A method for calibrating the three-dimensional depth modal deviation of a road surface

Abstract

The present invention discloses a method for calibrating the three-dimensional depth modal deviation of a road surface, which relates to the technical field of computer image processing. The calibration method includes data acquisition and storage, floating data repair, row granularity deviation calibration, and calibration data verification and storage. By combining a high-speed three-dimensional structured light measurement camera assembly with a movable detection platform, the present invention realizes the accurate acquisition of the three-dimensional depth modal of the road surface. At the same time, a floating data repair method using dynamic regression parameters and a sliding threshold region effectively repairs the floating outliers in the three-dimensional depth modal. In addition, a row granularity deviation calibration method based on a dynamic regression parameter reference plane is proposed to calibrate the three-dimensional depth modal of the road surface row by row, significantly improving the row granularity deviation problem. The present invention effectively improves the data deviation problem, provides reliable data support for road surface disease analysis, and improves the accuracy and efficiency of road surface detection.

Description

Technical Field

[0001] The present invention relates to the technical field of computer image processing, and specifically provides a method for calibrating the deviation of the three-dimensional depth mode of a road surface. Background Art

[0002] With the development of traffic infrastructure, using advanced imaging devices and computer vision algorithms for road surface disease detection has become the key to ensuring regular monitoring and maintenance of the road surface condition and maintaining its service performance. However, since most traditional road surface disease detection technologies rely on two-dimensional imaging devices, the lack of three-dimensional depth information of the road surface often makes it impossible to accurately perceive specific types of diseases, and it is easily affected by environmental light interference and shadows, resulting in inaccurate detection and low recall rate.

[0003] Structured light imaging technology, with its high-speed three-dimensional measurement and continuous data acquisition capabilities, has become an indispensable mainstream means in road surface disease detection, providing key data support for accurate disease analysis. However, when collecting the three-dimensional depth mode of the road surface, due to inevitable factors such as equipment vibration or unilateral longitudinal displacement deviation, the data often deviates, affecting the accuracy of the subsequent detection results, and severely restricting the development and application of road surface disease detection technology. Therefore, developing a method for calibrating the deviation of the three-dimensional depth mode of the road surface can not only improve the accuracy of subsequent road surface disease detection but also provide reliable data support for subsequent road surface maintenance and management. Summary of the Invention

[0004] The purpose of the present invention is to make up for the deficiencies of the prior art and provide a method for calibrating the deviation of the three-dimensional depth mode of a road surface. The specific steps of this method include the preparation and installation of three-dimensional depth mode imaging equipment, data collection and storage, data quality detection, floating data repair, row granularity deviation calibration, and data storage and reporting, solving the data deviation problem caused by vibration or unilateral longitudinal displacement deviation during the collection of the three-dimensional depth mode of the road surface by structured light imaging equipment, and improving the accuracy and efficiency of road surface disease detection.

[0005] To solve the above technical problems, the present invention provides the following technical solution: A method for calibrating the deviation of the three-dimensional depth mode of a road surface. The specific steps of this calibration method are as follows:

[0006] S100, Data Acquisition and Storage: Install the high-speed three-dimensional structured light measurement camera assembly on the movable detection platform and adjust it to the preset angle and height. Connect the camera assembly to the data transmission line through the RS422 interface, access the HES-2048 incremental encoder for control triggering, debug various parameters of the camera according to the test road surface environment illumination and road surface characteristics, and monitor whether the histogram and gray-scale distribution of the visualized image of the acquired data are normal. The camera assembly processor processes the optical signal, and the depth sensing module combines the structured light source to calculate the three-dimensional depth mode of the road surface and forms dual-mode data with the two-dimensional intensity image, which is transmitted to the computer in real time for road surface reconstruction processing and storage;

[0007] S200, Floating Data Repair: Read the three-dimensional depth mode, and use the floating data repair method based on dynamic regression parameters and sliding threshold regions to repair the three-dimensional depth floating outliers. Calculate the dynamic regression parameter DRP corresponding to the input three-dimensional depth mode. The dynamic regression parameter is jointly calculated from different weight combinations of the mean, median, and mode of the three-dimensional depth mode and the bias variable. Define the processing region and the sliding threshold region corresponding to the step size, traverse all the values in the processing region except the center point, and repair the three-dimensional depth mode values less than the threshold with the minimum value greater than DRP in this region. Here, exp() represents the exponential function, and n represents the three-dimensional depth mode value. Repeat this process until all three-dimensional data is processed, and finally output the processing result;

