Low-altitude inspection subway protection area intelligent detection method for curved track scene

Abstract

The invention discloses a curved track scene-oriented low-altitude inspection subway protection area intelligent detection method, which relates to the technical field of track traffic and unmanned aerial vehicles, and comprises a polygonal target detection module, a track curvature calculation and judgment module, a straight line fitting module and a curve fitting module, the polygon target detection module performs target detection on the track in a video stream mode; the track curvature calculation and judgment module is used for carrying out straight line and curve judgment on a track, the straight line fitting module and the curve fitting module are used for carrying out curve track fitting on a straight line track scene and a curved track scene respectively, and an optimized and improved scheme of fusing a polygon target detection algorithm and a Catmull-Rom curve fitting technology is adopted; the drawing precision of the subway protection area in the curved track scene is remarkably improved, and meanwhile the fitting effect of linear track design can be guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field, and in particular relates to an intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios. Background Art

[0002] The intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios aims to improve the safety and inspection efficiency of subway tracks. With the continuous development of subway construction, especially the expansion of urban rail transit networks, track inspection and maintenance have become particularly important. Traditional manual inspection methods are not only time-consuming and labor-intensive, but may also result in missed or false detections. Traditional track inspection methods mainly rely on manual inspections and vehicle inspections. Manual inspections require a lot of manpower and material resources and are easily restricted by weather, environmental factors, and other factors. While vehicle inspections can improve efficiency, they are less adaptable to special terrain and curved tracks and are prone to blind spots.

[0003] Curved sections of subway tracks present inspection challenges due to their small curve radius and complex track geometry. This is especially true during low-altitude inspections, where the complexity of the curved tracks can make it difficult for detection equipment to accurately collect information. Traditional inspection methods often struggle to adapt to this complex terrain and are unable to provide comprehensive, comprehensive inspections. Incorporating artificial intelligence technology, low-altitude inspection equipment not only collects data but also analyzes and evaluates track conditions through methods such as image recognition and machine learning. This intelligent detection method automatically identifies potential track issues such as cracks, deformation, and corrosion, significantly improving inspection accuracy and efficiency. Furthermore, the intelligent detection system can automatically adjust inspection strategies and routes based on historical data and environmental changes, further enhancing inspection quality.

[0004] The protection zone of the subway track is an area established to ensure the safety of train operation and avoid the impact of external interference and dangerous objects on the track. For the inspection of these protection zones, it is necessary not only to monitor the status of the track, but also to monitor its surrounding environment, including vegetation, buildings, facilities and other factors that may affect the safety of the track. Therefore, the intelligent detection method of the protection zone requires the equipment to have environmental perception and comprehensive analysis capabilities, and to be able to perform efficient inspections in complex environments. In summary, the proposal of the intelligent detection method of the subway protection zone for low-altitude inspection of curved track scenes combines drone inspection technology and intelligent analysis technology, which can overcome the shortcomings of traditional inspection methods, improve inspection efficiency and accuracy, and ensure the safe operation of subway tracks and protection zones. Based on this, the present invention designs an intelligent detection method of the subway protection zone for low-altitude inspection of curved track scenes. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems in the prior art and propose an intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios, comprising:

[0006] A polygonal target detection module, which performs target detection on the track using a video stream;

[0007] Track curvature calculation and judgment module, which determines whether the track is straight or curved;

[0008] The straight line fitting module and the curve fitting module perform curve trajectory fitting for straight line and curved track scenarios respectively;

[0009] And include the following detection steps:

[0010] S1: The drone conducts low-altitude inspections along a preset route, collects real-time video data along the route, uses a target detection algorithm to identify track feature information, extracts track centerline coordinates, and establishes spatial positioning of the track reference position. Then, through the polygon target detection module, it outputs the track polygon area boundary information, providing basic data for subsequent curvature analysis.

