An unmanned aerial vehicle multi-light imaging detection system based on multiple spectral information

Abstract

The application discloses a kind of unmanned aerial vehicle multi-optical imaging detection systems based on multiple spectral information, it is related to unmanned aerial vehicle detection technical field, including: obtaining multi-source detection data and coordinate conversion parameter table;Extract the component area of target to be measured, ultraviolet abnormal area and infrared abnormal area;Select coordinate conversion parameter group and determine second ultraviolet abnormal area and second infrared abnormal area;Cross-spectral anomaly pairing optimization and anomaly type output are carried out.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) detection technology, specifically to a UAV multi-light imaging detection system based on multiple spectral information. Background Technology

[0002] Unmanned aerial vehicle (UAV) inspections have become a common method for monitoring the condition of power transmission lines and electrical equipment. By equipping UAVs with visible light imaging units, thermal infrared imaging units, and solar-blind ultraviolet imaging units, information on equipment appearance, temperature distribution, and discharge can be acquired, thereby improving the efficiency of anomaly detection. While existing multi-spectral imaging inspection technologies can achieve simultaneous acquisition or fusion display of multiple spectral images, most solutions still focus on image acquisition, image overlay, and manual observation. They lack a stable method for mapping the location of anomalies under different spectra to the specific target components in the visible light image, and they also lack further joint analysis of the spatial relationship between ultraviolet and infrared anomalies. Therefore, in scenarios where the UAV's flight distance changes, the shooting attitude changes, or the target components are densely distributed, existing methods are prone to problems such as inaccurate anomaly location mapping, unclear anomaly attribution, and difficulty in coordinating the interpretation of multiple spectral detection results.

[0003] Chinese invention patent application CN119107347A, published on September 12, 2024, discloses a method, apparatus, device, and medium for dual-modal image registration of unmanned aerial vehicles (UAVs). It utilizes a visible-infrared dual-modal image pair acquired by the UAV, employs a feature extraction network to extract multi-scale features from the visible and infrared images, then uses a feature aggregation network to fuse common features, and combines an encoder, a decoder constructed from deformable convolutional networks and sparse sampling, a corner regression network, and a regression network to obtain transformation parameters for dual-modal image registration, thereby achieving online registration of the infrared image with the corresponding visible image.

[0004] However, the aforementioned technical solution primarily addresses the problem of online registration of dual-modal images between visible light and infrared images. Based on its published content, it processes visible light-infrared dual-modal image pairs, focusing on obtaining transformation parameters between images through a deep network to achieve online alignment of the dual-modal images. Therefore, this technical solution does not further disclose: how to unify ultraviolet, infrared, and visible light spectral images into a single component-level representation framework; how to identify the target component region and establish component numbers based on the visible light image; how to correlate and determine ultraviolet and infrared anomalous regions within the same target component; and how to output cross-spectral anomalous types based on the positional relationships between different spectral anomalies.

[0005] Therefore, the present invention provides a UAV multi-light imaging detection system based on multiple spectral information. Summary of the Invention

[0006] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a UAV multi-light imaging detection system based on multiple spectral information. By constructing a target component identification mechanism based on visible light images, a cross-spectral coordinate transformation and position correction mechanism based on distance values ​​and gimbal pitch angle values, and a cross-spectral anomaly association determination mechanism based on candidate combination relationships, it achieves unified localization, joint analysis, and anomaly type output for ultraviolet and infrared anomalies on the same target component.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a multi-spectral imaging detection method for unmanned aerial vehicles (UAVs) based on multiple spectral information. S1. Obtain multi-source detection data and coordinate transformation parameter table; S2. Extract the target component area, ultraviolet abnormal area, and infrared abnormal area; S3. Select the coordinate transformation parameter group and determine the second ultraviolet anomaly region and the second infrared anomaly region; S4. Perform cross-spectral anomaly pairing optimization and anomaly type output.

[0008] Preferably, the system acquires multi-source detection data, distance values, and gimbal pitch angle values ​​from the UAV at the same acquisition time. These data are assigned the same acquisition time number to form a data group for the same acquisition time. The multi-source detection data includes ultraviolet images, infrared images, and visible light images. Subsequent processing only calls ultraviolet images, infrared images, visible light images, distance values, and gimbal pitch angle values ​​with the same acquisition time number. The distance value refers to the distance between the UAV and the area it photographs. For any pixel in an ultraviolet, infrared, or visible light image, the column position of the pixel in the corresponding pixel matrix is ​​used as the horizontal coordinate value, and the row position of the pixel in the corresponding pixel matrix is ​​used as the vertical coordinate value, thus obtaining the coordinates of the pixel; the pixel matrix is ​​a two-dimensional set of pixels arranged in the row and column directions of the ultraviolet, infrared, or visible light image. The control processing unit stores a coordinate transformation parameter table, which records multiple sets of coordinate transformation parameters corresponding to different stored distance values ​​and different stored gimbal pitch angle values. During processing, the coordinate transformation parameter set corresponding to the current distance value and gimbal pitch angle value is read to transform the ultraviolet image coordinates and infrared image coordinates into a unified image pixel coordinate system. After synchronous acquisition is completed, the ultraviolet and infrared images are resampled according to the image size of the visible light image to obtain ultraviolet, infrared, and visible light images with the same number of pixel matrix rows and columns. The image resampling process only adjusts the pixel matrix specifications of the image and does not change the relative arrangement order of pixels in the original image.

[0009] Preferably, the grayscale values ​​of each pixel in the visible light image are first analyzed by grayscale difference analysis, and then edge detection processing is performed on the visible light image to obtain the corresponding edge pixels; then contour tracking processing is performed on adjacent edge pixels to obtain closed contours; then region filling processing is performed on the closed contours to obtain the filled regions surrounded by each closed contour, and each filled region is determined as a target component region to be tested; then a unique component number is assigned to each target component region to be tested, and the average position of all pixel coordinates in the target component region to be tested is taken to generate the component center coordinate value of the target component region to be tested; For the currently processed pixel, a ring of pixels adjacent to the edge and diagonally adjacent to the pixel is selected in the pixel matrix, with the pixel as the center. The ring of pixels constitutes the background neighborhood of the pixel. Each pixel in the ultraviolet image corresponds to a pixel response value output by the ultraviolet imaging unit. The pixel response value is used to characterize the intensity of ultraviolet radiation received at that pixel location. For each pixel in the ultraviolet image, the pixel response values ​​of all pixels in the background neighborhood of that pixel are read to calculate the average response value. When the pixel response value of a pixel is higher than the average response value of its background neighborhood, the pixel is marked as an ultraviolet anomalous pixel. Then, adjacent ultraviolet anomalous pixels are merged into an ultraviolet anomalous region. An isolated region containing only one ultraviolet anomalous pixel is not considered an ultraviolet anomalous region for subsequent processing. For each ultraviolet anomalous region, the average value of the coordinates of all pixels in the region is taken as the coordinate value of the ultraviolet anomalous center of the region. The ultraviolet region boundary and area of ​​the region are recorded, and a unique ultraviolet anomalous region number is assigned. The area of ​​the region is represented by the number of ultraviolet anomalous pixels in the region. Each pixel in the infrared image corresponds to a pixel temperature value output by the thermal infrared imaging unit. The pixel temperature value is used to characterize the thermal state corresponding to the pixel location. For each pixel in the infrared image, the average temperature value of all pixels in the background neighborhood of that pixel is read. When the pixel temperature value of that pixel is higher than the average temperature value of its background neighborhood, the pixel is marked as an infrared anomalous pixel. Then, adjacent infrared anomalous pixels are merged into an infrared anomalous region. An isolated region containing only one pixel is not considered an infrared anomalous region for subsequent processing. For each infrared anomalous region, the average value of the coordinates of all pixels in the region is taken as the infrared anomalous center coordinate value of the region. The infrared region boundary and area of ​​the region are recorded, and a unique infrared anomalous region number is assigned. The area of ​​the infrared region is represented by the number of infrared anomalous pixels in the region. After completing this step, you will obtain the target component area, component number, component center coordinates, ultraviolet anomalous area and its ultraviolet anomalous center coordinates and ultraviolet anomalous area number, infrared anomalous area and its infrared anomalous center coordinates and infrared anomalous area number.

