Intelligent detection method, system and equipment for highway pavement cracks based on lidar

Abstract

The present invention discloses a method, system, and device for intelligent detection of cracks in highway pavement based on laser radar, and relates to the field of highway detection technology. The method includes collecting pavement data, constructing a pavement material database, and recording and storing the designed pavement material, reflection type, and texture changes based on the pavement data to obtain comprehensive reflection characteristics; constructing a multi-feature fusion database; collecting real-time pavement data, determining the pavement material, and correcting the collected pavement data; pre-processing the data; comparing the pre-processed real-time pavement data with the preset threshold, and marking the data above the threshold to facilitate subsequent repeated detection and improve accuracy; and identifying cracks by comparing the real-time pavement data with the multi-feature fusion database. By constructing a pavement material database and a multi-feature fusion database, the present invention marks the data above the threshold, accurately identifies cracks, and improves the accuracy of crack detection.

Description

Technical Field

[0001] The present invention relates to the field of highway detection technology, and in particular to a laser radar-based intelligent detection method, system and equipment for highway pavement cracks. Background Art

[0002] With the rapid development of social economy, the construction of highway projects continues to grow. Due to the unique climate, hydrogeology and geomorphological conditions, a large number of roads built in the early days (especially lower-grade roads) have suffered damage. Among them, uneven settlement of the roadbed and low roadbed strength have led to a large number of cracks in the road surface. The continuous development of cracks has seriously affected the integrity of the road surface. In addition, water leakage has caused the road surface to become slippery, thereby reducing the adhesion between the tires and the road surface.

[0003] Application publication number CN115639353 A discloses a road surface detection method based on a high-precision map, comprising the following steps: Step 1: Installation of inspection equipment, placing the inspection equipment on the road surface to be inspected, and installing tools such as a laser radar, GPS, IMU, and a high-definition camera on the inspection equipment; Step 2: Information extraction, extracting the information transmitted by the inspection equipment using technologies such as GIS. The present invention utilizes technologies such as laser radar, GPS, IMU, and multi-channel high-definition cameras on the inspection equipment, combined with data fusion technology modules and GIS technology modules, to clearly present road surface cracks, highway assets, and the like on the road network map. By means of image enhancement, grayscale processing, and binary segmentation, the clarity of the image of the inspected road section is increased, thereby constructing a safe, reliable, economical, and efficient basic information system. The image output by the radar is analyzed using an AI algorithm module and a radar spectrum analysis module, thereby accurately presenting images of cracks on the detected highway.

[0004] Existing intelligent detection methods for road pavement cracks lack adaptability and are unable to adapt to various complex and changing road environments. For example, environmental factors such as shadows, water pollution, oil stains, as well as different lighting conditions and climatic conditions may affect the algorithm's recognition accuracy. When dealing with these complex environments, it is often impossible to accurately capture key features such as cracks, resulting in a significant decrease in recognition accuracy. At the same time, the diversity of road surface materials also increases the difficulty of recognition. The system has difficulty adapting to the reflective characteristics and texture changes of road surfaces of different materials, further reducing recognition accuracy. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, system and equipment for intelligent detection of road pavement cracks based on laser radar to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a laser radar-based intelligent detection method for road pavement cracks, the method comprising:

[0007] Collect road surface data and process historical road surface data;

[0008] Construct a pavement material database, record and store the designed pavement material, reflection type and texture changes based on pavement data, and obtain comprehensive reflection characteristics;

[0009] Build a multi-feature fusion database to collect and combine multiple layers of crack features based on pavement data;

[0010] Collect real-time road data, collect and record various road data information in real time;

[0011] Determine the road surface material by comparing the real-time road surface data with the road surface material database to confirm the road surface material;

[0012] Correct the collected road surface data and correct the real-time road surface data based on the road surface material;

[0013] Preprocess data to preprocess the corrected real-time road surface data to remove invalid, erroneous or abnormal information in the original data to ensure the accuracy and reliability of the data;

[0014] Preset threshold comparison: compare the pre-processed real-time road data with the preset threshold, and mark the data above the threshold to facilitate subsequent repeated detection and improve accuracy;

[0015] Identify cracks and compare real-time pavement data with a multi-feature fusion database to identify cracks and improve the accuracy of crack detection.

[0016] The calculation formula of the reflection characteristics is as follows:

[0017] ;

[0018] Where R(x,y) represents the comprehensive reflection characteristics at a specific position (x,y) on the road surface, n represents the total number of road surface material types, and M i Represents the material characteristic parameters of the i-th road material type, R i (x, y) represents the reflection characteristics of the i-th road material at position (x, y), T i (x,y) represents the texture change characteristics of the i-th road material at position (x,y).

