the classification results as pictures.

The image processing module 50 includes image preprocessing 51 and feature parameter extraction 52;

The image preprocessing 51 uses Gaussian filtering to clean the noise of the classified image and remove smaller particles in the image; the feature extraction 52 performs statistics on the mortar and native aggregate pixels in the classified image, and Calculate the proportion of mortar pixels, calculate the area and convex hull ratio of recycled aggregate. The feature extraction 52 is performed by finding the edge contour of the recycled aggregate after binarization of the image, and then counting the number of pixel points of the mortar and the original aggregate in the contour of the original image, and using the total number of pixel points of the mortar. Compared with the number of pixels of the entire recycled aggregate, the area ratio of the mortar can be determined. In addition, the area of recycled aggregate, convex hull ratio, etc. are also calculated.

The regression analysis module 60 includes parameter input 61 and acquisition of regression analysis model 62;

The parameter input 61 inputs the area ratio of the mortar, the area of the recycled aggregate and the convex hull ratio of the recycled aggregate into the regression model; the acquisition regression analysis model 62 will input the area ratio of the mortar, the area of the recycled aggregate and the convex hull ratio of the recycled aggregate into the regression model. The convex hull ratio of recycled aggregates is compared with water absorption and apparent density through multiple regression to obtain regression parameters and construct a regression equation for quality grade detection of recycled aggregates.

The installation method and work flow chart of the hyperspectral image acquisition module are shown in Figures 1 and 3. The hyperspectral camera acquires the light intensity reflected on the surface of the recycled aggregate through line scanning, thereby acquiring hyperspectral data. At the same time, the conveyor belt drives the recycled aggregate Move, and then obtain the surface light intensity of the entire recycled aggregate, and then calculate the correspondence between the three-dimensional vision systems to reconstruct the pseudo-color image of the recycled aggregate.

The workflow diagram of the deep learning module is shown in Figure 3. The KS algorithm is used to filter the collected hyperspectral data and select representative samples. Data preprocessing is to select the preprocessing that has the best differentiation effect on hyperspectral data features by comparing the degree of differentiation of the selected hyperspectral data features through methods such as first-order differential, logarithmic transformation of derivatives, and envelope removal. method. The feature extraction module uses principal component analysis PCA and wavelet transform WT to extract features from the preprocessed hyperspectral data to remove redundant data, and compares the obtained features for model training effects to select the model with the best training effect. Good feature extraction method. Model training compares the training model effects of extreme learning machines, Gaussian kernel extreme learning machines and wavelet kernel extreme learning machines, and selects the best training model for online detection. The model with the best generalization ability is used for online detection and classification of recycled aggregates.

The workflow diagram of the image processing module is shown in Figure 3. Gaussian filtering is used to remove noise from the classified images and remove small points in the images. The feature parameters are extracted by finding the edge contour of the recycled aggregate after image binarization, and then counting the number of pixel points of the mortar and the original aggregate in the contour of the original image, and using the total number of pixel points of the mortar to compare The number of pixels in the entire recycled aggregate can be determined by the area ratio of the mortar. In addition, the area of recycled aggregate, convex hull ratio, etc. are also calculated.

The workflow diagram of the regression analysis module is shown in Figure 3. Parameter input includes the mortar area ratio in the recycled aggregate, the area of the recycled aggregate, the convex hull ratio of the recycled aggregate, etc. The regression analysis model performs multiple regression on the proportion of mortar area in the recycled aggregate, the area of the recycled aggregate, the convex hull ratio of the recycled aggregate, etc., and the water absorption and apparent density of the recycled aggregate, and then obtains a multiple regression equation, thus Realize online detection of the quality grade of recycled aggregates.

The aggregate conveying module includes a vibration dispersion device, a conveyor belt device, and an encoder device. The controller device controls the conveying speed of the conveyor belt. The conveyor belt device sends the recycled aggregate dispersed by the vibration dispersing device to each image acquisition area in sequence, and the encoder device reads the current speed of the conveyor belt.

The light source module uses two 50W halogen lamps. The halogen lamps are symmetrically distributed on both sides of the hyperspectral camera to provide a stable and uniform ideal lighting environment for the measurement area of the hyperspectral camera, and to minimize the impact of the light source on the measurement area. influence on the spectral data.