[0010] Preferably, the control processing unit reads the distance value and gimbal pitch angle value from the current data group acquired at the same time, and searches the coordinate transformation parameter table for: the largest stored distance value whose distance value is not greater than the current distance value, the smallest stored distance value whose distance value is not less than the current distance value, the largest stored gimbal pitch angle value whose gimbal pitch angle value is not greater than the current gimbal pitch angle value, and the smallest stored gimbal pitch angle value whose gimbal pitch angle value is not less than the current gimbal pitch angle value; the found stored distance values ​​and stored gimbal pitch angle values ​​are combined, and the coordinate transformation parameter group corresponding to each combination is read; after removing duplicate combinations, one or more sets of coordinate transformation parameter groups are determined as candidate coordinate transformation parameter groups for the current acquisition time; Each set of candidate coordinate transformation parameters for the current acquisition time is applied to the coordinate values ​​of the ultraviolet anomaly center and the infrared anomaly center, respectively, and transformed to a unified image pixel coordinate system. For each set of candidate coordinate transformation parameters for the current acquisition time, the number of anomaly center coordinate values ​​located within the region of the target component after transformation is counted. The set of candidate coordinate transformation parameters for the current acquisition time with the largest number of anomaly center coordinate values ​​is selected as the current acquisition time selected coordinate transformation parameter set. If there are two or more sets of candidate coordinate transformation parameters for the current acquisition time with the same number of anomaly center coordinate values, the set with the smallest sum of distances from the anomaly center coordinate values ​​that do not fall within any region of the target component to the nearest boundary of the target component region is selected as the current acquisition time selected coordinate transformation parameter set. After determining the current acquisition time and selecting the coordinate transformation parameter group, the control processing unit selects the coordinate transformation parameter group according to the current acquisition time, transforms the coordinate values ​​of the ultraviolet anomaly center and the infrared anomaly center to a unified image pixel coordinate system, and obtains the first ultraviolet anomaly center coordinate value and the first infrared anomaly center coordinate value; then it determines which target component area each first ultraviolet anomaly center coordinate value and each first infrared anomaly center coordinate value is located in, and writes the corresponding component number; The control processing unit stores the component number, component center coordinates, second ultraviolet anomaly center coordinates, and second infrared anomaly center coordinates of each target component region in the previous acquisition time of the same acquisition time data group being processed, forming the previous acquisition time anomaly center storage result. If the previous acquisition time anomaly center storage result already exists before the current acquisition time, then for each first ultraviolet anomaly center coordinate value, it searches for the second ultraviolet anomaly center coordinate value in the previous acquisition time anomaly center storage result that has the same component number and the smallest distance. If a corresponding second ultraviolet anomaly center coordinate value is found, then the difference between the component center coordinates of the same component in the current acquisition time and the previous acquisition time is taken as the component displacement, and the found second ultraviolet anomaly center coordinate value is added to the component displacement to generate the predicted ultraviolet anomaly center coordinate value. If no corresponding second ultraviolet anomaly center coordinate value is found, then the first ultraviolet anomaly center coordinate value is directly determined as the predicted ultraviolet anomaly center coordinate value.

[0011] Preferably, for each first infrared anomaly center coordinate value, the second infrared anomaly center coordinate value with the same component number and the smallest distance is searched in the second infrared anomaly center coordinate value stored in the anomaly center storage result of the previous acquisition time; if a corresponding second infrared anomaly center coordinate value is found, the difference between the component center coordinate value of the same component at the current acquisition time and the previous acquisition time is taken as the component displacement, and the found second infrared anomaly center coordinate value is added to the component displacement to generate the predicted infrared anomaly center coordinate value; if no corresponding second infrared anomaly center coordinate value is found, the first infrared anomaly center coordinate value is directly determined as the predicted infrared anomaly center coordinate value. If there is no stored result of the anomaly center from the previous acquisition time before the current acquisition time, then the coordinates of the first ultraviolet anomaly center and the coordinates of the first infrared anomaly center are directly determined as the coordinates of the predicted ultraviolet anomaly center and the predicted infrared anomaly center, respectively. For each first ultraviolet anomaly center coordinate value, if the first ultraviolet anomaly center coordinate value is located within a certain target component area, then the first ultraviolet anomaly center coordinate value is determined as the second ultraviolet anomaly center coordinate value; if the first ultraviolet anomaly center coordinate value is not located within any target component area, and the corresponding predicted ultraviolet anomaly center coordinate value is located within a certain target component area, then the predicted ultraviolet anomaly center coordinate value is determined as the second ultraviolet anomaly center coordinate value; otherwise, the first ultraviolet anomaly center coordinate value is still determined as the second ultraviolet anomaly center coordinate value. For each first infrared anomaly center coordinate value, if the first infrared anomaly center coordinate value is located within a certain target component area, then the first infrared anomaly center coordinate value is determined as the second infrared anomaly center coordinate value; if the first infrared anomaly center coordinate value is not located within any target component area, and the corresponding predicted infrared anomaly center coordinate value is located within a certain target component area, then the predicted infrared anomaly center coordinate value is determined as the second infrared anomaly center coordinate value; otherwise, the first infrared anomaly center coordinate value is still determined as the second infrared anomaly center coordinate value. The distances between the predicted ultraviolet anomaly center coordinates and the second ultraviolet anomaly center coordinates, and the distances between the predicted infrared anomaly center coordinates and the second infrared anomaly center coordinates are calculated and determined as the ultraviolet time-series offset correction residuals and the infrared time-series offset correction residuals, respectively. If there is no stored result of the anomaly center from the previous acquisition time before the current acquisition time, the ultraviolet time-series offset correction residuals and the infrared time-series offset correction residuals are both recorded as zero. Subsequently, the control processing unit selects a set of coordinate transformation parameters based on the current acquisition time, transforms the boundaries of the ultraviolet and infrared abnormal regions to a unified image pixel coordinate system, and obtains the second ultraviolet abnormal region and the second infrared abnormal region respectively.

