Sensor position calibration system for gauge identification

Abstract

The invention relates to the technical field of gauge identification, and discloses a sensor position calibration system for gauge identification, which comprises a data acquisition module for acquiring position information data acquired by each sensor; the data processing module is used for acquiring position information data and respectively identifying position correlation characteristics; the data comparison module is used for carrying out consistency verification, stability verification and redundant information verification according to the position correlation characteristics, and sending verification result data to the credibility evaluation module; the credibility evaluation module is used for calculating the credibility score of each sensor according to a preset credibility scoring strategy, and the lower the credibility score is, the corresponding sensor is possibly abnormal; and the calibration early warning module is used for judging based on the distribution of the credibility scores, and if the deviation degree between the credibility score of one sensor and the mean value of the credibility scores of other sensors is greater than a preset score deviation threshold value, the sensor is used as a sensor to be verified and a verification instruction is generated.

Description

Technical Field

[0001] The present invention relates to the technical field of limit recognition, and in particular to a sensor position calibration system for limit recognition. Background Art

[0002] In areas such as railway transportation, clearance identification is crucial, as it can ensure the safe operation of trains and other means of transportation. Laser clearance detection is used for vehicle clearance detection. It can determine whether the vehicle's appearance exceeds the specified safety range, thereby avoiding collisions, scratches and other accidents during vehicle operation, and ensuring that the vehicle can safely pass through tunnels, bridges and other limited-limit structures. It is a necessary process for subway vehicles and engineering vehicles to detect and accept the vehicle's contour lines when they are newly built (or after maintenance). The laser clearance detection device uses a non-contact detection method, which can quickly collect data when the vehicle passes, and automatically complete the comparison with the standard limit without manual intervention, greatly improving the detection efficiency.

[0003] However, due to environmental factors, equipment aging and other reasons, the position of the sensor may shift, resulting in inaccurate data collected, which in turn affects the accuracy of limit recognition. At present, the existing sensor calibration system often lacks a comprehensive and accurate calibration mechanism, and cannot detect and handle abnormal conditions of the sensor in a timely manner, which poses a potential threat to the safety of industries such as railway transportation. Therefore, an efficient and reliable sensor position calibration system for limit recognition is needed. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a sensor position calibration system for limit identification, which can comprehensively process and verify the data collected by the sensor, accurately evaluate the credibility of the sensor, and promptly detect and handle sensors that may have abnormalities, thereby improving the accuracy and reliability of limit identification.

[0005] To achieve the above object, the present invention provides the following technical solutions: A sensor position calibration system for limit recognition, comprising: A data acquisition module obtains the location information data collected by each sensor; A data processing module, which acquires the position information data and identifies position-related features respectively, wherein the position-related features reflect features that meet corresponding relationships in the position information data collected by different sensors; The data comparison module performs consistency verification, stability verification and redundant information verification according to the position association features, and sends the verification result data to the credibility evaluation module. The credibility evaluation module calculates the credibility score of each sensor according to the preset credibility scoring strategy. The lower the credibility score, the more likely the corresponding sensor is to be abnormal. The calibration warning module makes a judgment based on the distribution of the credibility score. If the degree of deviation between the credibility score of one sensor and the average credibility score of other sensors is greater than the preset score deviation threshold, the sensor is used as a sensor to be verified and a verification instruction is generated.

[0006] Furthermore, the sensor includes several laser profilers and several area array cameras, the laser profilers obtain three-dimensional point cloud data of the train, the area array cameras obtain train image information of the train, the data processing module identifies point cloud key features and point cloud feature parameters in the three-dimensional point cloud data, the data processing module identifies image key features and image feature parameters in the train image information, and determines the similarity between the point cloud feature parameters and the image feature parameters. If the similarity meets a preset feature similarity threshold, the corresponding point cloud key features and image key features / several point cloud key features / several image key features are used as the position-associated features, and the position-associated features include corner points and contour points of the vehicle.

[0007] Furthermore, the calculation formula of the credibility scoring strategy is specifically as follows: , Where S represents the credibility evaluation score, and A represents the consistency score. represents the weight parameter of the consistency score, B represents the stability score, represents the weight parameter of the stability score, C represents the redundant information verification score, Represents the weight parameter of the redundant information verification score.

