A method and system for detecting blind seams of rails based on machine vision

Abstract

The present invention relates to the field of railway track detection technology, and specifically provides a method and system for detecting blind seams in rails based on machine vision. The method includes the steps of rail pair image acquisition, edge trajectory extraction, gap feature quantification, detection method selection and gap width analysis. Through multi-dimensional image feature analysis combined with an intelligent selection strategy, the optimal detection method is quickly matched, and efficient and accurate gap width assessment is achieved in a complex environment. The system includes a rail pair image acquisition module, an edge trajectory extraction module, a gap feature quantification module, a detection method selection module and a gap width analysis module. The present invention solves the problems of low detection efficiency, insufficient precision and poor environmental adaptability in the prior art, and significantly improves the accuracy and timeliness of gap width detection at track joints.

Description

Technical Field

[0001] The present invention relates to the technical field of railway track detection, and more particularly to a method and system for detecting blind seams in rails based on machine vision. Background Art

[0002] In the daily maintenance and safety assurance of the track system, detecting the gap width at the track joint is a crucial task. The gap width at the track joint must be strictly controlled within the specified range. This plays a decisive role in ensuring the smoothness and safety of train operation and the long-term reliability of the track system. If the gap is too wide, the train will produce severe bumps and vibrations during operation, which will not only reduce passenger comfort, but also cause additional wear and tear on vehicle components, and increase the risk of safety accidents such as derailment. If the gap is too narrow, the track may be deformed and squeezed due to thermal expansion and contraction.

[0003] Currently, a variety of technical methods have been developed for detecting gap widths at track joints. Common methods include line-to-line detection, point-to-surface detection, and three-dimensional feature-thermal imaging detection. The line-to-line detection method uses a specific sensor to obtain the position information of two lines at the track joint and then calculates the gap width. This method is relatively simple to operate, but is easily affected by factors such as uneven track surfaces and stains, resulting in poor detection accuracy. Point-to-surface detection uses a combination of line lasers and cameras to obtain point-to-surface data at the track joint and then calculates the gap width using a complex algorithm. This method requires high equipment installation precision and the data processing process is time-consuming, making it difficult to meet the needs of rapid detection. Three-dimensional feature-thermal imaging detection uses thermal imaging technology to obtain temperature distribution images of the track joint and infer the gap width by analyzing the three-dimensional characteristics reflected by the temperature differences. This method can detect hidden defects to a certain extent, but the equipment is expensive, the detection efficiency is low, and it is sensitive to changes in ambient temperature.

[0004] When faced with an image of a track joint, how to quickly and accurately select the appropriate method from the above-mentioned detection methods to determine the gap width has become a key issue that needs to be addressed. Existing technologies lack systematic guidance in this regard, and inspectors often make choices based on experience. This not only easily leads to inappropriate detection methods and affects the accuracy and timeliness of the detection results, but may also waste a lot of time and resources due to repeated attempts at different methods. Therefore, there is an urgent need for an intelligent detection method selection method based on image features that can quickly and accurately match the most suitable detection technology according to the characteristics of the track joint image, thereby achieving efficient and accurate detection of the gap width at the track joint. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a method and system for detecting blind seams in rails based on machine vision. This blind seam detection solves the problems of low efficiency, insufficient accuracy and poor environmental adaptability in the detection of gap width at track joints in the existing technology through multi-dimensional image feature analysis combined with intelligent selection strategies. This method and system can quickly match the optimal detection method according to the specific features of the image to be analyzed, and realize efficient and accurate gap width assessment in complex environments.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for detecting blind rail seams based on machine vision comprises the following steps:

[0008] The rail pair image acquisition step uses a visual camera installed on the bottom of the train to capture images of the rail joint as the image to be analyzed, and simultaneously records the surface temperature of the rail joint end;

[0009] an edge track extraction step, outlining the edge track of the gap at the rail joint end in the image to be analyzed using an edge detection algorithm, and determining the clarity of the edge track based on the edge point distribution density;

[0010] a gap feature quantification step, calculating the grayscale difference or normalized contrast between the gap area and the background area in the image to be analyzed as the gap contrast, analyzing the cleanliness of the gap surface, and generating a multidimensional feature vector by combining light intensity, surface temperature, and gap contrast;

[0011] a detection method selection step, inputting the multidimensional feature vector into a preset analysis method selection formula and outputting a judgment index, wherein the judgment index is used to select the optimal detection method among the line detection method, the point and surface detection method, and the three-dimensional feature-thermal imaging detection method;

[0012] The gap width analysis step performs the gap width analysis task according to the optimal detection method determined in the detection method selection step.

