Anti-interference image acquisition and detection method based on physical dislocation of light paths

Abstract

The present application relates to the field of rail transit, and discloses an anti-interference image acquisition and detection method based on physical light path offset, comprising the following steps: designing a multi-sensor system of rail transit without offset physical light path, which is used for acquiring anti-interference 2D images and 3D data, performing data registration and data preprocessing on the anti-interference 2D images and 3D data, and performing analysis to realize defect detection and identification of rail transit trains. The present application can eliminate direct interference between light sources through the form of physically offset light path at the source, improve data quality, and accurately fuse data of different modalities, so that the defect detection capability is stronger than that of a single modality, and the present application has self-adaptive preprocessing and online fusion quality evaluation capabilities, thereby improving the efficiency of acquisition and detection.

Description

Technical Field

[0001] This invention relates to the field of rail transit, and in particular to an anti-interference image acquisition and detection method based on physically staggered optical paths. Background Technology

[0002] In the daily operation and maintenance of rail transit electrical equipment, such as speed sensors, transponders, and connectors in the locomotive running gear, traditional manual visual inspection methods have the following fundamental bottlenecks: they rely on maintenance personnel to work in trenches or at heights, and each inspection is time-consuming, becoming a constraint on locomotive turnaround; due to the influence of personnel experience, fatigue, and lighting conditions, hidden dangers such as minor cracks and early loosening are easily missed; manual operation requires close proximity to high voltage and mobile equipment, posing safety risks, and cannot achieve dynamic and real-time detection.

[0003] To address the aforementioned issues, the industry has introduced automated image inspection systems, which deploy 2D / 3D sensors alongside the track to automatically acquire equipment images and perform intelligent analysis. However, in practical applications, simultaneously acquiring 2D and 3D information requires integrating an area scan camera and a 3D laser scanner. When the optical paths of both overlap on the equipment surface, the intense illumination from the camera light source can cause the laser lines to be overexposed, saturated, or even completely submerged in the image, leading to severe distortion or loss of 3D data. This phenomenon is particularly severe on highly reflective metal surfaces. Therefore, it is necessary to stagger the optical paths of the 2D and 3D equipment during inspection. Moreover, even if 2D and 3D data are acquired separately, it is crucial to align them with pixel-level precision within a unified coordinate system. Traditional calibration methods are complex, sensitive to environmental vibrations and equipment pose changes, and difficult to maintain stable high-precision alignment requirements in complex field environments over long periods. Therefore, the acquired images need to be interference-resistant. Thus, an interference-resistant image acquisition and inspection method based on physically staggered optical paths is proposed. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides an anti-interference image acquisition and detection method based on physically staggered optical paths.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides an anti-interference image acquisition and detection method based on physically staggered optical paths, comprising the following steps: Deploy a multi-sensor system for rail transit and design the physical optical path of the multi-sensor system; By combining a 2D area array camera system and a 3D line laser scanner, along with a high-precision time-division triggering mechanism, anti-interference images and data of the electrical equipment under inspection are acquired. Data registration is performed on the anti-interference image and anti-interference data of the electrical equipment to be inspected to obtain the registered anti-interference image and anti-interference data. The registered anti-interference images and anti-interference data are preprocessed, and the preprocessed data is combined to detect and identify defects in rail transit trains.

[0006] Furthermore, in a preferred embodiment of the present invention, the layout is used for a multi-sensor system for rail transit, and the physical optical path of the multi-sensor system is designed as follows: Determine the travel position of the rail transit train, use it as the workstation for the layout of the multi-sensor system for rail transit, and mark it as the target layout workstation; Measure the spatial parameters of the target layout workstation and the ambient light parameters of the target layout workstation. At the same time, determine the relative parameters of the electrical equipment to be inspected traveling on the target layout workstation. The electrical equipment to be inspected is the equipment on the rail transit train that needs to be fault identified. A 2D area scan camera system and a 3D line laser scanner are acquired. The 2D area scan camera system and the 3D line laser scanner are used to collect anti-interference images and data of the electrical equipment to be inspected at the target layout workstation. The coverage area of ​​the electrical equipment to be inspected during its movement is determined and marked as the target inspection area. The 2D area array camera system and the 3D line laser scanner are then installed on the target layout station using a universal bracket. Adjust the universal bracket, control the 2D area scan camera system and 3D line laser scanner to perform physical optical path misalignment design through Y-axis optical path interlacing method, and install mechanical light shields on the 2D area scan camera system and 3D line laser scanner.