[0019] The multi-layer features of the cracks are collected and combined, and the calculation formula is as follows:

[0020] ;

[0021] Among them, F represents the final feature fusion result, α, β, γ, δ and ϵ correspond to H, E, T, L and L respectively. cThe weight coefficient of the feature is used to adjust the contribution of different features in the final fusion result. N represents the number of image data, I i Represents the i-th image data, H(I i ) represents high-frequency filtering of the i-th image data, where H represents high-frequency feature extraction, E(I i ) represents edge detection processing for the i-th image data, where E represents edge detection, T(I i ) represents the texture analysis processing of the i-th image data, where T represents texture analysis, L(I i ) represents the brightness distribution calculation of the i-th image data, where L represents the brightness feature extraction, L c (I i ) represents the local contrast calculation for the i-th image data, where L c Represents local contrast calculation, and C represents the enhancement of the local contrast result.

[0022] The real-time road surface data is compared with the road surface material database, and the calculation formula is as follows:

[0023] ;

[0024] Among them, Ma is the determined road material type, Z real (x, y) is the comprehensive reflection characteristic of the real-time road surface data at the position (x, y), n is the total number of road surface material types, W j is the weight coefficient of the jth road material type, which is used to adjust the contribution of different materials in the comprehensive reflection characteristics. j is the material characteristic parameter of the jth pavement material type, representing the physical properties of the material, R j (x, y) is the reflection characteristic of the jth road material at position (x, y), T j (x, y) is the texture change characteristic of the jth road material at position (x, y), α is the weight coefficient of the texture change characteristic, which is used to adjust the contribution of texture change in the comprehensive reflection characteristic, β is the weight coefficient in the matching function, which is used to adjust the influence of Euclidean distance in the matching calculation, argmax i It means to search for the maximum matching degree of the i-th material type to find the best matching road material type. When j=i, it means the i-th material.

[0025] The real-time road surface data is corrected based on the road surface material, and the correction formula is as follows:

[0026] ;

[0027] Among them, Dcorrection(x,y) is the corrected pavement data at position (x,y), D(x,y) is the original pavement data at position (x,y), that is, the collected uncorrected pavement data, W is the material weight coefficient, which is used to adjust the influence of different material types on the correction results. The coefficient is determined according to the identified material type, B(x,y) is the material baseline value, which represents the baseline reflectivity or baseline characteristic value of a specific material type at position (x,y), O(x,y) is the superimposed correction effect, which represents the correction effect superimposed in the comprehensive correction step, which may include crack detection and optimization of correction parameter steps. Adjustment of the data, F(x,y) is the texture correction factor, which represents the correction effect of texture changes on the data at position (x,y), and A(x,y) is the application correction coefficient, which represents the further adjustment of the data in the final application correction step. The coefficient may be determined based on specific application requirements or environmental conditions.

[0028] The pre-processed real-time road data is compared with the preset threshold, and the data above the threshold is marked. The calculation formula is as follows:

[0029] ;

[0030] Among them, B(x,y) is the comparison result at position (x,y), D 预处理 (x, y) is the preprocessed road surface data at position (x, y), T is the preset threshold used for comparison with the preprocessed road surface data, γ is the gain coefficient used to amplify the difference between the preprocessed road surface data and the preset threshold, and δ is the offset used to adjust the benchmark of the comparison result.

[0031] The real-time road surface data is compared with the multi-feature fusion database. The similarity calculation formula between the real-time road surface data and the samples in the database is as follows:

[0032] ;

[0033] Among them, S(F 实 ,F 库 ) represents the real-time road data feature vector F 实 The cosine similarity between the feature vector F of the k-th sample in the database, F is the feature vector obtained after feature fusion of real-time road data, and F is the feature vector obtained after feature fusion of the k-th sample in the database, ∥F 实 ∥ represents the real-time norm of the eigenvector F, that is, the square root of the sum of the squares of the elements of the vector, ∥F 库 ∥ represents the eigenvector F 库 The norm of

[0034] Find the database sample with the highest similarity to the real-time road data. The calculation formula is as follows:

[0035] ;

[0036] Among them, k ∗ Indicates the index of the database sample with the highest similarity to the real-time road data, arg max Indicates taking the value of the independent variable corresponding to the maximum value, that is, finding the value that makes S(F 实 ,F 库 ) the largest k value, k=1,2,…,M represents the range of k, from 1 to M, where M is the total number of samples in the database, S(F 实 ,F 库 ) represents the cosine similarity between the feature vector of real-time road surface data and the feature vector of the kth sample in the database.