The hyperspectral camera in the hyperspectral image acquisition module acquires the light intensity reflected on the surface of the recycled aggregate through line scanning, thereby obtaining hyperspectral data. At the same time, the conveyor belt drives the recycled aggregate to move, thereby increasing the surface light intensity of the entire recycled aggregate. All are obtained, and then by calculating the correspondence between the stereoscopic vision systems, the pseudo-color image of the recycled aggregate is reconstructed.

Based on the above technical solution, further, the deep learning module includes sample selection, data preprocessing, feature extraction, model training and online detection. The sample selection uses the KS algorithm to screen the collected hyperspectral data and select representative samples. The data preprocessing is to compare the degree of differentiation of the selected hyperspectral data features by comparing the first-order differential, logarithmic transformation of the derivative, envelope removal and other methods, and select the one with the best differentiation effect on the hyperspectral data features. preprocessing method.

The above are only preferred embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any person familiar with the technical field can use this concept to carry out the present invention within the technical scope disclosed in the present invention. Non-substantive modifications are infringements of the protection scope of the present invention.

Claims (1)

1. A quality detection system for recycled aggregates containing mortar on the surface based on hyperspectral technology, which is characterized by including: a recycled aggregate transportation module (10), a light source module (20), a hyperspectral image acquisition module (30), and a deep learning module (40), image processing module (50), regression analysis module (60); The recycled aggregate transport module (10) is used to transport recycled aggregate to the acquisition area of the hyperspectral image acquisition module (30); The light source module (20) provides illumination for the collection area of the hyperspectral image collection module (30); The hyperspectral image acquisition module (30) is used to collect original data of spectral reflection intensity of recycled aggregate, and convert the original data into a pseudo-color image and output it to the deep learning module (40); The deep learning module (40) selects samples by using the KS algorithm, further differentiates the hyperspectral data features of the samples through data preprocessing, then performs dimensionality reduction processing on the preprocessed sample data features, and finally extracts The features are used for model training, and the trained model is used for online detection; The different objects detected online by the trained model are output in the form of pictures. The image processing module (50) counts the number of pixels of different objects on the output classified pictures, then counts the pixel ratio of the mortar, and calculates the regenerated bone. The area size and convex hull ratio of the material provide parameter input for the regression analysis module (60); The regression analysis module (60) constructs a regression model related to the water absorption and apparent density of the recycled aggregate; The recycled aggregate conveying module (10) includes a vibrating dispersing feeding device (11), an encoder (12), and a conveyor belt (13); the vibrating dispersing feeding device (11) is used to provide stability to the conveying belt (13) and dispersed recycled aggregate; the encoder (12) is used to read the current speed of the conveyor belt (13); the conveyor belt (13) transports the recycled aggregate dispersed by the vibration dispersion device to the hyperspectral image acquisition module (30) collection area; The light source module (20) includes two halogen lamps (21); The hyperspectral image acquisition module (30) is a hyperspectral camera (31); The deep learning module (40) includes a light sample selection module (41), a data preprocessing module (42), a feature extraction module (43), a model training module (44), and an online detection module (45); The sample selection module (41) screens the collected hyperspectral data and selects samples; the data preprocessing module (42) further expands the characteristics of the selected samples to make the characteristics of the samples more obvious. ; The feature extraction module (43) performs data dimensionality reduction on hyperspectral data and removes features with greater correlation; the model training module (44) trains the extracted features to obtain higher generalization capabilities. Strong model, and obtain model parameters for online detection; the online detection module (45) uses the trained model parameters for online classification of recycled aggregates, and outputs the classification results as pictures; The image processing module (50) includes image preprocessing (51) and feature parameter extraction (52); The image preprocessing (51) uses Gaussian filtering to clean the noise of the classified image and remove smaller particles in the image; the feature parameter extraction (52) removes the mortar and native aggregate in the classified image. Perform statistics on pixels, calculate the proportion of mortar pixels, and calculate the area and convex hull ratio of recycled aggregates; The regression analysis module (60) includes parameter input (61) and acquisition of the regression analysis model (62); The parameter input (61) inputs the area ratio of the mortar, the area of the recycled aggregate and the convex hull ratio of the recycled aggregate into the regression analysis model (62); the regression analysis model (62) obtains the area of the mortar. Proportion, recycled aggregate area and convex hull ratio of recycled aggregate are compared with water absorption and apparent density through multiple regression to obtain regression parameters and construct a regression equation for quality grade detection of recycled aggregate.