[0007] Furthermore, in a preferred embodiment of the present invention, the combination of a 2D area scan camera system and a 3D line laser scanner, along with a high-precision time-division triggering mechanism, to acquire anti-interference images and data of the electrical equipment under inspection, specifically involves: A photoelectric sensor is installed at the starting position of the target detection area. The photoelectric sensor is used to detect whether the rail transit train is blocking the photoelectric signal emitted by the photoelectric sensor. If so, it is determined that the rail transit train is passing through the target detection area, and the photoelectric sensor is controlled to emit a start trigger signal. An intelligent time-sharing controller is introduced and connected to a photoelectric sensor, a 2D area array camera system, and a 3D line laser scanner, respectively. The intelligent time-sharing controller receives the start trigger signal from the photoelectric sensor in real time. After receiving the start trigger signal from the photoelectric sensor, the intelligent time-sharing controller outputs the pre-start trigger signal for the 2D area scan camera system and the 3D line laser scanner, thereby realizing the pre-start processing of the 2D area scan camera system and the 3D line laser scanner. The intelligent time-sharing controller analyzes in real time the duration of the photoelectric signal emitted by the photoelectric sensor that is blocked by the rail transit train, the speed of the rail transit train, and the individual parameters of the rail transit train to generate the real-time displacement value of the rail transit train. A predetermined displacement threshold is set. When the real-time displacement value of the rail transit train reaches the displacement threshold, the 2D area array camera system and the 3D line laser scanner are activated by the intelligent time-sharing controller to achieve the purpose of acquiring 2D images and 3D laser data. Dynamic delay compensation is performed on the acquired 2D images and 3D laser data to obtain anti-interference images and anti-interference data of the electrical equipment under inspection.

[0008] Furthermore, in a preferred embodiment of the present invention, the step of performing dynamic delay compensation on the acquired 2D images and acquired 3D laser data to obtain the anti-interference image and anti-interference data of the electrical equipment under inspection specifically includes: A high-speed oscilloscope was introduced to measure the time from when the intelligent time-sharing controller receives the initial trigger signal from the photoelectric sensor to when it generates the real-time displacement value of the rail transit train, and this time was calibrated as the running trigger delay time. If the trigger delay time is not greater than the preset value, the acquired 2D image and 3D laser data will be directly calibrated as the anti-interference image and anti-interference data of the electrical equipment under inspection. If the operation trigger delay time is greater than the preset value, the operation trigger delay time will be combined with the speed of the rail transit train and the individual parameters of the rail transit train to generate the delay displacement value of the rail transit train, which will be calibrated as the displacement compensation value. Based on the displacement compensation value, the intelligent time-sharing controller is re-controlled to start the 2D area array camera system and the 3D line laser scanner respectively, so as to obtain the anti-interference image and anti-interference data of the electrical equipment under inspection.

[0009] Furthermore, in a preferred embodiment of the present invention, the process of data registration of the anti-interference image and anti-interference data of the electrical equipment to be inspected to obtain the registered anti-interference image and anti-interference data specifically involves: The anti-interference images and anti-interference data of the electrical equipment under inspection are calibrated as 2D anti-interference images and 3D anti-interference data. The standard natural feature points of the electrical equipment to be inspected are preset. The positions corresponding to the standard natural feature points are identified in the 2D anti-interference image and 3D anti-interference data, and 2D feature fingerprints and 3D feature fingerprints are generated. Among them, the 2D feature fingerprint records the pixel coordinates and texture features of standard natural feature points in the 2D anti-interference image, and the 3D feature fingerprint records the three-dimensional spatial coordinates and geometric features of standard natural feature points in the 3D anti-interference data. A consistency check is performed on the texture features in the 2D fingerprint and the geometric features in the 3D fingerprint. The consistency check involves randomly pairing the texture features in the 2D fingerprint and the geometric features in the 3D fingerprint to obtain different feature combinations. For each pair of feature combinations, a 2D-3D transformation relationship is calculated. The 2D-3D transformation relationship is calculated by mapping the geometric features in the 3D feature fingerprint onto the 2D anti-interference image. If the positional deviation pixel value between the texture features in the 2D feature fingerprint and the geometric features in the 3D feature fingerprint is less than a preset value, the corresponding feature combination is labeled as a qualified feature combination. By combining the qualified feature combinations, the registered anti-interference image and anti-interference data are obtained.