[0012] Preferably, for a second ultraviolet (UV) anomaly region and a second infrared (IR) anomaly region with the same component number, the control processing unit establishes multiple sets of UV-IR candidate combination relationships under that component number; for each set of UV-IR candidate combination relationships, the straight-line distance between the corresponding second UV anomaly center coordinates and the second IR anomaly center coordinates is calculated as the cross-spectral anomaly center offset value; the total number of pixels belonging to both the second UV anomaly region and the second IR anomaly region is counted, and the number of pixels is used to represent the overlap area of ​​the set of UV-IR candidate combination relationships; then, the UV region area and the IR region area of ​​the second UV anomaly region are read respectively, and the smaller of the two is taken as the smaller anomaly region area, and the ratio of the overlap area to the smaller anomaly region area is determined as the boundary envelope overlap rate of the set of UV-IR candidate combination relationships; at the same time, the sum of the corresponding UV time-series offset correction residual value and the IR time-series offset correction residual value is taken as the combination residual value; For any two sets of UV-IR candidate combination relationships under the same component number, first compare the cross-spectral anomaly center offset values ​​and select the set with the smaller value first; if the cross-spectral anomaly center offset values ​​are the same, then compare the combination residual values ​​and select the set with the smaller value first; if the combination residual values ​​are still the same, then compare the boundary envelope overlap rate and select the set with the larger value first; the control processing unit sorts all UV-IR candidate combination relationships under the same component number in the above order. The empty set is defined as the initial set of selected combination relationships under the component number; the control processing unit reads the ultraviolet-infrared candidate combination relationships one by one according to the sorted order, and determines the currently read set of ultraviolet-infrared candidate combination relationships as the current combination relationship to be inspected; if the second ultraviolet abnormal region and the second infrared abnormal region in the current combination relationship to be inspected do not appear in the set of selected combination relationships, then the current combination relationship to be inspected is added to the set of selected combination relationships; if the second ultraviolet abnormal region or the second infrared abnormal region in the current combination relationship to be inspected has already appeared in the set of selected combination relationships, then the combination relationship that uses the same second ultraviolet abnormal region or the same second infrared abnormal region as the current combination relationship to be inspected is extracted from the set of selected combination relationships, and the extracted combination relationship is determined as a conflicting combination relationship; After extracting the conflicting combination relationships, a set of combination relationships before replacement and a set of combination relationships after replacement are constructed respectively. The set of combination relationships before replacement is the second selected set of combination relationships, and the set of combination relationships after replacement is the set of combination relationships formed by deleting conflicting combination relationships from the second selected set of combination relationships and then adding the current combination relationship to be inspected. The sum of the transspectral anomaly center offset values ​​of all combination relationships in the set of combination relationships before replacement is calculated, and the sum is determined as the total offset value of the combination within the component before replacement. The sum of the transspectral anomaly center offset values ​​of all combination relationships in the set of combination relationships after replacement is calculated, and the sum is determined as the total offset value of the combination within the component after replacement. When the total offset value of the combination within the replaced component is less than the total offset value of the combination within the original component, the second selected combination relationship set is replaced with the replaced combination relationship set. If the total offset value of the combination within the replaced component is the same as the total offset value of the combination within the original component, the sum of the combination residual values ​​of all combination relationships in the original combination relationship set and the sum of the combination residual values ​​of all combination relationships in the replaced combination relationship set are calculated respectively. When the sum of the combination residual values ​​of the replaced combination relationship set is smaller, the second selected combination relationship set is replaced with the replaced combination relationship set. If the sum of the combination residual values ​​of the combination relationship sets before and after replacement is still the same, the sum of the boundary envelope coincidence rates of all combination relationships in the original combination relationship set and the sum of the boundary envelope coincidence rates of all combination relationships in the replaced combination relationship set are calculated respectively. When the sum of the boundary envelope coincidence rates of the replaced combination relationship set is larger, the second selected combination relationship set is replaced with the replaced combination relationship set; otherwise, the second selected combination relationship set remains unchanged.

[0013] Preferably, after all ultraviolet-infrared candidate combination relationships have completed one round of checks, if no replacement has occurred in that round of checks, the second selected combination relationship set is determined as the valid corresponding relationship set under the component number; if a replacement has occurred in that round of checks, the next round of checks is carried out on all ultraviolet-infrared candidate combination relationships in the above sorting order, until no replacement occurs in a certain round of checks. Anomaly types include transspectral anomalies, separated composite anomalies, ultraviolet-only anomalies, and infrared-only anomalies. For any set of ultraviolet-infrared correspondences in the valid correspondence set, if the boundary envelope overlap rate of the set of ultraviolet-infrared correspondences is greater than zero, then the second ultraviolet anomaly region and the second infrared anomaly region corresponding to the set of ultraviolet-infrared correspondences are determined to be transspectral anomalies. If the boundary envelope overlap rate of the set of ultraviolet-infrared correspondences is equal to zero, then the second ultraviolet anomaly region and the second infrared anomaly region corresponding to the set of ultraviolet-infrared correspondences are jointly determined to be separated composite anomalies within the same target component region. Second ultraviolet anomaly regions not included in the valid correspondence set are determined to be ultraviolet-only anomalies. Second infrared anomaly regions not included in the valid correspondence set are determined to be infrared-only anomalies. When outputting the results, the visible light image is used as the display base map. The boundaries of the second ultraviolet anomaly region and the second infrared anomaly region are superimposed on the corresponding locations of the target component regions to be measured. The corresponding component number, anomaly type, and cross-spectral anomaly center offset value are also output. At the same time, the component number and component center coordinates of each target component region to be measured at the current acquisition time, the coordinates of the second ultraviolet anomaly center at the current acquisition time, and the coordinates of the second infrared anomaly center at the current acquisition time are stored as the anomaly center storage results of the previous acquisition time for the next acquisition time.

[0014] A multi-spectral imaging detection system for unmanned aerial vehicles (UAVs) based on multiple spectral information includes an ultraviolet imaging unit, a thermal infrared imaging unit, a visible light imaging unit, a ranging unit, an attitude acquisition unit, a control and processing unit, and a storage unit. The ultraviolet imaging unit is used to acquire ultraviolet images of the target under test, so as to extract the coordinate values ​​of ultraviolet anomalous regions and the center of ultraviolet anomalous regions; The thermal infrared imaging unit is used to acquire infrared images of the target under test in order to extract the coordinates of infrared anomaly regions and the center of infrared anomalies. The visible light imaging unit is used to acquire visible light images of the target under test in order to identify the component area of ​​the target under test and generate component numbers; The ranging unit is used to measure the distance between the UAV and the target. The attitude acquisition unit is used to acquire the gimbal pitch angle value at the time of acquisition. The control and processing unit is used to synchronously process the data output by each unit and complete the extraction of abnormal regions, correction of the abnormal center position, cross-spectral pairing optimization, and output of detection results. The storage unit is used to store the abnormal central storage results from the previous acquisition time as well as the intermediate data required for current processing.

[0015] (III) Beneficial Effects This invention provides a UAV multi-light imaging detection system based on multiple spectral information, which has the following beneficial effects: 1. By integrating ultraviolet imaging unit, thermal infrared imaging unit and visible light imaging unit on the same UAV platform, and using the target component area in the visible light image as a unified positioning basis, the centralized attribution of multiple spectral anomaly information to the same target component is realized. Compared with the method of only outputting ultraviolet anomaly results and infrared anomaly results separately, this application can reduce the workload of inspection personnel in manually comparing different spectral images, and improve the consistency of anomaly object positioning and the readability of results.

