Historical and cultural city data acquisition, analysis and evaluation method based on eagle eye inspection

Abstract

The invention relates to the technical field of image information processing, in particular to a famous historical and cultural city data acquisition, analysis and evaluation method based on eagle eye inspection, which comprises the following steps of: distributing shooting points in a famous historical and cultural city protection area, planning a flight path, acquiring historical building images in each period, and extracting edge contours of historical buildings in each period to obtain historical and cultural city data; according to the method, the continuity evaluation coefficient of the edge contour of the historical building in each period is evaluated, the damage area of the historical building can be more accurately positioned by dividing the building into a building body and an accessory structure, the damage area image of the historical building in each period is detected and output, and the damage marking graph of the historical building in each period is generated; by evaluating the health evaluation index of the historical building in each period and determining and feeding back the reduction degree grade of the historical building in each period, the health condition and change trend of the historical building can be known in time, so that protection strategies and measures can be adjusted in time to deal with possible problems.

Description

Technical Field

[0001] The present invention relates to the technical field of image information processing. Specifically, it is a method for collecting, analyzing, and evaluating data of historical and cultural cities based on eagle-eye inspection. Background Art

[0002] As a precious heritage of human civilization, historical and cultural cities carry rich historical memories and cultural connotations. However, with the passage of time and the development of cities, these ancient buildings are facing multiple threats such as natural erosion, human damage, and environmental changes. Traditional monitoring methods such as manual inspection and satellite remote sensing have problems such as low efficiency, insufficient accuracy, and difficulty in capturing subtle changes. Although three-dimensional laser scanning technology can obtain high-precision data, it has high equipment costs, complex operations, and poor real-time performance, making it difficult to meet the large-scale and dynamic monitoring needs. There is an urgent need for more efficient and accurate technical means to protect and manage them.

[0003] The existing Chinese patent application with the application number 201711460075.8 discloses a method, device, server, and storage medium for detecting building changes. By comparing the outlines of buildings in the current satellite map with those in historical satellite images, this solution solves the consumption of time and investment of manpower in manually verifying building changes on-site, realizes the automatic detection of building changes, improves the detection efficiency of building changes and the user experience, and reduces the investment of manpower in later map updates.

[0004] However, there are the following problems in the above patent: This solution mainly focuses on map data updates, judging macroscopic changes such as the construction and demolition of buildings, and does not involve the attention to aspects such as building damage and appearance changes, making it difficult to meet the demand for comprehensive monitoring of buildings in some scenarios.

[0005] The existing Chinese patent application with the application number 201810219932.3 discloses an ancient building health monitoring method and device based on three-dimensional laser scanning technology. By obtaining the structural coordinate information and material information of ancient buildings, comprehensively analyzing various parameters reflecting the health of ancient buildings, and then judging each parameter separately and comprehensively comparing to obtain the health status of ancient buildings. When obtaining the structural coordinate information of ancient buildings, it is obtained separately through the first acquisition unit and the second acquisition unit, realizing the acquisition of detailed structural coordinate information of ancient buildings and overcoming the deficiency of dead angles that cannot be scanned by the scanner.

[0006] However, there are the following problems in the above patent: First, this solution uses three-dimensional laser scanning to scan ancient buildings, but three-dimensional laser needs to scan ancient buildings at close range. For large ancient building groups or ancient buildings with complex spaces, the movement of equipment and data collection takes a long time, making it difficult to quickly cover large areas or multiple buildings.

[0007] Second, this solution reconstructs the three-dimensional model and the standard model through surface algorithms and finite element analysis. The data processing process is complex, the amount of calculation is large, and the hardware equipment requirements are high. It mainly focuses on the structural coordinates and material information of ancient buildings, and lacks analysis and quantitative evaluation of building damage. Summary of the invention

[0008] In order to overcome the shortcomings of the background technology, an embodiment of the present invention provides a method for collecting, analyzing and evaluating data of historical and cultural cities based on Eagle Eye inspection, which can effectively solve the problems involved in the above-mentioned background technology.

[0009] The purpose of the present invention can be achieved through the following technical solutions: a method for data collection, analysis and evaluation of historical and cultural cities based on Eagle Eye inspection, the method comprising the following steps: S1. According to the distribution of each shooting point in the protection area of ​​the historical and cultural city, plan the flight path and obtain the images of historical buildings in each period.

[0010] S2. Extract the preliminary edge contours of historical buildings of each period through threshold segmentation, obtain the edge contours of historical buildings of each period using contour tracking algorithm, and remove invalid contours.

[0011] S3. Evaluate the continuity coefficient of the edge contours of historical buildings of each period.

[0012] S4. Based on the gray-level co-occurrence matrix and clustering algorithm, the building image area in the historical building images of each period is divided into the building body and the auxiliary structure.

[0013] S5. Use the deep learning semantic segmentation model to detect the damaged areas of historical buildings, output the damaged area images of historical buildings in each period, and then calculate and generate the damage marking map of historical buildings in each period.

[0014] S6. Evaluate the health evaluation index of historical buildings of each period, determine the degree of reduction of historical buildings of each period and provide feedback.

[0015] Preferably, the specific analysis method of step S1 is: using map software to draw the boundary of the historical and cultural city protection area on the map, within the boundary, select a number of historical buildings as shooting points according to the distribution of historical buildings, and plan the flight path of the drone according to the distribution of each shooting point.