[0010] Furthermore, in a preferred embodiment of the present invention, the step of performing data preprocessing on the registered anti-interference image and anti-interference data, and combining the preprocessed data to perform defect detection and identification of rail transit trains, specifically includes: In the registered anti-interference image and anti-interference data, three indicators are calculated for the anti-interference image: global contrast, information entropy, and peak signal-to-noise ratio. In the anti-interference data, three indicators are calculated: point density uniformity, noise ratio, and effective data integrity. Scoring thresholds are set for different indicators of the registered anti-interference image and anti-interference data. Based on the scoring thresholds, different indicators are scored on the registered anti-interference image and anti-interference data. A solution is introduced to address the issue of indicators whose scores in big data network retrieval are lower than preset values, until the scores of different indicators of the registered anti-interference image and anti-interference data are all greater than preset values, thus obtaining qualified anti-interference images and qualified anti-interference data. The method involves assigning the RGB color values ​​of 2D pixels from a qualified anti-interference image to a 3D point cloud of qualified anti-interference data to generate a preliminary pre-processed 3D point cloud. Semantic labels are applied to different pixels on qualified anti-interference images and simultaneously assigned to the pre-processed 3D point cloud to generate synchronous semantic 3D point cloud. The 3D feature values ​​of the synchronous semantic 3D point cloud are calculated to generate a high-dimensional fusion feature vector. By combining high-dimensional fusion feature vectors for analysis, defect detection and identification of rail transit trains can be achieved.

[0011] Furthermore, in a preferred embodiment of the present invention, the step of combining high-dimensional fusion feature vectors for analysis to achieve defect detection and identification of rail transit trains specifically includes: A knowledge base of defects in rail transit trains is acquired, and a three-layer defect identification model is constructed based on the defects in rail transit trains. The defects in rail transit trains include physical anomaly defects, state anomaly defects, and functional risk defects. Different layers in the three-layer defect identification model are used to identify different defect types. The method for constructing the three-layer defect identification model is as follows: input the sample vector from the high-dimensional fusion feature vector into the baseline model of the three-layer defect identification model, and introduce feature analysis rules into the baseline model to preliminarily analyze the fuzzy defect type of the electrical equipment to be inspected on the sample vector. Among them, the feature analysis rules are preset feature vector templates corresponding to the defect types of different electrical equipment to be inspected in rail transit trains; Different fuzzy defect types of electrical equipment under inspection are assigned to different layers of the baseline model of the three-layer defect recognition model. At the same time, a convolutional neural network is introduced to perform convolutional analysis on different fuzzy defect types of electrical equipment under inspection, so as to realize secondary classification training for different fuzzy defect types of electrical equipment under inspection. By using a three-layer defect identification model, defects are identified from high-dimensional fused feature vectors, and the defect types of different electrical equipment under inspection are output.

[0012] This invention addresses the technical deficiencies in the prior art and offers the following advantages: It designs a multi-sensor system for rail transit that does not physically separate optical paths, used to acquire interference-resistant 2D images and 3D data. This system performs data registration and preprocessing on the interference-resistant 2D images and 3D data, followed by analysis, to achieve defect detection and identification of rail transit trains. This invention eliminates direct interference between light sources at the source by physically separating optical paths, improving data quality and accurately fusing data from different modalities. This results in stronger defect detection capabilities compared to single-modal systems. Furthermore, it possesses adaptive preprocessing and online evaluation capabilities for fusion quality, improving the efficiency of data acquisition and detection.

[0013] S102: Layout of a multi-sensor system for rail transit, and design of the physical optical path of the multi-sensor system; S104: Combining a 2D area array camera system and a 3D line laser scanner, and incorporating a high-precision time-division triggering mechanism, it acquires anti-interference images and data of the electrical equipment under inspection. S106: Perform data registration on the anti-interference image and anti-interference data of the electrical equipment to be inspected to obtain the registered anti-interference image and anti-interference data; S108: Perform data preprocessing on the registered anti-interference image and anti-interference data, and combine the preprocessed data to perform defect detection and identification of rail transit trains. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0015] Figure 1 A flowchart of an anti-interference image acquisition and detection method based on physical optical path misalignment is shown; Figure 2 A flowchart illustrating a method for acquiring anti-interference images and anti-interference data of electrical equipment under inspection is shown. Detailed Implementation

[0016] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0018] Figure 1 A flowchart illustrating an anti-interference image acquisition and detection method based on physically staggered optical paths is shown, including the following steps: S102: Layout of a multi-sensor system for rail transit, and design of the physical optical path of the multi-sensor system; S104: Combining a 2D area array camera system and a 3D line laser scanner, and incorporating a high-precision time-division triggering mechanism, it acquires anti-interference images and data of the electrical equipment under inspection. S106: Perform data registration on the anti-interference image and anti-interference data of the electrical equipment to be inspected to obtain the registered anti-interference image and anti-interference data; S108: Perform data preprocessing on the registered anti-interference image and anti-interference data, and combine the preprocessed data to perform defect detection and identification of rail transit trains.

