Cement concrete pavement crack detection and classification method and system

Abstract

The invention provides a cement concrete pavement crack detection and classification method and system, and relates to the technical field of image processing, and the method comprises the following steps: collecting a cement concrete pavement image, and carrying out the feature extraction of the cement concrete pavement image, and obtaining the corresponding image features; constructing a recognition model, and optimizing the recognition model; detecting and classifying the image features by using the optimized recognition model to obtain a corresponding result; according to the method, large-scale data can be processed in a short time, so that the detection cost and the demand of human resources are remarkably reduced, and the efficiency and the cost effectiveness are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and more particularly to a method and system for detecting and classifying cracks in cement concrete pavement. Background Art

[0002] At present, cement concrete pavement has the advantages of long service life, high strength, good stability, good durability, low maintenance cost and high economic benefits, and is widely used in highway engineering and municipal engineering. However, with the extension of service time, cement concrete pavement often produces cracks, which not only affects the durability of concrete pavement, but also affects the comfort and safety of driving. Therefore, by timely detection and classification of cracks in cement concrete pavement, potential safety hazards can be discovered in time, providing a scientific basis for subsequent maintenance and reinforcement.

[0003] However, cracks on cement concrete pavements often have irregular shapes, including linear, mesh, bifurcated, etc., which makes them present complex geometric structures in the image. The width and length of cracks can vary in a wide range, from fine surface cracks to wide deep cracks, which leads to the scale diversity of cracks in the image. The degree of damage of cracks can vary, from surface cracks to deep cracks, and some cracks may have a greater impact on the structural stability of the pavement, so different levels of classification are required, and the existing technology lacks methods that can accurately detect and classify cracks in cement concrete pavements.

[0004] Therefore, how to provide a method for detecting and classifying cement concrete pavement cracks that can solve the above problems is an issue that technical personnel in this field urgently need to solve. Summary of the invention

[0005] In view of this, the present invention provides a method and system for detecting and classifying cracks in cement concrete pavement, which can process large-scale data in a shorter time, thereby significantly reducing the detection cost and human resource requirements, and improving efficiency and cost-effectiveness.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A method for detecting and classifying cracks in a cement concrete pavement comprises the following steps:

[0008] Collecting a cement concrete pavement image, and performing feature extraction on the cement concrete pavement image to obtain corresponding image features;

[0009] Constructing a recognition model and optimizing the recognition model;

[0010] The image features are detected and classified using the optimized recognition model to obtain corresponding results.

[0011] Preferably, the specific process of building the recognition model includes:

[0012] The recognition model is a fuzzy neural network model, which includes an input layer, a fuzzification layer, a fuzzy rule layer and a fuzzy decision layer which are connected in sequence.

[0013] Preferably, the specific process of optimizing the recognition model includes:

[0014] The recognition model is optimized using a particle swarm optimization algorithm.

[0015] Preferably, the specific process of optimizing the recognition model using the particle swarm optimization algorithm includes:

[0016] Initialize all particles and set the initial parameters corresponding to the particles;

[0017] Calculate the fitness function value of each particle and update the initial parameters of the particle;

[0018] Determine whether the end condition is met. If so, update the initial parameters of the particle and output the optimal solution.

[0019] Preferably, the specific process of extracting features from the cement concrete pavement image includes:

[0020] Preprocessing the cement concrete pavement image;

[0021] The pre-processed cement concrete pavement image is subjected to texture feature extraction to obtain corresponding image features.

[0022] Preferably, the process of preprocessing the cement concrete pavement image includes: image size adjustment, image grayscale, image noise reduction and image enhancement.

[0023] The present invention also provides a system for detecting and classifying cracks in cement concrete pavement, comprising:

[0024] The acquisition module is used to acquire cement concrete pavement images and perform feature extraction on the cement concrete pavement images to obtain corresponding image features;

[0025] A construction module, used for constructing a recognition model and optimizing the recognition model;

[0026] The recognition module is used to process the image features using the optimized recognition model to obtain corresponding detection results.