[0016] The inspection cycle is set. In each cycle, the drone is equipped with the Eagle Eye device to fly along the planned flight path. During the flight, the optical image of each shooting point is synchronously collected by the camera, which is recorded as the image of each shooting point in each cycle. The images of each shooting point in each cycle are integrated through image stitching to obtain a complete image covering the protection area of ​​the historical and cultural city, which is recorded as the image of the historical buildings in each cycle.

[0017] Preferably, the drone is also set with an inspection cycle, which includes a start time, an end time, and a time interval. The drone automatically conducts periodic inspections along its flight path according to the set inspection cycle.

[0018] Preferably, the specific analysis method for the edge contours of each cycle of historical buildings is as follows: Set a grayscale value threshold, and segment each cycle of historical building images into a building area and a background area through threshold segmentation to generate a binary image of each cycle of historical buildings. Calculate the gradient magnitude and direction of the binary image of each cycle of historical buildings, and perform non-maximum suppression on the binary image of each cycle of historical buildings based on this.

[0019] Set a high threshold and a low threshold for the gradient magnitude respectively. Traverse each pixel point of the binary image of each cycle of historical buildings after non-maximum suppression. Mark the pixel points with a gradient magnitude greater than the high threshold as strong edge points, mark the pixel points with a gradient magnitude between the low threshold and the high threshold as weak edge points, and mark the pixel points with a gradient magnitude less than the low threshold as non-edge points.

[0020] If a weak edge point is connected to at least one strong edge point, then retain the weak edge point as an edge point; otherwise, regard it as a noise point and remove it from the edge. By connecting the edge points that are higher than the low threshold and connected to the edges higher than the high threshold, obtain the preliminary edge contours of each cycle of historical buildings.

[0021] Preferably, the specific analysis method for the edge contours of each cycle of historical buildings is as follows: If there are discrete edge pixel points in the preliminary edge contours of each cycle of historical buildings, use a contour tracking algorithm to start from any discrete edge pixel point, search for edge pixel points in its neighborhood in a fixed direction to form a contour sequence until all edge pixels are traversed, thereby constructing the complete edge contours of each cycle of historical buildings and removing the invalid contours in the edge contours of each cycle of historical buildings.

[0022] Preferably, the specific analysis method for step S3 is as follows: Set a breakpoint judgment threshold, and calculate the Euclidean distance between each pair of adjacent pixel points in the contour sequence of the edge contours of each cycle of historical buildings respectively. If the Euclidean distance between a pair of adjacent pixel points in the edge contour of a certain cycle of historical buildings is greater than the breakpoint judgment threshold, then judge that there is a breakpoint at this position in the edge contour of this cycle of historical buildings, increment the breakpoint count by 1, and traverse the contour sequence of the edge contours of each cycle of historical buildings in this way. The obtained breakpoint count is the number of breakpoints in the edge contours of each cycle of historical buildings.

[0023] Starting from the starting point of the edge contour of each cycle of historical buildings, sequentially search for the breakpoint positions. Whenever a breakpoint is detected, divide the contour part before the breakpoint into a segment, record the starting and ending point coordinates of the segment, calculate the number of pixel points contained in the segment, and traverse the edge contours of each cycle of historical buildings in this way to obtain each contour segment of the edge contours of each cycle of historical buildings and their corresponding lengths, and calculate the proportion of the number of segments in different length intervals respectively.

[0024] Regard the edge contours of each cycle of historical buildings as composed of a series of line segments connected end to end. Use the distance formula between two points to sequentially calculate the lengths of the line segments between adjacent pixel points, and accumulate the lengths of all line segments to obtain the perimeter of the edge contours of each cycle of historical buildings. At the same time, calculate the area of the region enclosed by the edge contours of each cycle of historical buildings, and obtain the ratio of the perimeter to the area of the edge contours of each cycle of historical buildings through calculation.

[0025] Based on the number of breakpoints of the edge contours of each cycle of historical buildings, the proportion of the number of segments in different length intervals, and the ratio of the perimeter to the area of the edge contours, comprehensively evaluate to obtain the continuity evaluation coefficient of the edge contours of each cycle of historical buildings.

[0026] Preferably, the specific analysis method of step S4 is as follows: Slide and traverse on the images of each cycle of historical buildings with a window of a fixed pixel size, and the sliding distance each time is one pixel until the entire historical building image is covered. For each sliding window, set the distance parameter and direction parameter respectively, and calculate the gray-level co-occurrence matrix based on the distance parameter and direction parameter.

[0027] Extract texture feature parameters from the gray-level co-occurrence matrix, generate a set of 4D feature vectors according to the four extracted texture feature parameters in each direction, and combine the feature vectors in the four directions to form a 16D texture feature vector.

[0028] Set the number of clusters to 2, corresponding to the building body and the attached structure respectively. Randomly initialize 2 cluster centers, calculate the Euclidean distance between each 16D texture feature vector and the cluster centers, and assign each texture feature vector to the category of the cluster center with the closest distance.

[0029] Recalculate the mean value of the texture feature vectors in each cluster, and use this mean value as the new cluster center. Repeat the cluster assignment step and the cluster center update operation in this step until the cluster centers no longer change, so as to divide the building image areas in the images of each cycle of historical buildings into two categories: the building body and the attached structure.

[0030] Preferably, the specific analysis method for the historical building damage area images of each period is as follows: construct a deep learning semantic segmentation model, obtain a large number of building image data labeled with damage areas for training, input the historical building images of each period into the trained deep learning semantic segmentation model, and the model outputs the probability that each pixel of the historical building images of each period belongs to the damage area. By setting a threshold, the probability that each pixel belongs to the damage area is converted into a binary image, where 1 represents the damage area and 0 represents the non-damage area. Traverse the damage area prediction result image, record the coordinates of all pixels with a value of 1, and obtain the historical building damage area images of each period.

