A new track geometry parameter detection method

Abstract

This invention belongs to the field of track detection technology. It provides a novel method for detecting track geometric parameters. The method includes: determining and controlling a high-definition camera to rotate to a shooting angle based on the angle of sunlight illumination; performing track identification on the captured high-definition image data to obtain two track segments; performing foreign object detection on each track segment; predicting the attribute category of the foreign object based on its characteristics; dividing the track segments into smooth track segments and foreign object track segments based on the distribution location of the foreign object; extracting the first track geometric parameters of each smooth track segment using image recognition technology; calculating the second track geometric parameters of the foreign object track segment based on its attribute category; and aligning and integrating the first and second track geometric parameters to obtain the final track geometric parameters. This invention achieves dynamic adjustment of the camera's detection field of view and successful extraction of track geometric parameters from areas covered by foreign objects.

Description

Technical Field

[0001] This invention relates to the field of track inspection technology, and more specifically, to a novel method for detecting track geometric parameters. Background Technology

[0002] The detection of track geometry parameters is crucial for ensuring the safety, smoothness, and comfort of train operation. These parameters include track gauge, alignment, elevation, levelness, superelevation, and triangular irregularities, which directly reflect the quality condition of the track. With the increase in train speed, especially the development of high-speed railways, the requirements for track smoothness are becoming increasingly stringent, because even minor irregularities can lead to serious safety problems under high-speed operating conditions.

[0003] Detecting track geometric parameters using machine vision and image processing technology is a commonly used method. However, in practice, the detection angle of the camera is fixed, and the corresponding field of view is also fixed, which is detrimental to improving the efficiency of track geometric parameter detection. Although efficient detection methods based on dynamic fields of view are widely needed in the field, existing technologies have not yet provided relevant solutions. Furthermore, foreign objects often appear on the track, and their presence affects the image extraction of track geometric parameters; existing technologies also lack effective solutions to this problem. The technical objective of this invention is to solve at least one of the aforementioned technical problems. Summary of the Invention

[0004] To address this issue, the present invention provides a novel method, system, electronic device, computer storage medium, and computer program product for detecting orbital geometric parameters, thereby solving the aforementioned technical problems.

[0005] This invention discloses a novel method for detecting track geometry parameters, applied to an automatic inspection vehicle, which includes a high-definition camera and a light sensor; the method includes the following steps:

[0006] A light sensor detects the angle of sunlight in the current area, determines the shooting angle of the high-definition camera based on the angle of sunlight, and controls the high-definition camera to rotate to the shooting angle; wherein, different shooting angles correspond to different sizes of shooting field of view;

[0007] Track identification is performed on the high-definition image data captured by the high-definition camera at the shooting angle to obtain two track segments, and foreign object detection is performed on each track segment. Based on the detected foreign object features, the attribute category of the foreign object is predicted.

[0008] Based on the distribution location of the foreign object, the track segment is divided into a smooth track segment and a foreign object track segment. Image recognition technology is used to extract the first track geometric parameters of each smooth track segment, and the second track geometric parameters of the foreign object track segment are calculated based on the attribute category.

[0009] The first and second track geometry parameters are aligned and integrated to obtain the track geometry parameters.

[0010] In some embodiments, determining the shooting angle of the high-definition camera based on the illumination angle of sunlight includes:

[0011] Based on high-precision map data, the orientation data of the track is determined. Based on the orientation data and the illumination angle of the sunlight, the range of reflection angle of the sunlight reflected on the track at the illumination angle is calculated.

[0012] Based on the three-dimensional dimensions of the automatic inspection vehicle and the three-dimensional deployment position of the high-definition camera on the automatic inspection vehicle, the shooting angle that does not coincide with the reflection angle range is determined within the angle adjustment range of the high-definition camera, and the difference between the shooting angle and the current shooting angle of the high-definition camera is minimized.