[0011] S2: Analyze and identify the geometric features of the track boundary based on basic data to accurately distinguish between straight track and curved track scenarios;

[0012] S3: Perform trajectory fitting on the straight track and the curved track using the straight line fitting module and the curve fitting module respectively;

[0013] S4: performing the curve fitting process of steps S1 to S4 multiple times and performing overlap detection on the multiple fitted curves. When the overlap of the multiple fitted curves is insufficient, the number of coordinates extracted in step S1 is increased and the detection is performed again.

[0014] S5: Expand the fitted curve to both sides at a standard distance to generate the boundary outline of the subway protection zone, and draw and display the boundary outline of the protection zone in real time.

[0015] The beneficial effects of the present invention are as follows: the optimization and improvement scheme integrating polygonal target detection algorithm and curve fitting technology significantly improves the drawing accuracy of subway protection zone in curved track scenario, while ensuring the fitting effect of straight track design, ensuring the continuity and smoothness of the curve after trajectory fitting, having good local controllability and high computational efficiency, being able to accurately reflect the geometric characteristics of the track, and having a certain robustness to noise data. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is a flowchart of an intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios.

[0017] Figure 2 This is a working principle diagram of the track curvature calculation and discrimination module in the intelligent detection method of the subway protection zone for low-altitude inspection of curved track scenarios. DETAILED DESCRIPTION

[0018] An intelligent detection method for low-altitude inspection of subway protection areas for curved track scenarios, referring to Figure 1 The dotted box shows the improvements of this patent compared to the current UAV low-altitude inspection of subway protection zone solutions, which includes four parts: polygon target detection module, track curvature calculation and judgment module, straight line fitting module and curve fitting module.

[0019] 1) Polygonal target detection module

[0020] (1) Module function

[0021] Based on deep learning, real-time detection and positioning of track areas in video streams are achieved, and the polygonal area boundary information of the track is output to provide basic data for subsequent curvature analysis.

[0022] (2) Module working principle and process

[0023] Using the YOLOv5 object detection model, we trained the model to accurately identify track areas within video frames by performing polygonal annotation of track areas. The model output includes the coordinates of key points along the track boundary, which are used to construct a complete polygonal outline.

[0024] 2) Track curvature calculation and judgment module

[0025] (1) Module function

[0026] Responsible for intelligent identification of track types. By analyzing the geometric characteristics of the track boundary, it can accurately distinguish between straight track and curved track scenarios, providing a decision-making basis for subsequent trajectory fitting strategy selection.

[0027] Module working principle and process

[0028] The working principle and process of the module are as follows Figure 2 As shown in the figure, the boundaries of both sides of the detected track polygon area are extracted as follows:

[0029] Extract the boundaries of the detected track polygon area on both sides. The calculation formulas for the slopes k1 and k2 of the boundaries on both sides are as follows:

[0030]

[0031] Among them, (x1, y1) and (x2, y2) are the coordinates of the two endpoints of the track boundary on one side;

[0032] (x3, y3), (x4, y4) are the coordinates of the two endpoints of the track boundary on the other side.

[0033] When the absolute values ​​of the slopes on both sides are equal and have opposite signs (i.e., k1 + k2 = 0), the track is determined to be a straight track; otherwise, it is determined to be a curved track. This geometric feature-based identification method has the characteristics of high computational efficiency and high identification accuracy.

[0034] 3) Linear fitting module

[0035] (1) Module function

[0036] For straight track scenarios, an efficient trajectory fitting solution is provided to ensure accurate expression and fast processing of straight line segments.

[0037] (2) Module working principle and process

[0038] This method has a small computational load and is suitable for real-time processing applications.

[0039] Select several points on the boundary (x1, y1), (x2, y2)…(x n ,y n ), a linear fitting algorithm is used to fit the track boundary points by the least squares method to achieve an accurate mathematical description of the linear track. The slope k and intercept b of the fitted line are as follows:

[0040] k=(nΣxy-ΣxΣy) / (nΣx 2 -(Σx) 2 )

[0041] b=(Σy-kΣx) / n

[0042] 4) Curve fitting module

[0043] (1) Module function

[0044] Specifically for curved track scenarios, it provides high-precision curve trajectory fitting solutions to ensure accurate reconstruction of complex track shapes.