[0027] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method and system for detecting and classifying cracks in cement concrete pavement, which is implemented based on fuzzy neural network, GLCM method and particle swarm optimization algorithm, and has the following advantages:

[0028] (1) Automation and objectivity: The method provided by the present invention realizes automatic detection and classification of cement concrete cracks through computer vision and machine learning technology, eliminates subjectivity and inconsistency, and improves the objectivity and accuracy of classification results;

[0029] (2) Efficiency and cost-effectiveness: The method provided by the present invention can process large-scale data in a relatively short period of time, thereby significantly reducing the detection cost and human resource requirements, and improving efficiency and cost-effectiveness;

[0030] (3) Multi-feature fusion: By introducing the GLCM method and the particle swarm optimization algorithm into the fuzzy neural network model, the present invention can capture the characteristic information of the cracks more comprehensively and improve the accuracy and robustness of the classification; the GLCM method is used to extract texture features, while the particle swarm optimization algorithm can help optimize the parameters of the model to adapt to different types of cracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0032] Figure 1 An overall flow chart of a method for detecting and classifying cracks in a cement concrete pavement provided by the present invention;

[0033] Figure 2 A schematic diagram of the structure of a fuzzy neural network model provided by an embodiment of the present invention;

[0034] Figure 3 The overall flow chart of the particle swarm optimization algorithm provided by the embodiment of the present invention;

[0035] Figure 4 The present invention provides a structural principle block diagram of a cement concrete pavement crack detection and classification system. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] See also Figure 1 As shown, the embodiment of the present invention discloses a method for detecting and classifying cracks in a cement concrete pavement, comprising the following steps:

[0038] Collect cement concrete pavement images, and perform feature extraction on the cement concrete pavement images to obtain corresponding image features;

[0039] Build a recognition model and optimize it;

[0040] The optimized recognition model is used to detect and classify image features to obtain corresponding results.

[0041] Specifically, the flying carrier of the pavement crack image acquisition equipment is the DJI M300 RTK drone. Compared with traditional pavement inspection vehicles, choosing drones as pavement disease inspection equipment has the following advantages:

[0042] The acquisition speed is fast, its viewing angle is wide, it can quickly detect a large area of ​​roads, and can complete the inspection task faster; it has high accuracy, it can obtain global road area information, and can effectively reduce model misdetection from the root; it can reduce the risk of injury to workers and improve work safety; it is low-cost, which helps to reduce operating costs and inspection time. DJI's M300 RTK drone is the latest industry-level model under DJI, which can fully meet the data collection needs of various complex road sections. It has strong wind resistance stability and endurance, can withstand a maximum of 7 winds, and has a battery life of 50 minutes. The drone is equipped with a DJI Zenmuse P1 gimbal camera, which combines multiple types of sensors, a 45-megapixel zoom camera, an integrated full-frame image sensor, and a three-axis gimbal. It can achieve a high-performance, high-precision, and high-efficiency integrated aerial survey solution to meet the shooting accuracy requirements of the model. The flight altitude for collecting data is set at 30 meters, and the road image size taken is 8192x 5460.

[0043] In a specific embodiment, the specific process of building the recognition model includes:

[0044] The recognition model is a fuzzy neural network model, which includes an input layer, a fuzzification layer, a fuzzy rule layer and a fuzzy decision layer connected in sequence. The input layer is used to cache the input signal, the fuzzification layer is used to fuzzify the input signal, the fuzzy decision layer mainly classifies the quantities that meet certain conditions and defuzzifies the fuzzy quantities, and the output layer is used to output the operation results.

[0045] Specifically, Figure 2 The diagram is a structural diagram of the fuzzy neural network model. The fuzzy neural network model combines neural networks with fuzzy logic. In essence, it is to give fuzzy input signals and fuzzy weights to conventional neural networks. It has both the powerful learning ability of neural networks and the powerful structural knowledge expression ability of fuzzy logic reasoning. Fuzzy neural classification is equal to the process of grouping elements into fuzzy sets, and its membership function is defined by the truth value of fuzzy proposition functions. Fuzzy proposition functions are like expressions containing one or more variables. When these variables are assigned values, the expression becomes a fuzzy proposition.

[0046] "Membership function" is a mathematical function that specifies how many elements are included in a fuzzy set. This approach does not define the problem with precise numerical values, but describes the problem more naturally by using linguistic ideas. Gaussian membership function is currently the most commonly used membership function. In addition to the above, additional membership functions, such as generalized Bell membership function, triangular fuzzy numbers, trapezoidal fuzzy numbers, etc., can be used. The fuzzy layer contains ab nodes (where a is the number of fuzzy rules and b is the number of input features). There are nodes where y=1,2,..., and m is for each input feature x (x=1,2,...), n). Therefore, the model provided in this embodiment adopts Gaussian membership function, and the specific expression is:

[0047]

[0048] Where a, b, and c are trainable parameters that are applied to the input features to form the output. For the normal distribution, the weights are randomly assigned.