[0008] S300, Row Granularity Deviation Calibration: Read the three-dimensional depth mode after floating repair and calculate its dynamic regression parameter again. Use the row granularity deviation calibration method based on the dynamic regression parameter reference plane to calibrate the data. Strip the read data in row granularity units, read the three-dimensional depth mode data of a single row granularity, use the dynamic regression parameter as the reference plane, calculate the slope, deviation, and unary linear calibration function of the current row, traverse the positions of the depth data of the current row granularity, and the corresponding position index always ranges from I to the maximum width of the data size. Calculate the difference between the index position and the dynamic regression parameter using the slope and deviation of the unary linear calibration function, and determine the data point offset direction according to the positive or negative of the difference. If the difference is less than zero, it offsets upward, otherwise it offsets downward, and the offset amount is the absolute value of the difference, until all rows in the three-dimensional data are calibrated, and output the calibrated data;

[0009] S400, Calibration Data Verification and Storage: Visualize the calibrated data, observe the improvement of row-level deviation by comparing with the visualization effect of the data before calibration, and at the same time randomly extract some areas from the calibrated three-dimensional depth modality for detailed analysis, calculate the pavement evenness index and slope change parameters, compare them with the actual measured values to verify the accuracy of the calibrated data. After verification, classify and store the original, pre-calibrated, post-calibrated data and analysis results, and generate a report containing information about the test section, equipment status, and data quality assessment.

[0010] Further, in the S100, Data Acquisition and Storage, the high-speed three-dimensional structured light measurement camera assembly acquires data in a line array manner and is triggered for imaging by a HES-2048 incremental encoder.

[0011] Even further, the HES-2048 incremental encoder in the S100, Data Acquisition and Storage generates a profile trigger signal through an RS422 interface converter, which cooperates with the camera assembly for imaging triggering, avoiding data stretching and compression, and changing the displacement of one wheel rotation into 2048 periodic pulse signals.

[0012] Even further, the auxiliary software in the S100, Data Acquisition and Storage includes cxDiscover, cxExplorer, and cxShow3d. cxDiscover is used to monitor the working status of the camera assembly and modify and fix the corresponding IP address of the camera assembly. cxExplorer is used to debug the parameters of the camera assembly and monitor the quality of the acquired data. cxShow3d is used for pseudo-color visualization of the three-dimensional depth modality.

[0013] Even further, the specific steps of the floating data repair method based on dynamic regression parameters and sliding threshold regions in the S200, Floating Data Repair are as follows:

[0014] (1) Read the three-dimensional depth modality data I;

[0015] (2) Calculate its corresponding dynamic regression parameters (DRP) for I. The formula is: DRP = exp(0.3Avg(I) + 0.5Med(I) + 0.2Most(I) + σ), where Avg(·) represents mean calculation, Med(·) represents median statistics, Most(·) is mode statistics, σ is a small bias variable, and I represents the read three-dimensional depth modality;

[0016] (3) Define the processing area A (n×n) , where n can be dynamically adjusted, and use the center point of A as the floating data point I to be repaired a ;

[0017] (4) Traverse the numbers in area A except I a other than. If there is a three-dimensional depth floating anomaly value less than , it is repaired with the minimum value greater than DRP in this area;

[0018] (5) Repeat steps (3) to (4) with a step size of n until all three-dimensional depth modal data I is processed, and output the processing result

[0019] Furthermore, for the S300, the specific steps of the row granularity deviation calibration method based on the dynamic regression parameter reference plane in the row granularity deviation calibration are as follows:

[0020] (1) Read the three-dimensional depth mode after floating data repair

[0021] (2) Extract the row granularity of the read-in to obtain Q n ;

[0022] (3) Read Q n Using DRP as the reference plane, find the slope q n of the current Q n and the offset s n , and form the corresponding unary linear calibration function f (q,s) (x) = q n x i + s n ;

[0023] (4) Traverse the positions of the current Q n , and the corresponding position index i always ranges from I to the maximum width w of the data size;

[0024] (5) Use the slope q (q,s) of the unary linear calibration function f n and the deviation s n to calculate the difference err of the index position i relative to DRP, err = abs(DRP - f (q,s) (x i ));