[0045] (2) Module working principle and process

[0046] Based on the Catmull-Rom spline algorithm, key control points are first selected from the track boundary point sequence, and auxiliary control points are added at the beginning and end of the sequence to ensure the continuity of the entire curve. The added auxiliary points at the beginning and end are as follows:

[0047] P -1 =P0-(P1-P0)

[0048] P n+1 =P n -(P n -Pn-1 )

[0049] The uniform parameterization method is then used to assign parameter values ​​to each control point, and the tangent vector is calculated based on the control point position. The tangent vector and normal vector at each control point are calculated as follows, and a local coordinate system is constructed to determine the direction and shape of the curve segment:

[0050] Ti=α(P i+1 -P i-1 )

[0051] Ni=[-Ty,i,Tx,i]

[0052] During the curve interpolation phase, the Catmull-Rom interpolation formula is used to calculate the points on the curve. For each control point pair (Pi, Pi+1), the interpolation segment is calculated using the points before and after it (Pi-1, Pi+2), ensuring that the curve has C1 continuity and passes through all control points. A tension parameter α (usually set to 0.5) is also introduced to control the tightness of the curve, optimizing the curve shape to better fit the actual track profile.

[0053]

[0054] Where M is the Catmull-Rom basis matrix,

[0055] M=α*[-1,3,-3,1]

[0056] [2,-5,4,-1]

[0057] [-1,0,1,0]

[0058] [0,2,0,0]

[0059] At the junction of control points, the continuity of the tangent vector is guaranteed:

[0060] lim(t→1 - )P′(t)=lim(t→0 + )P′(t)

[0061] During the real-time inspection process, the algorithm will evaluate the fitting error in real time, dynamically adjust the control point density as needed, increase sampling points in key areas to improve accuracy, and smooth outliers to improve fitting stability.

[0062] This algorithm not only ensures the continuity and smoothness of the curve, but also has good local controllability and high computational efficiency. It can accurately reflect the geometric characteristics of the track and is robust to noisy data. It is very suitable for real-time track curve fitting applications.

[0063] When used, this method comprises the following steps:

[0064] S1: The drone conducts low-altitude inspections along a preset route, collects real-time video data along the route, uses a target detection algorithm to identify track feature information, extracts track centerline coordinates, and establishes spatial positioning of the track reference position. Then, through the polygon target detection module, it outputs the track polygon area boundary information, providing basic data for subsequent curvature analysis.

[0065] S2: Analyze and identify the geometric features of the track boundary based on basic data to accurately distinguish between straight track and curved track scenarios;

[0066] S3: Perform trajectory fitting on the straight track and the curved track using the straight line fitting module and the curve fitting module respectively;

[0067] S4: performing the curve fitting process of steps S1 to S4 multiple times and performing overlap detection on the multiple fitted curves. When the overlap of the multiple fitted curves is insufficient, the number of coordinates extracted in step S1 is increased and the detection is performed again.

[0068] S5: Expand the fitted curve to both sides at a standard distance to generate the boundary outline of the subway protection zone, and draw and display the boundary outline of the protection zone in real time.

[0069] It is understood from common technical knowledge that the present invention may be implemented by other embodiments that do not depart from its spirit or essential features. Therefore, the embodiments disclosed above are, in all respects, merely illustrative and not exclusive. All modifications within the scope of the present invention or equivalent to the scope of the present invention are intended to be encompassed by the present invention.

Claims

1. An intelligent detection method for low-altitude inspection of subway protection areas for curved track scenarios, characterized by: include: A polygonal target detection module, which performs target detection on the track using a video stream; Track curvature calculation and judgment module, which determines whether the track is straight or curved; The straight line fitting module and the curve fitting module perform curve trajectory fitting for straight line and curved track scenarios respectively; And include the following detection steps: S1: The drone conducts low-altitude inspections along a preset route, collects real-time video data along the route, uses a target detection algorithm to identify track feature information, extracts track centerline coordinates, and establishes spatial positioning of the track reference position. Then, through the polygon target detection module, it outputs the track polygon area boundary information, providing basic data for subsequent curvature analysis. S2: Analyze and identify the geometric features of the track boundary based on basic data to accurately distinguish between straight track and curved track scenarios; S3: Perform trajectory fitting on the straight track and the curved track using the straight line fitting module and the curve fitting module respectively; S4: performing the curve fitting process of steps S1 to S4 multiple times and performing overlap detection on the multiple fitted curves. When the overlap of the multiple fitted curves is insufficient, the number of coordinates extracted in step S1 is increased and the detection is performed again. S5: Expand the fitted curve to both sides at a standard distance to generate the boundary outline of the subway protection zone, and draw and display the boundary outline of the protection zone in real time.