[0016] 2. By combining the distance value and gimbal pitch angle value at the current acquisition time, the coordinate transformation relationship applicable to the current acquisition time is determined, and the coordinate values ​​of the ultraviolet anomaly center and the infrared anomaly center are uniformly transformed and the position is corrected. Even if there are changes in shooting distance and shooting attitude during the UAV inspection, the stability of the mapping of anomaly positions in different spectra can still be improved, and the offset error of the anomaly position in the cross-spectral correspondence process can be reduced.

[0017] 3. By establishing candidate combination relationships between ultraviolet and infrared anomalous regions under the same component number, and sorting, filtering and replacing the candidate combination relationships by combining the cross-spectral anomalous center offset value, combination residual value and boundary envelope overlap rate, it is possible to better distinguish between truly related cross-spectral anomalous regions and anomalous regions that are only close in spatial location but do not correspond on the same target component, thereby improving the accuracy of multispectral anomalous joint determination.

[0018] 4. Based on the effective correspondence set, it outputs various anomaly types such as cross-spectral anomalies, separated composite anomalies, ultraviolet-only anomalies, and infrared-only anomalies. It can not only complete the anomaly detection, but also further provide the correlation status and output category between anomalies. This improves the inspection results from "whether there is an anomaly" to "what type of anomaly it is and whether it has a cross-spectral correspondence", thus providing a more intuitive basis for subsequent maintenance judgment and handling. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the process of a UAV multi-light imaging detection method based on multiple spectral information according to the present invention; Figure 2 This is a schematic diagram of the structure of a UAV multi-light imaging detection system based on multiple spectral information according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 This invention provides a multi-spectral imaging detection system for unmanned aerial vehicles (UAVs) based on multiple spectral information, including an ultraviolet imaging unit, a thermal infrared imaging unit, a visible light imaging unit, a ranging unit, an attitude acquisition unit, a control and processing unit, and a storage unit. The ultraviolet imaging unit is used to acquire ultraviolet images of the target under test, so as to extract the coordinate values ​​of ultraviolet anomalous regions and the center of ultraviolet anomalous regions; The thermal infrared imaging unit is used to acquire infrared images of the target under test in order to extract the coordinates of infrared anomaly regions and the center of infrared anomalies. The visible light imaging unit is used to acquire visible light images of the target under test in order to identify the component area of ​​the target under test and generate component numbers; The ranging unit is used to measure the distance between the UAV and the target. The attitude acquisition unit is used to acquire the gimbal pitch angle value at the time of acquisition. The control and processing unit is used to synchronously process the data output by each unit and complete the extraction of abnormal regions, correction of the abnormal center position, cross-spectral pairing optimization, and output of detection results. The storage unit is used to store the abnormal central storage results from the previous acquisition time as well as the intermediate data required for current processing.

[0022] Please see Figure 2 This invention provides a multi-spectral imaging detection method for unmanned aerial vehicles (UAVs) based on multiple spectral information, comprising the following steps: S1. Obtain multi-source detection data and coordinate transformation parameter table.

[0023] The UAV is equipped with an ultraviolet imaging unit, a thermal infrared imaging unit, a visible light imaging unit, a ranging unit, an attitude acquisition unit, and a control and processing unit to simultaneously detect the target. The ultraviolet imaging unit outputs ultraviolet images, the thermal infrared imaging unit outputs infrared images, the visible light imaging unit outputs visible light images, the ranging unit outputs the distance between the UAV and the target, and the attitude acquisition unit outputs the gimbal pitch angle at the time of data acquisition. The gimbal pitch angle refers to the angle formed by the gimbal rotating around its lateral axis, and is used to characterize the pointing state of the imaging unit at the time of data acquisition. The system acquires multi-source detection data, distance values, and gimbal pitch angle values ​​from the same acquisition time and assigns them the same acquisition time number to form a data group with the same acquisition time. The multi-source detection data includes ultraviolet images, infrared images, and visible light images. Subsequent processing only calls ultraviolet images, infrared images, visible light images, distance values, and gimbal pitch angle values ​​with the same acquisition time number. The distance value refers to the distance between the UAV and the area captured by the UAV. For any pixel in an ultraviolet, infrared, or visible light image, the column position of the pixel in the corresponding pixel matrix is ​​used as the horizontal coordinate value, and the row position of the pixel in the corresponding pixel matrix is ​​used as the vertical coordinate value, thus obtaining the coordinates of the pixel; the pixel matrix is ​​a two-dimensional set of pixels arranged in the row and column directions of the ultraviolet, infrared, or visible light image. The control processing unit stores a coordinate transformation parameter table, which records multiple sets of coordinate transformation parameters corresponding to different stored distance values ​​and different stored gimbal pitch angle values. During processing, the coordinate transformation parameter set corresponding to the current distance value and gimbal pitch angle value is read to transform the ultraviolet image coordinates and infrared image coordinates into a unified image pixel coordinate system. After synchronous acquisition is completed, the ultraviolet and infrared images are resampled according to the image size of the visible light image to obtain ultraviolet, infrared, and visible light images with the same number of pixel matrix rows and columns. The image resampling process only adjusts the pixel matrix specifications of the image and does not change the relative arrangement order of pixels in the original image.

[0024] S2. Extract the target component area, ultraviolet abnormal area, and infrared abnormal area.

[0025] First, the grayscale values ​​of each pixel in the visible light image are analyzed using grayscale difference analysis. Then, edge detection processing is performed on the visible light image to obtain the corresponding edge pixels. Next, contour tracking processing is performed on adjacent edge pixels to obtain closed contours. Then, region filling processing is performed on the closed contours to obtain the filled regions enclosed by each closed contour, and each filled region is determined as a target component region to be tested. Then, a unique component number is assigned to each target component region to be tested, and the average position of all pixel coordinates within the target component region to be tested is taken to generate the component center coordinate value of the target component region to be tested. For the currently processed pixel, a ring of pixels adjacent to the edge and diagonally adjacent to the pixel is selected in the pixel matrix, with the pixel as the center. The ring of pixels constitutes the background neighborhood of the pixel. Each pixel in the ultraviolet image corresponds to a pixel response value output by the ultraviolet imaging unit. The pixel response value is used to characterize the intensity of ultraviolet radiation received at that pixel location. For each pixel in the ultraviolet image, the pixel response values ​​of all pixels in the background neighborhood of that pixel are read to calculate the average response value. When the pixel response value of a pixel is higher than the average response value of its background neighborhood, the pixel is marked as an ultraviolet anomalous pixel. Then, adjacent ultraviolet anomalous pixels are merged into an ultraviolet anomalous region. An isolated region containing only one ultraviolet anomalous pixel is not considered an ultraviolet anomalous region for subsequent processing. For each ultraviolet anomalous region, the average value of the coordinates of all pixels in the region is taken as the coordinate value of the ultraviolet anomalous center of the region. The ultraviolet region boundary and area of ​​the region are recorded, and a unique ultraviolet anomalous region number is assigned. The area of ​​the region is represented by the number of ultraviolet anomalous pixels in the region. Each pixel in the infrared image corresponds to a pixel temperature value output by the thermal infrared imaging unit. The pixel temperature value is used to characterize the thermal state corresponding to the pixel location. For each pixel in the infrared image, the average temperature value of all pixels in the background neighborhood of that pixel is read. When the pixel temperature value of that pixel is higher than the average temperature value of its background neighborhood, the pixel is marked as an infrared anomalous pixel. Then, adjacent infrared anomalous pixels are merged into an infrared anomalous region. An isolated region containing only one pixel is not considered an infrared anomalous region for subsequent processing. For each infrared anomalous region, the average value of the coordinates of all pixels in the region is taken as the infrared anomalous center coordinate value of the region. The infrared region boundary and area of ​​the region are recorded, and a unique infrared anomalous region number is assigned. The area of ​​the infrared region is represented by the number of infrared anomalous pixels in the region. After completing this step, you will obtain the target component area, component number, component center coordinates, ultraviolet anomalous area and its ultraviolet anomalous center coordinates and ultraviolet anomalous area number, infrared anomalous area and its infrared anomalous center coordinates and infrared anomalous area number.