[0025] (6) The correction value of each item i of the current Q n is x i ± err;

[0026] (7) Operate according to steps (4) to (6) until all rows in the three-dimensional data are calibrated, and output the calibrated data

[0027] Further, in the S300, the calculation formula for the difference err in the row granularity deviation calibration is: err = abs(DRP - f (q,s) (x i ))), where x i represents the position index of the current row data point, q n is the slope of the unary linear calibration function corresponding to the current row granularity Q n , and s n is the deviation of the unary linear calibration function corresponding to the current row granularity. q n and s n are determined according to the fitting straight line equation, and the formula is: f (q,s) (x) = q n x i + s n .

[0028] Further, in the S300, for the calculation of the data offset in the row granularity deviation calibration, the up and down offset is err, that is, the distance that the data point needs to be offset. The offset direction is determined by the positive or negative of DRP - f (q,s) (x i ). Up or down is relative to the reference plane. Up represents the direction of increasing depth value, and down represents the direction of decreasing depth value, that is, x i ± err.

[0029] Compared with the prior art, the three-dimensional depth modal deviation calibration method for road surfaces has the following beneficial effects:

[0030] First, in the data acquisition stage, the present invention uses a high-speed three-dimensional structured light measurement camera component to collect data in a line array manner and accurately trigger imaging with an incremental encoder, effectively avoiding the problems of data stretching and compression, ensuring the original accuracy of the data. In the data processing stage, through the floating data repair method combining dynamic regression parameters and a sliding threshold region, it can accurately identify and repair floating outliers in the three-dimensional depth modal, further improving the reliability of the data. In addition, based on the row granularity deviation calibration method of the dynamic regression parameter reference plane, it can effectively calibrate the row granularity deviation in the data, making the finally output three-dimensional depth modal more real and accurate, providing strong support for the accurate positioning and quantitative analysis of road surface diseases.

[0031] Second, through the integrated application of the acquisition assistance software, the present invention realizes the real-time monitoring and adjustment of the working state of the camera component and the data acquisition quality, ensuring the stability and efficiency of the data acquisition process. At the same time, the pseudo-color visualization processing technology is used to visualize the three-dimensional depth modality, which not only improves the readability of the data but also facilitates the user to intuitively observe the information on the location, scope, and type of road surface diseases. In addition, through the automated report generation function, the original data, the data before and after calibration, and the analysis results are classified and stored, and a detailed report containing information on the test section, equipment status, and data quality assessment is generated, providing convenient data support for subsequent decision-making analysis.

[0032] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0034] Figure 1 It is a flowchart of a method for calibrating the deviation of the three-dimensional depth modality of a road surface;

[0035] Figure 2 It is a schematic diagram of the distribution of data floating;

[0036] Figure 3 It is a schematic diagram of the data offset caused by the unilateral vertical displacement;

[0037] Figure 4 It is a schematic diagram of the process of calibrating the row granularity deviation;

[0038] Figure 5 It is a schematic diagram of the improvement effect of the test sample image quality by different filtering methods. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0040] Embodiment 1:

[0041] Urban road detection

[0042] Implementation environment and conditions: A section of urban road was selected as the test section. The road surface had common wear, repairs and slight unevenness, and the traffic flow was moderate, which was convenient for testing. The test was carried out during the day when the weather was clear and the ambient lighting met the normal working requirements of the camera to ensure the quality of the collected data. A specially configured mobile detection platform was used with a stable and moderate driving speed to provide stable motion conditions for data collection.

[0043] Specific implementation steps:

[0044] Data acquisition and storage (S100), properly install the high-speed 3D structured light measurement camera assembly on the detection platform, carefully adjust the camera angle to make it perpendicular to the road surface, and set the appropriate height to ensure that clear and accurate road surface images can be obtained. Connect the camera assembly and the data transmission line via the RS422 interface, and connect the HES-2048 incremental encoder. The encoder generates an accurate contour trigger signal through the RS422 interface converter, and works in conjunction with the camera assembly imaging trigger to accurately convert the displacement of one wheel rotation into 2048 pulse signals, thereby achieving precise control of the data acquisition frequency and position. The actual ambient lighting conditions of the road surface are taken into consideration, and various parameters of the camera, such as exposure time and gain, are carefully adjusted. At the same time, the quality of the collected data is closely monitored with the help of the auxiliary software cxExplorer to ensure that the histogram and grayscale distribution of the visualization image of the collected data are in normal state. The camera component processor quickly processes the light signal, and the depth sensing module combines the structured light source to calculate the three-dimensional depth mode of the road surface, which is formed into dual-modal data with the two-dimensional intensity image and transmitted to the computer in real time for road surface reconstruction and storage. In this process, the cxShow3d software can be used to perform pseudo-color visualization of the three-dimensional depth mode for intuitive observation of the data.

