An image calibration system based on dynamic train synchronous sampling

Abstract

The invention relates to the field of train image recognition, and in particular to an image calibration system based on dynamic train synchronous sampling. An image acquisition module is used to acquire an image to be tested, and the image is detected and compared by a primary comparison module, a secondary comparison module and a tertiary comparison module, and a calibration analysis module is used to perform image processing and calibration on the compared image. The invention solves the problems of deviation in image synchronous sampling, low precision of image splicing and three-dimensional modeling in the subsequent train detection process, and high cost of train detection data calculation in the prior art. By using the accurately calibrated image, the image can be used for subsequent foreign body detection, train scratch detection and train track defect detection, etc., thereby achieving the beneficial effects of reducing data detection cost, improving detection efficiency and ensuring the safety and stability of train operation.

Description

Technical Field

[0001] The invention relates to the field of train image recognition, and in particular to an image calibration system based on dynamic train synchronous sampling. Background Art

[0002] Train anomaly detection plays an important role in the safe operation of rail trains. Traditional train anomaly detection uses manual inspection. In the prior art, machine vision-based train anomaly detection has gradually replaced manual inspection. However, this detection method has very high requirements for image acquisition. In the prior art, when a train passes through a gantry for 360° detection, the image acquisition device needs to collect synchronized images of the train from multiple angles and splice the images or perform further image detection.

[0003] However, during the long-term use of the detection device, the left and right sampling of the train image sampling will be out of sync, which will lead to deviations in the train image, thus affecting the accuracy and effect of train abnormality detection. Therefore, in order to solve the sampling deviation problem of the train sampling system in the prior art, the present invention proposes an image calibration system based on dynamic train synchronous sampling, which collects synchronous train images through an image acquisition module, and undergoes detection and comparison through a primary comparison module, a secondary comparison module, and a tertiary comparison module, and performs image processing and calibration on the compared images through a calibration analysis module, thereby solving the problems of image synchronous sampling deviation, low image splicing and three-dimensional modeling accuracy in the subsequent train detection process, and high train detection data calculation cost in the prior art. By using the accurately calibrated image, it can be used for subsequent foreign body detection, train scratch detection, and train track defect detection, etc., achieving the beneficial effects of reducing data detection costs, improving detection efficiency, and ensuring the safety and stability of train operation. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned prior art, the first purpose of the present invention is to provide an image calibration system based on dynamic train synchronous sampling. By using the image calibration system based on dynamic train synchronous sampling, the effect of improving image acquisition accuracy and enhancing train monitoring and detection capabilities is achieved.

[0005] In order to solve the problems existing in the above-mentioned prior art, the second purpose of the present invention is to solve the problem in the prior art that the deviation of the images collected on the left and right sides when sampling the train image causes the image acquisition to be asynchronous, the image fusion to have errors, and the train detection effect to be affected.

[0006] In order to achieve the above object, the present invention provides a solution: an image calibration system based on dynamic train synchronous sampling, comprising:

[0007] An image acquisition module, the image acquisition module is used to acquire the image to be measured passing through the track section;

[0008] The first-level comparison module compares the image to be tested with the standard train image to determine whether the collected image has an angle deviation and obtain angle deviation information; thus, the influence of the angle deviation on the subsequent analysis and processing of the image is solved.

[0009] A secondary comparison module receives the image to be tested that has been judged and calibrated by the primary comparison module, and judges whether there is a position offset in the images to be tested collected on the left and right sides by comparing the coordinate deviations of the synchronous reference points selected from the train images on the left and right sides, thereby obtaining position deviation information; by using the secondary comparison module, the horizontal alignment of the images on the left and right sides can be ensured, thereby improving the accuracy of subsequent train image stitching, three-dimensional modeling or trajectory analysis.

[0010] The third-level comparison module receives the image to be tested after being judged and calibrated by the second-level comparison module, judges whether the collected image to be tested has an overall deviation caused by the train and the track, and obtains the overall deviation information; it is used to process the overall deformation of the image to ensure the accuracy and consistency of the image.

[0011] The calibration and correction module performs image calibration on the image to be measured based on the angle deviation information, the position deviation information and the integrity deviation information to obtain a calibrated image, thereby ensuring the high precision of the image.

[0012] A cloud database is used to store standard train images. The standard train images are train images collected without deviation. The standard train images are reference images for comparison and calibration.