[0019] Furthermore, in a preferred embodiment of the present invention, the layout is used for a multi-sensor system for rail transit, and the physical optical path of the multi-sensor system is designed as follows: Determine the travel position of the rail transit train, use it as the workstation for the layout of the multi-sensor system for rail transit, and mark it as the target layout workstation; Measure the spatial parameters of the target layout workstation and the ambient light parameters of the target layout workstation. At the same time, determine the relative parameters of the electrical equipment to be inspected traveling on the target layout workstation. The electrical equipment to be inspected is the equipment on the rail transit train that needs to be fault identified. A 2D area scan camera system and a 3D line laser scanner are acquired. The 2D area scan camera system and the 3D line laser scanner are used to collect anti-interference images and data of the electrical equipment to be inspected at the target layout workstation. The coverage area of ​​the electrical equipment to be inspected during its movement is determined and marked as the target inspection area. The 2D area array camera system and the 3D line laser scanner are then installed on the target layout station using a universal bracket. Adjust the universal bracket, control the 2D area scan camera system and 3D line laser scanner to perform physical optical path misalignment design through Y-axis optical path interlacing method, and install mechanical light shields on the 2D area scan camera system and 3D line laser scanner.

[0020] It should be noted that the design of a multi-sensor system for rail transit with staggered optical paths is crucial. This means that when the 2D area scan camera system and the 3D line laser scanner are operating, the illumination spot focused by the 2D area scan camera system and the laser line emitted by the 3D line laser scanner do not overlap, achieving optical path misalignment. Target layout workstations include, but are not limited to, fixed workstations within the section's inspection sheds or entry lines where locomotives pass at a constant speed and the lighting is relatively controllable, used for precise measurement of available installation space along the trackside. The electrical equipment to be inspected refers to devices on rail transit trains requiring fault identification, including but not limited to undercarriage speed sensors, roof pantograph plates, and connector housings. Measuring the spatial parameters of the target layout workstation, as well as the ambient light parameters, is essential for finding suitable installation locations for the 2D area scan camera system and the 3D line laser scanner. A universal bracket allows adjustment of the alignment direction of the optical paths of the 2D area scan camera system and the 3D line laser scanner, achieving physical optical path misalignment through a Y-axis optical path interlacing method.

[0021] The Y-axis optical path interlacing method involves using the direction along the rails where the train travels as the Y-axis. The bracket is adjusted so that the projection position of the 3D laser line on the surface of the electrical equipment under inspection maintains a fixed physical distance from the central area of ​​the 2D camera illumination spot in the direction of train travel. The advantage is that when a test point is scanned by the laser, it is located near the 2D camera illumination area, and there is no optical path interlacing. Mechanical light shields are installed on the 2D area array camera system and the 3D line laser scanner to limit their field of view and illumination range, strictly suppress stray light, and isolate the effects of sudden changes in external ambient light.

[0022] Furthermore, in a preferred embodiment of the present invention, the process of data registration of the anti-interference image and anti-interference data of the electrical equipment to be inspected to obtain the registered anti-interference image and anti-interference data specifically involves: The anti-interference images and anti-interference data of the electrical equipment under inspection are calibrated as 2D anti-interference images and 3D anti-interference data. The standard natural feature points of the electrical equipment to be inspected are preset. The positions corresponding to the standard natural feature points are identified in the 2D anti-interference image and 3D anti-interference data, and 2D feature fingerprints and 3D feature fingerprints are generated. Among them, the 2D feature fingerprint records the pixel coordinates and texture features of standard natural feature points in the 2D anti-interference image, and the 3D feature fingerprint records the three-dimensional spatial coordinates and geometric features of standard natural feature points in the 3D anti-interference data. A consistency check is performed on the texture features in the 2D fingerprint and the geometric features in the 3D fingerprint. The consistency check involves randomly pairing the texture features in the 2D fingerprint and the geometric features in the 3D fingerprint to obtain different feature combinations. For each pair of feature combinations, a 2D-3D transformation relationship is calculated. The 2D-3D transformation relationship is calculated by mapping the geometric features in the 3D feature fingerprint onto the 2D anti-interference image. If the positional deviation pixel value between the texture features in the 2D feature fingerprint and the geometric features in the 3D feature fingerprint is less than a preset value, the corresponding feature combination is labeled as a qualified feature combination. By combining the qualified feature combinations, the registered anti-interference image and anti-interference data are obtained.

[0023] It should be noted that 2D and 3D feature fingerprints record corresponding feature data. The standard natural feature points of the electrical equipment under inspection need to exhibit high contrast, clear edges, and regular shapes in the image, such as circular bolt heads and rectangular label corners, and should have obvious height abrupt changes, planar boundaries, or curvature variations. Examples include the hexagonal head of a mounting bolt and the rectangular corner of an interface connector. A consistency check is performed on the texture features in the 2D feature fingerprint and the geometric features in the 3D feature fingerprint to compare whether the 2D texture features and 3D geometric features describe the same type of object; for example, a bolt head should simultaneously exhibit circular texture and cylindrical geometry. The texture features in the 2D feature fingerprint and the geometric features in the 3D feature fingerprint are randomly paired to find all qualified feature combinations and achieve alignment. If the positional deviation pixel value between the texture features in the 2D feature fingerprint and the geometric features in the 3D feature fingerprint is less than a preset value, the corresponding feature combination is labeled as a qualified feature combination. The registered anti-interference image and anti-interference data contain qualified feature combinations.