[0045] Floating data repair (S200), the computer reads the collected three-dimensional depth modal data completely, and uses the floating data repair method based on dynamic regression parameters and sliding threshold areas to process the data. First, the dynamic regression parameter DRP is calculated according to the formula: DRP = exp(0.3Avg(I) + 0.5Med(I) + 0.2Most(I) + σ), then a processing area A of appropriate size is defined (its size is flexibly determined according to the local characteristics of the road surface and the data distribution), and the center point of A is set as the floating data point I to be repaired. a Then, traverse the area A except I a For all values other than For the three-dimensional depth floating outliers, they are repaired with the minimum value greater than DRP in this area, and this process is repeated with a step size equal to the side length of the processing area A until all three-dimensional depth modal data are processed, and finally the repaired three-dimensional depth modal data is output.

[0046] Row granularity deviation calibration (S300), read the three-dimensional depth modal data after floating repair, accurately strip the data according to the row granularity unit, read the three-dimensional depth modal data of a single row granularity, use the previously calculated dynamic regression parameter DRP as the reference plane, and find the current slope q through the corresponding algorithm n and offset s n , to form the corresponding unary linear calibration function, the formula is: f (q,s) (x) = q n x i + s n , then, perform a comprehensive position traversal on the depth data of the current row granularity, with the corresponding position index i ranging from I to the maximum width of the data size, and use the unary linear calibration function f (q,s) (x) slope q n and deviation s n to calculate the difference err of the index position i relative to DRP, the formula is: err = abs(DRP - f (q,s) (x i ))), determine the offset direction of the data point according to the positive and negative situation of the difference err. When DRP - f (q,s) (x i ) < 0, it offsets upward, otherwise it offsets downward, and the offset amount is err, and the correction value is x i ±err (accurately determine the positive and negative according to the offset direction). According to this operation process, until all rows in the three-dimensional data are calibrated, the calibrated data is output.

[0047] Calibration data verification and storage (S400), perform visualization processing on the calibrated data, and through a detailed comparison with the visualization effect of the data before calibration, intuitively observe the improvement of the row granularity deviation, such as whether the road surface texture becomes clearer and more continuous, and whether the contours of potholes and repair marks are more accurate. Randomly select a certain number and area of regions from the calibrated three-dimensional depth modal for in-depth and detailed analysis, calculate the road surface flatness index (such as the international roughness index IRI) and the slope change parameters, and strictly compare the calculated results with the actual measured values (measured using high-precision professional measuring instruments) to verify the accuracy of the calibrated data. The specific verification method is to check whether the error between the flatness index and slope change parameters calculated from the calibrated data and the actual measured values is within the allowable range.

[0048] Classify and properly store the original, pre-calibration, and post-calibration data, as well as the detailed analysis results, and generate a detailed report that includes detailed information about the test section (such as the specific location, length, and width of the section), equipment status (including various parameters of the camera and the driving speed of the detection platform), and data quality assessment (covering key aspects such as error analysis and visualization comparison results), providing comprehensive and reliable data support for the maintenance and management of urban roads.

[0049] In summary: This invention demonstrates good practicality and effectiveness in the urban road detection scenario. By carefully installing the camera components, accurately setting relevant parameters, and reasonably utilizing the auxiliary software, high-quality three-dimensional depth modal data of the road surface has been successfully collected and stored. In the floating data repair process, the method based on dynamic regression parameters and sliding threshold regions effectively processes outliers, improving the accuracy of the data. The row granularity deviation calibration further optimizes the data, significantly improving the visualization effect of road surface features. And by comparing with actual measurement values, the accuracy of the post-calibration data in calculating flatness indicators and slope change parameters is verified. The generated detailed report provides rich and reliable data basis for the maintenance and management of urban roads, helping to accurately judge the road surface conditions, reasonably plan maintenance strategies, improve the safety and comfort of urban roads, and at the same time providing a reference example for subsequent similar road detection work.