[0037] The LiDAR-based intelligent road pavement crack detection system uses the aforementioned LiDAR-based intelligent road pavement crack detection method, including:

[0038] Collection module: used to collect road surface data, collect and process historical road surface data, and collect real-time road surface data, and collect and record various road surface data information in real time;

[0039] Storage module: used to build a pavement material database, record and store the designed pavement material, reflection type and texture changes based on pavement data to obtain comprehensive reflection characteristics; and to build a multi-feature fusion database, collect and combine multi-layer features of cracks based on pavement data;

[0040] Data processing module: used to determine the pavement material, compare the real-time pavement data with the pavement material database, confirm the pavement material, and correct the collected pavement data, correct the real-time pavement data based on the pavement material, and pre-process the data, pre-process the corrected real-time pavement data, remove invalid, erroneous or abnormal information in the original data, ensure the accuracy and reliability of the data, and use the preset threshold comparison to compare the pre-processed real-time pavement data with the preset threshold, mark the data above the threshold, facilitate subsequent repeated detection, improve accuracy, and use the preset threshold comparison to compare the pre-processed real-time pavement data with the preset threshold, mark the data above the threshold, facilitate subsequent repeated detection, improve accuracy, and identify cracks, compare the real-time pavement data with the multi-feature fusion database, identify cracks, and improve the accuracy of crack detection.

[0041] The LiDAR-based intelligent road pavement crack detection device uses the aforementioned LiDAR-based intelligent road pavement crack detection method, including:

[0042] Data storage server, which collects and processes historical road surface data;

[0043] Millimeter-wave radar and laser scanners for collecting real-time road surface data;

[0044] High-performance computers are used to build a pavement material database to obtain comprehensive reflection characteristics; to build a multi-feature fusion database to collect and combine multi-layer features of cracks; to determine pavement material; to correct collected pavement data; to pre-process data to remove invalid, erroneous or abnormal information in the original data; to compare pre-processed real-time pavement data with preset thresholds, and to mark data above the threshold to facilitate subsequent repeated detection and improve accuracy; and to identify cracks, compare real-time pavement data with the multi-feature fusion database to improve the accuracy of crack detection.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] This LiDAR-based intelligent road pavement crack detection method builds a road material database, judges and corrects the collected real-time road surface data, compares the real-time road surface data with a preset threshold, and marks the data above the threshold to facilitate subsequent repeated detection and improve accuracy.

[0047] By building a multi-feature fusion database and comparing real-time pavement data with the multi-feature fusion database, cracks can be accurately identified and the accuracy of crack detection can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the principle structure of the present invention;

[0049] Figure 2 This is a schematic diagram of the principle structure of building a material database in the present invention;

[0050] Figure 3 Schematic diagram of the principle structure of data correction in the present invention;

[0051] Figure 4 This is a schematic structural diagram of the material judgment principle of the present invention;

[0052] Figure 5 Schematic diagram of the principle structure of gap feature extraction in the present invention;

[0053] Figure 6 Schematic diagram of the structure of the gap recognition principle in the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0055] In this application, for ease of understanding, the method steps used do not need to be executed in the order of the steps in this embodiment during actual operation. In other embodiments, these steps may be performed simultaneously or in a different order.

[0056] like Figures 1-6 As shown, the present invention provides a technical solution: a laser radar-based intelligent detection method for road pavement cracks, the method comprising:

[0057] Collect road surface data and process historical road surface data;

[0058] It should be noted that by collecting a large amount of historical pavement data, it is possible to cover information such as various pavement conditions, materials, reflective properties, and crack characteristics, making subsequent analysis and model construction more comprehensive and representative.

[0059] Construct a pavement material database, record and store the designed pavement material, reflection type and texture changes based on pavement data, and obtain comprehensive reflection characteristics;

[0060] It should be noted that accurate material identification is the basis for subsequent data correction and crack detection, which helps to improve the accuracy and pertinence of the entire detection process. Road surfaces of different materials have different reflective characteristics. Material judgment based on the database can avoid detection errors caused by material misjudgment.

[0061] Construct a multi-feature fusion database to collect and combine multiple layers of crack features based on pavement data;

[0062] It should be noted that by collecting and combining multiple layers of crack features, the shape, size, depth and other characteristics of the cracks can be described more comprehensively, reducing the misjudgment that may be caused by single feature identification and improving the accuracy of crack identification.

[0063] Collect real-time road data, collect and record various road data information in real time;

[0064] Determine the road surface material by comparing the real-time road surface data with the road surface material database to confirm the road surface material;

[0065] It should be noted that the database-based comparison method can quickly and accurately determine the road material, improving the efficiency and accuracy of the entire detection process.

[0066] Correct the collected road surface data and correct the real-time road surface data based on the road surface material;

[0067] It should be noted that the influence of reflection characteristics caused by differences in road materials should be eliminated so that the corrected data can more truly reflect the actual conditions of the road surface. Road surfaces of different materials reflect lasers in different ways. Correction can unify data standards and improve data quality.