[0031] Preferably, the specific operation method for generating the damage marking maps of the historical buildings of each period is as follows: according to the division results of the building body and the attached structures of the historical building images of each period, respectively screen the pixels belonging to the building body and the attached structures in the historical building damage area images of each period, and calculate the damage parameters of the building body and the attached structures of the historical buildings of each period accordingly. Each damage parameter includes the area, shape, and pixel density of the damage area.

[0032] Create a blank image with the same size as the historical building images of each period, use different colors to mark the damage parameters of the building body and the attached structures of the historical buildings of each period, and obtain the damage marking maps of the historical buildings of each period by superimposing and displaying them with the historical building images of each period.

[0033] Preferably, the specific analysis method for step S6 is as follows: assign corresponding weights to the building body and the attached structures respectively, calculate the comprehensive loss evaluation coefficients of the building body and the attached structures of the historical buildings of each period based on the damage parameters of the building body and the attached structures in the damage marking maps of the historical buildings of each period, and evaluate and obtain the health evaluation index of the historical buildings of each period in combination with the continuity evaluation coefficient of the edge contours of the historical buildings of each period. By comparing the health evaluation indices of adjacent historical buildings, obtain the health reduction degree of the historical buildings of each period.

[0034] Set the grading threshold for the health reduction degree, compare the health reduction degree of the historical buildings of each period with it, obtain the reduction degree level of the historical buildings of each period, and feedback it to the system.

[0035] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: First, through the distribution of each shooting point in the historical and cultural city protection area, the present invention plans the flight path and obtains the historical building images of each period, which can ensure comprehensive image acquisition and avoid missing important buildings or key parts of buildings.

[0036] II. By extracting the edge contours of historical buildings in each period and evaluating the continuity evaluation coefficient of the edge contours of historical buildings in each period, the change situation of the integrity and stability of the building structure can be intuitively reflected.

[0037] III. By dividing into the building body and accessory structures, the damage area of historical buildings can be more accurately located, the damage area images of historical buildings in each period can be detected and output, and the damage marking maps of historical buildings in each period can be generated, so as to accurately locate the damage.

[0038] IV. By evaluating the health evaluation index of historical buildings in each period, determining the reduction degree level of historical buildings in each period and giving feedback, the health status and change trend of historical buildings can be understood in time, so as to adjust the protection strategies and measures in time to deal with possible problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention will be further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to the following drawings without creative work.

[0040] Figure 1 It is a schematic flow chart of the method of the present invention.

[0041] Figure 2 is Figure 1 a schematic flow chart of step S2 in

[0042] Figure 3 is Figure 1 a schematic flow chart of step S3 in DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present invention.

[0044] Please refer to Figure 1 as shown, the present invention provides a method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection, and the method includes the following steps: S1. According to the distribution of each shooting point in the protection area of the historical and cultural city, plan the flight path and obtain the historical building images in each period.

[0045] The specific analysis method for step S1 is as follows: Use a map software to draw the boundary of the historical and cultural city protection area on the map. Within the boundary, select several historical buildings as each shooting point according to the distribution of historical buildings, and plan the flight path of the drone according to the distribution of each shooting point; ensure that the drone flight path covers all key historical buildings to prevent data omission, provide a complete and comprehensive data basis for subsequent analysis and evaluation, make the collected data more targeted and effective, and at the same time optimize the drone flight trajectory, reduce unnecessary flights, and improve the data collection efficiency.

[0046] Set the inspection cycle. In each cycle, the drone is equipped with an eagle eye device and flies along the planned flight path. During the flight, the optical images of each shooting point are synchronously collected through the camera, recorded as the images of each shooting point in each cycle, and the images of each shooting point in each cycle are integrated through image stitching to obtain a complete image covering the historical and cultural city protection area, recorded as the historical building images in each cycle; setting the inspection cycle can realize the long-term dynamic monitoring of historical buildings, timely capture the subtle changes generated by the buildings over time, and the image stitching technology integrates the scattered shooting point images into a complete image, presenting the overall style of the buildings within the historical and cultural city protection area, facilitating the analysis of the state of historical buildings from both macroscopic and microscopic levels, and providing detailed visual materials for protection decision-making.

[0047] The drone is also set with an inspection cycle, which includes a start time, an end time, and a time interval. The drone automatically conducts periodic inspections along the flight path of the drone according to the set inspection cycle; the automatic inspection mechanism greatly reduces the labor cost, eliminates the need for manual frequent operation of the drone to perform inspection tasks, reduces the possibility of human operation errors, and the drone flies automatically in strict accordance with the set inspection cycle and time nodes, ensuring the regularity and stability of data collection, making the data of different cycles comparable, facilitating the establishment of time series data on the state changes of historical buildings, thus more accurately analyzing the evolution trend of the building state, timely discovering potential problems, and providing strong support for the preventive protection of historical buildings.

[0048] S2. Extract the preliminary edge contours of historical buildings in each cycle through threshold segmentation, obtain the edge contours of historical buildings in each cycle using the contour tracking algorithm, and perform invalid contour elimination.