[0008] Furthermore, the consistency verification step includes: Define the current sensor as the target sensor, and obtain at least one target verification point based on the position association feature. Get the value of the target verification point in the world coordinate system to define it as the target verification value. A sensor having a key position feature with the target sensor is defined as a reference sensor, and a corresponding reference verification point is obtained based on the target verification point. Get the value of the reference verification point in the world coordinate system to define it as the reference verification value.

[0009] Furthermore, the consistency score calculation formula is specifically as follows: , , , D represents the total consistency deviation of the target sensor, m represents the number of reference sensors, and the j-th position-related feature has verification point, E is the kth target verification value of the key feature of the target sensor at the jth position in the ith measurement, and F is the kth target verification value compared with the target sensor. The reference verification value of a reference sensor under the same measurement conditions, W represents the weight of the verification times, Represents the total number of verifications of all reference sensor devices in the corresponding measurement.

[0010] Furthermore, the stability verification step includes: Acquire position information data and extract key dimensions, including width, length, and surface curvature, The historical dimensions are obtained and compared with the critical dimensions.

[0011] Furthermore, the calculation formula of the stability score is specifically: , , Where V represents the variance of the critical size and the historical size, q-1 represents the number of historical sizes, denoting the parameter value of the critical dimension, Indicates the parameter value of the corresponding historical dimension.

[0012] Furthermore, the redundant information verification step includes: Acquire depth data based on the three-dimensional point cloud data, and perform data preprocessing on the depth data; Acquire texture images based on train image information and perform image preprocessing; Convert the coordinates of the depth data to the same coordinate system as the texture image so that they have a corresponding relationship; Calculate the correlation between depth data and texture data. The greater the correlation, the higher the degree of fit and the higher the redundant information verification score.

[0013] Furthermore, the credibility evaluation module is configured with a weight dynamic adjustment strategy, and the formula of the weight dynamic adjustment strategy is specifically: , , in The values ​​of v are 1, 2, and 3, which represent the weight parameters of the consistency score, the stability score, and the redundant information verification score, respectively. represents the environmental parameter vector, represents the basic weight, represents the influence coefficient of environmental parameters on weight, A quantitative function that represents the influence of environmental parameters on weights.

[0014] Furthermore, the credibility evaluation module is configured with a stability weight parameter adjustment sub-strategy, including obtaining historical over-limit data, the historical over-limit data including the over-limit position of the train, judging the over-limit position of the train and the detection range of the sensor; if the number of train over-limit anomalies that occur within the detection range corresponding to the target sensor is greater, the basic weight of the stability score weight parameter is smaller; if the number of corresponding verifications of the target sensor is greater, the weight parameter of the consistency score is greater; if the number of sensors with position-related features with the target sensor is smaller, the basic weight of the redundant information verification score weight parameter is greater.

[0015] Beneficial effects of the present invention: The present invention uses sensors arranged in a vehicle clearance recognition system to verify each other, determines the association between data features collected by each sensor device by identifying position correlation features, uses sensors to verify each other, calculates credibility scores respectively, and sets several credibility score evaluation items, including consistency score, stability score and redundant information verification score. If the degree of deviation of the credibility score mean of one of the sensor devices is greater than a preset score deviation threshold, it indicates that the sensor device has a high degree of deviation from other devices and is more prone to errors. Therefore, it is determined whether the device needs to be verified and calibrated to ensure the detection accuracy of each sensor. The system setting of the present invention can perform detection by itself during routine detection, determine the credibility score and output verification instructions, which is convenient for timely detection of faults, has a high degree of automation and improves the reliability of sensor detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic diagram of the system architecture of the present invention; Figure 2 It is a flow chart of the consistency verification step in the present invention; Figure 3 It is a schematic flow chart of the stability verification step in the present invention; Figure 4 It is a flowchart of the redundant information verification step in the present invention. DETAILED DESCRIPTION

[0017] 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.

[0018] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may also be a component centered. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may also be a component centered. When a component is considered to be "set on" another component, it may be directly set on the other component or there may also be a component centered. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0020] like Figures 1 to 4 As shown, this embodiment provides a sensor position calibration system for limit recognition, including: A data acquisition module obtains the location information data collected by each sensor; A data processing module, which acquires the position information data and identifies position-related features respectively, wherein the position-related features reflect features that meet corresponding relationships in the position information data collected by different sensors; The data comparison module performs consistency verification, stability verification and redundant information verification according to the position association features, and sends the verification result data to the credibility evaluation module. The credibility evaluation module calculates the credibility score of each sensor according to the preset credibility scoring strategy. The lower the credibility score, the more likely the corresponding sensor is to be abnormal. The calibration warning module makes a judgment based on the distribution of the credibility score. If the degree of deviation between the credibility score of one sensor and the average credibility score of other sensors is greater than the preset score deviation threshold, the sensor is used as a sensor to be verified and a verification instruction is generated.