[0013] In some embodiments, the foreign object detection of each of the track segments and the prediction of the attribute category of the foreign object based on the detected foreign object features include:

[0014] Extract the first appearance features of each track segment, perform a continuity analysis of the appearance features based on the first appearance features, and identify areas where the appearance features are discontinuous as foreign objects.

[0015] The second appearance feature of the foreign object is extracted, which includes the appearance feature and the structural feature of the foreign object. Based on the appearance feature and the structural feature of the foreign object, the foreign object category is predicted. Based on the foreign object category and the structural feature of the foreign object, the geometric fullness and hardness of the foreign object are predicted.

[0016] The foreign object category, the geometric fullness, and the hardness are used as the attribute categories of the foreign object.

[0017] In some embodiments, predicting the geometric fullness and hardness of the foreign object based on the foreign object category and the foreign object structural features includes:

[0018] Based on the foreign object category and the foreign object structural features, the first geometric structure of the foreign object under normal conditions is predicted, and the real-time second geometric structure of the foreign object is extracted from the foreign object structural features.

[0019] The change range of the geometric structure of the foreign object is calculated based on the first geometric structure and the second geometric structure, and the geometric fullness is determined based on the change range.

[0020] The dynamic characteristics of the foreign object are extracted from its structural features, the dynamic value of the foreign object is calculated based on the dynamic characteristics, and the hardness of the foreign object is determined based on the dynamic value.

[0021] In some embodiments, the automated detection vehicle further includes an ultrasonic detector, then calculating the second track geometry parameters of the foreign object track segment based on the attribute category includes:

[0022] If the hardness is lower than the hardness threshold, the second track geometry parameters of the foreign object track segment are predicted based on the geometric parameters of the two first tracks adjacent to the upper and lower tracks of the foreign object track segment.

[0023] If the hardness is higher than the hardness threshold, then:

[0024] If the geometric fullness is lower than the fullness threshold, the second track geometric parameters of the foreign object track segment are predicted based on the geometric parameters of the two first tracks that are adjacent to each other above and below the foreign object track segment.

[0025] If the geometric fullness is higher than the fullness threshold, the ultrasonic detector is used to detect the foreign object track segment in order to extract the second track geometric parameters.

[0026] In some embodiments, predicting the first geometric structure of the foreign object under normal conditions based on the foreign object category and the foreign object structural features includes:

[0027] The foreign object category and the foreign object structural features are input into the geometric structure prediction model, and the geometric structure prediction model outputs its prediction of the first geometric structure of the foreign object under normal conditions; wherein, the geometric structure prediction model is obtained by fine-tuning the BERT large model.

[0028] This invention also discloses a novel track geometry parameter detection system applied to an automatic inspection vehicle, the automatic inspection vehicle including a high-definition camera and a light sensor; the system includes a processing module and a storage module, the processing module implementing the following method steps by calling and executing computer code in the storage module:

[0029] A light sensor detects the angle of sunlight in the current area, determines the shooting angle of the high-definition camera based on the angle of sunlight, and controls the high-definition camera to rotate to the shooting angle; wherein, different shooting angles correspond to different sizes of shooting field of view;

[0030] Track identification is performed on the high-definition image data captured by the high-definition camera at the shooting angle to obtain two track segments, and foreign object detection is performed on each track segment. Based on the detected foreign object features, the attribute category of the foreign object is predicted.

[0031] Based on the distribution location of the foreign object, the track segment is divided into a smooth track segment and a foreign object track segment. Image recognition technology is used to extract the first track geometric parameters of each smooth track segment, and the second track geometric parameters of the foreign object track segment are calculated based on the attribute category.

[0032] The first and second track geometry parameters are aligned and integrated to obtain the track geometry parameters.

[0033] The present invention also discloses an electronic device comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, the processor executing the computer program to implement the method as described in any of the preceding claims.

[0034] The present invention also discloses a computer storage medium storing a computer program that is executed by a processor to implement the methods described in any of the preceding methods.

