Spatial analysis platform for protection zone along subway line based on unmanned aerial vehicle orthographic map and GIS map fusion technology

Abstract

The application discloses a subway along-line protection area spatial analysis platform based on unmanned aerial vehicle orthographic map and GIS map fusion technology, belongs to the technical field of subway along-line protection area monitoring, and comprises a data acquisition module, a protection area image construction module, an abnormality judgment module and a GIS real-time data sharing module.The data acquisition module obtains a target area orthographic map by collecting target area video data through an unmanned aerial vehicle, combines ground control points set in the target area, calibrates the obtained orthographic image in precision by using RTK measurement coordinates and elevation, collects target subway line information, and marks the target subway line in the target area orthographic map.The subway along-line protection area spatial analysis platform based on unmanned aerial vehicle orthographic map and GIS map fusion technology simplifies complex geographic information through line processing, improves the accuracy, real-time performance and efficiency of risk determination, and provides strong support for the safety management of subway along-line protection areas.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of subway along line protection area monitoring, and particularly relates to a subway along line protection area spatial analysis platform based on unmanned aerial vehicle orthographic map and GIS map fusion technology. BACKGROUND

[0002] The subway along line protection area is a specific area defined around the subway line to protect the safe operation of the subway and the subway facilities from the influence of external activities, and is usually an area within a certain distance outside the outer line of the subway structure. The tunnels, stations and other structures of the subway are subjected to the load of train operation and the influence of the surrounding environment for a long time. The establishment of the subway along line protection area can avoid damage to the subway structure caused by surrounding construction activities.

[0003] Traditional subway along line protection area management relies on manual patrol and map analysis, which has obvious limitations. Manual patrol is low in efficiency and limited in coverage, and it is difficult to capture dynamic changes in the protection area in real time. Moreover, risk determination is easily affected by human experience, and the accuracy and reliability are insufficient. Through map analysis, a large amount of calculation is required, and the complex geographical information in the protection area will generate a large amount of redundant data, increasing the calculation amount of risk determination and affecting the analysis efficiency. SUMMARY

[0004] The present application aims to at least solve one of the problems in the prior art. To this end, the present application proposes a subway along line protection area spatial analysis platform based on unmanned aerial vehicle orthographic map and GIS map fusion technology, which improves the detection method and processing method to solve the above technical problems.

[0005] In order to achieve the above purpose, the application adopts the following technical scheme:

[0006] The subway along line protection area spatial analysis platform based on unmanned aerial vehicle orthographic map and GIS map fusion technology comprises a data acquisition module, a protection area image construction module, an abnormality determination module and a GIS real-time data sharing module.

[0007] The data acquisition module obtains the orthographic map of the target area by collecting video data of the target area by the unmanned aerial vehicle, combines the ground control points set in the target area, calibrates the obtained orthographic image in accuracy by using the RTK measurement coordinates and elevation, collects the target subway line information, and marks the target subway line in the target area orthographic map.

[0008] The protection area image construction module performs lineation processing on the identified features in the orthographic map based on the obtained target area orthographic map, preliminarily divides the protection area range of the subway line based on the marked subway line in the orthographic map, corrects the preliminarily divided protection area range according to the lineation-processed identified features, and maps the target area orthographic map into the abnormality judgment rectangular map to obtain the protection area range along the subway line in the abnormality judgment rectangular map.

[0009] The GIS real-time data sharing module collects the road condition information of the target area in real time through the GIS system, identifies the suspicious features in the road condition information, and performs lineation processing on the suspicious features in the road condition information and maps them into the abnormality judgment rectangular map in real time.

[0010] The abnormality analysis module performs abnormality judgment on the lines by combining the mapping of the suspicious features in the abnormality judgment rectangular map with the protection area range along the subway line, and identifies the real-time risk and potential risk of the subway protection area.

[0011] Further, the data collection module specifically includes the following steps:

[0012] According to the size, shape and terrain factors of the target area, a flight route of the unmanned aerial vehicle is planned, video image data of the target area is captured by the unmanned aerial vehicle on-board camera equipment, and the video image data captured by the unmanned aerial vehicle is subjected to image matching to obtain the target area orthographic map, specifically including the following steps:

[0013] The video image data captured by the unmanned aerial vehicle is subjected to image matching, the same-named points between different images are extracted by the SIFT algorithm, and then aerial triangulation is performed to solve the exterior orientation elements of each image, including three line elements and three angle elements , and a collinearity equation is constructed in combination with the camera equipment parameters:

[0014] ;

[0015] wherein, represents the image point coordinates, represents the ground point coordinates, represents the exterior orientation line elements, represents the exterior orientation angle elements for constructing a rotation matrix, is a rotation matrix element composed of the exterior orientation angle elements, represents the camera principal distance of the camera equipment, represents the image principal point coordinates, an error equation is established according to the collinearity equation to describe the relationship between the image point coordinates on the image and the ground point coordinates, the camera interior orientation elements and the exterior orientation elements, and a least square method is used to perform adjustment calculation on the error equation to solve the optimal solution of the exterior orientation elements.

[0016] According to the exterior orientation elements obtained by the aerial triangulation, the camera interior orientation elements and the digital elevation model, ortho-rectification is performed on each image, all ortho-images are spliced, the joints of the overlapping areas are processed, and an ortho-map of the target area is obtained;

[0017] Eight ground control points are uniformly set based on the length and width of the target area, and the three-dimensional coordinates of each ground control point are accurately measured by the RTK measurement device. The coordinates and elevation information of the measured ground control points are compared with the coordinates of the control points obtained by solving the ortho-map during the image matching process. Through adjustment calculation and solution by the least square method, the exterior orientation elements of the image are adjusted to calibrate the accuracy.