2. The intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios according to claim 1 is characterized in that: The polygonal target detection module adopts the YOLOv5 target detection model. By training the track area with polygonal annotation, the model can accurately identify the track area in the video frame. The model outputs the coordinates of the key points containing the track boundary and the polygonal area boundary information of the track.

3. The intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios according to claim 1 is characterized in that: The track curvature calculation and discrimination module is responsible for discriminating the track type and accurately distinguishing between straight track and curved track scenarios by analyzing the geometric features of the track boundary.

4. The intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios according to claim 3 is characterized in that: In the track curvature calculation and judgment module, the method is as follows: Extract the boundaries of the detected track polygon area on both sides. The calculation formulas for the slopes k1 and k2 of the boundaries on both sides are as follows: Among them, (x1, y1) and (x2, y2) are the coordinates of the two endpoints of the track boundary on one side; (x3, y3), (x4, y4) are the coordinates of the two endpoints of the track boundary on the other side; When the absolute values ​​of the slopes on both sides are equal and have opposite signs, that is, k1+k2=0, it is determined to be a straight track; otherwise, it is determined to be a curved track.

5. The intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios according to claim 1 is characterized in that: In the straight line fitting module, the specific method is as follows: Select several points on the boundary (x1, y1), (x2, y2)…(x n ,y n ), a linear fitting algorithm is used to fit the track boundary points to a straight line through the least squares method to achieve an accurate mathematical description of the straight track. The slope k and intercept b of the fitted line are calculated as follows: k=(nΣxy-ΣxΣy) / (nΣx 2 -(Sx) 2 ) b=(Σy-kΣx) / n 6. The intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios according to claim 1 is characterized in that: The curve fitting module is based on the Catmull-Rom spline algorithm, and the specific method includes the following steps: S1: Select key control points from the track boundary point sequence and add auxiliary control points at the beginning and end of the sequence to ensure the continuity of the entire curve. The added auxiliary points at the beginning and end are as follows: P -1 =P0-(P1-P0) P n+1 =P n -(P n -P n-1 ) S2: A uniform parameterization method is used to assign parameter values ​​to each control point, and the tangent vector is calculated based on the control point position. The tangent vector and normal vector at each control point are calculated as follows, and a local coordinate system is constructed to determine the direction and shape of the curve segment. Ti=α(P i+1 -P i-1 ) Ni=[-Ty,i,Tx,i] S3: In the curve interpolation stage, the points on the curve are calculated by the Catmull-Rom interpolation formula. For each control point pair (P i ,P i+1 ), using its front and back points (P i-1 ,P i+2 ) calculates the interpolation segment to ensure that the curve has C1 continuity and passes through all control points. At the same time, a tension parameter α (usually 0.5) is introduced to control the tightness of the curve and optimize the curve shape to make it better fit the actual track profile. Where M is the Catmull-Rom basis matrix, M＝α*[-1,3,-3,1] [2,-5,4,-1] [-1,0,1,0] [0,2,0,0] S4: Ensure the continuity of the tangent vector at the connection of the control points: lim(t→1 - )P ′ (t)=lim(t→0 + )P ′ (t) 7. The intelligent detection method for low-altitude inspection of subway protection zones for curved track scenarios according to claim 2 is characterized in that: In the curve fitting module, the control point density is dynamically adjusted, sampling points are added in key areas to improve accuracy, and outliers are smoothed to improve fitting stability.

Images