[0021] The credibility score can be set with multiple evaluation items. For example, consistency, stability and redundant verification items are set in this application. Consistency: Compare the same geometric feature parameters extracted by the two devices and calculate their degree of deviation. The smaller the deviation, the higher the consistency of the measurement results of the two devices and the higher the credibility. Stability index: Analyze the fluctuation of the measurement data of each device during multiple measurements. The smaller the data fluctuation, the better the stability of the device measurement and the higher the credibility. Redundant information verification index: Use the complementarity of the measurement information of the two devices, such as the texture information of the array camera and the depth information of the 3D laser profiler, to verify each other. The more the verification result is in line with expectations, the higher the credibility of the device measurement.

[0022] The following is an introduction to the modules of the sensor position calibration system for clearance recognition: 1. Data acquisition module The function of the data acquisition module is to obtain the location information data collected by each sensor. For example, in the railway clearance recognition scenario, the sensors may include several laser profilers and several array cameras. The laser profiler can obtain the three-dimensional point cloud data of the train, and the array camera can obtain the train image information of the train. Assume that multiple laser profilers and cameras are installed at the entrance of a railway tunnel. The laser profiler scans the train in real time to generate three-dimensional point cloud data of the train surface, and the array camera takes the image of the train. The data acquisition module collects this data.

[0023] 2. Data processing module This module obtains the position information data and identifies the position-related features respectively. The position-related features reflect the features that meet the corresponding relationship in the position information data collected by different sensors. Specifically, the data processing module identifies the point cloud key features and point cloud feature parameters in the three-dimensional point cloud data, and identifies the image key features and image feature parameters in the train image information. For example, in the three-dimensional point cloud data of the train, the point cloud key features such as the train corner points and contour points are identified, and their point cloud feature parameters, such as coordinates, angles, etc., are calculated; in the train image information, the image key features such as the train corner points and contour points are also identified, and the image feature parameters are calculated. Then the similarity between the point cloud feature parameters and the image feature parameters is determined. If the similarity meets the preset feature similarity threshold, the corresponding point cloud key features and image key features / several point cloud key features / several image key features are used as position-related features. For example, when the coordinate similarity of the train head corner point in the point cloud data and the corresponding corner point in the image is more than 90%, the two corner points are used as position-related features.

[0024] 3. Data comparison module This module performs consistency verification, stability verification and redundant information verification based on the position association features, and sends the verification result data to the credibility evaluation module.

[0025] Consistency verification: Consistency verification compares the same geometric feature parameters extracted by two devices and calculates their degree of deviation. The smaller the deviation, the higher the consistency of the measurement results of the two devices and the higher the credibility.

[0026] The consistency verification steps include: Define the current sensor as the target sensor, and obtain at least one target verification point based on the position association feature. Get the value of the target verification point in the world coordinate system to define it as the target verification value. A sensor having a key position feature with the target sensor is defined as a reference sensor, and a corresponding reference verification point is obtained based on the target verification point. Get the value of the reference verification point in the world coordinate system to define it as the reference verification value.

[0027] The consistency score calculation formula is specifically: , , , D represents the total consistency deviation of the target sensor, m represents the number of reference sensors, and the jth position-related feature has verification points, E is the kth target verification value of the key feature of the jth position of the target sensor in the i-th measurement, F is the reference verification value of the lth reference sensor compared with the target sensor under the same measurement conditions, W represents the weight of the verification times, Represents the total number of verifications of all reference sensor devices in the corresponding measurement. The smaller the deviation and the more times it has been verified, the higher the consistency score, which also reflects the reliability of the device as a reference for other devices.

[0028] Stability verification: During multiple measurements, stability verification analyzes the fluctuation of the measured data of each device. The smaller the data fluctuation, the better the stability of the device measurement and the higher the credibility.

[0029] The stability verification step comprises: Acquire position information data and extract key dimensions, including width, length, and surface curvature, The historical dimensions are obtained and compared with the critical dimensions.

[0030] The calculation formula of the stability score is specifically: , , Where V represents the variance of the critical size and the historical size, q-1 represents the number of historical sizes, denoting the parameter value of the critical dimension, Indicates the parameter value of the corresponding historical dimension.