[0026] S3. Select the coordinate transformation parameter group and determine the second ultraviolet anomaly region and the second infrared anomaly region.

[0027] The control processing unit reads the distance value and gimbal pitch angle value from the current data group acquired at the same time, and searches the coordinate transformation parameter table for: the largest stored distance value whose distance value is not greater than the current distance value, the smallest stored distance value whose distance value is not less than the current distance value, the largest stored gimbal pitch angle value whose gimbal pitch angle value is not greater than the current gimbal pitch angle value, and the smallest stored gimbal pitch angle value whose gimbal pitch angle value is not less than the current gimbal pitch angle value; it combines the found stored distance values ​​with the stored gimbal pitch angle values, and reads the coordinate transformation parameter group corresponding to each combination; after removing duplicate combinations, it determines one or more sets of coordinate transformation parameter groups as the candidate coordinate transformation parameter groups for the current acquisition time; Each set of candidate coordinate transformation parameters for the current acquisition time is applied to the coordinate values ​​of the ultraviolet anomaly center and the infrared anomaly center, respectively, and transformed to a unified image pixel coordinate system. For each set of candidate coordinate transformation parameters for the current acquisition time, the number of anomaly center coordinate values ​​located within the region of the target component after transformation is counted. The set of candidate coordinate transformation parameters for the current acquisition time with the largest number of anomaly center coordinate values ​​is selected as the current acquisition time selected coordinate transformation parameter set. If there are two or more sets of candidate coordinate transformation parameters for the current acquisition time with the same number of anomaly center coordinate values, the set with the smallest sum of distances from the anomaly center coordinate values ​​that do not fall within any region of the target component to the nearest boundary of the target component region is selected as the current acquisition time selected coordinate transformation parameter set. After determining the current acquisition time and selecting the coordinate transformation parameter group, the control processing unit selects the coordinate transformation parameter group according to the current acquisition time, transforms the coordinate values ​​of the ultraviolet anomaly center and the infrared anomaly center to a unified image pixel coordinate system, and obtains the first ultraviolet anomaly center coordinate value and the first infrared anomaly center coordinate value; then it determines which target component area each first ultraviolet anomaly center coordinate value and each first infrared anomaly center coordinate value is located in, and writes the corresponding component number; The control processing unit stores the component number, component center coordinates, second ultraviolet anomaly center coordinates, and second infrared anomaly center coordinates of each target component region in the previous acquisition time of the same acquisition time data group being processed, forming the previous acquisition time anomaly center storage result. If the previous acquisition time anomaly center storage result already exists before the current acquisition time, then for each first ultraviolet anomaly center coordinate value, it searches for the second ultraviolet anomaly center coordinate value in the previous acquisition time anomaly center storage result that has the same component number and the smallest distance. If a corresponding second ultraviolet anomaly center coordinate value is found, then the difference between the component center coordinates of the same component in the current acquisition time and the previous acquisition time is taken as the component displacement, and the found second ultraviolet anomaly center coordinate value is added to the component displacement to generate the predicted ultraviolet anomaly center coordinate value. If no corresponding second ultraviolet anomaly center coordinate value is found, then the first ultraviolet anomaly center coordinate value is directly determined as the predicted ultraviolet anomaly center coordinate value. For each first infrared anomaly center coordinate value, search for the second infrared anomaly center coordinate value with the same component number and the smallest distance from the anomaly center coordinate value stored in the previous acquisition time. If a corresponding second infrared anomaly center coordinate value is found, the difference between the component center coordinate value of the same component at the current acquisition time and the previous acquisition time is taken as the component displacement, and the found second infrared anomaly center coordinate value is added to the component displacement to generate the predicted infrared anomaly center coordinate value. If no corresponding second infrared anomaly center coordinate value is found, the first infrared anomaly center coordinate value is directly determined as the predicted infrared anomaly center coordinate value. If there is no stored result of the anomaly center from the previous acquisition time before the current acquisition time, then the coordinates of the first ultraviolet anomaly center and the coordinates of the first infrared anomaly center are directly determined as the coordinates of the predicted ultraviolet anomaly center and the predicted infrared anomaly center, respectively. For each first ultraviolet anomaly center coordinate value, if the first ultraviolet anomaly center coordinate value is located within a certain target component area, then the first ultraviolet anomaly center coordinate value is determined as the second ultraviolet anomaly center coordinate value; if the first ultraviolet anomaly center coordinate value is not located within any target component area, and the corresponding predicted ultraviolet anomaly center coordinate value is located within a certain target component area, then the predicted ultraviolet anomaly center coordinate value is determined as the second ultraviolet anomaly center coordinate value; otherwise, the first ultraviolet anomaly center coordinate value is still determined as the second ultraviolet anomaly center coordinate value. For each first infrared anomaly center coordinate value, if the first infrared anomaly center coordinate value is located within a certain target component area, then the first infrared anomaly center coordinate value is determined as the second infrared anomaly center coordinate value; if the first infrared anomaly center coordinate value is not located within any target component area, and the corresponding predicted infrared anomaly center coordinate value is located within a certain target component area, then the predicted infrared anomaly center coordinate value is determined as the second infrared anomaly center coordinate value; otherwise, the first infrared anomaly center coordinate value is still determined as the second infrared anomaly center coordinate value. The distances between the predicted ultraviolet anomaly center coordinates and the second ultraviolet anomaly center coordinates, and the distances between the predicted infrared anomaly center coordinates and the second infrared anomaly center coordinates are calculated and determined as the ultraviolet time-series offset correction residuals and the infrared time-series offset correction residuals, respectively. If there is no stored result of the anomaly center from the previous acquisition time before the current acquisition time, the ultraviolet time-series offset correction residuals and the infrared time-series offset correction residuals are both recorded as zero. Subsequently, the control processing unit selects a set of coordinate transformation parameters based on the current acquisition time, transforms the boundaries of the ultraviolet and infrared abnormal regions to a unified image pixel coordinate system, and obtains the second ultraviolet abnormal region and the second infrared abnormal region respectively.