[0068] Preprocess the data to preprocess the corrected real-time road surface data to remove invalid, erroneous or abnormal information in the original data to ensure the accuracy and reliability of the data;

[0069] Preset threshold comparison: compare the pre-processed real-time road data with the preset threshold, and mark the data above the threshold to facilitate subsequent repeated detection and improve accuracy;

[0070] It should be noted that by setting thresholds, areas where cracks or anomalies may exist can be screened out and marked to facilitate subsequent key detection and verification, thereby improving detection efficiency.

[0071] Identify cracks and compare real-time pavement data with a multi-feature fusion database to identify cracks and improve the accuracy of crack detection.

[0072] The designed road surface material, reflection type and texture change are stored based on the road surface data. The calculation formula is as follows:

[0073] ;

[0074] Where R(x,y) represents the comprehensive reflection characteristics at a specific position (x,y) on the road surface, n represents the total number of road surface material types, and M i Represents the material characteristic parameters of the i-th road material type, R i (x,y) represents the reflection characteristics of the i-th road material at position (x,y), T i (x,y) represents the texture change characteristics of the i-th road material at position (x,y).

[0075] Bring in data:

[0076] The total number of pavement material types n=3;

[0077] The material characteristic parameters Mi are M1=0.5, M2=0.3, M3=0.2;

[0078] The reflection characteristics Ri(x,y) at position (x,y)=(1,1) are:

[0079] R1(1,1)=0.4;

[0080] R2(1,1)=0.6;

[0081] R3(1,1)=0.5;

[0082] The texture change characteristics Ti(x,y) at position (x,y)=(1,1) are:

[0083] T1(1,1)=0.8;

[0084] T2(1,1)=0.7;

[0085] T3(1,1)=0.9;

[0086] Now we can plug this data into the formula to calculate:

[0087] R(1,1)==0.376;

[0088] Therefore, the integrated reflection characteristic R(1,1) at the position (1,1) is 0.376.

[0089] The multi-layer features of the cracks are collected and combined, and the calculation formula is as follows:

[0090] ;

[0091] Among them, F represents the final feature fusion result, α, β, γ, δ, ϵ represent the weight coefficients of each feature, which are used to adjust the contribution of different features in the final fusion result, N represents the number of image data, I i Represents the i-th image data, H(I i ) represents high-frequency filtering of the i-th image data, where H represents high-frequency feature extraction, E(I i ) represents edge detection processing for the i-th image data, where E represents edge detection, T(I i ) represents the texture analysis processing of the i-th image data, where T represents texture analysis, L(I i ) represents the calculation of the brightness distribution of the i-th image data, where L represents the brightness feature extraction, Lc(I i ) represents the local contrast calculation for the i-th image data, where L c Represents local contrast calculation, and C represents the enhancement of the local contrast result.

[0092] Bring in data:

[0093] The number of image data N=3;

[0094] α is the weight of high-frequency filter features, α=0.2;

[0095] β is the weight of edge detection feature, β=0.2;

[0096] γ is the weight of texture analysis features, γ = 0.2;

[0097] δ is the weight of the brightness distribution feature, δ=0.2;

[0098] ϵ is the weight of the local contrast feature, ϵ=0.2;

[0099] For each image data Ii, we have the following processing results:

[0100] The results of high-frequency filtering processing H(Ii) are H(I1)=0.5, H(I2)=0.6, H(I3)=0.7; the results of edge detection processing E(Ii) are E(I1)=0.4, E(I2)=0.5, E(I3)=0.6; the results of texture analysis processing T(Ii) are T(I1)=0.3, T(I2)=0.4, T(I3)=0.5; the results of brightness distribution calculation L(Ii) are L(I1)=0.8, L(I2)=0.7, L(I3)=0.6; the results of local contrast calculation Lc(Ii) are: L c (I1)=0.2,L c (I2)=0.3,L c (I3)=0.4; the enhancement coefficient of the local contrast enhancement process C is set to 1.5 (i.e., the enhanced result is 1.5 times the original result).

[0101] Calculate the sum of each feature: high-frequency filtering sum: ∑H(Ii)=1.8; edge detection sum: ∑E(Ii)=1.5; texture analysis sum: ∑T(Ii)=1.2; brightness distribution sum: ∑L(Ii)=2.1; local contrast sum (after enhancement): ∑C×Lc(Ii)=1.35.

[0102] Calculate the weighted sum: high-frequency filtering contribution: α×∑H(Ii)=0.36; edge detection contribution: β×∑E(Ii)=0.3; texture analysis contribution: γ×∑T(Ii)=0.24; brightness distribution contribution: δ×∑L(Ii)=0.42; local contrast contribution: ϵ×∑C×Lc(Ii)=0.27.

[0103] Calculate the final feature fusion result, F=1.59, therefore, the final feature fusion result F is 1.59.