[0049] Please refer to Figure 2As shown in the figure, the specific analysis method for the edge contour of each cycle of historical buildings is as follows: Set the grayscale value threshold, segment each cycle of historical building images into building areas and background areas through threshold segmentation to generate binary images of each cycle of historical buildings, calculate the gradient magnitude and direction of the binary images of each cycle of historical buildings, and perform non-maximum suppression on the binary images of each cycle of historical buildings; The generation of binary images makes the main structure of the building clearer and facilitates the subsequent accurate recognition of building edges.

[0050] It should be noted that the specific analysis method for the non-maximum suppression is as follows: Starting from the top-left pixel of the binary image of each cycle of historical buildings, traverse each pixel point row by row and column by column. For each pixel point traversed, determine the gradient direction interval where it is located according to the gradient magnitude and direction of the binary image of each cycle of historical buildings. On the gradient direction of this pixel point, compare its gradient magnitude with that of adjacent pixel points. If the gradient magnitude of the current pixel point is greater than the gradient magnitudes of the two adjacent pixel points in the gradient direction, retain this pixel point as an edge point; otherwise, judge that this pixel point is not an edge point and set its gradient magnitude to 0.

[0051] Set high and low thresholds for the gradient magnitude respectively. Traverse each pixel point of the binary image of each cycle of historical buildings after non-maximum suppression. Mark the pixel points with gradient magnitudes greater than the high threshold as strong edge points, mark the pixel points with gradient magnitudes between the low threshold and the high threshold as weak edge points, and mark the pixel points with gradient magnitudes less than the low threshold as non-edge points; The setting of double thresholds provides a more detailed standard for the classification of edge pixels. Strong edge points have higher gradient magnitudes and are the more obvious and definite edge parts in the image, which can be used as the core basis for edge detection. Although the gradient magnitudes of weak edge points are relatively low, they may be the extensions of strong edges or some weaker but still meaningful edge information. By marking them separately, it is possible to further determine whether they are real edges in subsequent processing. The marking of non-edge points helps to exclude noise and interference information in the image, making subsequent edge connection and contour extraction more accurate and efficient, and improving the reliability of edge detection.

[0052] If a weak edge point is connected to at least one strong edge point, retain this weak edge point as an edge point; otherwise, regard it as a noise point and remove it from the edges. By connecting the edge points that are higher than the low threshold and connected to the edges higher than the high threshold, obtain the preliminary edge contour of each cycle of historical buildings; The acquisition of the preliminary edge contour provides a basic framework for the subsequent detailed analysis of historical buildings, can reflect the general shape and structural characteristics of the building, and helps to further identify the damage, deformation, etc. of the building.

[0053] The specific analysis method for the edge contour of each cycle of historical buildings is as follows: If there are discrete edge pixel points in the preliminary edge contour of each cycle of historical buildings, the contour tracking algorithm is used to start from any discrete edge pixel point and search for edge pixel points in its neighborhood in a fixed direction to form a contour sequence until all edge pixels are traversed, thereby constructing a complete edge contour of each cycle of historical buildings and removing invalid contours from the edge contour of each cycle of historical buildings; The contour tracking algorithm can effectively handle discrete edge pixel points existing in the preliminary edge contour. By searching for edge pixel points in the neighborhood starting from any discrete point, the scattered edge pixels are connected into a complete contour sequence, making the edge contour of the historical building more coherent and complete, and being able to accurately reflect the actual shape of the building. Removing invalid contours can eliminate unnecessary edge information generated by noise, image interference, or misdetection, making the finally obtained edge contour more concise and accurate.

[0054] It should be noted that the specific analysis method of the contour tracking algorithm is as follows: Traverse the binary image after edge detection, scan pixel points row by row and column by column starting from the upper left corner of the image, select the first detected edge pixel point as the starting point and record its coordinates. The eight-direction chain code is used to assign numbers 0-7 to the 8 neighborhood directions centered on the current pixel point.

[0055] In the eight-direction chain code, 0 represents the positive right direction, 1 represents the upper right direction, 2 represents the positive upper direction, 3 represents the upper left direction, 4 represents the positive left direction, 5 represents the lower left direction, 6 represents the positive lower direction, and 7 represents the lower right direction.

[0056] Centered on the starting point, sequentially check its 8-neighborhood pixels in the preset direction. If the pixel in the 0 direction is an edge pixel, use it as the next contour point, record the coordinate information, and set this point as the current point. If the pixel in the 0 direction is not an edge pixel, sequentially check the pixel in the 1 direction, and so on until the next edge pixel point is found.

[0057] Use the newly found edge pixel point as the current point, repeat the above operation, continuously search for the next edge pixel point and add it to the contour sequence. When the coordinates of the found pixel point are exactly the same as the coordinates of the starting point, it is judged that a complete contour tracking is completed, record the coordinate information of each pixel point, and form a complete contour sequence.

[0058] It should be noted that during the tracking process, to avoid repeatedly accessing the traversed pixel points, a marker matrix is set, initialized as a matrix of all 0s. Whenever a pixel point is accessed, the value of the corresponding position in the marker matrix is set to 1. When searching for neighborhood pixels, skip the marked pixel points to prevent falling into an infinite loop.

[0059] It should be noted that the specific analysis method for removing the invalid contours from the edge contours of historical buildings in each period is as follows: Set the distance threshold between the starting point and the ending point of the contour and the minimum pixel point number threshold of the contour. For the contour sequence obtained by contour tracking in the edge contours of historical buildings in each period, obtain their starting point coordinates and ending point coordinates respectively, and use the Euclidean distance formula to calculate the distance between the starting point coordinates and the ending point coordinates. If the distance between the starting point coordinates and the ending point coordinates in the edge contour of a certain period of historical building is less than or equal to the set ending distance threshold, it is determined that the edge contour of this period of historical building is close to being closed in terms of spatial position; otherwise, it is determined that the edge contour of this period of historical building is an unclosed contour.