[0031] Redundant information verification: Redundant information verification uses the complementarity of the measurement information of two devices, such as the texture information of the array camera and the depth information of the 3D laser profiler, to verify each other. The more the verification result conforms to expectations, the higher the credibility of the device measurement.

[0032] The redundant information verification step comprises: Acquire depth data based on the three-dimensional point cloud data, and perform data preprocessing on the depth data; Acquire texture images based on train image information and perform image preprocessing; Convert the coordinates of the depth data to the same coordinate system as the texture image so that they have a corresponding relationship; Calculate the correlation between depth data and texture data. The greater the correlation, the higher the degree of fit and the higher the redundant information verification score.

[0033] Calculate the local surface normal vector of the depth data, which can reflect the orientation information of the surface. You can also extract the depth change rate, that is, the difference in the depth values ​​of adjacent points, to characterize the curvature of the surface. For example, in the case of the surface shape of a train, the surface normal vector can indicate the tilt direction of the body surface, and the depth change rate can reflect the degree of surface undulation.

[0034] For texture images, we extract scale-invariant feature transform (SIFT) features, speeded-up robust features (SURF) features, etc. These features are invariant to image rotation, scaling, and illumination changes, and can effectively describe the uniqueness of textures.

[0035] Use the extracted features for matching. For example, use a data structure such as a KD tree to quickly match the feature points of the depth data with the feature points of the texture data. For example, match the surface normal feature points in the depth data with the SIFT feature points in the texture data to find the corresponding relationship between the two at the feature level.

[0036] Calculate the correlation between the depth value and texture feature strength of the matching point pair. For example, count whether the texture feature strength corresponding to the point with a larger depth value in a certain number of matching point pairs also shows a certain regular change. You can also calculate the consistency of the change trend of depth data and texture data in the same area, such as calculating the consistency of the gradient direction of the two in a certain local area, and evaluate the degree of fit between the two through quantitative indicators.

[0037] 4. Credibility Evaluation Module The credibility score of each sensor is calculated according to the preset credibility score strategy, where the lower the credibility score, the more likely the corresponding sensor is abnormal. The specific calculation formula of the credibility score strategy is: , Where S represents the credibility evaluation score, and A represents the consistency score. represents the weight parameter of the consistency score, B represents the stability score, represents the weight parameter of the stability score, C represents the redundant information verification score, Represents the weight parameter of the redundant information verification score.

[0038] The credibility evaluation module is configured with a weight dynamic adjustment strategy, and the formula of the weight dynamic adjustment strategy is specifically: , , in The values ​​of v are 1, 2, and 3, which represent the weight parameters of the consistency score, the stability score, and the redundant information verification score, respectively. represents the environmental parameter vector, represents the basic weight, Indicates the influence coefficient of the preset environmental parameters on the weight, A quantitative function that represents the impact of environmental parameters on weights. For example, if the environment is highly vibrating, the stability score may be more important, and the weight of stability should be increased; or in the case of insufficient light, redundant information: such as the correlation between depth data and texture may be more critical, and the weight of redundant information needs to be increased.

[0039] For example, this time a total of three environmental parameters are introduced, including light intensity, vibration and humidity, which require the setting of photosensors, accelerometers and humidity sensors. The light intensity will affect the quality of texture data, vibration will affect the stability of the point cloud, and humidity will affect the laser reflectivity. The quantitative function that represents the influence of environmental parameters on weights is a pre-set function. A specific example is given below: Set as the quantization function of light intensity, ; Set as a quantization function of vibration intensity, ; Set as the quantization function of humidity, ; in The weight parameter representing the light intensity, Indicates the light intensity, () is specifically a trigonometric function formula, The weight parameter representing the vibration intensity, Indicates the vibration intensity. represents the weight parameter of humidity, Indicates humidity.

[0040] The credibility evaluation module is configured with a stability weight parameter adjustment sub-strategy, including obtaining historical overrun data, the historical overrun data including the overrun position of the train, judging the overrun position of the train and the detection range of the sensor. If the number of train overrun anomalies that occur within the detection range corresponding to the target sensor is greater, the basic weight of the stability score weight parameter is The smaller it is, the more times the target sensor is verified, the larger the weight parameter of the consistency score. If the number of sensors with position-related features with the target sensor is smaller, the basic weight of the redundant information verification score weight parameter is The bigger.