[0013] Furthermore, the edge trajectory extraction step includes an edge point screening strategy, which includes an edge point density calculation step and an edge point continuity judgment step.

[0014] The edge point density calculation step is to count the number of edge points of the gap edge track in the image to be analyzed in units of pixels, and calculate the edge point density within a unit length;

[0015] The edge point continuity judgment step judges the continuity of the edge track according to a distance threshold between edge points. If the continuity is lower than a preset threshold, the edge track is marked as a fuzzy track.

[0016] Furthermore, the gap feature quantification step also includes a surface cleanliness assessment strategy.

[0017] The surface cleanliness assessment strategy analyzes the texture features and brightness distribution of the gap area in the image to be analyzed, calculates the coverage rate of stains accumulated on the gap surface, and generates a surface cleanliness score by combining the stain coverage rate and the gap contrast.

[0018] Furthermore, the method includes a real-time monitoring step, which starts a real-time monitoring timer. If the gap width analysis task is not completed within the maximum allowed processing time, another detection method is switched to re-execute the analysis task.

[0019] Furthermore, the real-time monitoring step includes a method switching strategy, which includes a time threshold setting step and a switching analysis method step.

[0020] The time threshold setting step analyzes the maximum processing time of each detection method based on historical data and sets the maximum allowable processing time;

[0021] The step of switching the analysis method selects the next detection method according to the proximity between the judgment value and the detection method value range.

[0022] Furthermore, the gap width analysis step includes an analysis result verification strategy, which includes a repeated measurement step and an error correction step.

[0023] Repeat the measurement step to measure the gap width of the same rail joint multiple times, record the results of each measurement and calculate the average value;

[0024] The error correction step determines the stability of the measurement based on the standard deviation of multiple measurement results. If the standard deviation exceeds a preset threshold, the measurement task is re-executed or the acquisition angle of the visual camera is adjusted.

[0025] Furthermore, the detection method selection step includes an extreme condition triggering strategy, which includes directly triggering an ignore instruction when the gap area is blocked by stains and the color is the same as the rust in the background area of ​​the surface of the rail docking end, and using a three-dimensional feature-thermal imaging detection method for analysis.

[0026] Furthermore, the gap width analysis step analyzes two or more consecutive images to be analyzed to determine the change in the gap width, and compares it with a preset change threshold, and outputs a normal instruction or an abnormal instruction based on the comparison result.

[0027] Furthermore, the edge track extraction step includes an image preprocessing strategy, which includes intercepting the track joint area in the image to be analyzed through a target detection algorithm, and retaining the track section A and the track section B on both sides of the gap.

[0028] A rail blind seam detection system based on machine vision, comprising:

[0029] The rail pair image acquisition module uses a visual camera installed on the bottom of the train to capture images of the rail joints as the images to be analyzed, and simultaneously records the surface temperature of the rail joints;

[0030] An edge track extraction module, which uses an edge detection algorithm to outline the edge track of the gap at the rail joint end in the image to be analyzed, and determines the clarity of the edge track based on the distribution density of edge points;

[0031] A gap feature quantification module calculates the grayscale difference or normalized contrast between the gap area and the background area in the image to be analyzed as the gap contrast, analyzes the cleanliness of the gap surface, and generates a multidimensional feature vector based on the light intensity, surface temperature, and gap contrast;

[0032] a detection method selection module that inputs the multidimensional feature vector into a preset analysis method selection formula and outputs a judgment index for selecting the optimal detection method among the line detection method, the point and surface detection method, and the three-dimensional feature-thermal imaging detection method;

[0033] The gap width analysis module performs the gap width analysis task according to the optimal detection method determined in the detection method selection step.