[0035] The present invention also discloses a computer program product that, when run on a terminal, causes the terminal to execute in order to implement the method described in any of the preceding methods.

[0036] The beneficial effects of this invention are as follows:

[0037] (1) This invention selects the appropriate shooting angle of a high-definition camera by analyzing the angle of sunlight, thereby realizing the dynamic adjustment of its shooting field of view, which can effectively improve the extraction efficiency of orbital geometric parameters.

[0038] (2) This invention analyzes and identifies the attribute categories of foreign objects on the track, and calculates the track geometry parameters of the track segment covered by the foreign object based on the attribute category, thereby realizing the smooth detection of track geometry parameters when there are foreign objects covering the track, which can greatly improve the adaptability of track geometry parameter detection. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of a scenario where an automatic inspection vehicle is used to detect track geometric parameters, as disclosed in an embodiment of the present invention.

[0041] Figure 2 This is a flowchart illustrating a novel method for detecting track geometric parameters disclosed in an embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of the structure of a novel track geometry parameter detection system disclosed in an embodiment of the present invention. Detailed Implementation

[0043] The following specific embodiments illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0044] Furthermore, the technical features involved in the different embodiments of this application described below can be combined with each other as long as they do not conflict with each other.

[0045] like Figure 1 , Figure 2 As shown in the figure, this invention discloses a novel method for detecting track geometry parameters, applied to an automatic inspection vehicle. The automatic inspection vehicle includes a high-definition camera and a light sensor. The method includes the following steps:

[0046] A light sensor detects the angle of sunlight in the current area, determines the shooting angle of the high-definition camera based on the angle of sunlight, and controls the high-definition camera to rotate to the shooting angle; wherein, different shooting angles correspond to different sizes of shooting field of view;

[0047] Track identification is performed on the high-definition image data captured by the high-definition camera at the shooting angle to obtain two track segments, and foreign object detection is performed on each track segment. Based on the detected foreign object features, the attribute category of the foreign object is predicted.

[0048] Based on the distribution location of the foreign object, the track segment is divided into a smooth track segment and a foreign object track segment. Image recognition technology is used to extract the first track geometric parameters of each smooth track segment, and the second track geometric parameters of the foreign object track segment are calculated based on the attribute category.

[0049] The first and second track geometry parameters are aligned and integrated to obtain the track geometry parameters.

[0050] This invention constructs an automatic inspection vehicle equipped with a high-definition camera and a light sensor. The optimal shooting angle for the high-definition camera is determined based on the angle of sunlight detected by the light sensor. When the high-definition camera adjusts to this optimal shooting angle, its field of view is adjusted. Specifically, the smaller the shooting angle, the closer the upper edge of the high-definition camera's field of view is to the automatic inspection vehicle, resulting in a smaller field of view (i.e., close-up view), covering only the nearby track. Conversely, the larger the shooting angle, the farther the upper edge of the high-definition camera's field of view is from the automatic inspection vehicle, resulting in a larger field of view (i.e., distant view), covering a longer and more distant section of the track.

[0051] Simultaneously, after capturing high-definition image data from the aforementioned shooting angle, the high-definition camera identifies the track contained within, obtaining two track segments, left and right. In addition, foreign object detection is performed on each track segment, and the attribute category of the detected foreign object is analyzed based on its characteristics. Based on the distribution location of the foreign object, the track segments are divided into smooth track segments and foreign object track segments; the foreign object track segment is the area on the track where a foreign object exists.

[0052] Next, image recognition technology is used to extract the first track geometry parameters of each smooth track segment, and the second track geometry parameters of the foreign object track segment are calculated based on the aforementioned attribute categories of the foreign object. The track geometry parameters include at least one of track gauge, track surface flatness, and track orientation.

[0053] Finally, by aligning and integrating the first and second track geometric parameters obtained above, the track geometric parameters of the left and right tracks corresponding to the current high-definition image data are obtained.