[0024] Furthermore, in a preferred embodiment of the present invention, the step of performing data preprocessing on the registered anti-interference image and anti-interference data, and combining the preprocessed data to perform defect detection and identification of rail transit trains, specifically includes: In the registered anti-interference image and anti-interference data, three indicators are calculated for the anti-interference image: global contrast, information entropy, and peak signal-to-noise ratio. In the anti-interference data, three indicators are calculated: point density uniformity, noise ratio, and effective data integrity. Scoring thresholds are set for different indicators of the registered anti-interference image and anti-interference data. Based on the scoring thresholds, different indicators are scored on the registered anti-interference image and anti-interference data. A solution is introduced to address the issue of indicators whose scores in big data network retrieval are lower than preset values, until the scores of different indicators of the registered anti-interference image and anti-interference data are all greater than preset values, thus obtaining qualified anti-interference images and qualified anti-interference data. The method involves assigning the RGB color values ​​of 2D pixels from a qualified anti-interference image to a 3D point cloud of qualified anti-interference data to generate a preliminary pre-processed 3D point cloud. Semantic labels are applied to different pixels on qualified anti-interference images and simultaneously assigned to the pre-processed 3D point cloud to generate synchronous semantic 3D point cloud. The 3D feature values ​​of the synchronous semantic 3D point cloud are calculated to generate a high-dimensional fusion feature vector. By combining high-dimensional fusion feature vectors for analysis, defect detection and identification of rail transit trains can be achieved.

[0025] It should be noted that three metrics—global contrast, information entropy, and peak signal-to-noise ratio—are calculated for anti-interference images, and three metrics—point density uniformity, noise ratio, and effective data integrity—are calculated for anti-interference data. These metrics are all used to evaluate the data quality of 2D images or 3D point clouds. Data preprocessing methods are tailored to data quality, including: for high-quality 2D images, adaptive histogram equalization and mild enhancement with a desharpening mask are performed; for high-quality 3D point clouds, statistical filtering is used to remove outliers; for 2D images with average data quality, the Retinex multi-scale illumination compensation algorithm is used; for 3D point clouds, bilateral filtering is used to smooth noise while preserving edges; for poor-quality data, such as areas with excessive darkness or severe noise, 2D images use a deep learning-based local enhancement model to restore low-brightness areas; and 3D point clouds are repaired using normal-vector guided filtering and moving least squares surface reconstruction. This completes the data preprocessing. Assigning RGB values ​​and semantic information to a 3D point cloud as a 2D image can produce an intuitive 3D point cloud that describes the semantics of different locations and generates a high-dimensional fusion feature vector, which facilitates subsequent defect detection.

[0026] Furthermore, in a preferred embodiment of the present invention, the step of combining high-dimensional fusion feature vectors for analysis to achieve defect detection and identification of rail transit trains specifically includes: A knowledge base of defects in rail transit trains is acquired, and a three-layer defect identification model is constructed based on the defects in rail transit trains. The defects in rail transit trains include physical anomaly defects, state anomaly defects, and functional risk defects. Different layers in the three-layer defect identification model are used to identify different defect types. The method for constructing the three-layer defect identification model is as follows: input the sample vector from the high-dimensional fusion feature vector into the baseline model of the three-layer defect identification model, and introduce feature analysis rules into the baseline model to preliminarily analyze the fuzzy defect type of the electrical equipment to be inspected on the sample vector. Among them, the feature analysis rules are preset feature vector templates corresponding to the defect types of different electrical equipment to be inspected in rail transit trains; Different fuzzy defect types of electrical equipment under inspection are assigned to different layers of the baseline model of the three-layer defect recognition model. At the same time, a convolutional neural network is introduced to perform convolutional analysis on different fuzzy defect types of electrical equipment under inspection, so as to realize secondary classification training for different fuzzy defect types of electrical equipment under inspection. By using a three-layer defect identification model, defects are identified from high-dimensional fused feature vectors, and the defect types of different electrical equipment under inspection are output.