[0018] Further, the target subway line information is collected, and the target subway line is marked in the target area ortho-map, and the steps are as follows:

[0019] The target subway line information is obtained through the official website of urban rail transit, including the passing places of the subway line and the length of the subway line. The target subway line is marked in the target area ortho-map based on the passing places of the subway line, and all the passing positions of the target subway line are marked.

[0020] Further, the protected area image construction module is based on the obtained target area ortho-map, and the identified features in the ortho-map are line processing, and the subway line protection area is preliminarily divided based on the marked subway line in the ortho-map, which includes the following steps:

[0021] The image is grayed to reduce the data amount, the noise in the image is removed by mean filtering to smooth the image, and linear contrast enhancement is used to highlight the identified features, that is, , wherein represents the pixel value of the input image, represents the pixel value of the output image, respectively represent the adjustment parameters of contrast and brightness;

[0022] The features that need to be line processed are identified on the preprocessed target area ortho-map, and the identified features in the ortho-map are line processed, and the specific steps are as follows:

[0023] The image is further smoothed using a Gaussian filter, and the gradient of the image in the x direction and y direction is calculated by a Sobel operator , and the gradient amplitude and gradient direction are calculated, wherein the gradient amplitude , the gradient direction , the local maximum value of the gradient amplitude is detected in the gradient direction, and the local maximum value is retained and other non-maximum points are suppressed;

[0024] Confirming strong edge points by double threshold detection, by setting a strong threshold and a weak threshold, marking the pixel points with gradient amplitude greater than the strong threshold as strong edge points, marking the pixel points with gradient amplitude less than the weak threshold as weak edge points, for the pixel points between the strong threshold and the weak threshold, when they are connected with weak edge points, they are determined as weak edge points, when they are connected with strong edge points, they are determined as strong edge points, connecting the strong edge points and the weak edge points connected therewith to form complete edges, and separating the identified features from the background by dividing the image into different regions according to the gray values of the image by Otsu algorithm;

[0025] Refining the identified feature region into a single-pixel-wide line by Zhang-Suen thinning algorithm, and removing the boundary pixels by iteration to obtain the line of the identified feature, and removing the line with a length < wherein represents the line length threshold;

[0026] Mapping an abnormality judgment rectangular map with the same length and width according to the size of the target area orthographic map, mapping the line of the identified feature into the two-dimensional coordinates in the abnormality judgment rectangular map according to the two-dimensional coordinates of the identified feature in the target area orthographic map, extracting the two-dimensional coordinates of each point on the target subway line from the target area orthographic map, mapping the two-dimensional coordinates of each point on the target subway line to the same two-dimensional coordinate position of the abnormality judgment rectangular map as they are, taking a circle as the range of the subway line protection zone, placing the center of the circle on the subway line, i.e. taking each point on the subway line as a center of the circle, mapping the two-dimensional coordinates of each center of the circle on the subway line to the same two-dimensional coordinate position of the abnormality judgment rectangular map as they are, and drawing a circle with the center of each mapped center as the center and the radius of the subway protection zone in the abnormality judgment rectangular map to obtain the preliminary division range of the subway line protection zone.

[0027] Further, the preliminary division range of the protection zone is corrected according to the line of the identified feature after the line processing, and the target area orthographic map is mapped into the abnormality judgment rectangular map to obtain the range of the subway line protection zone, specifically including the following steps:

[0028] Deleting the interference position in the abnormality judgment rectangular map according to the preliminary division range of the subway line protection zone and the mapped line of the identified feature to correct the range of the subway protection zone and obtain the range of the subway line protection zone, including the following steps:

[0029] For each point on the line of the identified feature mapped into the abnormality judgment rectangular map, interference checking is performed, wherein , is the number of points on the line of the identified feature.

[0030] For each center point on the subway line ,in , To determine the number of centers of circles along the subway line, calculate and identify points on the feature lines. To the center distance ,when When the radius of the subway protection zone is less than or equal to the current point, it represents the current point. If a point falls within the subway protection zone, interference exists. For circular protection zones where interference occurs, their boundaries are redefined, retaining only the portion that does not interfere with the identification feature lines. The remaining protection zones are then merged to form a continuous subway line protection zone, thus obtaining the subway line protection zone.

[0031] Furthermore, the GIS real-time data sharing module collects road condition information of the target area in real time through the GIS system, processes suspicious features in the road condition information into lines, and maps them onto the orthophoto map of the target area in real time. Specifically, this includes the following steps:

[0032] In the orthophoto map of the target area, the road condition information of the target area is obtained in real time through the GIS system. A neural network model is built through historical data to identify the foundation pits and construction vehicles in the road condition information in real time. The identified foundation pits and construction vehicles are marked as suspicious features, and the coordinate range of the suspicious features is recorded.

[0033] The coordinate range of suspicious features is transmitted to the protected area image construction module. The Sobel operator is used to identify the boundaries of suspicious features within the coordinate range. The Zhang-Suen thinning algorithm is used to refine the identified suspicious feature boundaries into lines with a width of one pixel. Based on the coordinates of the suspicious feature lines in the orthophoto of the target area, the suspicious features are mapped to the anomaly determination rectangle.