[0054] Therefore, on the one hand, this invention selects a suitable shooting angle for a high-definition camera by analyzing the angle of sunlight, thereby dynamically adjusting its field of view and effectively improving the extraction efficiency of track geometry parameters. On the other hand, this invention analyzes and identifies the attribute categories of foreign objects on the track, and calculates the track geometry parameters of the track segment covered by the foreign object based on these attribute categories. This enables the successful detection of track geometry parameters when foreign objects are covering the track, greatly improving the adaptability of track geometry parameter detection.

[0055] It should be noted that the automatic detection vehicle used in this invention can be fully automated or operated under manual remote control; this invention does not limit its operation. Furthermore, Figure 1Although the power system, processing system, communication system, and other functional components of the automatic detection vehicle are not shown in the diagram, the configuration of these functional components is a mature existing technology in this field, and will not be described in detail here. The shooting angle adjustment of the high-definition camera can be achieved based on a multi-axis gimbal.

[0056] In some embodiments, determining the shooting angle of the high-definition camera based on the illumination angle of sunlight includes:

[0057] Based on high-precision map data, the orientation data of the track is determined. Based on the orientation data and the illumination angle of the sunlight, the range of reflection angle of the sunlight reflected on the track at the illumination angle is calculated.

[0058] Based on the three-dimensional dimensions of the automatic inspection vehicle and the three-dimensional deployment position of the high-definition camera on the automatic inspection vehicle, the shooting angle that does not coincide with the reflection angle range is determined within the angle adjustment range of the high-definition camera, and the difference between the shooting angle and the current shooting angle of the high-definition camera is minimized.

[0059] In this embodiment of the invention, the automatic detection vehicle can determine the orientation data of the currently detected track based on high-precision map data to establish a three-dimensional coordinate system with the track as the reference. Combining the sunlight's illumination angle with the three-dimensional coordinate system, the range of reflection angles of the sunlight reflected on the track at that illumination angle can be inferred. Due to constant frictional contact with train wheels, the top surface of the track is very smooth and has a strong light-reflecting ability. Excessively strong reflected sunlight will cause overexposure in the high-definition image data captured by the high-definition camera, making it impossible to extract the track's geometric parameters.

[0060] In response, this invention uses the aforementioned three-dimensional coordinate system as a basis. Based on the three-dimensional dimensions of the automatic inspection vehicle and the three-dimensional deployment position of the high-definition camera on the automatic inspection vehicle, the coordinates of the high-definition camera lens can be determined. Then, from the angle adjustment range of the high-definition camera, several shooting angles that do not coincide with the aforementioned reflection angle range can be selected. From these, the shooting angle with the smallest difference from the current shooting angle of the high-definition camera can be selected. This can effectively reduce the travel distance of the high-definition camera in adjusting the shooting angle, reduce the waiting time of the automatic inspection vehicle (it needs to stop or slow down when adjusting the shooting angle), and improve the detection efficiency of track geometry parameters.

[0061] In some embodiments, the foreign object detection of each of the track segments and the prediction of the attribute category of the foreign object based on the detected foreign object features include:

[0062] Extract the first appearance features of each track segment, perform a continuity analysis of the appearance features based on the first appearance features, and identify areas where the appearance features are discontinuous as foreign objects.

[0063] The second appearance feature of the foreign object is extracted, which includes the appearance feature and the structural feature of the foreign object. Based on the appearance feature and the structural feature of the foreign object, the foreign object category is predicted. Based on the foreign object category and the structural feature of the foreign object, the geometric fullness and hardness of the foreign object are predicted.

[0064] The foreign object category, the geometric fullness, and the hardness are used as the attribute categories of the foreign object.