[0060] Count the number of pixel points included in the contour sequence of the edge contours of historical buildings in each period. If the number of pixel points in the edge contour of a certain period of historical building is greater than or equal to the set minimum pixel point number threshold of the contour, it is determined that the edge contour of this period of historical building is a valid contour; otherwise, it is determined that the number of pixel points in the edge contour of the historical building in this period is too small and it is an invalid contour, and it is removed.

[0061] For unclosed contours, calculate the straight-line equation between the starting point and the ending point of the contour. According to the straight-line equation, evenly insert a certain number of pixel points between the starting point and the ending point to close the contour, and obtain the newly generated closed contour. If the number of pixel points in the newly generated closed contour still cannot meet the requirement of being greater than or equal to the minimum pixel point number threshold of the contour, it is determined that this contour is an invalid contour and it is removed.

[0062] S3. Evaluate the continuity evaluation coefficient of the edge contours of historical buildings in each period.

[0063] Please refer to Figure 3 As shown, the specific analysis method for step S3 is as follows: Set the breakpoint judgment threshold, and calculate the Euclidean distance between each pair of adjacent pixel points in the contour sequence of the edge contours of historical buildings in each period respectively. If the Euclidean distance between a certain pair of adjacent pixel points in the edge contour of a certain period of historical building is greater than the breakpoint judgment threshold, it is determined that there is a breakpoint at this position in the edge contour of this period of historical building, and the breakpoint count is incremented by 1. Traverse the contour sequence of the edge contours of historical buildings in each period in this way, and the obtained breakpoint count is the number of breakpoints of the edge contours of historical buildings in each period; By setting a threshold to quantify the distance between adjacent pixel points, the breakpoints in the edge contours of historical buildings can be accurately identified, and the discontinuous situation of the contour can be digitally expressed. The number of breakpoints can intuitively reflect the degree of breakage of the edge contour, helping to quickly judge whether there are damages, deformations, etc. in the structure of historical buildings, and providing a key indicator for subsequent evaluation of the building state.

[0064] Starting from the starting point of the edge contour of the historical buildings of each period, the breakpoint positions are searched in sequence. Whenever a breakpoint is detected, the contour part before the breakpoint is divided into a segment, and the coordinates of the starting and ending points of the segment are recorded. The number of pixels contained in the segment is calculated, and the edge contours of the historical buildings of each period are traversed to obtain the edge contour segments of the historical buildings of each period and their corresponding lengths, and the proportion of the number of segments in different length intervals is calculated respectively. This helps to discover subtle differences in the edge contours of buildings, capture the morphological evolution trends of historical buildings in different periods, and provide data support for studying the impact of factors such as building aging and environmental erosion on architectural morphology.

[0065] The edge contour of each period historical building is regarded as consisting of a series of line segments connected end to end. The distance formula between two points is used to calculate the length of the line segments between adjacent pixel points in sequence, and all the line segment lengths are accumulated to obtain the perimeter of the edge contour of each period historical building. At the same time, the area of ​​the area enclosed by the edge contour of each period historical building is calculated. The ratio of the perimeter to the area of ​​the edge contour of each period historical building is obtained by calculation; the perimeter reflects the total length of the building edge, and the area reflects the spatial range occupied by the building. The ratio of the two can be used as an indicator to measure the compactness or complexity of the building.

[0066] The continuity evaluation coefficient of the edge contour of historical buildings in each period is obtained through a comprehensive evaluation based on the number of breakpoints in the edge contour of historical buildings in each period, the proportion of the number of fragments in different length intervals, and the ratio of the perimeter of the edge contour to the area; the continuity evaluation coefficient is generated by integrating multi-dimensional indicators, which avoids the one-sidedness of a single indicator and can comprehensively and objectively evaluate the continuity of the edge contour of historical buildings.

[0067] In a preferred embodiment of the present invention, the specific analysis method of the continuity evaluation coefficient of the edge contour of each period of historical buildings is as follows: extract the number of breakpoints of the edge contour of each period of historical buildings, the proportion of the number of fragments in different length intervals, and the ratio of the edge contour perimeter to the area, and then sum them up according to the weights to obtain the continuity evaluation coefficient of the edge contour of each period of historical buildings.

[0068] For example, the weights corresponding to the number of breakpoints in the edge contour of historical buildings of each period, the proportion of the number of fragments in different length intervals, and the ratio of the edge contour perimeter to the area are: .

[0069] S4. Based on the gray-level co-occurrence matrix and clustering algorithm, the building image area in the historical building images of each period is divided into the building body and the auxiliary structure.

[0070] The specific analysis method for step S4 is as follows: On each periodic historical building image, a window with a fixed pixel size is used for sliding traversal, and the sliding distance each time is one pixel until the entire historical building image is covered. For each sliding window, distance parameters and direction parameters are respectively set, and a gray-level co-occurrence matrix is calculated based on the distance parameters and direction parameters; by respectively setting distance parameters and direction parameters to calculate the gray-level co-occurrence matrix, the gray-level co-occurrence matrix can reflect the texture features of the image at different distances and directions, which helps to capture various complex texture patterns in the historical building image and provides a rich data basis for subsequent texture feature analysis.