[0034] Furthermore, the step of acquiring road condition information of the target area in real time through a GIS system in the orthophoto map of the target area, constructing a neural network model through historical data, identifying excavation pits and construction vehicles in the road condition information in real time, marking the identified excavation pits and construction vehicles as suspicious features, and recording the coordinate range of the suspicious features includes the following steps:

[0035] The target area's orthophoto map and corresponding road condition information are collected in real time through a GIS system. Past road condition image data and corresponding annotation information of the target area are collected. The annotation content covers the categories of foundation pits, soil piles, and construction vehicles. The collected historical data is divided into training set, validation set, and test set. The Faster R-CNN model is trained to identify suspicious features. The model parameters are initialized through Xavier initialization. Cross-entropy loss is selected as the loss function. The training set data is input into the model in batches. The value of the loss function is calculated. The model parameters are updated using the Adam optimizer. The trained model is evaluated using the validation set and test set to build a neural network model.

[0036] The orthophoto map of the target area is acquired in real time through the GIS system and input into a trained neural network model. The neural network model analyzes the input orthophoto map, identifies the foundation pit mounds and construction vehicles, and marks them as suspicious features. The coordinate range of the suspicious features is recorded according to the coordinate system in the orthophoto map.

[0037] Furthermore, the anomaly analysis module, by mapping suspicious features in the anomaly determination rectangle and combining them with the protected area along the subway line, performs anomaly determination on the lines, identifying real-time and potential risks within the subway protected area, including the following steps:

[0038] Based on the mapped coordinates of suspicious features in the anomaly detection rectangle, and combined with the coordinates of the protected area along the subway line, anomaly detection is performed on the current abnormal feature lines, including the following steps:

[0039] For each point on the abnormal feature line The method of ray casting is used to determine whether it is located within the polygon of the subway protection zone. Inside, through from point A ray is emitted horizontally to the right; the relationship between the ray and the polygon of the protected area is statistically analyzed. The number of intersection points of the edges;

[0040] When the number of intersection points is odd, it means that the current suspicious feature is located within the polygon of the protected area. Within the protected area, where a real-time risk exists, a red circle is drawn to mark it in the anomaly detection rectangle. When the number of intersection points is even, it means that the currently suspicious feature is not within the protected area polygon. Within this range, suspicious features are marked as pending.

[0041] For suspicious features of pending markers, based on the type of suspicious feature, at fixed time intervals... Analyze changes in suspicious characteristics to determine potential risks:

[0042] For the excavation pit soil pile, continuous statistics were collected. interval The average expansion rate of the current foundation pit soil pile is calculated, the edge line position after expansion is predicted, and the ray method is judged. When the number of intersection points is odd, the current foundation pit soil pile is marked as a potential risk, and a blue circle is marked in the abnormal judgment rectangular diagram, and when the number of intersection points is even, the current foundation pit soil pile is ignored.

[0043] For the construction vehicle, the edge line center point position of the interval is continuously counted. The edge line center point position of every two intervals is connected, and the meeting point of all intervals is counted. Based on the ray method, when more than four-fifths of the meeting points are odd, the current construction vehicle is marked as a potential risk, and a blue circle is marked in the abnormal judgment rectangular diagram, otherwise, it is ignored.

[0044] Compared with the prior art, the beneficial effects of the present application are:

[0045] 1. In the present application, by combining the unmanned aerial vehicle orthographic map with the GIS system, the line treatment of the subway along the protection area is carried out, which is helpful to intuitively determine the real-time risk and potential risk of the subway along the protection area, and reduces the calculation amount of risk determination.

[0046] 2. In the present application, the real image of the subway along the protection area can be provided by the unmanned aerial vehicle orthographic map, which clearly presents the actual situation of the topography, landform, building and the like of the protection area.

[0047] 3. In the present application, the complex geographical information of the subway along the protection area is simplified to lines and graphics by line treatment, which reduces the redundancy and complexity of data. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1The figure is a platform block diagram of a spatial analysis platform for a protection area along a subway line based on a UAV orthographic map and GIS map fusion technology. DETAILED DESCRIPTION

[0049] The technical solutions of the present application will be described clearly and completely below in conjunction with embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0050] Embodiment 1

[0051] As shown in the figure, the spatial analysis platform for the protection area along the subway line based on the UAV orthographic map and GIS map fusion technology comprises a data acquisition module, a protection area image construction module, an abnormality determination module and a GIS real-time data sharing module. Figure 1 The data acquisition module obtains a target area orthographic map by collecting target area video data by using a UAV, combines the ground control points set in the target area, calibrates the obtained orthographic image in precision by using RTK measurement coordinates and elevation, collects target subway line information and marks the target subway line in the target area orthographic map.

[0052] According to the size, shape and terrain factors of the target area, a flight route of the UAV is planned, target area video image data is shot by using a UAV on-board camera device, the target area orthographic map is obtained by performing image matching on the video image data shot by the UAV, and the specific steps include the following steps.