[0013] The image acquisition module is used to acquire images of trains passing through the train detection section, and the first-level, second-level, and third-level comparison modules are used to perform deviation detection on the acquired images, so as to determine whether the acquired train images have angle deviations, sampling point position offsets, and image symmetry deviations. After determining the image deviation, image processing is performed to calibrate the image deviations.

[0014] As a further improvement of the present invention, the image acquisition module includes a first image acquisition device and a second image acquisition device, the image acquisition module includes a first image acquisition device and a second image acquisition device, the first image acquisition device is arranged on the left side of the train, the second image acquisition device is arranged on the right side of the train, and the first image acquisition device and the second image acquisition device are arranged at symmetrical positions on both sides of the train; the image acquisition module acquires video images of the train traveling in the calibration interval, and acquires the train image to be tested by extracting frames of the video image. The two image acquisition devices are arranged at symmetrical positions on the left and right sides of the train, which is helpful for synchronous sampling of the train image, that is, sampling the symmetrical positions of the passing train at the same time; the video image of the train is acquired by the image acquisition device, and the image deviation comparison is performed by extracting the synchronously acquired key frame images in the video image, which can reduce the amount of data processing while ensuring the image acquisition quality, and provide a basis for acquiring continuous images for the three-level comparison, which is helpful for improving the accuracy of deviation comparison and calibration.

[0015] As a further improvement of the present invention, the first-level comparison module receives the image to be tested, and calls the standard train image in the cloud database, extracts the geometric features of the train contour in the image through straight line detection, and respectively calculates the angles between the contour extraction straight line and the horizontal axis in the image to be tested and the standard train image. If the angle α between the contour extraction straight line and the horizontal axis in the image to be tested is not equal to the angle β between the contour extraction straight line and the horizontal axis in the standard train image, it is determined that there is an angle deviation in the image to be tested, and the angle deviation information is obtained, and the calibration correction module performs a first-level calibration. By comparing the train image to be tested with the standard train image through the first-level comparison module, the angle deviation of the image to be tested can be detected, which can not only detect the angle deviation of the train image, but also detect whether the image acquisition device has an angle offset, so as to adjust the image sampling device in time.

[0016] As a further improvement of the present invention, the calibration correction module receives the image to be tested after the judgment of the primary comparison module, and locates the center point of the marker contour therein, rotates the image to be tested with the contour center point as the rotation center, makes the angle α equal to the angle β, and fills in the missing pixels of the image by interpolation, crops the image to the size of the standard train image, and completes the primary calibration. By selecting the center point of the positioning marker, it can be ensured that the image aspect ratio and pixel position are affected as much as possible during the image angle correction process; by using the method of interpolating and filling in pixels, the edge pixels of the image can be expanded and filled after the image angle is rotated, so as to facilitate image cropping and re-acquire the corrected image; the marker is a marker with a clear contour such as a window or door in the image, so that it can be clearly positioned.

[0017] As a further improvement of the present invention, the secondary comparison module receives the image to be tested after the primary calibration, selects points for the outline of the marker in the image to be tested, evenly selects eight outline reference points and one outline center reference point and obtains the coordinates of each point, and calculates the coordinate pixel deviation value of the corresponding reference point pair in the left image to be tested and the right image to be tested. If the deviation value is higher than the set threshold value of 1 pixel value, it is judged that there is a position deviation in the image to be tested, and the position deviation information is obtained, and the calibration correction module performs secondary calibration. The secondary comparison module can compare the position deviation of the image to be tested, so as to determine the position deviation of the image to be tested. By selecting the reference point to calculate the pixel value deviation, the comparison result can be made more accurate, and each reference point has a different weight, which increases the reliability of the deviation comparison.

[0018] As a further improvement of the present invention, the coordinate pixel deviation value calculation in the secondary comparison module is provided as follows:

[0019] ;

[0020] Among them, k represents the coordinate pixel deviation value, x li ,y li Indicates the horizontal and vertical coordinates of the i-th contour reference point of the image to be tested on the left, x ri ,y ri represents the horizontal and vertical coordinates of the i-th contour reference point of the image to be tested on the right side. The weight of the contour reference point is 0.1, x l0 ,y l0 Indicates the horizontal and vertical coordinates of the center reference point of the contour of the image to be tested on the left, x r0 ,y r0 Indicates the horizontal and vertical coordinates of the center reference point of the contour of the image to be tested on the right, and the weight of the center point is 0.2.