[0104] The real-time road surface data is compared with the road surface material database, and the calculation formula is as follows:

[0105] ;

[0106] Among them, Ma is the determined road material type, Z real (x, y) is the comprehensive reflection characteristic of the real-time road surface data at the position (x, y), n is the total number of road surface material types, W j is the weight coefficient of the jth road material type, which is used to adjust the contribution of different materials in the comprehensive reflection characteristics. j is the material characteristic parameter of the jth pavement material type, representing the physical properties of the material, R j (x, y) is the reflection characteristic of the jth road material at position (x, y), T j (x, y) is the texture change characteristic of the jth road material at position (x, y), α is the weight coefficient of the texture change characteristic, which is used to adjust the contribution of texture change in the comprehensive reflection characteristic, β is the weight coefficient in the matching function, which is used to adjust the influence of Euclidean distance in the matching calculation, argmax i It means to search for the maximum matching degree of the i-th material type to find the best matching road material type. When j=i, it means the i-th material.

[0107] Bring in data:

[0108] The comprehensive reflection characteristics of real-time road surface data at the position (x, y) are Zreal(x, y), the total number of road surface material types n=3, the weight coefficient Wj of the j-th road surface material type, the material characteristic parameter Mj of the j-th road surface material type, the reflection characteristics Rj(x, y) of the j-th road surface material at the position (x, y), the texture change characteristics Tj(x, y) of the j-th road surface material at the position (x, y), and the weight coefficient α of the texture change characteristics is 0.3.

[0109] The weight coefficient β in the matching function is 0.5, Zreal(x,y)=0.6, W1=0.4, W2=0.3, W3=0.3, M1=0.2, M2=0.5, M3=0.7, R1(x,y)=0.5, R2(x,y)=0.6, R3(x,y)=0.4, T1(x,y)=0.3, T2(x,y)=0.2, T3(x,y)=0.4.

[0110] Calculate the comprehensive reflective properties of each material type;

[0111] For the first material:

[0112] Z1(x,y)=[0.4×(0.2×0.5+0.3×0.3)] / (0.4+0.3+0.3)=0.076;

[0113] For the second material:

[0114] Z2(x,y)=[0.3×(0.5×0.6+0.3×0.2)] / 1=0.108;

[0115] For the third material:

[0116] Z3(x,y)=[0.3×(0.7×0.4+0.3×0.4)] / 1=0.12;

[0117] Calculate the squared distance;

[0118] For the first material:

[0119] D1 2 =(Z real (x,y)−Z1(x,y)) 2 =(0.6−0.076) 2 =0.274576;

[0120] For the second material:

[0121] D2 2 =(Z real (x,y)−Z2(x,y)) 2 =(0.6−0.108) 2 =0.242064;

[0122] For the third material:

[0123] D3 2 =(Z real (x,y)−Z3(x,y)) 2 =(0.6−0.12) 2 =0.2304;

[0124] Calculate the matching degree (using squared distance);

[0125] For the first material:

[0126] Matching degree 1 = 1 / (1 + β × D1 2 );

[0127] =1 / (1+0.5×0.274576);

[0128] ≈0.879;

[0129] For the second material:

[0130] Matching degree 2 = 1 / (1+β×D2 2 );

[0131] =1 / (1+0.5×0.242064);

[0132] ≈0.892;

[0133] For the third material:

[0134] Matching degree 3 = 1 / (1 + β × D3 2 );

[0135] =1 / (1+0.5×0.2304);

[0136] ≈0.897;

[0137] Find the maximum matching degree;

[0138] Ma=argimax(matching degree i);

[0139] Therefore, the third material has the highest matching degree, so Ma=3.

[0140] The real-time road surface data is corrected based on the road surface material, and the correction formula is as follows:

[0141] ;

[0142] Among them, Dcorrection(x,y) is the corrected road surface data at the position (x,y), D(x,y) is the original road surface data at the position (x,y), that is, the collected uncorrected road surface data, W is the material weight coefficient, which is used to adjust the influence of different material types on the correction results. The coefficient is determined according to the identified material type, B(x,y) is the material reference value, which represents the reference reflectivity or reference characteristic value of a specific material type at the position (x,y), O(x,y) is the superimposed correction effect, which represents the correction effect superimposed in the comprehensive correction step, which may include crack detection, optimization of correction parameter steps, and adjustment of data. In general, F(x,y) is the texture correction factor, which indicates the correction effect of texture changes on the data at position (x,y). A(x,y) is the application correction coefficient, which indicates further adjustment of the data in the final application correction step. The coefficient may be determined based on specific application requirements or environmental conditions. Specifically, A(x,y) can adjust the direction of the overall result. When the application scenario requires the overall data to be larger, A(x,y) can be adjusted to a higher value. When the application scenario requires the overall data to be smaller, A(x,y) can be adjusted to a smaller value. The adjustment can be customized according to the usage scenario to achieve the desired result.