[0041] 5. Calibration warning module Based on the distribution of the credibility score, if the credibility score of one sensor deviates from the average credibility score of other sensors by more than the preset score deviation threshold, the sensor will be used as a sensor to be verified and a verification instruction will be generated. For example, if the preset score deviation threshold is 20%, when the credibility score of a laser profiler deviates from the average credibility score of other laser profilers and cameras by more than 20%, the laser profiler will be used as a sensor to be verified, and the system will generate a verification instruction to prompt the staff to further check and calibrate the sensor.

[0042] Working principle: The operating principle of this system is introduced according to the work flow: 1. Data collection stage Different types of sensors are used to collect data respectively. The area array camera and laser profiler used in this application are used for measurement. The data acquisition module collects the three-dimensional point cloud data of the train collected by the laser profiler and the image information of the train collected by the area array camera in real time. For example, every 1 second, the laser profiler scans the passing train once to generate three-dimensional point cloud data of the train surface; the area array camera simultaneously captures the image information of the train.

[0043] 2. Data processing stage The data processing module processes the collected data. The data collected by the array camera and the 3D laser profiler are unified into the same global coordinate system. Through the precise calibration process, the conversion relationship between the camera coordinate system, the laser profiler coordinate system and the global coordinate system is determined to ensure the consistency of the spatial position of the measurement data of different devices. Representative feature points, such as the corner points and key contour points of the vehicle, are extracted from the array camera image and the 3D laser profiler point cloud data. The feature point matching algorithm is used to find the corresponding feature point pairs in the two data. If the similarity meets the preset feature similarity threshold (such as 85%), the corresponding point cloud key features and image key features are used as position association features. Based on the matched feature point pairs, the position mapping relationship between the array camera image and the 3D laser profiler point cloud data is established. Through this mapping relationship, the position information measured by the two devices can be correlated to facilitate subsequent abnormal analysis.

[0044] (III) Data comparison stage 1. Consistency verification Take a laser profiler as the target sensor, select a corner point of the train head as the target verification point according to the position association feature, and obtain its target verification value in the world coordinate system. Take another laser profiler with the same position key feature as the laser profiler as the reference sensor, and obtain the reference verification value of the same corner point of the train head collected by it. Calculate the consistency score according to the consistency score calculation formula.

[0045] 2. Stability verification The key dimensions of the train carriage, such as width, length and surface curvature, are extracted from the train 3D point cloud data collected by the laser profilometer and compared with the historical dimensions. The stability score is calculated according to the stability score calculation formula.

[0046] 3. Redundant information verification Process the 3D point cloud data to obtain depth data and perform preprocessing; process the train image information to obtain the texture image and perform preprocessing. Convert the coordinates of the depth data to the same coordinate system as the texture image, calculate the correlation between the depth data and the texture data, and obtain the redundant information verification score.

[0047] 4. Credibility Evaluation Stage The credibility evaluation module calculates the credibility score of each sensor according to the preset credibility scoring strategy. At the same time, according to the weight parameter adjustment sub-strategy, the weight parameters of each verification indicator are adjusted in combination with the historical overrun data. For example, if a certain laser profiler has a large number of train overrun abnormalities in the detection range recently, the weight parameter of its stability score will be appropriately reduced.

[0048] (V) Calibration warning stage The calibration warning module makes judgments based on the distribution of credibility scores. If the degree of deviation between the credibility score of a sensor and the average credibility score of other sensors is greater than the preset score deviation threshold (such as 20%), the sensor will be treated as a sensor to be verified, and a verification instruction will be generated to notify the staff to check and calibrate the sensor.

[0049] The laser profilometer can be used to obtain 3D point cloud data, and the area array camera can obtain image information of the train. The 3D point cloud data and image information can be associated based on position features, such as contour features such as edges and corners, and then the credibility score can be calculated.

[0050] Through the above steps, the system of the present invention can calibrate and monitor the position of the sensor in real time to ensure the accuracy and reliability of limit recognition. The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A sensor position calibration system for boundary recognition, characterized in that: include: A data acquisition module obtains the location information data collected by each sensor; A data processing module, which acquires the position information data and identifies position-related features respectively, wherein the position-related features reflect features that meet corresponding relationships in the position information data collected by different sensors; The data comparison module performs consistency verification, stability verification and redundant information verification according to the position association features, and sends the verification result data to the credibility evaluation module. The credibility evaluation module calculates the credibility score of each sensor according to the preset credibility scoring strategy. The lower the credibility score, the more likely the corresponding sensor is to be abnormal. The calibration warning module makes a judgment based on the distribution of the credibility score. If the degree of deviation between the credibility score of one sensor and the average credibility score of other sensors is greater than the preset score deviation threshold, the sensor is used as a sensor to be verified and a verification instruction is generated.