[0021] By selecting reference points in the train image and obtaining the reference point coordinates, and comparing the corresponding reference point coordinates in the train images collected on the left and right sides to see if there is any deviation, it is possible to compare whether there is a symmetry deviation in the collected train images on the left and right sides, including whether the left and right camera sampling time and sampling points are synchronized, ensuring the accuracy of the deviation comparison. In addition, the symmetry deviation is determined by the reference point coordinates, which facilitates the subsequent correction efficiency.

[0022] As a further improvement of the present invention, the calibration correction module receives the image to be tested after the judgment of the secondary comparison module, determines the synchronization reference points in the two images, records their coordinate information, calculates the deviation value of the synchronization reference point coordinates in the left and right images to be tested, obtains the image translation amount through the calculated synchronization reference point coordinate deviation value, and performs secondary calibration on the left and right images to be tested. By calibrating the asymmetric problem of the left and right train images caused by the position deviation in the secondary comparison module, the accuracy and efficiency of image calibration are improved.

[0023] As a further improvement of the present invention, the three-level comparison module extracts train images of continuous frames before and after the image to be tested, and compares the overall deviation of the image to be tested. When there are regular deviations in the train images captured by continuous frames, it is judged that the overall deviation is caused by the track; when there is no regular deviation in the train images captured by continuous frames, it is judged that the overall deviation is caused by the train, and the overall deviation information is obtained.

[0024] As a further improvement of the present invention, the three-level comparison module extracts four train images of consecutive frames before and after the image to be tested, selects three triangular feature positioning points in each image, and uses this to calculate the deviation of the coordinates of the three feature positioning points of every two images. The weight of each deviation calculation is 1 / 6, and the total deviation weight value is added. If the deviation weight value is less than 1 after comparison, it is judged to be an overall deviation caused by the track; if the deviation weight value is greater than 1 after comparison, it is judged to be an overall deviation caused by train differences. By using the three-level comparison module to compare the collected train images, it is possible to detect whether there is an overall deviation in the image, thereby saving a lot of data detection costs in the subsequent foreign body detection, train scratch detection and train track defect detection of the train, and improving the efficiency of subsequent train detection.

[0025] The beneficial effects of the present invention are:

[0026] 1. Improve the accuracy of image calibration. By using the hierarchical comparison module to calibrate the image, the collected train images can be gradually compared and calibrated in detail, thereby effectively solving problems such as angle deviation, position offset and overall deviation, and significantly improving the accuracy of image calibration.

[0027] 2. Optimize the synchronously sampled images. Setting image acquisition devices symmetrically on the left and right sides of the train helps to achieve synchronous sampling of images. Calibrating the synchronously sampled images through the image calibration system helps to achieve time synchronization and symmetry of the sampled images, providing rapid detection conditions for subsequent image stitching or scratch detection of train inspection, and improving the accuracy of image splicing and three-dimensional modeling.

[0028] 3. Enhance the subsequent inspection effect of the train. By using the accurately calibrated images, it can be used for subsequent foreign object detection, train scratch detection, and train track defect detection, etc., reducing data detection costs, improving detection efficiency, and ensuring the safety and stability of train operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a system block diagram of an image calibration system based on dynamic train synchronous sampling of the present invention;

[0030] Figure 2 It is a control flow chart of an image calibration system based on dynamic train synchronous sampling of the present invention;

[0031] Figure 3 It is a working schematic diagram of a primary comparison module of an image calibration system based on dynamic train synchronous sampling of the present invention;

[0032] Figure 4 It is a working schematic diagram of a secondary comparison module of an image calibration system based on dynamic train synchronous sampling of the present invention;

[0033] Figure 5 It is a working schematic diagram of a three-level comparison module of an image calibration system based on dynamic train synchronous sampling of the present invention. DETAILED DESCRIPTION

[0034] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described here are only part of the embodiments of the present invention, not all of the embodiments. Based on the following embodiments, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] The embodiment of the present invention provides an image calibration system based on dynamic train synchronous sampling, comprising:

[0036] An image acquisition module is used to acquire images of trains passing through the track section to be tested. The image acquisition module needs to be set on both sides of the track and has the characteristics of high resolution, high sensitivity and low latency, so that it can acquire images of trains passing at high speed.

[0037] The first-level comparison module compares the image to be tested with the standard train image to determine whether the collected image has an angle deviation and obtain angle deviation information; thus, the influence of the angle deviation on the subsequent analysis and processing of the image is solved.