[0143] The input data are as follows: original road surface data D(x,y)=100; material weight coefficient W=0.8; material reference value B(x,y)=90; superposition correction effect O(x,y)=5; texture correction factor F(x,y)=1.1; application correction coefficient A(x,y)=1.05; W×[D(x,y)−B(x,y)]=0.8×[100−90]=8;

[0144] Superposition correction effect: after superposition = 13;

[0145] Apply the texture correction factor and apply the correction coefficient: Dcorrected(x,y)=13×1.1×1.05=15.015; for simplicity, we can round the result to two decimal places, Dcorrected(x,y)≈15.02:, therefore, the corrected road surface data Dcorrected(x,y) is 15.02.

[0146] The pre-processed real-time road data is compared with the preset threshold, and the data above the threshold is marked. The calculation formula is as follows:

[0147] ;

[0148] Among them, B(x,y) is the comparison result at position (x,y), D 预处理 (x, y) is the preprocessed road surface data at position (x, y), T is the preset threshold used for comparison with the preprocessed road surface data, γ is the gain coefficient used to amplify the difference between the preprocessed road surface data and the preset threshold, and δ is the offset used to adjust the benchmark of the comparison result.

[0149] Bring in data:

[0150] Preprocessed road surface data Dpreprocessed (x, y) = 120; preset threshold T = 100; gain coefficient γ = 1.5; offset δ = 10;

[0151] Calculate the difference between the pre-processed road surface data and the preset threshold: D 预处理 (x,y)−T=120−100=20;

[0152] Zoom in on the difference:

[0153] γ×[Dpreprocessing(x,y)−T]=30;

[0154] Adjust the baseline of the comparison results:

[0155] B(x,y)=40, so the alignment result B(x,y) at position (x,y) is 40;

[0156] This result shows that at position (x, y), the preprocessed road surface data is 20 higher than the preset threshold. After being amplified by the gain coefficient and adding the offset, the final comparison result is 40.

[0157] The real-time road surface data is compared with the multi-feature fusion database. The similarity calculation formula between the real-time road surface data and the samples in the database is as follows:

[0158] ;

[0159] Among them, S(F 实 ,F 库 ) represents the real-time road data feature vector F 实 The cosine similarity between the feature vector F of the k-th sample in the database, F is the feature vector obtained after feature fusion of real-time road data, and F is the feature vector obtained after feature fusion of the k-th sample in the database, ∥F 实 ∥ represents the real-time norm of the eigenvector F, that is, the square root of the sum of the squares of the elements of the vector, ∥F 库 ∥ represents the eigenvector F 库 The norm of .

[0160] Bring in data for calculation:

[0161] Freal=[1,2,3];

[0162] Fbank=[4,5,6];

[0163] Compute the dot product:

[0164] Freal × Fbank = 1 × 4 + 2 × 5 + 3 × 6 = 32;

[0165] Compute the square of the norm:

[0166] Freal×Freal=12+22+32=14;

[0167] F library × F library = 42 + 52 + 62 = 77;

[0168] Calculate S(Freal, Flix)2:

[0169] S(F entity, F library) 2 =(Factory × Flier) 2 / (Freal×Freal)×(Flibrary×Flibrary);=32 2 / (14×77);

[0170] =1024 / 1078≈0.950;

[0171] Taking the square root yields S(Freal, Fbase):

[0172] S(Freal, Flier)≈0.975;

[0173] Therefore, the cosine similarity S(Freal, Fbank) between the real-time road data feature vector Freal and the k-th sample feature vector Fbank in the database is approximately 0.975.

[0174] Find the database sample with the highest similarity to the real-time road data. The calculation formula is as follows:

[0175] ;

[0176] Among them, k ∗ Indicates the index of the database sample with the highest similarity to the real-time road data, arg max Indicates taking the value of the independent variable corresponding to the maximum value, that is, finding the value that makes S(F 实 ,F 库 ) the largest k value, k=1,2,…,M represents the range of k, from 1 to M, where M is the total number of samples in the database, S(F 实 ,F 库 ) represents the cosine similarity between the feature vector of real-time road surface data and the feature vector of the kth sample in the database.