[0053] The video image data shot by the UAV is subjected to image matching, the same-named points between different images are extracted by using a SIFT algorithm, and then aerial triangulation is performed to solve the exterior orientation elements of each image, including three line elements

[0054] and three angle elements , a collineation equation is constructed in combination with the camera device parameters:

[0055] ;

[0056] wherein, represents a pixel point coordinate, represents a ground point coordinate, represents an exterior orientation line element, represents an exterior orientation angle element used to construct a rotation matrix, is a rotation matrix element composed of the exterior orientation angle elements, represents a camera principal distance of the camera device, ​Representing the principal point coordinates, the error equation is established according to the collinearity equation to describe the relationship between the image point coordinates on the image, the ground point coordinates, the camera interior orientation elements and the exterior orientation elements, and the least square method is used to perform adjustment calculation on the error equation to solve the optimal solution of the exterior orientation elements;

[0057] It should be noted that before image processing, it is necessary to ensure that the collected video data is of good quality and contains sufficient overlapping areas. In the present application, the heading overlap degree is 60%-80%, the lateral overlap degree is 30%-60%, the SIFT feature descriptor is a 128-dimensional vector obtained by calculating the gradient direction histogram in the neighborhood around the feature point. The specific steps include calculating the gradient direction and gradient amplitude in the neighborhood around the feature point, dividing the neighborhood into 4x4 sub-regions, and calculating the gradient direction histogram of 9 directions in each sub-region. Finally, all the histograms of the sub-regions are combined into a 128-dimensional vector.

[0058] The error equation is established by the collinearity equation, wherein the error equation is , respectively represent the error vector, the coefficient matrix, the correction number vector of the exterior orientation elements and the constant term vector. When the least square method is adjusted, , wherein is the weight matrix.

[0059] According to the exterior orientation elements obtained by aerial triangulation, the camera interior orientation elements and the digital elevation model, each image is orthorectified, all orthographic images are spliced, the seams of the overlapping areas are processed, and the target area orthographic map is obtained.

[0060] Based on the length and width of the target area, 8 ground control points are uniformly set, each ground control point is accurately measured by the RTK measurement device to obtain three-dimensional coordinates, the measured ground control point coordinates and elevation information are compared with the control point coordinates obtained by solving the orthographic map in the image matching process, and the exterior orientation elements of the image are adjusted by adjustment calculation and least square method to calibrate the accuracy.

[0061] It should be noted that the ground control points are set on obvious ground feature points, including road intersections, building corner points, etc., and the number is between 5-10. In the present application, the number of ground control points is 8.

[0062] Collect the target subway line information, and mark the target subway line in the target area orthographic map, the steps are as follows:

[0063] Obtaining the target subway line information through the official website of the urban rail transit, including the passing places of the subway line and the length of the subway line, marking the target place line based on the passing places of the subway line in the target area orthographic map, and marking all the passing places of the target subway line.

[0064] It should be noted that all the passing places of the target subway line are all the areas to be passed through in the subway running process, so as to facilitate the protection area mapping based on the marked passing places subsequently.

[0065] Embodiment 2:

[0066] The protection area image construction module performs line processing on the identified features in the obtained target area orthographic map, preliminarily divides the subway line protection area range based on the marked subway line in the orthographic map, corrects the preliminarily divided protection area range according to the line processed identified features, and maps the target area orthographic map to the anomaly judgment rectangular map to obtain the subway line protection area range in the anomaly judgment rectangular map.

[0067] , and specifically includes the following steps:

[0068] For the obtained target area orthographic map, the image is grayed to reduce the data amount, the noise in the image is removed by mean filtering to smooth the image, and linear contrast enhancement is adopted to highlight the identified features, that is, , wherein represents the pixel value of the input image, represents the pixel value of the output image, respectively represent the adjustment parameters of contrast and brightness;

[0069] It should be noted that the noise in the image is removed by mean filtering to smooth the image, which can avoid the interference of the noise on the subsequent feature recognition and line processing. For a 3x3 filter window, the mean value calculation formula of the pixels in the window is , wherein represents the pixel value of the i-th row and j-th column in the window, is the new value of the center pixel after filtering, and a and b need to be mapped to the desired minimum gray value and maximum gray value of the output image according to the minimum gray value and maximum gray value of the input image to calculate the values of a and b, wherein the desired minimum gray value of the output image is a x the minimum gray value of the image + b, and the desired maximum gray value of the output image is a x the maximum gray value of the image + b.

[0070] Identifying the features that need to be line processed on the preprocessed target area orthographic map, performing line processing on the identified features in the orthographic map, and the specific steps are as follows:

[0071] The image is further smoothed using a Gaussian filter, and the Sobel operator is used to calculate the image's smoothness. gradient in direction The gradient magnitude and gradient direction are calculated, where the gradient magnitude is... gradient direction Local maxima are detected in the gradient magnitude along the gradient direction, local maxima are preserved, and other non-maxima points are suppressed.

[0072] It should be noted that during non-maximum suppression, for each pixel in the image, two adjacent pixels are selected based on its gradient direction. For example, if the gradient direction is 0°, the pixels to the left and right of the pixel are selected; if the gradient direction is 45°, the pixels to the upper right and lower left of the pixel are selected. The gradient magnitude of the pixel is compared with the gradient magnitudes of the two adjacent pixels. If the gradient magnitude of the pixel is greater than that of the two adjacent pixels, the pixel is considered a local maximum and its gradient magnitude is retained; otherwise, the gradient magnitude of the pixel is set to 0. The features to be identified include buildings, trees, etc., which need to be determined based on the specific urban environment of the application.

[0073] Strong edge points are identified through dual threshold detection. By setting strong and weak thresholds, pixels with gradient magnitudes greater than the strong threshold are marked as strong edge points, and pixels with gradient magnitudes less than the weak threshold are marked as weak edge points. For pixels between the strong and weak thresholds, they are determined to be weak edge points when connected to weak edge points and strong edge points when connected to strong edge points. Strong edge points and their connected weak edge points are connected to form complete edges. The Otsu algorithm is used to segment the image into different regions based on the image's grayscale values, separating the recognition features from the background.