[0049] In a specific embodiment, the specific process of optimizing the recognition model includes:

[0050] The particle swarm optimization algorithm is used to optimize the recognition model.

[0051] In a specific embodiment, the specific process of optimizing the recognition model using the particle swarm optimization algorithm includes:

[0052] Initialize all particles and set the initial parameters corresponding to the particles;

[0053] Calculate the fitness function value of each particle and update the initial parameters of the particle;

[0054] Determine whether the end condition is met. If so, update the initial parameters of the particle and output the optimal solution.

[0055] For details, please refer to the specific process of particle swarm optimization algorithm. Figure 3 As shown in the figure, the particle swarm optimization algorithm (PSO) is a branch of evolutionary computing and a random search algorithm that simulates biological activities in nature. Particle swarm optimization algorithm (PSO) is the most successful method for solving complex high-dimensional nonlinear problems. It is an effective global optimization algorithm. PSO simulates the process of bird flocks and fish swarms preying in nature. The global optimal solution to the problem is found through collaboration in the group. The basic process is:

[0056] (1) Initialize all particles, that is, assign values ​​to their velocities and positions, and set the individual's historical best position (pBest) as the current position, and the best individual position in the group as the current group's best position (gBest);

[0057] (2) In each generation of evolution, the fitness function value of each particle is calculated.

[0058] (3) If the current fitness function value is better than the historical optimal value, the individual optimal position is updated.

[0059] (4) If the current fitness function value is better than the global historical optimal value, the group optimal position is updated.

[0060] (5) The velocity and position of the d-th dimension of each particle i are updated according to equations (2) and (3) respectively:

[0061]

[0062]

[0063] In the formula, ω represents the inertia factor, c1 and c2 are empirical weights, Represents the discount factor.

[0064] (6) Determine whether the end condition has been met. If not, go to the second step to continue execution.

[0065] In a specific embodiment, the specific process of extracting features from a cement concrete pavement image includes:

[0066] Preprocess cement concrete pavement images;

[0067] The texture features of the preprocessed cement concrete pavement image are extracted to obtain the corresponding image features.

[0068] In a specific embodiment, the process of preprocessing the cement concrete pavement image includes: image size adjustment, image grayscale, image noise reduction and image enhancement.

[0069] Specifically, the pavement crack image preprocessing step is mainly to eliminate noise distortion and increase the important information in the original pavement crack image. By performing preprocessing methods such as image grayscale processing, image noise reduction processing, and image enhancement processing on the original image, a suitable image is provided for subsequent disease segmentation and feature extraction.

[0070] S1: Resize the pavement crack images: In the first step of preprocessing, we require all images to be of the same size, so we need to resize all dataset images to 256x256.

[0071] S2: Image grayscale processing: Use weighted average method to obtain information from 11% blue component, 59% green component, and 30% red component; according to the importance level of different components and other indicators that conform to human vision, the three components in the RGB image are weighted averaged according to different weights and used as the grayscale value of the pixel at that point; use color layer information to improve the image contrast level. The human eye is most sensitive to green (G) in the real world and least sensitive to blue (B), so the weight ratio of the three components of the RGB image is distributed as follows:

[0072] im(i,j)=0.3R(i,j)+0.59G(i,j)+0.11B(i,j) (4)

[0073] S3: Image noise reduction: Common image noises include salt and pepper noise, Poisson noise, multiplicative noise, and Gaussian noise. Salt and pepper noise is more common in road images. Salt and pepper noise is also called impulse noise. It is caused by the sudden strong signal interference encountered by the image signal during the transmission process. Median filtering is a typical nonlinear filtering technology. It uses the median of the grayscale value of the pixel neighborhood to replace the grayscale value of the pixel. Median filtering has a more obvious inhibitory effect on salt and pepper noise.

[0074] S4: Image enhancement processing: During the image acquisition process, the road surface is affected by the lighting, and the acquired images will have the defect of partial underexposure, which often affects the capture of feature targets due to the imbalance of contrast. The Laplace image grayscale equalization method averages the grayscale of the image and retains the pixel distribution of the darker and brighter areas, highlighting the contrast between images.

[0075] The preprocessed cement concrete pavement images are subjected to texture feature extraction, mainly through the GLCM-based texture combining the adjacent gray-level co-occurrence matrix and its frequency in the pavement crack data. These features are generated by performing matrix evaluation on the gray-level dimensions specified in the pavement crack data.