[0028] S4. Perform cross-spectral anomaly pairing optimization and anomaly type output.

[0029] For a second ultraviolet (UV) anomaly region and a second infrared (IR) anomaly region with the same component number, the control processing unit establishes multiple sets of UV-IR candidate combination relationships under that component number. For each set of UV-IR candidate combination relationships, the straight-line distance between the corresponding coordinates of the second UV anomaly center and the coordinates of the second IR anomaly center is calculated as the cross-spectral anomaly center offset value. The total number of pixels belonging to both the second UV anomaly region and the second IR anomaly region is counted, and this number is used to represent the overlap area of ​​the set of UV-IR candidate combination relationships. Then, the UV area of ​​the second UV anomaly region and the IR area of ​​the second IR anomaly region are read respectively, and the smaller of the two is taken as the smaller anomaly region area. The ratio of the overlap area to the smaller anomaly region area is determined as the boundary envelope overlap rate of the set of UV-IR candidate combination relationships. At the same time, the sum of the corresponding UV time-series offset correction residual value and the IR time-series offset correction residual value is taken as the combination residual value. For any two sets of UV-IR candidate combination relationships under the same component number, first compare the cross-spectral anomaly center offset values ​​and select the set with the smaller value first; if the cross-spectral anomaly center offset values ​​are the same, then compare the combination residual values ​​and select the set with the smaller value first; if the combination residual values ​​are still the same, then compare the boundary envelope overlap rate and select the set with the larger value first; the control processing unit sorts all UV-IR candidate combination relationships under the same component number in the above order. The empty set is defined as the initial set of selected combination relationships under the component number; the control processing unit reads the ultraviolet-infrared candidate combination relationships one by one according to the sorted order, and determines the currently read set of ultraviolet-infrared candidate combination relationships as the current combination relationship to be inspected; if the second ultraviolet abnormal region and the second infrared abnormal region in the current combination relationship to be inspected do not appear in the set of selected combination relationships, then the current combination relationship to be inspected is added to the set of selected combination relationships; if the second ultraviolet abnormal region or the second infrared abnormal region in the current combination relationship to be inspected has already appeared in the set of selected combination relationships, then the combination relationship that uses the same second ultraviolet abnormal region or the same second infrared abnormal region as the current combination relationship to be inspected is extracted from the set of selected combination relationships, and the extracted combination relationship is determined as a conflicting combination relationship; After extracting the conflicting combination relationships, a set of combination relationships before replacement and a set of combination relationships after replacement are constructed respectively. The set of combination relationships before replacement is the second selected set of combination relationships, and the set of combination relationships after replacement is the set of combination relationships formed by deleting conflicting combination relationships from the second selected set of combination relationships and then adding the current combination relationship to be inspected. The sum of the transspectral anomaly center offset values ​​of all combination relationships in the set of combination relationships before replacement is calculated, and the sum is determined as the total offset value of the combination within the component before replacement. The sum of the transspectral anomaly center offset values ​​of all combination relationships in the set of combination relationships after replacement is calculated, and the sum is determined as the total offset value of the combination within the component after replacement. When the total offset value of the combination within the replaced component is less than the total offset value of the combination within the original component, the second selected combination relationship set is replaced with the replaced combination relationship set. If the total offset value of the combination within the replaced component is the same as the total offset value of the combination within the original component, the sum of the combination residual values ​​of all combination relationships in the original combination relationship set and the sum of the combination residual values ​​of all combination relationships in the replaced combination relationship set are calculated respectively. When the sum of the combination residual values ​​of the replaced combination relationship set is smaller, the second selected combination relationship set is replaced with the replaced combination relationship set. If the sum of the combination residual values ​​of the combination relationship sets before and after replacement is still the same, the sum of the boundary envelope coincidence rates of all combination relationships in the original combination relationship set and the sum of the boundary envelope coincidence rates of all combination relationships in the replaced combination relationship set are calculated respectively. When the sum of the boundary envelope coincidence rates of the replaced combination relationship set is larger, the second selected combination relationship set is replaced with the replaced combination relationship set; otherwise, the second selected combination relationship set remains unchanged. After all the ultraviolet-infrared candidate combination relationships have been checked in one round, if no replacement has occurred in the round of checks, the second selected combination relationship set is determined as the valid corresponding relationship set under the component number; if a replacement has occurred in the round of checks, the next round of checks on all ultraviolet-infrared candidate combination relationships is carried out again according to the above sorting order, until no replacement occurs in a certain round of checks. Anomaly types include transspectral anomalies, separated composite anomalies, ultraviolet-only anomalies, and infrared-only anomalies. For any set of ultraviolet-infrared correspondences in the valid correspondence set, if the boundary envelope overlap rate of the set of ultraviolet-infrared correspondences is greater than zero, then the second ultraviolet anomaly region and the second infrared anomaly region corresponding to the set of ultraviolet-infrared correspondences are determined to be transspectral anomalies. If the boundary envelope overlap rate of the set of ultraviolet-infrared correspondences is equal to zero, then the second ultraviolet anomaly region and the second infrared anomaly region corresponding to the set of ultraviolet-infrared correspondences are jointly determined to be separated composite anomalies within the same target component region. Second ultraviolet anomaly regions not included in the valid correspondence set are determined to be ultraviolet-only anomalies. Second infrared anomaly regions not included in the valid correspondence set are determined to be infrared-only anomalies. When outputting the results, the visible light image is used as the display base map. The boundaries of the second ultraviolet anomaly region and the second infrared anomaly region are superimposed on the corresponding locations of the target component regions to be measured. The corresponding component number, anomaly type, and cross-spectral anomaly center offset value are also output. At the same time, the component number and component center coordinates of each target component region to be measured at the current acquisition time, the coordinates of the second ultraviolet anomaly center at the current acquisition time, and the coordinates of the second infrared anomaly center at the current acquisition time are stored as the anomaly center storage results of the previous acquisition time for the next acquisition time.

[0030] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0031] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0032] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A multi-spectral imaging detection method for unmanned aerial vehicles (UAVs) based on multiple spectral information, characterized in that, include: S1. Obtain multi-source detection data and coordinate transformation parameter table; S2. Extract the target component area, ultraviolet abnormal area, and infrared abnormal area; S3. Select the coordinate transformation parameter group and determine the second ultraviolet anomaly region and the second infrared anomaly region; S4. Perform cross-spectral anomaly pairing optimization and anomaly type output.