[0074] It should be noted that during the edge connection process, the process starts from a strong edge point and checks the pixels in its 8-neighborhood. If a weak edge point exists in the neighborhood, it is marked as a strong edge point, and the 8-neighborhood of the newly marked strong edge point is checked until no new weak edge points are marked. After edge connection, all weak edge points connected to the strong edge point are merged into a strong edge point, forming a complete edge. During the setting of the strong and weak thresholds, two thresholds need to be manually set based on experience and prior knowledge of the image. The operator observes the image features, contrast, and expected edge detection conditions, initially sets the thresholds, and then checks the edge detection results. If the results are not ideal, the thresholds are continuously adjusted until a satisfactory edge detection effect is obtained, thus obtaining the strong and weak thresholds.

[0075] The identified feature region is thinned into a single-pixel wide line by a Zhang-Suen thinning algorithm, and boundary pixels are iteratively deleted to obtain a line of the identified feature, and the length of the line is represents a line length threshold value;

[0076] It should be noted that in the iteration process of the Zhang-Suen thinning algorithm for feature line, two sub-steps are included, wherein step one is to traverse each pixel p in the image, and the pixel p can be considered for deletion when the following conditions are met: the value of the pixel p is 1, the number of 1 in the 8-neighborhood of the pixel p is between 2-6, wherein the 8-neighborhood refers to 8 pixels adjacent to the pixel p, the number of transitions from 0 to 1 in the 8-neighborhood of the pixel p is 1, wherein the number of transitions refers to the number of times of transition from 0 to 1 when traversing the 8-neighborhood pixels in a clockwise direction, and at least one of the upper, right, and lower positions in the 8-neighborhood of the pixel p is 0, and the pixel p satisfying the above conditions is marked as a pixel to be deleted;

[0077] Step two is to traverse each pixel p in the image again, and the pixel p can be considered for deletion when the following conditions are met, wherein the determination conditions are the same as those in step one, after completing steps one and two, the value of all pixels marked for deletion is set to 0, and when no pixel is marked for deletion in a certain iteration, the iteration stops, and the image obtained at this time is a single-pixel wide line image after thinning, wherein The actual image scale parameter needs to be consulted by a relevant field expert for setting, and short line segments with a length less than a certain threshold value can be noise or misidentified results, and the quality of the line is optimized by removing the short line segments.

[0078] An abnormality judgment rectangle map with the same length and width is mapped according to the size of the target area orthographic map, the identified feature line is mapped into the two-dimensional coordinates in the abnormality judgment rectangle map according to the two-dimensional coordinates of the identified feature in the target area orthographic map, the two-dimensional coordinates of each point on the target subway line are extracted from the target area orthographic map, and the two-dimensional coordinates of each point on the target subway line are mapped to the same two-dimensional coordinate position of the abnormality judgment rectangle map, a circle is taken as the range of the subway line protection zone, the center of the circle is placed on the subway line, that is, each point on the subway line is taken as a circle center, the two-dimensional coordinates of each circle center on the subway line are mapped to the same two-dimensional coordinate position of the abnormality judgment rectangle map, and a circle is drawn with each mapped circle center as the center and with the radius of the subway protection zone to obtain a preliminary division range of the subway line protection zone.

[0079] ​​It should be noted that the radius of the subway protection zone needs to be determined according to actual needs. The subway line protection zone is represented by a continuous circle and mapped to the anomaly detection rectangle. This can simplify the calculations for subsequent anomaly identification and more clearly present the actual location of the anomaly.

[0080] The initially defined protected area is corrected based on the identified features after line processing, and the orthophoto map of the target area is mapped onto the anomaly detection rectangle map. The protected area along the subway line is then obtained from the anomaly detection rectangle map. The specific steps include:

[0081] Based on the preliminary delineation of the subway line protection zone and the mapped identification feature lines in the anomaly detection rectangle, interference locations are deleted to correct the subway protection zone range, thus obtaining the subway line protection zone range. This includes the following steps:

[0082] For each point mapped onto the identification feature lines in the anomaly detection rectangle... Interference checks were performed, among which , To identify the number of points on the feature lines;

[0083] For each center point on the subway line ,in , To determine the number of centers of circles along the subway line, calculate and identify points on the feature lines. To the center distance ,when When the radius of the subway protection zone is less than or equal to the current point, it represents the current point. If a point falls within the subway protection zone, interference exists. For circular protection zones where interference occurs, their boundaries are redefined, retaining only the portion that does not interfere with the identification feature lines. The remaining protection zones are then merged to form a continuous subway line protection zone, thus obtaining the subway line protection zone.

[0084] It should be noted that when determining interference in the protected area, the circle is divided into multiple small regions. Each small region is assessed for interference with the identified feature lines, and then the interfering regions are removed. This process involves calculating the points on the identified feature lines. To the center distance It can be calculated using the Euclidean distance formula, that is... .