[0076] The methods for extracting texture features can be divided into: (1) structure-based methods, which model the texture to be detected and search for repeated patterns in the image. (2) statistical data-based methods. Eight features were obtained, and each feature is described as follows:

[0077] Let M be the N-dimensional co-occurrence matrix, i,j be the coordinates of the object;

[0078] 1. Autocorrelation: The fitness and roughness of the texture features in the image are defined by Auc. In the road crack image of a smooth or normal surface, the autocorrelation method shows the troughs and peaks. The “Auc” property is shown in Equation (5):

[0079]

[0080] 2. Contrast: Con monitors the intensity of each pixel and its neighboring pixels throughout the image. It also affects the amount of geographical differences in the image. Equation (6) describes the "Con" feature in detail:

[0081] Con=Σ l Σ m |lm| 2 p(l,m) (6)

[0082] 3. Correlation: The correlation of an image checks the correlation between a pixel in the entire image and its neighboring pixels. The correlation value is -1 to 1. The "CR" attribute is shown in formula (7):

[0083]

[0084] 4. Clustering shading: The sha feature is a concept of image homogeneity and is used to evaluate the skewness of the GLCM matrix. A higher score indicates the presence of asymmetry, as shown in Equation (8):

[0085]

[0086] 5. Dissimilarity (DS): Calculates the average grayscale variance in the image distribution. A larger number means a greater difference in intensity levels between adjacent pixels. The “DS” feature is described in detail:

[0087] DS=Σ l Σ m |lm|p(l,m) (9)

[0088] 6. Energy (En): En represents the number of times a pixel pair is repeated. It is a measure of the texture irregularity of the input data. Equation (10) describes the “En” characteristic in detail:

[0089] Ene=Σ l，m p(l,m) 2 (10)

[0090] 7. Maximum Probability (MP): The maximum probability of generating a pixel of interest is measured by MP. The “MP” characteristic is shown in equation (11):

[0091] MP=max.p(l,m)for all(l,m) (11)

[0092] 8.SA (Sum average): SA feature evaluates the average value of pavement crack grayscale and distribution, as shown in the following formula:

[0093]

[0094] See also Figure 4 As shown, an embodiment of the present invention further provides a system for detecting and classifying cement concrete pavement cracks using any one of the above embodiments, comprising:

[0095] The acquisition module is used to acquire cement concrete pavement images and perform feature extraction on the cement concrete pavement images to obtain corresponding image features;

[0096] A construction module is used to construct a recognition model and optimize the recognition model;

[0097] The recognition module is used to process image features using an optimized recognition model to obtain corresponding detection results.

[0098] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0099] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting and classifying cracks in cement concrete pavement, characterized in that: The following steps are involved: Collecting a cement concrete pavement image, and performing feature extraction on the cement concrete pavement image to obtain corresponding image features; Constructing a recognition model and optimizing the recognition model; The image features are detected and classified using the optimized recognition model to obtain corresponding results.

2. A method for detecting and classifying cracks in cement concrete pavement according to claim 1, characterized in that: The specific process of building a recognition model includes: The recognition model is a fuzzy neural network model, which includes an input layer, a fuzzification layer, a fuzzy rule layer and a fuzzy decision layer which are connected in sequence.

3. A method for detecting and classifying cracks in cement concrete pavement according to claim 1, characterized in that: The specific process of optimizing the recognition model includes: The recognition model is optimized using a particle swarm optimization algorithm.

4. A method for detecting and classifying cracks in cement concrete pavement according to claim 3, characterized in that: The specific process of optimizing the recognition model using the particle swarm optimization algorithm includes: Initialize all particles and set the initial parameters corresponding to the particles; Calculate the fitness function value of each particle and update the initial parameters of the particle; Determine whether the end condition is met. If so, update the initial parameters of the particle and output the optimal solution.

5. The method for detecting and classifying cracks in cement concrete pavement according to claim 1, characterized in that: The specific process of extracting features from the cement concrete pavement image includes: Preprocessing the cement concrete pavement image; The pre-processed cement concrete pavement image is subjected to texture feature extraction to obtain corresponding image features.

6. A method for detecting and classifying cracks in cement concrete pavement according to claim 5, characterized in that: The process of preprocessing the cement concrete pavement image includes: image size adjustment, image grayscale, image noise reduction and image enhancement.

7. A system using the method for detecting and classifying cement concrete pavement cracks according to any one of claims 1 to 6, characterized in that: include: The acquisition module is used to acquire cement concrete pavement images and perform feature extraction on the cement concrete pavement images to obtain corresponding image features; A construction module, used for constructing a recognition model and optimizing the recognition model; The recognition module is used to process the image features using the optimized recognition model to obtain corresponding detection results.

Images