[0027] It should be noted that a three-layer defect identification model is trained to analyze the defect types represented in the high-dimensional fused feature vectors. The rail transit train defect knowledge base records vectors corresponding to all rail transit train defects. Physical anomalies include missing parts, deformation, cracks, loosening, contamination, corrosion, and color anomalies; state anomalies include improper installation, stress concentration, fatigue signs, and sealing failure; and functional risk defects include electrical connection risks, mechanical failure risks, and thermal runaway risks. The three-layer defect identification model is constructed, and combined with feature analysis rules, the sample vectors of the high-dimensional fused feature vectors are analyzed to train the model. The model can screen for significant defects in the vectors and further refine the analysis. A convolutional neural network is introduced to perform convolutional analysis on fuzzy defect types of different electrical equipment under inspection. The neural network can determine the presence of defect vectors by comparing them with standard vectors, find the corresponding locations, and generate the defect type of the electrical equipment under inspection.

[0028] Figure 2 A flowchart illustrating a method for acquiring anti-interference images and anti-interference data of electrical equipment under inspection is shown, including the following steps: S202: Combining a 2D area array camera system and a 3D line laser scanner, and incorporating a high-precision time-division triggering mechanism, it acquires anti-interference images and data of the electrical equipment under inspection; S204: Perform dynamic delay compensation on the acquired 2D images and acquired 3D laser data to obtain the anti-interference image and anti-interference data of the electrical equipment under inspection.

[0029] Furthermore, in a preferred embodiment of the present invention, the combination of a 2D area scan camera system and a 3D line laser scanner, along with a high-precision time-division triggering mechanism, to acquire anti-interference images and data of the electrical equipment under inspection, specifically involves: A photoelectric sensor is installed at the starting position of the target detection area. The photoelectric sensor is used to detect whether the rail transit train is blocking the photoelectric signal emitted by the photoelectric sensor. If so, it is determined that the rail transit train is passing through the target detection area, and the photoelectric sensor is controlled to emit a start trigger signal. An intelligent time-sharing controller is introduced and connected to a photoelectric sensor, a 2D area array camera system, and a 3D line laser scanner, respectively. The intelligent time-sharing controller receives the start trigger signal from the photoelectric sensor in real time. After receiving the start trigger signal from the photoelectric sensor, the intelligent time-sharing controller outputs the pre-start trigger signal for the 2D area scan camera system and the 3D line laser scanner, thereby realizing the pre-start processing of the 2D area scan camera system and the 3D line laser scanner. The intelligent time-sharing controller analyzes in real time the duration of the photoelectric signal emitted by the photoelectric sensor that is blocked by the rail transit train, the speed of the rail transit train, and the individual parameters of the rail transit train to generate the real-time displacement value of the rail transit train. A predetermined displacement threshold is set. When the real-time displacement value of the rail transit train reaches the displacement threshold, the 2D area array camera system and the 3D line laser scanner are activated by the intelligent time-sharing controller to achieve the purpose of acquiring 2D images and 3D laser data. Dynamic delay compensation is performed on the acquired 2D images and 3D laser data to obtain anti-interference images and anti-interference data of the electrical equipment under inspection.

[0030] It should be noted that the photoelectric sensor is used to detect whether the rail transit train is blocking the photoelectric signal emitted by the sensor. When the rail transit train coincides with the photoelectric signal, the 2D area scan camera system and 3D line laser scanner can begin operation. A time-sharing controller can be set to calculate the train's displacement and receive photoelectric signals, while simultaneously controlling the operation of the 2D area scan camera system and 3D line laser scanner. Pre-startup processing for the 2D area scan camera system and 3D line laser scanner can reduce startup delays and allow them to enter a ready state. Real-time analysis of the duration of the photoelectric signal emitted by the rail transit train blocking the sensor, the train's speed, and individual parameters generates a real-time displacement value for the train. This real-time displacement value can determine the timing and alignment position of the camera and laser emitter.

[0031] Furthermore, in a preferred embodiment of the present invention, the step of performing dynamic delay compensation on the acquired 2D images and acquired 3D laser data to obtain the anti-interference image and anti-interference data of the electrical equipment under inspection specifically includes: A high-speed oscilloscope was introduced to measure the time from when the intelligent time-sharing controller receives the initial trigger signal from the photoelectric sensor to when it generates the real-time displacement value of the rail transit train, and this time was calibrated as the running trigger delay time. If the trigger delay time is not greater than the preset value, the acquired 2D image and 3D laser data will be directly calibrated as the anti-interference image and anti-interference data of the electrical equipment under inspection. If the operation trigger delay time is greater than the preset value, the operation trigger delay time will be combined with the speed of the rail transit train and the individual parameters of the rail transit train to generate the delay displacement value of the rail transit train, which will be calibrated as the displacement compensation value. Based on the displacement compensation value, the intelligent time-sharing controller is re-controlled to start the 2D area array camera system and the 3D line laser scanner respectively, so as to obtain the anti-interference image and anti-interference data of the electrical equipment under inspection.