[0085] The GIS real-time data sharing module collects road condition information of a target area in real time through a GIS system, identifies suspicious features in the road condition information, and performs line processing on the suspicious features in the road condition information, and maps the suspicious features in real time to an abnormality judgment rectangular diagram, and specifically includes the following steps:

[0086] The road condition information of the target area is obtained in real time in the target area orthographic map through the GIS system, a neural network model is constructed through historical data, and the foundation pit soil pile and the construction vehicle in the road condition information are identified in real time, the identified foundation pit soil pile and construction vehicle are marked as suspicious features, and the coordinate range of the suspicious features is recorded, including the following steps:

[0087] The target area orthographic map and corresponding road condition information are collected in real time through the GIS system, past road condition image data and corresponding annotation information of the target area are collected, the annotation content covers the categories of foundation pit soil piles and construction vehicles, the collected historical data is divided into a training set, a validation set and a test set, a Faster R-CNN model is trained to identify suspicious features, the parameters of the model are initialized through Xavier initialization, the cross-entropy loss is selected as the loss function, the training set data is input into the model in batches, the value of the loss function is calculated, and the parameters of the model are updated using the Adam optimizer, the trained model is evaluated using the validation set and the test set, and a neural network model is constructed;

[0088] It should be noted that when evaluating the model, accuracy, recall rate and F1 value and other indicators need to be calculated, if the performance of the model does not meet the expected performance, more training data needs to be collected, or the existing data needs to be data enhanced, to improve the generalization ability of the model, until the performance of the trained model meets the standard.

[0089] The orthographic map of the target area is obtained in real time through the GIS system and input into the trained neural network model, the input orthographic map is analyzed through the neural network model, the foundation pit soil pile and the construction vehicle are identified, and are marked as suspicious features, and the coordinate range of the suspicious features is recorded according to the coordinate system in the orthographic map.

[0090] The coordinate range of the suspicious feature is transmitted to the protection area image construction module, the suspicious feature boundary in the coordinate range is identified through the Sobel operator, the identified suspicious feature boundary is thinned to a single-pixel-wide line through the Zhang-Suen thinning algorithm, and the suspicious feature is mapped to the abnormality judgment rectangular diagram according to the coordinates of the suspicious feature line in the target area orthographic map.

[0091] Embodiment 3:

[0092] The anomaly analysis module, by mapping suspicious features in the anomaly detection rectangle and combining this with the boundaries of the subway protection zone, performs anomaly detection on lines, identifying real-time and potential risks within the subway protection zone. This includes the following steps:

[0093] Based on the mapped coordinates of suspicious features in the anomaly detection rectangle, and combined with the coordinates of the protected area along the subway line, anomaly detection is performed on the current abnormal feature lines, including the following steps:

[0094] For each point on the abnormal feature line The method of ray casting is used to determine whether it is located within the polygon of the subway protection zone. Inside, through from point A ray is emitted horizontally to the right; the relationship between the ray and the polygon of the protected area is statistically analyzed. The number of intersection points of the edges;

[0095] When the number of intersection points is odd, it means that the current suspicious feature is located within the polygon of the protected area. Within the protected area, where a real-time risk exists, a red circle is drawn to mark it in the anomaly detection rectangle. When the number of intersection points is even, it means that the currently suspicious feature is not within the protected area polygon. Within this range, suspicious features are marked as pending.

[0096] For suspicious features of pending markers, based on the type of suspicious feature, at fixed time intervals... Analyze changes in suspicious characteristics to determine potential risks:

[0097] For the excavation pit soil pile, continuous statistics were collected. interval The position of the lower edge line is used to calculate the average expansion rate of the current foundation pit soil mound, and predict its position. The edge line position is determined by ray casting. When the number of intersection points is odd, the current foundation pit mound is marked as a potential risk and marked with a blue circle in the anomaly determination rectangle. When the number of intersection points is even, the current foundation pit mound is ignored.

[0098] It should be noted that, for each time point, the area of ​​the excavation pit mound is calculated using the shoelace formula based on the position of its corresponding edge line, starting from the initial time point. arrive average expansion rate ,in Represent and Area at time, Represent the total time, assuming that the foundation soil pile continues to expand at the current average expansion rate, according to the relationship between the area and the edge line, the future edge line position is obtained, the interval t is usually set to 30 minutes, N is set to 15, and can also be flexibly adjusted according to different use cases.

[0099] For construction vehicles, the edge line center point positions in the consecutive intervals are continuously counted The edge line center point positions in each two consecutive intervals are connected, and the meeting points of all the connected lines in the intervals are counted. The meeting points are determined based on the ray method, and when the intersection of more than four-fifths of the meeting points is odd, the current construction vehicle is marked as a potential risk, and is marked with a blue circle in the abnormal judgment rectangle diagram, otherwise it is ignored.

[0100] It should be noted that by continuously counting the edge line center point positions of the construction vehicle in multiple intervals, and connecting the edge line center point positions of each two consecutive intervals, the motion trajectory and change of the construction vehicle can be captured to determine whether it is a potential risk.

[0101] In the embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic, and the division of the modules is merely a logical function division. In actual implementation, another division mode can be used. The modules illustrated as separate components can be or can not be physically separate, and the components illustrated as modules can be or can not be physical units. They can be located in one place or distributed on multiple network units. According to actual needs, some or all of the modules can be selected to achieve the purpose of the method of the embodiments.