2. The UAV multi-spectral imaging detection system based on multiple spectral information according to claim 1, characterized in that: The multi-source detection data and coordinate transformation parameter table obtained in S1 are as follows: The system acquires multi-source detection data, distance values, and gimbal pitch angle values ​​from the UAV at the same acquisition time. It assigns the same acquisition time number to these multi-source detection data, distance values, and gimbal pitch angle values ​​at the same acquisition time, forming a data group for the same acquisition time. The multi-source detection data includes ultraviolet images, infrared images, and visible light images. Subsequent processing only calls ultraviolet images, infrared images, visible light images, distance values, and gimbal pitch angle values ​​with the same acquisition time number. The distance value refers to the distance between the UAV and the area it photographs. For any pixel in an ultraviolet, infrared, or visible light image, the column position of the pixel in the corresponding pixel matrix is ​​used as the horizontal coordinate value, and the row position of the pixel in the corresponding pixel matrix is ​​used as the vertical coordinate value, thus obtaining the coordinates of the pixel; The control processing unit stores a coordinate transformation parameter table, which records multiple sets of coordinate transformation parameters corresponding to different stored distance values ​​and different stored gimbal pitch angle values. During processing, the coordinate transformation parameter set corresponding to the current distance value and gimbal pitch angle value is read to transform the ultraviolet image coordinates and infrared image coordinates into a unified image pixel coordinate system. The ultraviolet and infrared images are resampled according to the image size of the visible light image to obtain ultraviolet, infrared, and visible light images with the same number of rows and columns of pixel matrix.

3. The UAV multi-spectral imaging detection system based on multiple spectral information according to claim 1, characterized in that: In S2, the target component region, ultraviolet anomaly region, and infrared anomaly region are extracted as follows: First, the grayscale values ​​of each pixel in the visible light image are analyzed by grayscale difference analysis. Then, edge detection processing is performed on the visible light image to obtain the corresponding edge pixels. Next, contour tracking processing is performed on adjacent edge pixels to obtain closed contours. Then, region filling processing is performed on the closed contours to obtain the filled regions surrounded by each closed contour, and each filled region is determined as a target component region to be tested. Then, assign a unique component number to each target component area to be tested, and take the average position of all pixel coordinates in the target component area to generate the component center coordinate value of the target component area to be tested; For the currently processed pixel, a ring of pixels adjacent to the edge and diagonally adjacent to the pixel is selected in the pixel matrix, with the pixel as the center. The ring of pixels constitutes the background neighborhood of the pixel. Each pixel in an ultraviolet image corresponds to a pixel response value output by an ultraviolet imaging unit. The pixel response value is used to characterize the intensity of ultraviolet radiation received at that pixel location. For each pixel in the ultraviolet image, the average response value is calculated by reading the pixel response values ​​of all pixels in the background neighborhood of that pixel. When the pixel response value of a pixel is higher than the average response value of its background neighborhood, the pixel is marked as an ultraviolet anomalous pixel; then adjacent ultraviolet anomalous pixels are merged into an ultraviolet anomalous region. Isolated regions containing only one UV anomalous pixel are not considered UV anomalous regions for subsequent processing. For each UV anomalous region, the average of the coordinates of all pixels within the region is taken as the UV anomalous center coordinate value, and the UV region boundary and UV region area are recorded, while a unique UV anomalous region number is assigned. The UV region area is represented by the number of UV anomalous pixels in the UV anomalous region. Each pixel in the infrared image corresponds to a pixel temperature value output by the thermal infrared imaging unit. The pixel temperature value is used to characterize the thermal state corresponding to the pixel position. For each pixel in the infrared image, the average temperature value of all pixels in the background neighborhood of that pixel is read. When the pixel temperature value of a pixel is higher than the average temperature value of its background neighborhood, the pixel is marked as an infrared anomalous pixel; then, adjacent infrared anomalous pixels are merged into an infrared anomalous region; isolated regions containing only one pixel are not considered infrared anomalous regions for subsequent processing; for each infrared anomalous region, the average value of all pixel coordinates within the region is taken as the infrared anomalous center coordinate value of the region, and the infrared region boundary and area are recorded, while a unique infrared anomalous region number is assigned; the area of ​​the infrared region is represented by the number of infrared anomalous pixels in the region. The following information is obtained: the target component region, component number, component center coordinates, ultraviolet anomalous region and its ultraviolet anomalous center coordinates and ultraviolet anomalous region number, infrared anomalous region and its infrared anomalous center coordinates and infrared anomalous region number.

4. The UAV multi-spectral imaging detection system based on multiple spectral information according to claim 1, characterized in that: In S3, the coordinate transformation parameter set is selected and the second ultraviolet anomaly region and the second infrared anomaly region are determined as follows: The control processing unit reads the distance value and gimbal pitch angle value from the current data group acquired at the same time, and searches the coordinate transformation parameter table for: the largest stored distance value whose distance value is not greater than the current distance value, the smallest stored distance value whose distance value is not less than the current distance value, the largest stored gimbal pitch angle value whose gimbal pitch angle value is not greater than the current gimbal pitch angle value, and the smallest stored gimbal pitch angle value whose gimbal pitch angle value is not less than the current gimbal pitch angle value; it combines the found stored distance values ​​with the stored gimbal pitch angle values, and reads the coordinate transformation parameter group corresponding to each combination; after removing duplicate combinations, it determines one or more sets of coordinate transformation parameter groups as the candidate coordinate transformation parameter groups for the current acquisition time; Each set of candidate coordinate transformation parameters for the current acquisition time is applied to the coordinate values ​​of the ultraviolet anomaly center and the infrared anomaly center, respectively, and transformed to a unified image pixel coordinate system. For each set of candidate coordinate transformation parameters for the current acquisition time, the number of anomaly center coordinate values ​​located within the region of the target component after transformation is counted. The set of candidate coordinate transformation parameters for the current acquisition time with the largest number of anomaly center coordinate values ​​is selected as the current acquisition time selected coordinate transformation parameter set. If there are two or more sets of candidate coordinate transformation parameters for the current acquisition time with the same number of anomaly center coordinate values, the set with the smallest sum of distances from the anomaly center coordinate values ​​that do not fall within any region of the target component to the nearest boundary of the target component region is selected as the current acquisition time selected coordinate transformation parameter set. After determining the current acquisition time and selecting the coordinate transformation parameter group, the control processing unit selects the coordinate transformation parameter group based on the current acquisition time and transforms the coordinate values ​​of the ultraviolet anomaly center and the infrared anomaly center to a unified image pixel coordinate system to obtain the first ultraviolet anomaly center coordinate value and the first infrared anomaly center coordinate value. Next, determine which target component region each first ultraviolet anomaly center coordinate value and each first infrared anomaly center coordinate value are located in, and write the corresponding component number. The control processing unit stores the component number, component center coordinates, second ultraviolet anomaly center coordinates, and second infrared anomaly center coordinates of each target component region in the previous acquisition time of the same acquisition time data group being processed, forming the previous acquisition time anomaly center storage result. If the previous acquisition time anomaly center storage result already exists before the current acquisition time, then for each first ultraviolet anomaly center coordinate value, it searches for the second ultraviolet anomaly center coordinate value in the previous acquisition time anomaly center storage result that has the same component number and the smallest distance. If a corresponding second ultraviolet anomaly center coordinate value is found, then the difference between the component center coordinates of the same component in the current acquisition time and the previous acquisition time is taken as the component displacement, and the found second ultraviolet anomaly center coordinate value is added to the component displacement to generate the predicted ultraviolet anomaly center coordinate value. If no corresponding second ultraviolet anomaly center coordinate value is found, then the first ultraviolet anomaly center coordinate value is directly determined as the predicted ultraviolet anomaly center coordinate value.