[0038] A secondary comparison module receives the image to be tested that has been judged and calibrated by the primary comparison module, and judges whether there is a position offset in the images to be tested collected on the left and right sides by comparing the coordinate deviations of the synchronous reference points selected from the train images on the left and right sides, thereby obtaining position deviation information; by using the secondary comparison module, the horizontal alignment of the images on the left and right sides can be ensured, thereby improving the accuracy of subsequent train image stitching, three-dimensional modeling or trajectory analysis.

[0039] The third-level comparison module receives the image to be tested after being judged and calibrated by the second-level comparison module, judges whether the collected image to be tested has an overall deviation caused by the train and the track, and obtains the overall deviation information; it is used to process the overall deformation of the image to ensure the accuracy and consistency of the image.

[0040] The calibration and correction module performs image calibration on the image to be measured based on the angle deviation information, the position deviation information and the integrity deviation information to obtain a calibrated image, thereby ensuring the high precision of the image.

[0041] A cloud database is used to store standard train images. The standard train images are train images collected without deviation. The standard train images are used as reference images in comparison modules and calibration modules at all levels.

[0042] The image acquisition module is used to acquire images of trains passing through the train detection section, and the first-level, second-level, and third-level comparison modules are used to perform deviation detection on the acquired images, so as to determine whether the acquired train images have angle deviations, sampling point position offsets, and image symmetry deviations. After determining the image deviation, image processing is performed to calibrate the image deviations.

[0043] In this embodiment, refer to Figure 1-Figure 5As shown, the image acquisition module includes a first image acquisition device and a second image acquisition device, the first image acquisition device is arranged on the left side of the train, the second image acquisition device is arranged on the right side of the train, and the first image acquisition device and the second image acquisition device are arranged at symmetrical positions on both sides of the train; when the train travels into the calibration interval, the two image acquisition devices start to work, and collect video images of the train in real time during the travel process. The image acquisition module performs frame extraction operation on the collected video, and extracts clear and stable train images by analyzing the key frames with obvious features in the video images as representatives, and ensures that the frames selected by the image acquisition devices on the left and right sides are image frames collected at the same time. The two image acquisition devices are arranged at symmetrical positions on the left and right sides of the train, which is helpful for synchronous sampling of train images, that is, sampling the symmetrical positions of the passing train at the same time; the video images of the train are collected by the image acquisition devices, and the image deviation comparison is performed by extracting the synchronously collected key frame images in the video images, which can reduce the amount of data processing while ensuring the image acquisition quality, provide a basis for obtaining continuous images for the three-level comparison, and help improve the accuracy of deviation comparison and calibration.

[0044] In this embodiment, refer to Figure 1-Figure 5 As shown, the primary comparison module receives the image to be tested, which is synchronously captured by the first image acquisition device and the second image acquisition device arranged on the left and right sides of the train when the train passes through a specific track section; and calls the standard train image in the cloud database, and applies the straight line detection technology to the received image to be tested and the standard train image, and the primary comparison module extracts the geometric features of the train contour. These geometric features mainly include the straight line part of the train contour, including the edge of the car body and the boundary of the window. When the external light is sufficient, the edge of the car body is selected as the geometric feature of the extracted train contour. When the external light is insufficient, there is a light source in the window of the moving train. At this time, the window boundary is more obvious. The window boundary is selected as the geometric feature of the extracted train contour, and the angle between the contour extraction line and the horizontal axis in the image to be tested and the standard train image is calculated respectively. In the image to be tested, the angle is recorded as α; in the standard train image, the angle is recorded as β; the angle α is compared with the angle β. If the angle α between the contour extraction line and the horizontal axis in the image to be tested is not equal to the angle β between the contour extraction line and the horizontal axis in the standard train image, it is judged that there is an angle deviation in the image to be tested. After detecting the angle deviation, the first-level comparison module will output the result to the calibration correction module, and the calibration correction module will perform a first-level calibration. By comparing the train image with the standard train image through the first-level comparison module, the angle deviation of the image to be tested can be detected. Not only can the angle deviation of the train image to be tested be detected, but also whether the image acquisition device has an angle offset, so as to adjust the image sampling device in time.