[0034] Beneficial effects of the present invention: The present invention solves the problems of low efficiency, insufficient accuracy and poor environmental adaptability in the detection of gap width at the track joint in the prior art through multi-dimensional image feature analysis combined with an intelligent selection strategy. That is, it can quickly match the optimal detection method according to the specific features of the image to be analyzed, and realize efficient and accurate gap width assessment in a complex environment. Specifically, the edge detection algorithm is combined with the edge point screening strategy, and the gap edge trajectory is accurately outlined through density calculation and continuity judgment, and the fuzzy trajectory is effectively identified, thereby improving the accuracy of edge extraction. When quantifying the gap features, not only the grayscale difference and contrast are considered, but also the surface cleanliness assessment is introduced. The stain coverage is analyzed by texture and brightness, and a multi-dimensional feature vector is generated by combining multiple parameters, laying the foundation for the intelligent selection of the detection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is the overall flow chart of the present invention;

[0036] Figure 2It is a top view captured vertically downward by the visual camera in the present invention;

[0037] Figure 3 is the axonometric view captured by the visual camera in the present invention;

[0038] Figure 4 It is the system fluid diagram of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be described in further detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom," "top," "inner," and "outer" refer to directions toward or away from the geometric center of a particular component, respectively.

[0040] When faced with an image of a track joint, how to quickly and accurately select a suitable method from the above-mentioned detection methods to determine the gap width has become a key issue that needs to be resolved. The existing technology lacks systematic guidance in this regard, and detection personnel often make choices based on experience, which not only easily leads to improper detection methods and affects the accuracy and timeliness of the detection results, but also may waste a lot of time and resources due to repeated attempts at different methods. Therefore, there is an urgent need for an intelligent detection method selection method based on image features, which can quickly and accurately match the most suitable detection technology according to the characteristics of the track joint image, thereby realizing efficient and accurate detection of the gap width at the track joint. Therefore, the present invention designs this rail blind seam detection method and system based on machine vision, combined with the attached Figure 1-3 As shown, the following steps are included:

[0041] In the step of acquiring the rail pair image, a visual camera installed at the bottom of the train is used to capture the image of the rail joint as the image to be analyzed, and a temperature sensor or thermal imaging is used to synchronously record the surface temperature of the rail joint end. In the initial state, Figure 2 As shown, the installation position of the vision camera must ensure that its lens optical axis is perpendicular to the track docking surface. The vision camera can be adjusted. Its pitch angle 4 and yaw angle 5 are adjustable within the range of ±5° and ±3° respectively. They are precisely controlled by a stepper motor drive to reduce image distortion caused by shooting angle deviation. The temperature sensor or thermal imaging and the vision camera are fixed on the same bracket, and the distance between the two does not exceed 10 cm to ensure the time synchronization of surface temperature data and image acquisition;

[0042] In the edge trajectory extraction step, a deep learning-based convolutional neural network model is used to roughly locate the track junction area. After generating candidate frames, the optimal candidate frame is selected using a non-maximum suppression algorithm. Subsequently, the image within the candidate frame is finely segmented to retain the complete structural information of track segments A and B. During this process, the light intensity in the image to be analyzed is calculated as an important reference parameter for subsequent analysis. The calculation of light intensity is based on the overall grayscale value distribution of the image and is obtained by statistically analyzing the average and standard deviation of the pixel brightness values. This process provides high-quality basic data for subsequent edge trajectory extraction and feature parameter extraction.

[0043] In the image to be analyzed, the edge detection algorithm (Canny operator) is used to outline the gap edge trajectory of the track docking end, and the clarity of the edge trajectory is judged according to the edge point distribution density. Specifically, the edge trajectory extraction process includes two steps: edge point density calculation and continuity judgment. In the edge point density calculation step, the number of edge points on the gap edge trajectory is counted in pixels, and the edge point density per unit length is calculated. In the continuity judgment step, the continuity of the edge trajectory is judged according to the distance threshold between edge points. In the subjective judgment stage, if the continuity is lower than the preset threshold, the edge trajectory is marked as a fuzzy trajectory. Among them, the edge trajectory continuity is based on the edge point distribution density and the distance threshold. The formula configuration is:

[0044] ,

[0045] in, is the edge trajectory continuity, is the number of edge points (the number of countable pixels obtained from the edge detection algorithm), is the average distance between points (calculated as the total trajectory length divided by -1), is the standard deviation of the distance between points (measure of distance dispersion), is the distance between the i-th points (Euclidean distance between adjacent edge points), is the distance threshold (preset parameter, used to determine continuity interruption), is the density scaling factor (a preset constant that controls the contribution of density to continuity), is the penalty coefficient (a preset constant that amplifies the effect of discontinuity), is the Heaviside step function (defined as , filtering out distances exceeding a threshold), using Normalized coefficient of variation ( reflects the density uniformity), the exponential term penalizes points whose distance exceeds the threshold ( Sum the number of discontinuous points), range , close to 1 indicates high continuity (clear edges), and close to 0 indicates low continuity (fuzzy or interrupted edges).

[0046] In the gap feature quantification step, the grayscale difference or normalized contrast between the gap area and the background area in the image to be analyzed is calculated as the gap contrast. The gap contrast formula is configured as follows:

[0047] ,

[0048] in, is the gap contrast, the range , close to 1 means high contrast,(clear gap), close to 0 means low contrast (blurred gap), is the average grayscale of the gap area (calculated from the image pixel grayscale value), is the average grayscale of the background area (the average grayscale of the non-gap area), is the grayscale standard deviation of the gap area (measure of grayscale fluctuation), is the grayscale standard deviation of the background area, is the contrast gain parameter (preset constant), is the area of ​​the gap region (number of pixels), is the integration domain of the gap region (image coordinate range), Pixel The gray value of is the hyperbolic tangent function (normalize the contrast to [-1, 1] and take the absolute value to [0, 1]), is the double integral (calculating the regional mean absolute deviation).

[0049] Analyze the surface cleanliness of the gap. Specifically, the surface cleanliness assessment strategy is included. By analyzing the texture characteristics and brightness distribution of the gap area in the image to be analyzed, the coverage rate of the accumulated stains on the gap surface is calculated. The surface cleanliness score is generated by combining the stain coverage rate and the gap contrast. The surface cleanliness scoring algorithm is configured as follows:

[0050] ,

[0051] in, Score the cleanliness of the gap surface, the value range , close to 1 means high cleanliness, close to 0 means low cleanliness, is the stain coverage (stain pixel area ratio, calculated from image segmentation), is the texture entropy (calculated based on the gray-level co-occurrence matrix, , which measures texture complexity, is the stain sensitivity coefficient (preset constant), is the texture scaling factor (a preset constant), is the contrast coupling parameter (preset constant), is the gap contrast, is an exponential function (suppressing the influence of stains), is the Gaussian error function.

[0052] Generate a multi-dimensional feature vector based on edge track clarity, crack surface cleanliness, light intensity, surface temperature and crack contrast;

[0053] The detection method selection step inputs the multidimensional feature vector into a preset analysis method selection formula and outputs a judgment index, wherein the analysis method selection formula is configured as:

[0054] ,

[0055] in, To judge the index, is the edge trajectory continuity, is the gap contrast, Score the cleanliness of the crack surface. is the light intensity, is the surface temperature, is the feature weight parameter (preset constant), is the integral offset parameter (preset constant), is the integration variable (imaginary variable), is the sigmoid function (defined as , normalized to [0,1]), is the Gaussian integral (introducing noise robustness and filtering outliers).