[0065] In this embodiment of the invention, the upper surface of the track is typically smooth. The geometric features of the track extracted using image detection technology (e.g., the edge lines of the track's upper surface) should be continuous and smooth. However, if foreign objects are present, the corresponding area will become discontinuous and rough. To address this, the present invention performs a continuity analysis on the first appearance features of the extracted track segment. When a discontinuity in the appearance features is detected, the corresponding area is identified as a foreign object. This foreign object identification method of the present invention has a particularly low computational load and does not require pre-setting features for various types of foreign objects, which can greatly improve the efficiency of foreign object identification. It can support automatic detection vehicles to perform track geometric parameter detection and extraction at higher operating speeds, thereby improving the detection efficiency of track geometric parameters.

[0066] After identifying the foreign object, its secondary appearance features are extracted. These features include both physical and structural characteristics. Physical features primarily include the object's color, texture, and reflectivity, while structural features refer to its geometric dimensions, the length of the portion protruding from the track surface, and its dynamic characteristics, such as whether it exhibits significant swaying. Based on these physical and structural features, the object's category can be predicted, such as a metal sheet, paper, plastic bottle, or milk carton. Then, based on this category and structural features, the object's geometric fullness and hardness are predicted. Finally, the obtained object category, geometric fullness, and hardness are used as the object's attribute category.

[0067] In some embodiments, predicting the geometric fullness and hardness of the foreign object based on the foreign object category and the foreign object structural features includes:

[0068] Based on the foreign object category and the foreign object structural features, the first geometric structure of the foreign object under normal conditions is predicted, and the real-time second geometric structure of the foreign object is extracted from the foreign object structural features.

[0069] The change range of the geometric structure of the foreign object is calculated based on the first geometric structure and the second geometric structure, and the geometric fullness is determined based on the change range.

[0070] The dynamic characteristics of the foreign object are extracted from its structural features, the dynamic value of the foreign object is calculated based on the dynamic characteristics, and the hardness of the foreign object is determined based on the dynamic value.

[0071] In this embodiment of the invention, to facilitate subsequent analysis of the geometric parameters of the track in the area where the foreign object exists, it is necessary to predetermine the geometric fullness and hardness of the foreign object. Geometric fullness refers to the degree of change between the current geometric structure of the foreign object and its original geometric structure; hardness refers to the degree of softness or hardness of the foreign object itself, for example, metal sheets are high hardness and paper is low hardness.

[0072] Specifically, firstly, based on the aforementioned attribute categories and structural features of the foreign object, the first geometric structure of the foreign object under normal conditions is predicted. For example, if a foreign object is identified as a milk carton, further consideration of its geometric dimensions (which belong to the structural features of the foreign object) allows for the prediction of the first geometric structure of the milk carton when it is in good condition. The real-time second geometric structure of the foreign object can also be directly extracted from its structural features. Then, by comparing the magnitude of geometric change between the first and second geometric structures, the geometric fullness of the foreign object can be determined based on this magnitude. Obviously, the larger the magnitude of change, the lower the geometric fullness, and vice versa. This magnitude of change can, for example, be equal to the ratio of the volume corresponding to the second geometric structure to the volume corresponding to the first geometric structure.

[0073] Next, the dynamic characteristics of the foreign object are extracted from its structural features. Based on these dynamic characteristics, the dynamic value of the foreign object can be calculated. The dynamic characteristics refer to the amplitude and direction of the positional variation of the reference point of the foreign object. A larger amplitude and more varied directions of variation indicate lower hardness, such as in paper; conversely, lower amplitude and more varied directions indicate higher hardness, such as in metal sheets. The hardness of the foreign object can then be determined based on this dynamic value.

[0074] In some embodiments, the automated detection vehicle further includes an ultrasonic detector, then calculating the second track geometry parameters of the foreign object track segment based on the attribute category includes:

[0075] If the hardness is lower than the hardness threshold, the second track geometry parameters of the foreign object track segment are predicted based on the geometric parameters of the two first tracks adjacent to the upper and lower tracks of the foreign object track segment.

[0076] If the hardness is higher than the hardness threshold, then:

[0077] If the geometric fullness is lower than the fullness threshold, the second track geometric parameters of the foreign object track segment are predicted based on the geometric parameters of the two first tracks that are adjacent to each other above and below the foreign object track segment.