[0032] It should be noted that the high-speed oscilloscope accurately measures the time from when the intelligent time-sharing controller receives the initial trigger signal from the photoelectric sensor to when it generates the real-time displacement value of the rail transit train, thus enabling delay analysis. Because of this delay, there may be a specific distance the train travels before image and data acquisition begins. Therefore, displacement compensation is necessary to prevent image and data deviations caused by displacement errors.

[0033] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An anti-interference image acquisition and detection method based on physically staggered optical paths, characterized in that, Includes the following steps: Deploy a multi-sensor system for rail transit and design the physical optical path of the multi-sensor system; By combining a 2D area array camera system and a 3D line laser scanner, along with a high-precision time-division triggering mechanism, anti-interference images and data of the electrical equipment under inspection are acquired. Data registration is performed on the anti-interference image and anti-interference data of the electrical equipment to be inspected to obtain the registered anti-interference image and anti-interference data. The registered anti-interference images and anti-interference data are preprocessed, and the preprocessed data is combined to detect and identify defects in rail transit trains.

2. The anti-interference image acquisition and detection method based on physically staggered optical paths as described in claim 1, characterized in that, The layout is used for a multi-sensor system in rail transit, and the physical optical path of the multi-sensor system is designed as follows: Determine the travel position of the rail transit train, use it as the workstation for laying out the multi-sensor system for rail transit, and mark it as the target layout workstation; Measure the spatial parameters of the target layout workstation and the ambient light parameters of the target layout workstation. At the same time, determine the relative parameters of the electrical equipment to be inspected traveling on the target layout workstation. The electrical equipment to be inspected is the equipment on the rail transit train that needs to be fault identified. A 2D area scan camera system and a 3D line laser scanner are acquired. The 2D area scan camera system and the 3D line laser scanner are used to collect anti-interference images and data of the electrical equipment to be inspected at the target layout workstation. The coverage area of ​​the electrical equipment to be inspected during its movement is determined and marked as the target inspection area. The 2D area array camera system and the 3D line laser scanner are then installed on the target layout station using a universal bracket. Adjust the universal bracket, control the 2D area scan camera system and 3D line laser scanner to perform physical optical path misalignment design through Y-axis optical path interlacing method, and install mechanical light shields on the 2D area scan camera system and 3D line laser scanner.

3. The anti-interference image acquisition and detection method based on physically staggered optical paths as described in claim 1, characterized in that, The system combines a 2D area array camera system and a 3D line laser scanner, along with a high-precision time-division triggering mechanism, to acquire anti-interference images and data of the electrical equipment under inspection. Specifically: A photoelectric sensor is installed at the starting position of the target detection area. The photoelectric sensor is used to detect whether the rail transit train is blocking the photoelectric signal emitted by the photoelectric sensor. If so, it is determined that the rail transit train is passing through the target detection area, and the photoelectric sensor is controlled to emit a start trigger signal. An intelligent time-sharing controller is introduced and connected to a photoelectric sensor, a 2D area array camera system, and a 3D line laser scanner, respectively. The intelligent time-sharing controller receives the start trigger signal from the photoelectric sensor in real time. After receiving the start trigger signal from the photoelectric sensor, the intelligent time-sharing controller outputs the pre-start trigger signal for the 2D area scan camera system and the 3D line laser scanner, thereby realizing the pre-start processing of the 2D area scan camera system and the 3D line laser scanner. The intelligent time-sharing controller analyzes in real time the duration of the photoelectric signal emitted by the photoelectric sensor that is blocked by the rail transit train, the speed of the rail transit train, and the individual parameters of the rail transit train to generate the real-time displacement value of the rail transit train. A predetermined displacement threshold is set. When the real-time displacement value of the rail transit train reaches the displacement threshold, the 2D area array camera system and the 3D line laser scanner are activated by the intelligent time-sharing controller to achieve the purpose of acquiring 2D images and 3D laser data. Dynamic delay compensation is performed on the acquired 2D images and 3D laser data to obtain anti-interference images and anti-interference data of the electrical equipment under inspection.

4. The anti-interference image acquisition and detection method based on physical optical path misalignment as described in claim 3, characterized in that, The process of performing dynamic delay compensation on the acquired 2D images and 3D laser data to obtain the anti-interference image and anti-interference data of the electrical equipment under inspection is specifically as follows: A high-speed oscilloscope was introduced to measure the time from when the intelligent time-sharing controller receives the initial trigger signal from the photoelectric sensor to when it generates the real-time displacement value of the rail transit train, and this time was calibrated as the running trigger delay time. If the trigger delay time is not greater than the preset value, the acquired 2D image and 3D laser data will be directly calibrated as the anti-interference image and anti-interference data of the electrical equipment under inspection. If the operation trigger delay time is greater than the preset value, the operation trigger delay time will be combined with the speed of the rail transit train and the individual parameters of the rail transit train to generate the delay displacement value of the rail transit train, which will be calibrated as the displacement compensation value. Based on the displacement compensation value, the intelligent time-sharing controller is re-controlled to start the 2D area array camera system and the 3D line laser scanner respectively, so as to obtain the anti-interference image and anti-interference data of the electrical equipment under inspection.