[0056] This judgment index is used to select the optimal detection method among the line-line detection method, point-surface detection method, and three-dimensional feature-thermal imaging detection method. Line-line detection obtains the position information of two lines at the track joint and calculates the distance between the two lines as the gap width. It is suitable for scenarios where the track surface is flat and there are no obvious stains. It is easily affected by factors such as uneven track surface, stains, and uneven lighting, resulting in reduced detection accuracy. For example, when there is rust or oil on the track surface, the line position may not be accurately captured, resulting in measurement errors; the point-surface detection method is to project a line onto the track joint to form a light strip, the camera collects the point-surface data of the light strip, and then calculates the gap width through a complex algorithm. The detection accuracy is high, but the installation position accuracy of sensors such as cameras is high, and the data processing process is time-consuming; the three-dimensional feature-thermal imaging detection method uses thermal imaging technology to obtain the temperature distribution image of the track joint, combined with a three-dimensional reconstruction algorithm, to analyze the three-dimensional structural features corresponding to the temperature difference, infer the gap width, and can penetrate some stains or rust. The detection accuracy is high, but it is sensitive to ambient temperature and is suitable for extreme conditions (such as the gap is covered by stains, the color is confused with the background) or scenes requiring deep structural analysis (triggering extreme condition strategies).

[0057] The data examples are as follows:

[0058] 1. Edge trajectory continuity calculation

[0059] Number of edge points =100 (edge ​​points), average distance between points =2.0 (pixel), standard deviation of distance between points =0.5, sequence( =1.8, =2.1,...), distance threshold =2.5, density scaling factor =1.0, penalty coefficient =0.1;

[0060] calculate: (Assuming 10 points exceed the threshold, ),but ,Right now .

[0061] 2. Calculation of gap contrast

[0062] Average grayscale of the gap area =50, average grayscale of background area =100, grayscale standard deviation of the gap area =10, background area grayscale standard deviation =15, contrast gain parameter =0.5, gap area =200 (pixels), pixel points Grayscale value Example: The grayscale uniformity of the gap area is 50, then ;

[0063] calculate: , integral item = 10000 / 200 = 50, then , (after normalization , assuming the maximum grayscale difference is 100).

[0064] 3. Calculation of surface cleanliness score

[0065] Stain coverage =0.2 (20% stain), texture entropy =1.5, stain sensitivity coefficient =1.0, texture scaling factor =1.0, contrast coupling parameter =0.5, gap contrast =0.5,

[0066] calculate: , , .

[0067] 4. Calculation and selection of analysis methods

[0068] Edge trajectory continuity , gap contrast , gap surface cleanliness score , light intensity =80 (lux), surface temperature =30℃, ;

[0069] Calculation: Linear part = , the integral is approximately Gaussian expectation (numerical integration) of (Because the linear part is large, sigmoid is saturated), and the initial rule is: Indicates that the line detection method is optimal (high edge definition, low temperature), Indicates that the line and surface detection method is optimal (medium parameters), Indicates that the three-dimensional feature-thermal imaging detection method is optimal (low edge definition, high temperature or low cleanliness), the results of the embodiment are , indicating that the 3D feature-thermal image detection method is optimal (due to low continuity and high temperature =30℃).

[0070] The gap width analysis step performs the gap width analysis task according to the optimal detection method determined in the detection method selection step.

[0071] It also includes a real-time monitoring step. When a certain detection method is started, a real-time monitoring timer is started. If the gap width analysis task is not completed within the maximum allowed processing time, another detection method is switched to re-execute the analysis task. The switching basis is determined by the method switching strategy. The method switching strategy includes a time threshold setting step and a switching analysis method step. In order to make the judgment value closer to the value range corresponding to the detection method, the specific selection rules are as follows: if the judgment value is close to the value range of the line-line detection method, it is prioritized to switch to line-line detection; if the judgment value is close to the value range of the point-surface detection method, it is prioritized to switch to point-surface detection; if the judgment value is close to the value range of the three-dimensional feature-thermal imaging detection method, it is prioritized to switch to three-dimensional feature-thermal imaging detection; the timer is tightly coupled with the execution logic of the detection method to ensure the timeliness and accuracy of the switching process.

[0072] The gap width analysis step includes an analysis result verification strategy, which includes repeated measurement steps and error correction steps to ensure the accuracy and stability of the gap width detection results.

[0073] Repeat the measurement steps and measure the gap width of the same track joint more than twice and independently. For example, use the visual cameras at the bottom of the second and third carriages to capture images of the same track joint, record the results of each measurement, and calculate the average value.