[0078] If the geometric fullness is higher than the fullness threshold, the ultrasonic detector is used to detect the foreign object track segment in order to extract the second track geometric parameters.

[0079] In this embodiment of the invention, the automatic inspection vehicle is also equipped with an ultrasonic detector (such as...). Figure 1 As shown, the ultrasonic detector and high-definition camera are integrated. The ultrasonic waves emitted by the detector can penetrate foreign objects covering the track, thereby detecting the track's geometric parameters and determining whether there are pits or cracks in the track (possibly caused by train wheels running over the hard foreign object). Therefore, this invention provides a method for detecting and determining the second track geometric parameters of the track segment with foreign objects based on the hardness and geometric fullness of the foreign object.

[0080] Specifically, if the hardness of the foreign object is below the hardness threshold, it indicates that it is a soft material such as paper or a milk carton. Even if the train wheels run over it, it will not damage the track. In this case, the second track geometry parameters of the foreign object track segment can be obtained by directly performing line fitting on the geometric parameters of the two adjacent first track segments. For line fitting, for example, the edge lines of the upper surfaces of the two adjacent smooth track segments can be detected in advance, and the lower endpoint of the upper edge line can be connected to the upper endpoint of the lower edge line to form the edge line of the foreign object track segment.

[0081] If the hardness of the foreign object exceeds the hardness threshold, it may cause defects such as dents or cracks on the track due to being crushed by train wheels, affecting track safety. To address this, further analysis is needed to determine if the geometric fullness of the foreign object is below the fullness threshold. If it is below the fullness threshold, it indicates that the foreign object underwent significant deformation after being crushed by the train wheels. While its hardness is above the hardness threshold (which is usually set too low), it is still relatively low for high-hardness tracks and insufficient to cause damage. In this case, the second track geometric parameters of the foreign object track segment are indirectly calculated using the aforementioned method. Specifically, the second track geometric parameters of the foreign object track segment can be obtained by directly performing line fitting on the two adjacent first track geometric parameters.

[0082] However, if the value exceeds the saturation threshold, it indicates that the foreign object did not undergo significant deformation after being run over by the train wheels, suggesting that its hardness is sufficiently high, and the train wheels' crushing of it could damage the track. In this case, the above calculation method is not advisable. Instead, an ultrasonic detector should be activated to probe the track segment covered by the foreign object, and the second track geometry parameters of the track segment under the foreign object should be directly extracted using ultrasonic guided wave technology.

[0083] Since the scanning operation for detecting track geometric parameters based on ultrasonic guided wave technology needs to be performed one by one, the detection efficiency is relatively low. Therefore, this invention makes a decision based on the hardness and geometric fullness of the foreign object, and uses image recognition technology or ultrasonic guided wave technology to detect the track geometric parameters in the area where the foreign object is located. This can significantly improve the efficiency of track detection, and effectively reduce the probability of missing damaged track segments.

[0084] In some embodiments, predicting the first geometric structure of the foreign object under normal conditions based on the foreign object category and the foreign object structural features includes:

[0085] The foreign object category and the foreign object structural features are input into the geometric structure prediction model, and the geometric structure prediction model outputs its prediction of the first geometric structure of the foreign object under normal conditions; wherein, the geometric structure prediction model is obtained by fine-tuning the BERT large model.

[0086] In this embodiment of the invention, to improve the accuracy of predicting the first geometric structure of foreign objects under normal conditions, a geometric structure prediction model is constructed. This geometric structure prediction model can be a dedicated vertical model, such as a model built based on convolutional networks, Transformers, or other algorithms. However, the construction and training of such models require significant manpower and resources, resulting in high costs. Therefore, this invention preferably uses the existing, general-purpose BERT large model to construct the aforementioned geometric structure prediction model. That is, it uses collected small sample data to fine-tune and train the BERT large model, thereby obtaining a local model capable of analyzing and predicting the first geometric structure of foreign objects under normal conditions. Furthermore, knowledge distillation techniques can be used to further slim down this local model, reducing its size and thus improving the convenience of embedding the model into automated detection vehicles.