[0045] Specifically, the above steps may be:

[0046] Image preprocessing: converting the collected train image and the standard train image into grayscale images, filtering the grayscale image by using Gaussian filtering or median filtering, and using edge detection algorithm to extract edge information in the image;

[0047] Feature extraction: Use Hough transform to further detect straight line features in the edge image and obtain their direction and position information; select one or more reference lines in the reference image and the acquired image respectively, and the reference lines are selected from the edge line of the car, the edge line of the door or the frame of the window;

[0048] Angle calculation: For each reference line, calculate its direction angle, that is, the angle between it and the horizontal axis; compare the direction angle of the reference line in the acquired image with the corresponding direction angle of the reference line in the reference image, and the difference between them is the angle deviation.

[0049] Angle deviation judgment is performed, and the maximum angle deviation at which the train carriage image can appear is set to 3°, that is, the threshold is set to 3°. The deviation result of the above angle calculation is compared with the set threshold. If the angle deviation is less than or equal to the threshold, it is considered that the collected image is within the acceptable angle deviation range; otherwise, it is considered that there is a significant angle deviation.

[0050] In this embodiment, refer to Figure 1-Figure 5 As shown, the calibration correction module locates the center point of the marker contour in the image to be tested after the first-level comparison module determines, and rotates the image to be tested with the center point of the contour as the rotation center, so that the angle α is equal to the angle β. Due to the rotation of the image, the edge of the image may have missing pixels. The calibration correction module uses an interpolation algorithm to fill in the missing pixels. In this embodiment, the nearest neighbor interpolation can be used to select the value of the known pixel closest to the pixel to be found as the value of the pixel to be found; it can also be selected according to the values ​​of the four nearest neighboring pixels around the pixel to be found, and the value of the pixel to be found can be obtained by linear interpolation; it can also be used to use the values ​​of the 16 nearest neighboring pixels around the pixel to be found, and the value of the pixel to be found can be obtained by bicubic polynomial interpolation. The interpolation algorithm can estimate and fill in the missing pixel values ​​according to the information of the known pixels in the image, so as to maintain the integrity and continuity of the image. Finally, the calibration correction module will crop the rotated and filled image to the size of the standard train image. By selecting the center point of the marker positioning frame, it is possible to ensure that the impact on the image aspect ratio and pixel position is minimized during the image angle correction process; by using the method of interpolating and completing pixels, the image edge pixels can be expanded and completed after the image angle is rotated, thereby facilitating image cropping and re-acquisition of the corrected image.

[0051] In this embodiment, refer to Figure 1-Figure 5 As shown, the secondary comparison module receives the image to be tested output by the primary comparison module, selects points on the collected train image landmark contour, evenly selects eight contour reference points and one contour center reference point and obtains the coordinates of each point, and calculates the coordinate pixel deviation value of the corresponding reference point pair in the left image to be tested and the right image to be tested. If the deviation value is higher than the set threshold value of 1 pixel value, it is determined that the image to be tested has a position deviation, and the calibration correction module performs secondary calibration.

[0052] In this embodiment, refer to Figure 1-Figure 5 As shown, the coordinate pixel deviation value calculation in the secondary comparison module is provided as follows:

[0053] ;

[0054] Among them, k represents the coordinate pixel deviation value, x li ,y li represents the horizontal and vertical coordinates of the i-th contour reference point of the left image to be tested, xri and yri represent the horizontal and vertical coordinates of the i-th contour reference point of the right image to be tested, and the weight of the contour reference point is 0.1, xl0 and yl0 represent the horizontal and vertical coordinates of the contour center reference point of the left train image, xr0 and yr0 represent the horizontal and vertical coordinates of the contour center reference point of the right train image, and the weight of the center point is 0.2.

[0055] If the secondary comparison module detects that there is a position deviation in the image, it means that the sampling synchronization time of the sampling systems on both sides of the train image sampling system is not unified, or there is a horizontal displacement of the image acquisition device or sampling point, so that the problems in the train detection can be quickly judged. By selecting a reference point in the train image and obtaining the reference point coordinates, and comparing the corresponding reference point coordinates in the train images collected on the left and right sides to see if there is a deviation, it is possible to compare whether there is a symmetry deviation in the collected left and right train images, including whether the sampling time and sampling points of the left and right cameras are synchronized, ensuring the accuracy of the deviation comparison. In addition, the symmetry deviation is determined by the reference point coordinates, which facilitates the subsequent correction efficiency.

[0056] In this embodiment, refer to Figure 1-Figure 5 As shown, the calibration and correction module receives the image to be tested after being judged by the secondary comparison module, determines the synchronization reference points in the two images, records their coordinate information, calculates the deviation value of the synchronization reference point coordinates in the left and right train images, obtains the image translation amount through the calculated synchronization reference point coordinate deviation value, and calibrates the left and right images to be tested.