5. The anti-interference image acquisition and detection method based on physical optical path misalignment as described in claim 1, characterized in that, The process of registering the anti-interference image and anti-interference data of the electrical equipment under inspection to obtain the registered anti-interference image and anti-interference data is as follows: The anti-interference images and anti-interference data of the electrical equipment under inspection are calibrated as 2D anti-interference images and 3D anti-interference data. The standard natural feature points of the electrical equipment to be inspected are preset. The positions corresponding to the standard natural feature points are identified in the 2D anti-interference image and 3D anti-interference data, and 2D feature fingerprints and 3D feature fingerprints are generated. Among them, the 2D feature fingerprint records the pixel coordinates and texture features of standard natural feature points in the 2D anti-interference image, and the 3D feature fingerprint records the three-dimensional spatial coordinates and geometric features of standard natural feature points in the 3D anti-interference data. A consistency check is performed on the texture features in the 2D fingerprint and the geometric features in the 3D fingerprint. The consistency check involves randomly pairing the texture features in the 2D fingerprint and the geometric features in the 3D fingerprint to obtain different feature combinations. For each pair of feature combinations, a 2D-3D transformation relationship is calculated. The 2D-3D transformation relationship is calculated by mapping the geometric features in the 3D feature fingerprint onto the 2D anti-interference image. If the positional deviation pixel value between the texture features in the 2D feature fingerprint and the geometric features in the 3D feature fingerprint is less than a preset value, the corresponding feature combination is labeled as a qualified feature combination. By combining the qualified feature combinations, the registered anti-interference image and anti-interference data are obtained.

6. The anti-interference image acquisition and detection method based on physical optical path misalignment as described in claim 1, characterized in that, The process involves preprocessing the registered anti-interference image and anti-interference data, and then combining the preprocessed data to perform defect detection and identification of rail transit trains. Specifically: In the registered anti-interference image and anti-interference data, three indicators are calculated for the anti-interference image: global contrast, information entropy, and peak signal-to-noise ratio. In the anti-interference data, three indicators are calculated: point density uniformity, noise ratio, and effective data integrity. Scoring thresholds are set for different indicators of the registered anti-interference image and anti-interference data. Based on the scoring thresholds, different indicators are scored on the registered anti-interference image and anti-interference data. A solution is introduced to address the issue of indicators whose scores in big data network retrieval are lower than preset values, until the scores of different indicators of the registered anti-interference image and anti-interference data are all greater than preset values, thus obtaining qualified anti-interference images and qualified anti-interference data. The method involves assigning the RGB color values ​​of 2D pixels from a qualified anti-interference image to a 3D point cloud of qualified anti-interference data to generate a preliminary pre-processed 3D point cloud. Semantic labels are applied to different pixels on qualified anti-interference images and simultaneously assigned to the pre-processed 3D point cloud to generate synchronous semantic 3D point cloud. The 3D feature values ​​of the synchronous semantic 3D point cloud are calculated to generate a high-dimensional fusion feature vector. By combining high-dimensional fusion feature vectors for analysis, defect detection and identification of rail transit trains can be achieved.

7. The anti-interference image acquisition and detection method based on physical optical path misalignment as described in claim 6, characterized in that, The method of combining high-dimensional fusion feature vectors for analysis to achieve defect detection and identification of rail transit trains is as follows: A knowledge base of defects in rail transit trains is acquired, and a three-layer defect identification model is constructed based on the defects in rail transit trains. The defects in rail transit trains include physical anomaly defects, state anomaly defects, and functional risk defects. Different layers in the three-layer defect identification model are used to identify different defect types. The method for constructing the three-layer defect identification model is as follows: input the sample vector from the high-dimensional fusion feature vector into the baseline model of the three-layer defect identification model, and introduce feature analysis rules into the baseline model to preliminarily analyze the fuzzy defect type of the electrical equipment to be inspected on the sample vector. Among them, the feature analysis rules are preset feature vector templates corresponding to the defect types of different electrical equipment to be inspected in rail transit trains; Different fuzzy defect types of electrical equipment under inspection are assigned to different layers of the baseline model of the three-layer defect recognition model. At the same time, a convolutional neural network is introduced to perform convolutional analysis on different fuzzy defect types of electrical equipment under inspection, so as to realize secondary classification training for different fuzzy defect types of electrical equipment under inspection. By using a three-layer defect identification model, defects are identified from high-dimensional fused feature vectors, and the defect types of different electrical equipment under inspection are output.

Images