[0087] like Figure 3 As shown, this embodiment of the invention also discloses a novel track geometry parameter detection system applied to an automatic inspection vehicle. The automatic inspection vehicle includes a high-definition camera and a light sensor. The system includes a processing module and a storage module. The processing module implements the following method steps by calling and executing computer code in the storage module:

[0088] A light sensor detects the angle of sunlight in the current area, determines the shooting angle of the high-definition camera based on the angle of sunlight, and controls the high-definition camera to rotate to the shooting angle; wherein, different shooting angles correspond to different sizes of shooting field of view;

[0089] Track identification is performed on the high-definition image data captured by the high-definition camera at the shooting angle to obtain two track segments, and foreign object detection is performed on each track segment. Based on the detected foreign object features, the attribute category of the foreign object is predicted.

[0090] Based on the distribution location of the foreign object, the track segment is divided into a smooth track segment and a foreign object track segment. Image recognition technology is used to extract the first track geometric parameters of each smooth track segment, and the second track geometric parameters of the foreign object track segment are calculated based on the attribute category.

[0091] The first and second track geometry parameters are aligned and integrated to obtain the track geometry parameters.

[0092] This invention also discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method described in the foregoing embodiments.

[0093] This invention also discloses a computer storage medium storing a computer program that is executed by a processor to implement the methods described in the foregoing embodiments.

[0094] This invention also discloses a computer program product that, when run on a terminal, causes the terminal to execute the method described in the foregoing embodiments.

[0095] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0096] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0097] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0098] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention.

Claims

1. A novel track geometry parameter detection method, characterized in that The method is applied to an automated inspection vehicle, which includes a high-definition camera and a light sensor; the method includes the following steps: A light sensor detects the angle of sunlight in the current area, and determines the shooting angle of a high-definition camera based on the angle of sunlight. This includes: determining the orientation data of the track based on high-precision map data; calculating the reflection angle range of the sunlight reflected on the track at the specified angle based on the orientation data and the angle of sunlight; determining the shooting angle that does not coincide with the reflection angle range within the angle adjustment range of the high-definition camera, based on the three-dimensional dimensions of the automatic inspection vehicle and the three-dimensional deployment position of the high-definition camera on the automatic inspection vehicle, with the difference between the shooting angle and the current shooting angle of the high-definition camera being minimized; and controlling the high-definition camera to rotate to the shooting angle; wherein different shooting angles correspond to different sizes of shooting fields of view. Track identification is performed on the high-definition image data captured by the high-definition camera at the shooting angle to obtain two track segments, and foreign object detection is performed on each track segment. Based on the detected foreign object features, the attribute category of the foreign object is predicted. Based on the distribution location of the foreign object, the track segment is divided into a smooth track segment and a foreign object track segment. Image recognition technology is used to extract the first track geometric parameters of each smooth track segment, and the second track geometric parameters of the foreign object track segment are calculated based on the attribute category. The first and second track geometry parameters are aligned and integrated to obtain the track geometry parameters.

2. The novel track geometry parameter detection method according to claim 1, characterized in that: Foreign object detection is performed on each of the aforementioned track segments, and the attribute category of the foreign object is predicted based on the detected foreign object features, including: Extract the first appearance features of each track segment, perform a continuity analysis of the appearance features based on the first appearance features, and identify areas where the appearance features are discontinuous as foreign objects. The second appearance feature of the foreign object is extracted, which includes the appearance feature and the structural feature of the foreign object. Based on the appearance feature and the structural feature of the foreign object, the foreign object category is predicted. Based on the foreign object category and the structural feature of the foreign object, the geometric fullness and hardness of the foreign object are predicted. The foreign object category, the geometric fullness, and the hardness are used as the attribute categories of the foreign object.