[0071] It should be noted that the specific analysis method for the gray-level co-occurrence matrix is as follows: In the horizontal direction, for two pixels with a distance of , if their gray values are respectively , then the value at the corresponding position in the matrix is incremented by 1. The specific implementation steps are as follows: Traverse each pixel in the window , check whether it is within the window. If it is within the window, obtain the gray values of these two pixels , and increment the value of by 1.

[0072] In the 45° direction, for two pixels with a distance of , if their gray values are respectively , then the value at the corresponding position in the matrix is incremented by 1.

[0073] In the vertical direction, for two pixels with a distance of , if their gray values are respectively , then the value at the corresponding position in the matrix is incremented by 1.

[0074] In the 135° direction, for two pixels with a distance of , if their gray values are respectively , then the value at the corresponding position in the matrix is incremented by 1.

[0075] After calculating the occurrence times of all pixel pairs, the gray-level co-occurrence matrix is normalized so that the sum of its elements is 1.

[0076] Texture feature parameters are extracted from the gray-level co-occurrence matrix, a set of 4-dimensional feature vectors are generated in each direction based on the four extracted texture feature parameters, and the feature vectors in the four directions are combined to form a 16-dimensional texture feature vector; the texture features in different directions may be different, and multi-directional feature extraction can better capture these differences.

[0077] The number of clusters is set to 2, corresponding to the building body and the ancillary structure respectively, and two cluster centers are randomly initialized. The Euclidean distance between each of the 16-dimensional texture feature vectors and the cluster center is calculated, and each texture feature vector is assigned to the category to which the cluster center is closest. The classification method based on Euclidean distance is simple and intuitive, and can quickly make a preliminary division of the image area, thereby improving the efficiency of classification, and can reflect the differences in texture features between the building body and the ancillary structure to a certain extent.

[0078] The mean of the texture feature vectors in each cluster is recalculated, and the mean is used as the new cluster center. The cluster assignment step and the cluster center update operation of this step are repeated until the cluster center no longer changes, thereby dividing the building image area in each period of historical building images into two categories: building body and auxiliary structure; through multiple iterations, each texture feature vector can be accurately assigned to the appropriate category, thereby achieving accurate division of the building body and auxiliary structure in the historical building image.

[0079] S5. Use the deep learning semantic segmentation model to detect the damaged areas of historical buildings, output the damaged area images of historical buildings in each period, and then calculate and generate the damage marking map of historical buildings in each period.

[0080] The specific analysis method of the damaged area images of historical buildings in each period is as follows: constructing a deep learning semantic segmentation model, obtaining a large amount of building image data marked with damaged areas for training, inputting the historical building images of each period into the trained deep learning semantic segmentation model, the model outputs the probability that each pixel of the historical building image of each period belongs to the damaged area, and converts the probability of each pixel belonging to the damaged area into a binary image by setting a threshold, wherein 1 represents a damaged area and 0 represents a non-damaged area, traversing the damaged area prediction result image, recording the coordinates of all pixel points with a value of 1, and obtaining the damaged area images of the historical buildings in each period; avoiding the subjectivity and errors of manual judgment, and realizing the pixel-level positioning and extraction of the damaged area.

[0081] The specific operation method for generating the damage marking maps of historical buildings in each period is as follows: According to the division results of the building body and attached structures in the historical building images in each period, each pixel belonging to the building body and attached structures in the damage area images of historical buildings in each period is screened respectively, and each damage parameter of the building body and attached structures of historical buildings in each period is calculated accordingly. Each damage parameter includes the area, shape, and pixel density of the damage area. By making a vertical comparison of the damage parameters in different periods, the damage development trend can be clearly presented.

[0082] It should be noted that the specific analysis method for screening each pixel belonging to the building body and attached structures in the damage area images of historical buildings in each period is as follows: Based on the existing division results of the building body and attached structures, logical AND operations are respectively performed between them and the damage area images of historical buildings in each period. In the logical AND operation, the result is 1 only when the corresponding pixel positions of the building body / attached structure division image and the damage area image are both 1, and the result is 0 in other cases. In this way, the pixels belonging to the damage area in the building body and attached structures are screened out respectively.

[0083] The calculation method for the damage area is as follows: Count the number of pixels with a value of 1 in the pixels of the damage area of the building body, multiply it by the actual area represented by a single pixel to obtain the area of the damage area of the building body. Using the same method, count the number of pixels with a value of 1 in the pixels of the damage area of the attached structure, and then multiply it by the actual area of a single pixel to obtain the area of the damage area of the attached structure.

[0084] The calculation method for the damage shape is as follows: For the binary images of the damage areas of the building body and attached structures, the edge detection algorithm is used to extract the contours of their respective damage areas respectively. The approximate perimeters of the damage areas of the building body and attached structures are obtained by accumulating the distances between the contour points on the contours of the damage areas of the building body and attached structures. Using the formula calculate the circularity of the damage areas of the building body and attached structures respectively. The closer the circularity is to 1, the closer the shape of the damage area is to a circle, and the smaller the value, the more irregular the shape.

[0085] Find the circumscribed rectangles of the contours of the damage areas of the building body and attached structures respectively, measure the length and width of the circumscribed rectangles, and calculate the ratio of the length to the width to obtain the aspect ratio of the damage areas of the building body and attached structures.