[0102] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.​

Claims

1. A spatial analysis platform for subway line protection zones based on the fusion technology of UAV orthophotos and GIS maps, characterized by: It includes a data acquisition module, a protected area image construction module, an anomaly analysis module, and a GIS real-time data sharing module; The data acquisition module obtains an orthophoto of the target area by collecting video data of the target area through a drone. Combined with the ground control points set in the target area, the module performs accuracy calibration on the obtained orthophoto by measuring coordinates and elevation using RTK, collects target subway line information, and marks the target subway line in the orthophoto of the target area. The protected area image construction module performs line processing on the identification features in the obtained target area orthophoto map, performs preliminary division of the protected area of ​​the subway line based on the subway line marked in the orthophoto map, corrects the preliminary division of the protected area based on the identification features after line processing, and maps the target area orthophoto map to the anomaly detection rectangle map to obtain the protected area along the subway line in the anomaly detection rectangle map. The GIS real-time data sharing module collects road condition information of the target area in real time through the GIS system, identifies suspicious features in the road condition information, and processes the suspicious features in the road condition information into lines and maps them to the anomaly judgment rectangle in real time. The anomaly analysis module, by mapping suspicious features in the anomaly detection rectangle and combining them with the boundaries of the subway protection zone, performs anomaly detection on lines, identifying real-time and potential risks within the subway protection zone. This includes the following steps: Based on the mapped coordinates of suspicious features in the anomaly detection rectangle, and combined with the coordinates of the protected area along the subway line, anomaly detection is performed on the current abnormal feature lines, including the following steps: For each point on the abnormal feature line , judge whether it is in the polygon of the protection area along the subway line by the ray method , shoot a ray horizontally to the right from the point , and count the number of intersections of the ray with the edges of the polygon of the protection area ; When the number of intersection points is odd, it means that the current suspicious feature is located in the protection polygon , i.e. there is real-time risk, and a red circle mark is made in the abnormality determination histogram; when the number of intersection points is even, it means that the current suspicious feature is not located in the protection polygon , i.e. there is no real-time risk, and a pending mark is made for the suspicious feature; For suspicious features to be labeled, based on the type of suspicious features, at fixed time intervals Statistical changes in suspicious features to determine potential risks: For the excavation pit soil pile, continuous statistics were collected. Interval The position of the lower edge line is used to calculate the average expansion rate of the current foundation pit soil mound, and predict its position. The edge line position is determined by ray casting. When the number of intersection points is odd, the current foundation pit mound is marked as a potential risk and marked with a blue circle in the anomaly determination rectangle. When the number of intersection points is even, the current foundation pit mound is ignored. For construction vehicles, continuously statistics The edge line center point position under each interval Connect the edge line center point positions of every two consecutive intervals, and statistics all The meeting points of the lines under all intervals, determine the meeting points based on the ray method, when the intersection of more than four-fifths of the meeting points is odd, mark the current construction vehicle as a potential risk, and mark it with a blue circle in the abnormal judgment rectangle diagram, otherwise ignore it.

2. The spatial analysis platform for the protection zone along the subway line based on the fusion of the UAV orthographic map and the GIS map according to claim 1, characterized in that: The data acquisition module specifically includes the following steps: Based on the size, shape, and terrain of the target area, a drone flight path is planned. Video images of the target area are captured using the drone's onboard camera equipment. Image matching is then performed on the video images captured by the drone to obtain an orthophoto of the target area. The specific steps include: Image matching is performed on video image data captured by drones. Corresponding points between different images are extracted using the SIFT algorithm. Then, aerial triangulation is performed to calculate the exterior orientation elements of each image, including three line elements. and three corner elements By combining the camera equipment parameters, a collinearity equation is constructed: ； in, Represents the image point coordinates. Represents the coordinates of a ground point. Represents the outer orientation line element. The exterior orientation elements are used to construct the rotation matrix. It is a rotation matrix element composed of exterior orientation elements. Represents the camera's main distance in video recording equipment. Representing the principal point coordinates, an error equation is established based on the collinearity equation to describe the relationship between the image point coordinates and the ground point coordinates, the camera interior orientation elements and exterior orientation elements. The least squares method is used to adjust the error equation and solve for the optimal solution of the exterior orientation elements. Based on the exterior orientation elements and camera interior orientation elements obtained from aerial triangulation, as well as the digital elevation model, each image is orthorectified, all orthorectified images are stitched together, and the seams of overlapping areas are processed to obtain an orthorectified map of the target area. Eight ground control points are uniformly set based on the length and width of the target area. The three-dimensional coordinates of each ground control point are accurately measured using RTK measurement equipment. The coordinates and elevation information of the measured ground control points are compared with the coordinates of the control points obtained by orthophoto calculation during image matching. The exterior orientation elements of the image are adjusted and the accuracy is calibrated by adjusting the least squares method through adjustment calculation.

3. The spatial analysis platform for subway line protection zones based on the fusion technology of UAV orthophotos and GIS maps as described in claim 1, characterized in that: The steps for collecting target subway line information and marking the target subway lines in the orthophoto map of the target area are as follows: Obtain the target subway line information through the official website of the urban rail transit system, including the locations the subway line passes through and the length of the subway line. Mark the target location line in the orthophoto map of the target area based on the locations the subway line passes through, and mark all the locations along the target subway line.