3. The novel track geometry parameter detection method according to claim 2, characterized in that: Based on the foreign object category and the foreign object structural characteristics, the geometric fullness and hardness of the foreign object are predicted, including: Based on the foreign object category and the foreign object structural features, the first geometric structure of the foreign object under normal conditions is predicted, and the real-time second geometric structure of the foreign object is extracted from the foreign object structural features. The change range of the geometric structure of the foreign object is calculated based on the first geometric structure and the second geometric structure, and the geometric fullness is determined based on the change range. The dynamic characteristics of the foreign object are extracted from its structural features, the dynamic value of the foreign object is calculated based on the dynamic characteristics, and the hardness of the foreign object is determined based on the dynamic value.

4. The novel track geometry parameter detection method according to claim 3, characterized in that: The automated detection vehicle also includes an ultrasonic detector. The calculation of the second track geometry parameters of the foreign object track segment based on the attribute category includes: If the hardness is lower than the hardness threshold, the second track geometry parameters of the foreign object track segment are predicted based on the geometric parameters of the two first tracks adjacent to the upper and lower tracks of the foreign object track segment. If the hardness is higher than the hardness threshold, then: If the geometric fullness is lower than the fullness threshold, the second track geometric parameters of the foreign object track segment are predicted based on the geometric parameters of the two first tracks that are adjacent to each other above and below the foreign object track segment. If the geometric fullness is higher than the fullness threshold, the ultrasonic detector is used to detect the foreign object track segment in order to extract the second track geometric parameters.

5. The novel track geometry parameter detection method according to claim 4, characterized in that: Based on the foreign object category and the foreign object structural characteristics, the first geometric structure of the foreign object under normal conditions is predicted, including: The foreign object category and the foreign object structural features are input into the geometric structure prediction model, and the geometric structure prediction model outputs its prediction of the first geometric structure of the foreign object under normal conditions; wherein, the geometric structure prediction model is obtained by fine-tuning the BERT large model.

6. A novel track geometry parameter detection system, applied to an automatic inspection vehicle, the automatic inspection vehicle including a high-definition camera and a light sensor; the system including a processing module and a storage module; characterized in that: The processing module implements the following method steps by calling and executing the computer code in the storage module: A light sensor detects the angle of sunlight in the current area, and determines the shooting angle of a high-definition camera based on the angle of sunlight. This includes: determining the orientation data of the track based on high-precision map data; calculating the reflection angle range of the sunlight reflected on the track at the specified angle based on the orientation data and the angle of sunlight; determining the shooting angle that does not coincide with the reflection angle range within the angle adjustment range of the high-definition camera, based on the three-dimensional dimensions of the automatic inspection vehicle and the three-dimensional deployment position of the high-definition camera on the automatic inspection vehicle, with the difference between the shooting angle and the current shooting angle of the high-definition camera being minimized; and controlling the high-definition camera to rotate to the shooting angle; wherein different shooting angles correspond to different sizes of shooting fields of view. Track identification is performed on the high-definition image data captured by the high-definition camera at the shooting angle to obtain two track segments, and foreign object detection is performed on each track segment. Based on the detected foreign object features, the attribute category of the foreign object is predicted. Based on the distribution location of the foreign object, the track segment is divided into a smooth track segment and a foreign object track segment. Image recognition technology is used to extract the first track geometric parameters of each smooth track segment, and the second track geometric parameters of the foreign object track segment are calculated based on the attribute category. The first and second track geometry parameters are aligned and integrated to obtain the track geometry parameters.

7. An electronic device, comprising: At least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, characterized in that: the processor executes the computer program to implement the method as claimed in any one of claims 1-5.

8. A computer storage medium storing a computer program, characterized in that: The computer program is executed by a processor to implement the method as described in any one of claims 1-5.

9. A computer program product comprising a computer program stored on a non-transitory computer-readable medium, characterized in that: When the computer program is executed by a processor, it implements the method as described in any one of claims 1-5.

Images