[0057] In this embodiment, the steps are:

[0058] Select reference points and mark them. Select synchronization reference points that are easy to identify and match in the left and right images. Mark the synchronization reference points respectively and record their coordinate information. Read the coordinate information of the synchronization reference points from the marks of the left and right images and calculate the deviation value of the coordinates of the synchronization reference points in the left and right images.

[0059] A correction algorithm is selected. In this embodiment, the image is rotated or translated by affine transformation or perspective transformation. The selected correction algorithm is applied to correct the left and right images. During the correction process, the asymmetry caused by the position deviation is eliminated by adjusting the translation amount in the algorithm parameters. The asymmetry of the left and right train images caused by the position deviation in the secondary comparison module is calibrated to improve the accuracy and efficiency of image calibration.

[0060] In this embodiment, refer to Figure 1-Figure 5 As shown, the three-level comparison module extracts four train images of continuous frames before and after the train image is collected. These four images need to maintain temporal continuity to ensure that slight changes in the train and the track can be captured. Then the module compares the overall deviations in the four images. When there are regular deviations in the train images collected by continuous frames, that is, the deviations show a predictable and repetitive pattern in time and space, then this deviation is likely to be caused by unevenness, deformation or maintenance problems of the track, and it is judged to be an overall deviation caused by the track. When there is no regular deviation in the train images collected by continuous frames, but a random and irregular deviation is shown, then this deviation is likely to be caused by factors of the train itself, such as vibration, suspension system problems, etc., and it is judged to be an overall deviation caused by the train.

[0061] In this embodiment, refer to Figure 1-Figure 5 As shown, the three-level comparison module extracts four train images of consecutive frames before and after the image to be tested, selects three triangular feature positioning points in each image, and uses this to calculate the deviation of the coordinates of the three feature positioning points of every two images. The weight of each deviation calculation is 1 / 6, and the total deviation weight value is added. If the deviation weight value after comparison is less than 1, it is judged to be an overall deviation caused by the track; if the deviation weight value after comparison is greater than 1, it is judged to be an overall deviation caused by train differences. By using the three-level comparison module to compare the collected train images, it is possible to detect whether there is an overall deviation in the image, thereby saving a lot of data detection costs in the subsequent foreign body detection, train scratch detection and train track defect detection of the train, improving the efficiency of subsequent train detection, and thus ensuring the safety and stability of train operation.

[0062] By using the image calibration system based on dynamic train synchronous sampling mentioned in this embodiment, when the three-level comparison deviation is detected in the image, the type of the overall deviation of the image is obtained, and the three-level comparison deviation is matched. The cloud database also stores track abnormality images and train abnormality images, and the corresponding track abnormality images or train abnormality images in the cloud database are called for comparison, so as to quickly determine whether there are track abnormalities, or train abnormalities such as scratches, stains or foreign objects.

[0063] This embodiment proposes an image calibration system based on dynamic train synchronous sampling. By using a hierarchical comparison module to calibrate the image, the collected train images can be gradually compared and calibrated in detail, thereby effectively solving the problems of sampling time difference, angle deviation, position offset and integrity deviation, and significantly improving the accuracy of image calibration; by using the accurately calibrated image, it can be used for subsequent foreign object detection, train scratch detection and train track defect detection, etc., reducing data detection costs, improving detection efficiency, and ensuring the safety and stability of train operation.

[0064] The above shows and describes the basic features, principles and advantages of the present invention. It should be noted that the present invention is not limited to the above embodiments, which are only partial embodiments. Under the premise of not departing from the spirit and scope of the present invention, several improvements and supplements made are considered as the protection scope of the present invention.