[0074] Error correction step: determine the stability of the measurement based on the standard deviation of multiple measurement results. The standard deviation calculation formula of the measurement results is: ,in is the standard deviation, is the number of measurements, is the i-th measurement result, The mean of the measurement results is taken. If the standard deviation exceeds the preset threshold, the measurement task is re-executed, that is, the same position is detected again to eliminate accidental interference (such as sudden changes in illumination and image blur caused by vibration) or the acquisition angle of the visual camera is adjusted, from the original vertical acquisition of an upper surface to the oblique acquisition of the upper surface and side surfaces, such as a pitch angle of ±5° and a yaw angle of ±3°, to obtain the side structure information of the gap and avoid feature occlusion caused by a single shooting angle. Figure 3 The image shown is collected from the side, and the line detection method should be selected for gap width detection.

[0075] The detection method selection step includes an extreme condition triggering strategy. The extreme condition triggering strategy includes ignoring the analysis method selection formula when the gap is blocked by stains and the background area on the surface of the rail joint is rusty, and directly calling the three-dimensional feature-thermal imaging detection method for analysis. The judgment that the gap is blocked by stains is confirmed by color histogram analysis to confirm the color consistency of the background area and the gap area. If the overlap rate of the color histograms of the two exceeds 90%, the three-dimensional feature-thermal imaging detection method is triggered. The three-dimensional feature-thermal imaging detection method combines thermal imaging technology with a three-dimensional reconstruction algorithm to achieve accurate assessment of the gap width in complex environments.

[0076] The gap width analysis step analyzes two or more consecutive images to be analyzed to determine the change in gap width, and compares it with the preset change threshold. Based on the comparison result, a normal instruction or an abnormal instruction is output. Dynamic gap abnormalities, such as instantaneous expansion of gaps caused by loose rail fasteners, can be discovered in advance by several hours to several days compared to traditional static detection (one inspection per day).

[0077] According to the rail blind seam detection method based on machine vision, a detection system is designed. Figure 4 As shown, the system includes an image acquisition module installed on the underside of the train near the track junction. It captures the image to be analyzed at the track junction and simultaneously records the surface temperature of the track junction. An edge trajectory extraction module, connected to the image acquisition module, receives the image to be analyzed and uses an edge detection algorithm to extract the edge trajectory of the gap at the track junction. The edge trajectory clarity is determined based on the density of edge point distribution. A gap feature quantification module, connected to the edge trajectory extraction module, calculates the grayscale difference or normalized contrast between the gap and background areas as the gap contrast. It also combines light intensity, surface temperature, and gap contrast to generate a multidimensional feature vector. A detection method selection module receives the multidimensional feature vector from the gap feature quantification module, outputs a judgment index using a preset analysis method selection formula, and selects the optimal detection method based on the judgment index. A gap width analysis module, connected to the detection method selection module, executes the selected detection method to complete the gap width analysis task and starts a real-time monitoring timer. A real-time monitoring module, connected to the gap width analysis module, monitors the execution time of the analysis task and, if a timeout occurs, switches to another detection method and re-executes the analysis task.

[0078] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that improvements and modifications that do not depart from the principles of the present invention are within the scope of protection of the present invention.

Claims

1. A method for detecting blind rail seams based on machine vision, characterized in that: The steps include: The rail pair image acquisition step uses a visual camera installed on the bottom of the train to capture images of the rail joint as the image to be analyzed, and simultaneously records the surface temperature of the rail joint end; an edge track extraction step, outlining the edge track of the gap at the rail joint end in the image to be analyzed using an edge detection algorithm, and judging the clarity of the edge track based on the edge point distribution density. The edge track continuity is calculated based on the edge point distribution density and the distance threshold; a gap feature quantification step, calculating the grayscale difference or normalized contrast between the gap area and the background area in the image to be analyzed as the gap contrast, analyzing the cleanliness of the gap surface, and generating a multidimensional feature vector by combining edge trajectory continuity, light intensity, surface temperature, and gap contrast; a detection method selection step, inputting the multidimensional feature vector into a preset analysis method selection formula and outputting a judgment index, wherein the judgment index is used to select the optimal detection method among the line detection method, the point and surface detection method, and the three-dimensional feature-thermal imaging detection method; The gap width analysis step performs the gap width analysis task according to the optimal detection method determined in the detection method selection step.