[0050] Embodiment 2:

[0051] Highway Detection

[0052] Implementation environment and conditions: Select a section of highway as the test section. This section has a large traffic flow and high vehicle speeds, with relatively uniform road surface wear but typical problems such as some ruts and minor cracks. The detection is carried out during a period with relatively low traffic flow at night to minimize the impact on traffic and ensure the safety and efficiency of the detection operation. The weather conditions are good, without precipitation and fog interference, but the environmental light is weak, and it is necessary to rely fully on the lighting system of the detection equipment itself to assist in data collection work. Use a professionally modified high-speed detection vehicle as the mobile detection platform. This vehicle has high-speed driving ability and excellent stability, can operate safely and smoothly on the highway, and is equipped with a high-performance shock absorption system to effectively reduce the bumps during vehicle driving, creating favorable conditions for data collection.

[0053] Specific implementation steps

[0054] Data acquisition and storage (S100): Precisely install a high-speed three-dimensional structured light measurement camera assembly at a specific position on the inspection vehicle. Adjust the camera angle to an appropriate angle with the road surface (such as a 45-degree angle), and set a reasonable height to detect a wider area of the road surface and obtain more detailed information. Connect the camera assembly to the in-vehicle data transmission line through the RS422 interface, access the HES-2048 incremental encoder, and reasonably set the encoder parameters to ensure accurate triggering of camera imaging in the high-speed driving state. Precisely convert the displacement of one wheel rotation into 2048 pulse signals to achieve precise synchronous data acquisition. According to the actual situation of weak ambient light at night, carefully adjust the key parameters of the camera's exposure time, gain, and structured light source intensity. Use the auxiliary software cxExplorer to strictly monitor the quality of the acquired data to ensure that the histogram and gray distribution of the visualized image of the acquired data are normal. The camera assembly processor quickly processes the optical signal, and the depth sensing module combines the structured light source to calculate the three-dimensional depth mode of the road surface, forming dual-modal data with the two-dimensional intensity image, which is transmitted to the computer in real time for road surface reconstruction processing and storage. During this period, use the cxShow3d software to perform pseudo-color visualization on the three-dimensional depth mode to assist data observation and analysis.

[0055] Floating data repair (S200): Completely read the acquired three-dimensional depth mode data, and process the data using a floating data repair method based on dynamic regression parameters and sliding threshold regions. Calculate the dynamic regression parameter DRP according to the formula: DRP = exp(0.3Avg(I) + 0.5Med(I) + 0.2Most(I) + σ). Define a processing region A with an appropriate size (determine its size according to the road surface continuity and data correlation), and set the center point of A as the floating data point I to be repaired. a , Traverse all the values in the region A except I a . If there are three-dimensional depth floating outliers less than , then repair them with the minimum value greater than DRP in this region. Repeat this process with a step size equal to the side length of the processing region A until all the three-dimensional depth mode data is processed, and output the repaired three-dimensional depth mode data.

[0056] Row granularity deviation calibration (S300): Read the three-dimensional depth mode data after floating repair, precisely strip the data by row granularity units, read the three-dimensional depth mode data of a single row granularity, use the dynamic regression parameter DRP as the reference plane, and calculate the slope q n and offset s n of the current row through relevant algorithms, and form a corresponding unary linear calibration function. The formula is: f (q,s) (x) = q n x i + s n, perform a comprehensive position traversal on the depth data at the current line granularity. The corresponding position index i always ranges from I to the maximum width w of the data size, and utilize the slope q of the unary linear calibration function n and the deviation s n Calculate the difference err between the index position i and DRP. The formula is: err = abs(DRP - f (q,s) (x i ))). Determine the offset direction of the data point according to the positive or negative value of the difference err. When DRP - f (q,s) (x i ) < 0, offset upward; otherwise, offset downward. The offset amount is err, and the correction value is x i ±err (determine the positive or negative according to the offset direction). Follow this operation process until all rows in the three-dimensional data are calibrated, and then output the calibrated data.