[0086] The calculation method of the pixel density of the damaged area is as follows: Statistically calculate the total area of the building body (obtained by multiplying the number of pixels with a value of 1 in the building body division image by the actual area of a single pixel), divide the number of pixels in the damaged area of the building body by the total area of the building body to obtain the pixel density of the damaged area of the building body. Similarly, statistically calculate the total area of the accessory structure (obtained by multiplying the number of pixels with a value of 1 in the accessory structure division image by the actual area of a single pixel), divide the number of pixels in the damaged area of the accessory structure by the total area of the accessory structure to obtain the pixel density of the damaged area of the accessory structure.

[0087] Create a blank image with the same size as the historical building images of each period, use different colors to mark each damage parameter of the building body and accessory structure of the historical building of each period, and obtain a damage marking map of the historical building of each period by superimposing and displaying it with the historical building images of each period; superimpose the damage parameters marked with colors on the original image, convert the abstract damage data into intuitive visual graphics, which can quickly locate the damage position, understand the damage distribution and severity, and greatly improve the information transmission efficiency.

[0088] S6. Evaluate the health evaluation index of the historical building of each period, determine the reduction degree level of the historical building of each period and give feedback.

[0089] The specific analysis method of the step S6 is as follows: Assign corresponding weights to the building body and the accessory structure respectively. Based on each damage parameter of the building body and the accessory structure in the damage marking map of the historical building of each period, calculate the comprehensive loss evaluation coefficients of the building body and the accessory structure of the historical building of each period respectively. Combine the continuity evaluation coefficient of the edge contour of the historical building of each period to evaluate and obtain the health evaluation index of the historical building of each period. By comparing the health evaluation indexes of adjacent historical buildings, obtain the health reduction degree of the historical building of each period; it can clearly understand the health change trend of the historical building in different time stages, and help to timely discover the rapid deterioration or abnormal change of the building health condition.

[0090] In a preferred embodiment of the present invention, the specific analysis method of the health evaluation index of the historical building of each period is as follows: Extract the comprehensive loss evaluation coefficients of the building body and the accessory structure of the historical building of each period, and the continuity evaluation coefficient of the edge contour, and then perform a weighted summation calculation to obtain the health evaluation index of the historical building of each period.

[0091] Exemplarily, the weights corresponding to the comprehensive loss evaluation coefficients of the building body and the accessory structure of the historical building of each period, and the continuity evaluation coefficient of the edge contour are .

[0092] Set the grading threshold for the degree of health reduction, compare the degree of health reduction of historical buildings in each period with it, obtain the reduction degree level of historical buildings in each period, and feedback it to the system; adopt corresponding protection measures and intervention means according to different levels to achieve precise protection and effective management of historical buildings in the protection area of historical and cultural cities.

[0093] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention, and still be covered by the protection scope of the present invention.

Claims

1. A method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection, characterized in that, It includes the following steps: S1. According to the distribution of each shooting point in the historical and cultural city protection area, plan the flight path and obtain historical building images for each period; S2. Extract the preliminary edge contours of historical buildings for each period through threshold segmentation, use the contour tracking algorithm to obtain the edge contours of historical buildings for each period, and eliminate invalid contours; S3. Evaluate the continuity evaluation coefficient of the edge contours of historical buildings for each period; S4. Based on the gray-level co-occurrence matrix and clustering algorithm, divide the building image area in the historical building images for each period into the building main body and accessory structures; S5. Use the deep learning semantic segmentation model to detect the damaged areas of historical buildings, output the damaged area images of historical buildings for each period, and thus calculate and generate the damage marking maps of historical buildings for each period; S6. Evaluate the health evaluation index of historical buildings for each period, determine the reduction degree level of historical buildings for each period and give feedback.

2. The method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 1, characterized in that: The specific analysis method of step S1 is as follows: Use map software to draw the boundary of the historical and cultural city protection area on the map. Inside the boundary, select several historical buildings as each shooting point according to the distribution of historical buildings, and plan the flight path of the drone according to the distribution of each shooting point; Set the inspection cycle. In each cycle, the drone is equipped with an eagle eye device and flies along the planned flight path. During the flight, the optical images of each shooting point are synchronously collected through the camera, which are recorded as the images of each shooting point for each period. The images of each shooting point for each period are integrated through image stitching to obtain a complete image covering the historical and cultural city protection area, which is recorded as the historical building images for each period.

3. The method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 2, characterized in that: The drone is also set with an inspection cycle. The inspection cycle includes a start time, an end time, and a time interval. The drone automatically conducts periodic inspections along the flight path of the drone according to the set inspection cycle.

4. The method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 1, characterized in that: The specific analysis method of the edge contours of historical buildings for each period is as follows: Set the gray value threshold, segment the historical building images for each period through threshold segmentation into the building area and the background area, generate the binary images of historical buildings for each period, calculate the gradient amplitude and direction of the binary images of historical buildings for each period, and perform non-maximum suppression on the binary images of historical buildings for each period accordingly; Set a high threshold and a low threshold for the gradient amplitude respectively. Traverse each pixel point of the binary images of historical buildings for each period after non-maximum suppression. Mark the pixel points with a gradient amplitude greater than the high threshold as strong edge points, mark the pixel points with a gradient amplitude between the low threshold and the high threshold as weak edge points, and mark the pixel points with a gradient amplitude less than the low threshold as non-edge points; If a weak edge point is connected to at least one strong edge point, then retain the weak edge point as an edge point, otherwise regard it as a noise point and remove it from the edge. By connecting the edge points above the low threshold and connected to the edge above the high threshold, obtain the preliminary edge contours of historical buildings for each period.