2. The sensor position calibration system for limit recognition according to claim 1, characterized in that: The sensor includes several laser profilers and several area array cameras. The laser profilers obtain three-dimensional point cloud data of the train, and the area array cameras obtain train image information of the train. The data processing module identifies point cloud key features and point cloud feature parameters in the three-dimensional point cloud data. The data processing module identifies image key features and image feature parameters in the train image information, and determines the similarity between the point cloud feature parameters and the image feature parameters. If the similarity meets the preset feature similarity threshold, the corresponding point cloud key features and image key features / several point cloud key features / several image key features are used as the position-related features, and the position-related features include corner points and contour points of the vehicle.

3. The sensor position calibration system for limit recognition according to claim 1, characterized in that: The calculation formula of the credibility scoring strategy is specifically: ， Where S represents the credibility evaluation score, and A represents the consistency score. represents the weight parameter of the consistency score, B represents the stability score, represents the weight parameter of the stability score, C represents the redundant information verification score, Represents the weight parameter of the redundant information verification score.

4. The sensor position calibration system for limit recognition according to claim 3, characterized in that: The consistency verification step includes: Define the current sensor as the target sensor, and obtain at least one target verification point based on the position association feature. Get the value of the target verification point in the world coordinate system to define it as the target verification value. A sensor having a key position feature with the target sensor is defined as a reference sensor, and a corresponding reference verification point is obtained based on the target verification point. Get the value of the reference verification point in the world coordinate system to define it as the reference verification value.

5. The sensor position calibration system for limit recognition according to claim 4, characterized in that: The consistency score calculation formula is specifically: ， ， ， D represents the total consistency deviation of the target sensor, m represents the number of reference sensors, and the jth position-related feature has verification points, E is the kth target verification value of the key feature of the jth position of the target sensor in the i-th measurement, F is the reference verification value of the lth reference sensor compared with the target sensor under the same measurement conditions, W represents the weight of the verification times, Represents the total number of validations of all reference sensor devices in the corresponding measurement.

6. The sensor position calibration system for boundary recognition according to claim 3, characterized in that: The stability verification step comprises: Acquire position information data and extract key dimensions, including width, length, and surface curvature, The historical dimensions are obtained and compared with the critical dimensions.

7. The sensor position calibration system for limit recognition according to claim 6, characterized in that: The calculation formula of the stability score is specifically: ， ， Where V represents the variance of the critical size and the historical size, q-1 represents the number of historical sizes, denoting the parameter value of the critical dimension, Indicates the parameter value of the corresponding historical dimension.

8. The sensor position calibration system for limit recognition according to claim 2, characterized in that: The redundant information verification step comprises: Acquire depth data based on the three-dimensional point cloud data, and perform data preprocessing on the depth data; Acquire texture images based on train image information and perform image preprocessing; Convert the coordinates of the depth data to the same coordinate system as the texture image so that they have a corresponding relationship; Calculate the correlation between depth data and texture data. The greater the correlation, the higher the degree of fit and the higher the redundant information verification score.

9. The sensor position calibration system for boundary recognition according to claim 3, characterized in that: The credibility evaluation module is configured with a weight dynamic adjustment strategy, and the formula of the weight dynamic adjustment strategy is specifically: ， ， in The values ​​of v are 1, 2, and 3, which represent the weight parameters of the consistency score, the stability score, and the redundant information verification score, respectively. represents the environmental parameter vector, represents the basic weight, represents the influence coefficient of environmental parameters on weight, A quantitative function that represents the influence of environmental parameters on weights.

10. The sensor position calibration system for boundary recognition according to claim 3, characterized in that: The credibility evaluation module is configured with a stability weight parameter adjustment sub-strategy, including obtaining historical over-limit data, the historical over-limit data including the over-limit position of the train, judging the over-limit position of the train and the detection range of the sensor. If the number of train over-limit anomalies that occur within the detection range corresponding to the target sensor is greater, the basic weight of the stability score weight parameter is smaller; if the number of corresponding verifications of the target sensor is greater, the weight parameter of the consistency score is greater; if the number of sensors with position-related features with the target sensor is smaller, the basic weight of the redundant information verification score weight parameter is greater.

Images