[0057] Calibration data verification and storage (S400). Perform visualization processing on the calibrated data, carefully compare its visualization effect with that of the data before calibration, and observe the improvement of the line granularity deviation. For example, whether the shapes of ruts and cracks are clearer and whether the depth measurement is more accurate. Randomly extract a certain number and area of regions from the calibrated three-dimensional depth modality for in-depth analysis, calculate the pavement evenness index and slope change parameters, and compare the calculation results with the actual measurement values (obtained using high-precision professional measurement instruments) to verify the accuracy of the calibrated data, that is, check whether the errors between the evenness index and slope change parameters calculated from the calibrated data and the actual measurement values are within a reasonable range.

[0058] Classify and properly store the original, pre-calibration, post-calibration data and analysis results, and generate a detailed report containing test section information (section location, length, number of lanes), equipment status (camera parameters, detection vehicle driving speed), data quality assessment (error analysis, visualization comparison results), providing strong data support for the maintenance and management of highway pavements.

[0059] In summary, in the highway detection scenario, the present invention fully exerts its advantages. Considering the actual situation of large traffic flow, high vehicle speed, and road surface characteristics on highways, appropriate detection time periods and equipment installation methods are selected to ensure the feasibility and accuracy of data collection. During the data collection process, the collaborative work of the camera component and the encoder, as well as the reasonable debugging of parameters, provide a solid data foundation for subsequent processing. The floating data repair and line-granularity deviation calibration processes effectively improve the data quality. From the visualization comparison and the comparison results with actual measurement values, the calibrated data can more accurately reflect the rutting and crack conditions of the road surface, and the accuracy of the flatness index and slope change parameters is effectively verified. The finally generated report contains comprehensive information, providing strong data support for highway maintenance decision-making and safety assessment work, which is of great significance for ensuring the efficient operation and driving safety of highways and also provides a practical reference for the further development of highway detection technology.

[0060] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A three-dimensional depth modal deviation calibration method for road surfaces, characterized in that, The specific steps of this calibration method are as follows: S100, Data acquisition and storage: Install the high-speed three-dimensional structured light measurement camera component on the movable detection platform and adjust it to the preset angle and height. Connect the camera component and the data transmission line through the RS422 interface, access the HES-2048 incremental encoder to control the trigger, debug various parameters of the camera according to the test road surface environment illumination and road surface characteristics, and monitor whether the histogram and gray distribution of the visualized image of the collected data are normal. The camera component processor processes the optical signal, and the depth sensing module combines the structured light source to calculate the three-dimensional depth mode of the road surface and forms dual-mode data with the two-dimensional intensity image, which is transmitted to the computer in real time for road surface reconstruction processing and storage; S200, Floating data repair: Read the three-dimensional depth modality, and use the floating data repair method based on dynamic regression parameters and sliding threshold regions to repair the three-dimensional depth floating outliers. Calculate the dynamic regression parameters DRP corresponding to the input three-dimensional depth modality. The dynamic regression parameters are calculated jointly by different weight combinations of the mean, median, and mode of the three-dimensional depth modality and the bias variable. Define the processing region and the sliding threshold region corresponding to the step size. Traverse all the values in the processing region except the center point, and repair the three-dimensional depth modality values less than the threshold with the minimum value greater than DRP in this region. Here, exp() represents the exponential function, and n represents the three-dimensional depth modality value. Repeat this process until all three-dimensional data processing is completed, and finally output the processing result; S300, Row granularity deviation calibration: Read the three-dimensional depth mode after floating repair and calculate its dynamic regression parameters again. Use the row granularity deviation calibration method based on the reference plane of the dynamic regression parameters to calibrate the data. Strip the read data in row granularity units, read the three-dimensional depth mode data of a single row granularity, use the dynamic regression parameters as the reference plane, calculate the slope, deviation and unary linear calibration function of the current row, traverse the positions of the depth data of the current row granularity, and the corresponding position index always ranges from I to the maximum width of the data size. Calculate the difference between the index position and the dynamic regression parameters using the slope and deviation of the unary linear calibration function, determine the offset direction of the data point according to the positive or negative of the difference. If the difference is less than zero, it offsets upward, otherwise it offsets downward, and the offset amount is the absolute value of the difference, until all rows in the three-dimensional data are calibrated, and the calibrated data is output; S400, Calibration data verification and storage: Perform visualization processing on the calibrated data. By comparing with the visualization effect of the data before calibration, observe the improvement of the row granularity deviation. At the same time, randomly select some areas from the calibrated three-dimensional depth mode for detailed analysis, calculate the flatness index and slope change parameters of the road surface, and compare them with the actual measured values to verify the accuracy of the calibrated data. Classify and store the original, pre-calibrated, post-calibrated data and analysis results, and generate a report containing information about the test section, equipment status, and data quality assessment.