5. The method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 4, characterized in that: The specific analysis method of the edge contours of historical buildings for each period is as follows: If there are discrete edge pixel points in the preliminary edge contours of historical buildings in each period, the contour tracking algorithm is used to start from any discrete edge pixel point and search for edge pixel points in its neighborhood in a fixed direction to form a contour sequence until all edges are traversed, thereby constructing the complete edge contours of historical buildings in each period and removing the invalid contours in the edge contours of historical buildings in each period.

6. The method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 1, characterized in that: The specific analysis method of step S3 is as follows: Set a breakpoint judgment threshold, and calculate the Euclidean distance between each pair of adjacent pixel points in the contour sequence of the edge contour of historical buildings in each period. If the Euclidean distance between a pair of adjacent pixel points in the edge contour of a certain period of historical building is greater than the breakpoint judgment threshold, it is judged that there is a breakpoint at this position in the edge contour of this period of historical building, and the breakpoint count is incremented by 1. Traverse the contour sequence of the edge contour of historical buildings in each period in this way, and the obtained breakpoint count is the number of breakpoints in the edge contour of historical buildings in each period; Starting from the starting point of the edge contour of historical buildings in each period, find the breakpoint positions in turn. Whenever a breakpoint is detected, the contour part before the breakpoint is divided into a segment, and the starting point and ending point coordinates of the segment are recorded. Calculate the number of pixel points included in the segment. Traverse the edge contours of historical buildings in each period in this way to obtain each contour segment of the edge contour of historical buildings in each period and its corresponding length, and calculate the proportion of the number of segments in different length intervals respectively; Regard the edge contour of historical buildings in each period as composed of a series of line segments connected end to end. Use the distance formula between two points to calculate the length of the line segment between adjacent pixel points in turn, and accumulate the lengths of all line segments to obtain the perimeter of the edge contour of historical buildings in each period. At the same time, calculate the area of the region enclosed by the edge contour of historical buildings in each period, and calculate the ratio of the perimeter to the area of the edge contour of historical buildings in each period; Based on the number of breakpoints in the edge contour of historical buildings in each period, the proportion of the number of segments in different length intervals, and the ratio of the perimeter to the area of the edge contour, a comprehensive evaluation is made to obtain the continuity evaluation coefficient of the edge contour of historical buildings in each period.

7. The method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 1, characterized in that: The specific analysis method of step S4 is as follows: On the images of historical buildings in each period, a window with a fixed pixel size is used for sliding traversal, and the sliding distance each time is one pixel until the entire historical building image is covered. For each sliding window, a distance parameter and a direction parameter are set respectively, and the gray-level co-occurrence matrix is calculated based on the distance parameter and the direction parameter; Extract texture feature parameters from the gray-level co-occurrence matrix, generate a set of 4D feature vectors according to the four extracted texture feature parameters in each direction, and combine the feature vectors in the four directions to form a 16D texture feature vector; Set the number of clusters to 2, corresponding to the building body and the attached structure respectively. Randomly initialize 2 cluster centers, calculate the Euclidean distance between each 16D texture feature vector and the cluster centers, and assign each texture feature vector to the category of the cluster center with the closest distance; Recalculate the mean of the texture feature vectors in each cluster, use this mean as the new cluster center, and repeat the clustering assignment step and the operation of updating the cluster center in this step until the cluster center no longer changes, so as to divide the building image areas in the historical building images of each period into two categories: the building body and the attached structure.

8. A method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 1, characterized in that: The specific analysis method for the damaged area images of the historical buildings in each period is as follows: Construct a deep learning semantic segmentation model, obtain a large number of building image data marked with damaged areas for training, input the historical building images of each period into the trained deep learning semantic segmentation model, and the model outputs the probability that each pixel in the historical building images of each period belongs to the damaged area. Convert the probability that each pixel belongs to the damaged area into a binary image by setting a threshold, where 1 represents the damaged area and 0 represents the non-damaged area. Traverse the predicted result image of the damaged area and record the coordinates of all pixel points with a value of 1 to obtain the damaged area images of the historical buildings in each period.

9. A method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 8, characterized in that: The specific operation method for generating the damage marking maps of the historical buildings in each period is as follows: According to the division results of the building body and the attached structure in the historical building images of each period, respectively screen the pixels belonging to the building body and the attached structure in the damaged area images of the historical buildings in each period, and thus calculate the damage parameters of the building body and the attached structure of the historical buildings in each period. Each damage parameter includes the area, shape, and pixel density of the damaged area. Create a blank image with the same size as the historical building images of each period, mark the damage parameters of the building body and the attached structure of the historical buildings in each period with different colors, and obtain the damage marking maps of the historical buildings in each period by superimposing and displaying them with the historical building images of each period.

10. A method for data collection, analysis and evaluation of historical and cultural cities based on eagle-eye inspection according to claim 1, characterized in that: The specific analysis method for step S6 is as follows: Assign corresponding weights to the building body and the attached structure respectively. Based on the damage parameters of the building body and the attached structure in the damage marking maps of the historical buildings in each period, calculate the comprehensive loss evaluation coefficients of the building body and the attached structure of the historical buildings in each period respectively, and evaluate the health evaluation index of each historical building in combination with the continuity evaluation coefficient of the edge contour of each historical building. By comparing the health evaluation indices of adjacent historical buildings, obtain the health reduction degree of each historical building. Set the classification threshold of the health reduction degree, compare the health reduction degree of each historical building with it, obtain the reduction degree level of each historical building, and feedback it to the system.

Images