4. The spatial analysis platform for subway line protection zones based on the fusion technology of UAV orthophotos and GIS maps as described in claim 2, characterized in that: The protected area image construction module, based on the obtained target area orthophoto, performs line-based processing on the identification features in the orthophoto, and preliminarily delineates the scope of the subway line protected area based on the subway lines marked in the orthophoto. Specifically, this includes the following steps: For the obtained orthophoto of the target region, the image is converted to grayscale to reduce the amount of data. Mean filtering is used to remove noise and smooth the image. Linear contrast enhancement is then employed to highlight the recognition features. ,in Represents the pixel values ​​of the input image. Represents the pixel values ​​of the output image. These represent the adjustment parameters for contrast and brightness, respectively. Features requiring line-based processing are identified on the preprocessed orthophoto of the target region. The specific steps for performing line-based processing on the identified features in the orthophoto are as follows: The image is further smoothed using a Gaussian filter, and the Sobel operator is used to calculate the image's smoothness. gradient in direction The gradient magnitude and gradient direction are calculated, where the gradient magnitude is... gradient direction Local maxima are detected in the gradient magnitude along the gradient direction, local maxima are preserved, and other non-maxima points are suppressed. Strong edge points are identified through dual threshold detection. By setting strong and weak thresholds, pixels with gradient magnitudes greater than the strong threshold are marked as strong edge points, and pixels with gradient magnitudes less than the weak threshold are marked as weak edge points. For pixels between the strong and weak thresholds, they are determined to be weak edge points when connected to weak edge points and strong edge points when connected to strong edge points. Strong edge points and their connected weak edge points are connected to form complete edges. The Otsu algorithm is used to segment the image into different regions based on the image's grayscale values, separating the recognition features from the background. The Zhang-Suen thinning algorithm refines the identified feature regions into lines of single-pixel width. Boundary pixels are iteratively removed to obtain the lines representing the identified features. The line length is then adjusted. < The lines were removed, among which... Represents the line length threshold; Based on the size of the orthophoto map of the target area, an anomaly detection rectangle with the same length and width is mapped. Based on the two-dimensional coordinates of the identified features in the orthophoto map of the target area, the identified feature lines are mapped to the two-dimensional coordinates in the anomaly detection rectangle. The two-dimensional coordinates of each point on the target subway line are extracted from the orthophoto map of the target area. The two-dimensional coordinates of each point on the target subway line are mapped to the same two-dimensional coordinate position in the anomaly detection rectangle. A circle is used as the scope of the subway line protection zone. The center of the circle is placed on the subway line. That is, each point on the subway line is used as a center. The two-dimensional coordinates of each center of the circle on the subway line are mapped to the same two-dimensional coordinate position in the anomaly detection rectangle. In the anomaly detection rectangle, a circle is drawn with each mapped center as the center and the radius of the subway protection zone, thus obtaining the preliminary division of the subway line protection zone.

5. The spatial analysis platform for subway line protection zones based on the fusion technology of UAV orthophotos and GIS maps as described in claim 4, characterized in that: The process of refining the initially defined protected area based on the identified features after line processing, and mapping the orthophoto map of the target area onto the anomaly detection rectangle to obtain the protected area along the subway line from the anomaly detection rectangle, specifically includes the following steps: Based on the preliminary delineation of the subway line protection zone and the mapped identification feature lines in the anomaly detection rectangle, interference locations are deleted to correct the subway protection zone range, thus obtaining the subway line protection zone range. This includes the following steps: For each point mapped onto the identification feature lines in the anomaly detection rectangle... Interference checks were performed, among which , To identify the number of points on the feature lines; For each center point on the subway line ,in , To determine the number of centers of circles along the subway line, calculate and identify points on the feature lines. To the center distance ,when When the radius of the subway protection zone is less than or equal to the current point, it represents the current point. If a point falls within the subway protection zone, interference exists. For circular protection zones where interference occurs, their boundaries are redefined, retaining only the portion that does not interfere with the identification feature lines. The remaining protection zones are then merged to form a continuous subway line protection zone, thus obtaining the subway line protection zone.

6. The spatial analysis platform for subway line protection zones based on the fusion technology of UAV orthophotos and GIS maps as described in claim 5, characterized in that: The GIS real-time data sharing module collects road condition information of the target area in real time through the GIS system, processes suspicious features in the road condition information into lines, and maps them onto the orthophoto map of the target area in real time. Specifically, it includes the following steps: In the orthophoto map of the target area, the road condition information of the target area is obtained in real time through the GIS system. A neural network model is built through historical data to identify the foundation pits and construction vehicles in the road condition information in real time. The identified foundation pits and construction vehicles are marked as suspicious features, and the coordinate range of the suspicious features is recorded. The coordinate range of suspicious features is transmitted to the protected area image construction module. The Sobel operator is used to identify the boundaries of suspicious features within the coordinate range. The Zhang-Suen thinning algorithm is used to refine the identified suspicious feature boundaries into lines with a width of one pixel. Based on the coordinates of the suspicious feature lines in the orthophoto of the target area, the suspicious features are mapped to the anomaly determination rectangle.

7. The spatial analysis platform for subway line protection zones based on the fusion technology of UAV orthophotos and GIS maps as described in claim 6, characterized in that: The process of acquiring real-time road condition information of the target area through a GIS system in the orthophoto map of the target area, constructing a neural network model using historical data, identifying excavation pits and construction vehicles in the road condition information in real time, marking the identified excavation pits and construction vehicles as suspicious features, and recording the coordinate range of the suspicious features includes the following steps: The target area's orthophoto map and corresponding road condition information are collected in real time through a GIS system. Past road condition image data and corresponding annotation information of the target area are collected. The annotation content covers the categories of foundation pits, soil piles, and construction vehicles. The collected historical data is divided into training set, validation set, and test set. The Faster R-CNN model is trained to identify suspicious features. The model parameters are initialized through Xavier initialization. Cross-entropy loss is selected as the loss function. The training set data is input into the model in batches. The value of the loss function is calculated. The model parameters are updated using the Adam optimizer. The trained model is evaluated using the validation set and test set to build a neural network model. The orthophoto map of the target area is acquired in real time through the GIS system and input into a trained neural network model. The neural network model analyzes the input orthophoto map, identifies the foundation pit mounds and construction vehicles, and marks them as suspicious features. The coordinate range of the suspicious features is recorded according to the coordinate system in the orthophoto map.

Images