[0177] Bring in data for calculation:

[0178] Freal=[1,2,3];

[0179] There are 3 sample feature vectors in the database:

[0180] Fbank1=[4,5,6];

[0181] Fbank2=[1,0,0];

[0182] Fbank3=[2,2,2];

[0183] Calculate S(Freal, Flix1)2:

[0184] Freal × Fbank1 = 1 × 4 + 2 × 5 + 3 × 6 = 32;

[0185] Freal×Freal=12+22+32=14;

[0186] F bank 1×F bank 1=42+52+62=77;

[0187] S(F real, F library 1)2=32 2 / (14×77)=1024 / 1078;

[0188] Calculate S(Freal, Flix2)2:

[0189] Freal × Fbank2 = 1 × 1 + 2 × 0 + 3 × 0 = 1;

[0190] F library 2 × F library 2 = 1 2 +0 2 +0 2 = 1;

[0191] S(F actual, F library 2)² = 1 2 / (14 × 1) = 1 / 14;

[0192] Calculate S(F actual, F library 3)²:

[0193] F actual × F library 3 = 1 × 2 + 2 × 2 + 3 × 2 = 12;

[0194] F library 3 × F library 3 = 2 2 +2 2 +2 2 = 12;

[0195] S(F actual, F library 3)² = 12 2 / (14 × 12) = 144 / 168;

[0196] Then compare the values of these three S(F actual, F library k)², and find the index k* corresponding to the maximum value;

[0197] Through calculation, we can find that:

[0198] S(F actual, F library 1)² ≈ 0.950;

[0199] S(F actual, F library 2)² ≈ 0.071;

[0200] S(F actual, F library 3)² ≈ 0.857;

[0201] Therefore, k ∗ = 1, that is, the database sample with the highest similarity to the real-time road surface data is the first sample.[[ID=K]]

[0202] By comparing the values of S(F actual, F library k)², we can avoid using square roots in the calculation process and ensure the consistency of the calculation results.

[0203] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended embodiments and their equivalents.

Claims

1. A laser radar-based intelligent detection method for road pavement cracks, characterized by: The method comprises: Collect road surface data and process historical road surface data; Construct a pavement material database, record and store the designed pavement material, reflection type and texture changes based on pavement data, and obtain comprehensive reflection characteristics; Construct a multi-feature fusion database to collect and combine multiple layers of crack features based on pavement data; Collect real-time road data, collect and record various road data information in real time; Determine the road surface material by comparing the real-time road surface data with the road surface material database to confirm the road surface material; Correct the collected road surface data and correct the real-time road surface data based on the road surface material; Preprocess data to preprocess the corrected real-time road surface data to remove invalid, erroneous or abnormal information in the original data to ensure the accuracy and reliability of the data; Preset threshold comparison: compare the pre-processed real-time road data with the preset threshold, and mark the data above the threshold to facilitate subsequent repeated detection and improve accuracy; Identify cracks by comparing real-time pavement data with a multi-feature fusion database to identify cracks and improve the accuracy of crack detection; The calculation formula of the reflection characteristics is as follows: ， Where R(x,y) represents the comprehensive reflection characteristics at a specific position (x,y) on the road surface, n represents the total number of road surface material types, and M i Represents the material characteristic parameters of the i-th road material type, R i (x,y) represents the reflection characteristics of the i-th road material at position (x,y), T i (x,y) represents the texture change characteristics of the i-th road material at position (x,y); The multi-layer features of the cracks are collected and combined, and the calculation formula is as follows: ， Among them, F represents the final feature fusion result, α, β, γ, δ and ϵ correspond to H, E, T, L and L respectively. c The weight coefficient of the feature is used to adjust the contribution of different features in the final fusion result. N represents the number of image data, I i Represents the i-th image data, H(I i ) represents high-frequency filtering of the i-th image data, where H represents high-frequency feature extraction, E(I i ) represents edge detection processing for the i-th image data, where E represents edge detection, T(I i ) represents the texture analysis processing of the i-th image data, where T represents texture analysis, L(I i ) represents the brightness distribution calculation of the i-th image data, where L represents the brightness feature extraction, L c (I i ) represents the local contrast calculation for the i-th image data, where L c Represents local contrast calculation, and C represents the enhancement of local contrast results; The real-time road surface data is corrected based on the road surface material, and the correction formula is as follows: ， Among them, Dcorrection(x,y) is the corrected road surface data at position (x,y), D(x,y) is the original road surface data at position (x,y), that is, the collected uncorrected road surface data, W is the material weight coefficient, which is used to adjust the influence of different material types on the correction results. The coefficient is determined according to the identified material type, B(x,y) is the material baseline value, which represents the baseline reflectivity or baseline characteristic value of a specific material type at position (x,y), O(x,y) is the superposition correction effect, which represents the correction effect superimposed in the comprehensive correction step, F(x,y) is the texture correction factor, which represents the correction influence of texture change on the data at position (x,y), and A(x,y) is the application correction coefficient, which represents the further adjustment of the data in the final application correction step. The coefficient is determined based on application requirements or environmental conditions.