2. A method for calibrating the three-dimensional depth mode deviation of a road surface according to claim 1, characterized in that, In the S100, Data acquisition and storage:, the high-speed three-dimensional structured light measurement camera component acquires data in a line array mode and is triggered for imaging through the HES-2048 incremental encoder.

3. A method for calibrating the three-dimensional depth modal deviation of a road surface according to claim 1, characterized in that, The HES-2048 incremental encoder in the S100, Data acquisition and storage generates a profile trigger signal through the RS422 interface converter, coordinates with the camera component for imaging trigger, avoids data stretching and compression, and changes the displacement of one wheel rotation into 2048 periodic pulse signals.

4. A method for calibrating the three-dimensional depth modal deviation of a road surface according to claim 1, characterized in that, The S100, the auxiliary software in data acquisition and storage includes cxDiscover, cxExplorer and cxShow3d. cxDiscover is used to monitor the working status of the camera component and modify and fix the IP address corresponding to the camera component. cxExplorer is used to debug the parameters of the camera component and monitor the quality of the acquired data. cxShow3d is used for pseudo-color visualization of the three-dimensional depth mode.

5. A method for calibrating the three-dimensional depth modal deviation of a road surface according to claim 1, characterized in that, The S200, the specific steps of the floating data repair method based on dynamic regression parameters and sliding threshold region in floating data repair are as follows: (1) Read the three-dimensional depth mode data I; (2) Calculate its corresponding dynamic regression parameter DRP for I. The formula is: DRP = exp(0.3Avg(I)+0.5Med(I)+0.2Most(I)+σ), where Avg(·) represents mean calculation, Med(·) represents median statistics, Most(·) is mode statistics, σ is a small bias variable, and I represents the read three-dimensional depth mode; (3) Define the processing area A (n×n) , where n can be dynamically adjusted, and use the center point of A as the floating data point I to be repaired a ; (4) Traverse the numbers in area A except I a If there exists a three-dimensional depth floating anomaly value less than , repair it with the minimum value greater than DRP in this area; (5) Repeat steps (3) to (4) with a step size of n until all three-dimensional depth modal data I is processed, and output the processing result.

6. A method for calibrating the three-dimensional depth modal deviation of a road surface according to claim 1, characterized in that, The S300, the specific steps of the row granularity deviation calibration method based on the dynamic regression parameter reference plane in row granularity deviation calibration are as follows: (1) Read the three-dimensional depth modality after floating data repair (2) Read in Perform line-granularity extraction to obtain Q n ; (3) Read Q n Using DRP as the reference plane, find the current Q n with a slope of q n and an offset of s n to form a corresponding unary linear calibration function, the formula is: f (q,s) (x) = q n x i + s n ; (4)Traverse the current Q n Perform a position traversal, and the corresponding position index i always ranges from I to the maximum width w of the data size; (5) Using the slope q of the unary linear calibration function f (q,s) (x) n and the deviation s n Calculate the difference err between the index position i and the DRP. The formula is: err = abs(DRP - f (q,s) (x i )); (6) Current Q n The correction value for each item i of i is x ± err; (7)Operate according to steps (4) to (6) until all rows in the three-dimensional data are calibrated, and output the calibrated data 7. A method for calibrating the three-dimensional depth mode deviation of a road surface according to claim 1, characterized in that In the S300, the calculation formula for the difference err in the line granularity deviation calibration is: err = abs(DRP - f (q,s) (x i ))), where x i represents the position index of the current line data point, q n is the slope of the unary linear calibration function corresponding to the current line granularity Q n , s n is the deviation of the unary linear calibration function corresponding to the current line granularity, q n and s n are determined according to the fitting line equation, and the formula is: f (q,s) (x) = q n x i + s n .

8. A method for calibrating the three-dimensional depth modal deviation of a road surface according to claim 1, characterized in that, For the S300, in the line granularity deviation calibration, the data offset is calculated, the up and down offset is err, and the offset direction is based on DRP-f (q,s) (x i ) is determined by its positive or negative value. Upward or downward is relative to the reference plane. Upward indicates the direction of increasing depth value, and downward indicates the direction of decreasing depth value.

Images