2. The method for detecting blind rail seams based on machine vision according to claim 1, wherein: The edge trajectory extraction step includes an edge point screening strategy, which includes an edge point density calculation step and an edge point continuity judgment step. The edge point density calculation step is to count the number of edge points of the gap edge track in the image to be analyzed in units of pixels, and calculate the edge point density within a unit length; The edge point continuity judgment step judges the edge track continuity according to a distance threshold between edge points. If the continuity is lower than a preset threshold, the edge track is marked as a fuzzy track.

3. The method for detecting blind rail seams based on machine vision according to claim 2, wherein: The gap feature quantification step also includes a surface cleanliness assessment strategy. The surface cleanliness assessment strategy analyzes the texture features and brightness distribution of the gap area in the image to be analyzed, calculates the coverage rate of stains accumulated on the gap surface, and generates a surface cleanliness score by combining the stain coverage rate and the gap contrast.

4. A rail blind seam detection method based on machine vision according to claim 1 or 3, characterized in that: The method also includes a real-time monitoring step, which starts a real-time monitoring timer. If the gap width analysis task is not completed within the maximum allowed processing time, another detection method is switched to re-execute the analysis task.

5. The method for detecting blind rail seams based on machine vision according to claim 4, characterized in that: The real-time monitoring step includes a method switching strategy, which includes a time threshold setting step and a switching analysis method step. The time threshold setting step analyzes the maximum processing time of each detection method based on historical data and sets the maximum allowable processing time; The step of switching the analysis method selects the next detection method according to the proximity between the judgment value and the detection method value range.

6. The method for detecting blind rail seams based on machine vision according to claim 5, characterized in that: The gap width analysis step includes an analysis result verification strategy, which includes a repeated measurement step and an error correction step. Repeat the measurement step to measure the gap width of the same rail joint multiple times, record the results of each measurement and calculate the average value; The error correction step determines the stability of the measurement based on the standard deviation of multiple measurement results. If the standard deviation exceeds a preset threshold, the measurement task is re-executed or the acquisition angle of the visual camera is adjusted.

7. The method for detecting blind rail seams based on machine vision according to claim 1, wherein: The detection method selection step includes an extreme condition triggering strategy, which includes directly triggering an ignore instruction when the gap area is blocked by stains and the color is the same as the rust in the background area of ​​the rail butt end surface, and using a three-dimensional feature-thermal imaging detection method for analysis.

8. The method for detecting blind rail seams based on machine vision according to claim 1, wherein: The gap width analysis step analyzes two or more consecutive images to be analyzed to determine the change in the gap width, compares it with a preset change threshold, and outputs a normal instruction or an abnormal instruction based on the comparison result.

9. The method for detecting blind rail seams based on machine vision according to claim 1, wherein: The edge track extraction step includes an image preprocessing strategy, which includes intercepting the track joint area in the image to be analyzed using a target detection algorithm and retaining track section A and track section B on both sides of the gap.

10. A rail blind seam detection system based on machine vision, characterized by: include: The rail pair image acquisition module uses a visual camera installed on the bottom of the train to capture images of the rail joints as the images to be analyzed, and simultaneously records the surface temperature of the rail joints; An edge track extraction module, which uses an edge detection algorithm to outline the edge track of the gap at the rail joint end in the image to be analyzed, and determines the clarity of the edge track based on the edge point distribution density. The edge track continuity is calculated based on the edge point distribution density and the distance threshold; A gap feature quantification module calculates the grayscale difference or normalized contrast between the gap area and the background area in the image to be analyzed as the gap contrast, analyzes the cleanliness of the gap surface, and generates a multidimensional feature vector by combining edge trajectory continuity, light intensity, surface temperature, and gap contrast; a detection method selection module that inputs the multidimensional feature vector into a preset analysis method selection formula and outputs a judgment index for selecting the optimal detection method among the line detection method, the point and surface detection method, and the three-dimensional feature-thermal imaging detection method; The gap width analysis module performs the gap width analysis task according to the optimal detection method determined in the detection method selection step.

Images