2. The method for intelligent detection of road pavement cracks based on laser radar according to claim 1, characterized in that: The real-time road surface data is compared with the road surface material database, and the calculation formula is as follows: ， Among them, Ma is the determined road material type, Z real (x, y) is the comprehensive reflection characteristic of the real-time road surface data at the position (x, y), n is the total number of road surface material types, W j is the weight coefficient of the jth road material type, which is used to adjust the contribution of different materials in the comprehensive reflection characteristics. j is the material characteristic parameter of the jth pavement material type, which represents the physical properties of the material, R j (x, y) is the reflection characteristic of the jth road material at position (x, y), T j (x, y) is the texture change characteristic of the jth road material at position (x, y), α is the weight coefficient of the texture change characteristic, which is used to adjust the contribution of texture change in the comprehensive reflection characteristic, β is the weight coefficient in the matching function, which is used to adjust the influence of Euclidean distance in the matching calculation, argmax i It means to search for the maximum matching degree of the i-th material type to find the best matching road material type. When j=i, it means the i-th material.

3. The method for intelligent detection of road pavement cracks based on laser radar according to claim 1, characterized in that: The pre-processed real-time road data is compared with the preset threshold, and the data above the threshold is marked. The calculation formula is as follows: ， Among them, B(x,y) is the comparison result at position (x,y), D 预处理 (x, y) is the preprocessed road surface data at position (x, y), T is the preset threshold used for comparison with the preprocessed road surface data, γ is the gain coefficient used to amplify the difference between the preprocessed road surface data and the preset threshold, and δ is the offset used to adjust the benchmark of the comparison result.

4. The method for intelligent detection of road pavement cracks based on laser radar according to claim 1, characterized in that: The real-time road surface data is compared with the multi-feature fusion database. The similarity calculation formula between the real-time road surface data and the samples in the database is as follows: ， Among them, S(F 实 ,F 库 ) represents the real-time road data feature vector F 实 The cosine similarity between the feature vector F of the k-th sample in the database, F is the feature vector obtained after feature fusion of real-time road data, and F is the feature vector obtained after feature fusion of the k-th sample in the database, ∥F 实 ∥ represents the real-time norm of the eigenvector F, that is, the square root of the sum of the squares of the elements of the vector, ∥F 库 ∥ represents the eigenvector F 库 The norm of Find the database sample with the highest similarity to the real-time road data. The calculation formula is as follows: ， Among them, k ∗ Indicates the index of the database sample with the highest similarity to the real-time road data, arg max Indicates taking the value of the independent variable corresponding to the maximum value, that is, finding the value that makes S(F 实 ,F 库 ) the largest k value, k=1,2,…,M represents the range of k, from 1 to M, where M is the total number of samples in the database, S(F 实 ,F 库 ) represents the cosine similarity between the feature vector of real-time road surface data and the feature vector of the kth sample in the database.

5. A laser radar-based intelligent road pavement crack detection system, characterized by: A method for intelligent detection of cracks in a highway pavement based on a laser radar as claimed in any one of claims 1 to 4 is used, comprising: Collection module: used to collect road surface data, collect and process historical road surface data, and collect real-time road surface data, and collect and record various road surface data information in real time; Storage module: used to build a pavement material database, record and store the designed pavement material, reflection type and texture changes based on pavement data to obtain comprehensive reflection characteristics; and to build a multi-feature fusion database, collect and combine multi-layer features of cracks based on pavement data; Data processing module: used to determine the pavement material, compare the real-time pavement data with the pavement material database, confirm the pavement material, and correct the collected pavement data, correct the real-time pavement data based on the pavement material, and pre-process the data, pre-process the corrected real-time pavement data, remove invalid, erroneous or abnormal information in the original data, ensure the accuracy and reliability of the data, and use the preset threshold comparison to compare the pre-processed real-time pavement data with the preset threshold, mark the data above the threshold, facilitate subsequent repeated detection, improve accuracy, and use the preset threshold comparison to compare the pre-processed real-time pavement data with the preset threshold, mark the data above the threshold, facilitate subsequent repeated detection, improve accuracy, and identify cracks, compare the real-time pavement data with the multi-feature fusion database, identify cracks, and improve the accuracy of crack detection.

6. A laser radar-based intelligent device for detecting road cracks, characterized by: A method for intelligent detection of cracks in a highway pavement based on a laser radar according to any one of claims 1 to 4 comprises: Data storage server, which collects and processes historical road surface data; Millimeter-wave radar and laser scanners for collecting real-time road surface data; High-performance computers are used to build a pavement material database to obtain comprehensive reflection characteristics; to build a multi-feature fusion database to collect and combine multi-layer features of cracks; to determine pavement material; to correct collected pavement data; to pre-process data to remove invalid, erroneous or abnormal information in the original data; to compare pre-processed real-time pavement data with preset thresholds, and to mark data above the threshold to facilitate subsequent repeated detection and improve accuracy; and to identify cracks, compare real-time pavement data with the multi-feature fusion database to improve the accuracy of crack detection.

Images