Claims

1. An image calibration system based on dynamic train synchronous sampling, characterized in that: include: An image acquisition module, the image acquisition module is used to acquire the image to be measured passing through the track section; A primary comparison module, wherein the primary comparison module determines whether there is an angle deviation in the collected image by comparing the image to be tested with the standard train image. The primary comparison module receives the image to be tested and calls the standard train image in the cloud database, extracts the geometric features of the train contour in the image by straight line detection, and respectively calculates the angles between the contour extraction straight line and the horizontal axis in the image to be tested and the standard train image. If the angle α between the contour extraction straight line and the horizontal axis in the image to be tested is not equal to the angle β between the contour extraction straight line and the horizontal axis in the standard train image, it is determined that there is an angle deviation in the image to be tested, and the angle deviation information is obtained, and the calibration correction module performs primary calibration; A secondary comparison module, wherein the secondary comparison module receives the image to be tested that has been judged and calibrated by the primary comparison module, and judges whether there is a positional offset in the images to be tested collected on the left and right sides by comparing the coordinate deviations of the synchronous reference points selected from the train images on the left and right sides. The secondary comparison module receives the image to be tested after the primary calibration, selects points for the outline of the marker in the image to be tested, evenly selects eight outline reference points and one outline center reference point, and obtains the coordinates of each point. By calculating the coordinate pixel deviation value of the corresponding reference point pair in the left image to be tested and the right image to be tested, if the deviation value is higher than the set threshold value of 1 pixel value, it is judged that the image to be tested has a positional deviation, and the positional deviation information is obtained, and the calibration correction module performs secondary calibration; A three-level comparison module, wherein the three-level comparison module receives the image to be tested after being judged and calibrated by the two-level comparison module, and judges whether the collected image to be tested has an overall deviation caused by the train and the track. The three-level comparison module extracts four train images of consecutive frames before and after the image to be tested, selects three triangular feature positioning points in each image, and calculates the deviation of the coordinates of the three feature positioning points of every two images. The weight of each deviation calculation is 1 / 6, and the total deviation weight value is added. If the deviation weight value after comparison is less than 1, it is judged to be an overall deviation caused by the track; If the deviation weight value after comparison is greater than 1, it is judged to be an overall deviation caused by train differences, and the overall deviation information is obtained; A calibration correction module, wherein the calibration correction module performs image calibration on the image to be measured based on the angle deviation information, the position deviation information and the integrity deviation information to obtain a calibration image; A cloud database is used to store standard train images, where the standard train images are train images collected without deviation.

2. The image calibration system based on dynamic train synchronous sampling according to claim 1 is characterized in that: The image acquisition module includes a first image acquisition device and a second image acquisition device, the first image acquisition device is arranged on the left side of the train, and the second image acquisition device is arranged on the right side of the train, and the first image acquisition device and the second image acquisition device are arranged at symmetrical positions on both sides of the train; the image acquisition module acquires video images of the train traveling in the calibration section, and obtains the train image to be tested by extracting frames of the video image.

3. The image calibration system based on dynamic train synchronous sampling according to claim 1, characterized in that: The calibration and correction module receives the image to be tested after being judged by the first-level comparison module, locates the center point of the marker contour in it, rotates the image to be tested with the contour center point as the rotation center, makes the angle α equal to the angle β, and fills in the missing pixels of the image by interpolation, crops the image to the size of the standard train image, and completes the first-level calibration.

4. The image calibration system based on dynamic train synchronous sampling according to claim 1, characterized in that: The coordinate pixel deviation value calculation in the secondary comparison module is provided by: ； Among them, k represents the coordinate pixel deviation value, x li ,y li Indicates the horizontal and vertical coordinates of the i-th contour reference point of the image to be tested on the left, x ri ,y ri represents the horizontal and vertical coordinates of the i-th contour reference point of the image to be tested on the right side. The weight of the contour reference point is 0.1, x l0 ,y l0 Indicates the horizontal and vertical coordinates of the center reference point of the contour of the image to be tested on the left, x r0 ,y r0 Indicates the horizontal and vertical coordinates of the center reference point of the contour of the image to be tested on the right, and the weight of the center point is 0.

2.

5. The image calibration system based on dynamic train synchronous sampling according to claim 1, characterized in that: The calibration and correction module receives the image to be tested after being judged by the secondary comparison module, determines the synchronization reference points in the two images, records their coordinate information, calculates the deviation value of the synchronization reference point coordinates in the left and right images to be tested, obtains the image translation amount through the calculated synchronization reference point coordinate deviation value, and performs secondary calibration on the left and right images to be tested.

6. The image calibration system based on dynamic train synchronous sampling according to claim 1, characterized in that: The three-level comparison module extracts train images of continuous frames before and after the image to be tested, and compares the overall deviation of the image to be tested. When there are regular deviations in the train images collected in continuous frames, it is judged that the overall deviation is caused by the track. When there are no regular deviations in the train images collected in continuous frames, it is judged that the overall deviation is caused by the train, and the overall deviation information is obtained.

Images