5. The UAV multi-spectral imaging detection system based on multiple spectral information according to claim 1, characterized in that: In S3, the coordinate transformation parameter set is selected and the second ultraviolet anomaly region and the second infrared anomaly region are determined as follows: For each first infrared anomaly center coordinate value, find the second infrared anomaly center coordinate value with the same component number and the smallest distance from the second infrared anomaly center coordinate value in the anomaly center storage result of the previous acquisition time. If the corresponding second infrared anomaly center coordinate value is found, the difference between the current acquisition time and the component center coordinate value of the same component at the previous acquisition time is taken as the component displacement, and the found second infrared anomaly center coordinate value is added to the component displacement to generate the predicted infrared anomaly center coordinate value. If the corresponding second infrared anomaly center coordinates are not found, the first infrared anomaly center coordinates are directly determined as the predicted infrared anomaly center coordinates. If there is no stored result of the anomaly center from the previous acquisition time before the current acquisition time, then the coordinates of the first ultraviolet anomaly center and the coordinates of the first infrared anomaly center are directly determined as the coordinates of the predicted ultraviolet anomaly center and the predicted infrared anomaly center, respectively. For each first ultraviolet anomaly center coordinate value, if the first ultraviolet anomaly center coordinate value is located within a certain target component area, then the first ultraviolet anomaly center coordinate value is determined as the second ultraviolet anomaly center coordinate value. If the coordinates of the first ultraviolet anomaly center are not located within any target component area, and the corresponding predicted ultraviolet anomaly center coordinates are located within a target component area, then the predicted ultraviolet anomaly center coordinates are determined as the second ultraviolet anomaly center coordinates. Otherwise, the coordinates of the first ultraviolet anomaly center shall still be determined as the coordinates of the second ultraviolet anomaly center; For each first infrared anomaly center coordinate value, if the first infrared anomaly center coordinate value is located within a certain target component area, then the first infrared anomaly center coordinate value is determined as the second infrared anomaly center coordinate value; if the first infrared anomaly center coordinate value is not located within any target component area, and the corresponding predicted infrared anomaly center coordinate value is located within a certain target component area, then the predicted infrared anomaly center coordinate value is determined as the second infrared anomaly center coordinate value. Otherwise, the coordinates of the first infrared anomaly center will still be determined as the coordinates of the second infrared anomaly center; The distances between the predicted ultraviolet anomaly center coordinates and the second ultraviolet anomaly center coordinates, and the distances between the predicted infrared anomaly center coordinates and the second infrared anomaly center coordinates are calculated and determined as the ultraviolet time-series offset correction residuals and the infrared time-series offset correction residuals, respectively. If there is no stored result of the anomaly center from the previous acquisition time before the current acquisition time, the ultraviolet time-series offset correction residuals and the infrared time-series offset correction residuals are both recorded as zero. The boundaries of the ultraviolet and infrared anomalous regions are transformed into a unified image pixel coordinate system to obtain the second ultraviolet anomalous region and the second infrared anomalous region, respectively.

6. The UAV multi-light imaging detection system based on multiple spectral information according to claim 1, characterized in that: In S4, cross-spectral anomaly pairing optimization and anomaly type output are performed as follows: For the second ultraviolet anomaly region and the second infrared anomaly region with the same component number, the control processing unit establishes multiple sets of ultraviolet-infrared candidate combination relationships under the component number; for each set of ultraviolet-infrared candidate combination relationships, the straight-line distance between the corresponding second ultraviolet anomaly center coordinate value and the second infrared anomaly center coordinate value is calculated as the cross-spectral anomaly center offset value. The total number of pixels belonging to both the second ultraviolet (UV) anomaly region and the second infrared (IR) anomaly region is counted, and this number is used to represent the overlap area of ​​the UV-IR candidate combination relationship. Then, the UV area of ​​the second UV anomaly region and the IR area of ​​the second IR anomaly region are read separately, and the smaller of the two is taken as the smaller anomaly region area. The ratio of the overlap area to the smaller anomaly region area is determined as the boundary envelope overlap rate of the UV-IR candidate combination relationship. Simultaneously, the sum of the corresponding UV temporal offset correction residual and the IR temporal offset correction residual is used as the combination residual value. For any two sets of UV-IR candidate combination relationships under the same component number, first compare the cross-spectral anomaly center offset values, and take the set with the smaller value first; If the cross-spectral anomaly center offset values ​​are the same, the combined residual values ​​are compared, and the group with the smaller value is ranked first; if the combined residual values ​​are still the same, the boundary envelope overlap rate is compared, and the group with the larger value is ranked first; the control processing unit sorts all ultraviolet-infrared candidate combination relationships under the same component number in the above order. The empty set is defined as the initial set of selected combination relationships under the component number; the control processing unit reads the ultraviolet-infrared candidate combination relationships one by one according to the sorted order, and determines the currently read set of ultraviolet-infrared candidate combination relationships as the current combination relationship to be inspected; if the second ultraviolet abnormal region and the second infrared abnormal region in the current combination relationship to be inspected do not appear in the set of selected combination relationships, then the current combination relationship to be inspected is added to the set of selected combination relationships; if the second ultraviolet abnormal region or the second infrared abnormal region in the current combination relationship to be inspected has already appeared in the set of selected combination relationships, then the combination relationship that uses the same second ultraviolet abnormal region or the same second infrared abnormal region as the current combination relationship to be inspected is extracted from the set of selected combination relationships, and the extracted combination relationship is determined as a conflicting combination relationship; After extracting the conflicting combination relationships, a set of combination relationships before replacement and a set of combination relationships after replacement are constructed respectively. The set of combination relationships before replacement is the second selected set of combination relationships, and the set of combination relationships after replacement is the set of combination relationships formed by deleting conflicting combination relationships from the second selected set of combination relationships and then adding the current combination relationship to be inspected. The sum of the transspectral anomaly center offset values ​​of all combination relationships in the set of combination relationships before replacement is calculated, and the sum is determined as the total offset value of the combination within the component before replacement. The sum of the transspectral anomaly center offset values ​​of all combination relationships in the set of combination relationships after replacement is calculated, and the sum is determined as the total offset value of the combination within the component after replacement. When the total offset value of the combination within the replaced component is less than the total offset value of the combination within the original component, the second selected combination relationship set is replaced with the replaced combination relationship set. If the total offset value of the combination within the replaced component is the same as the total offset value of the combination within the original component, the sum of the combination residual values ​​of all combination relationships in the original combination relationship set and the sum of the combination residual values ​​of all combination relationships in the replaced combination relationship set are calculated respectively. When the sum of the combination residual values ​​of the replaced combination relationship set is smaller, the second selected combination relationship set is replaced with the replaced combination relationship set. If the sum of the combination residual values ​​of the combination relationship sets before and after replacement is still the same, the sum of the boundary envelope coincidence rates of all combination relationships in the original combination relationship set and the sum of the boundary envelope coincidence rates of all combination relationships in the replaced combination relationship set are calculated respectively. When the sum of the boundary envelope coincidence rates of the replaced combination relationship set is larger, the second selected combination relationship set is replaced with the replaced combination relationship set; otherwise, the second selected combination relationship set remains unchanged.

7. The UAV multi-spectral imaging detection system based on multiple spectral information according to claim 1, characterized in that: In S4, cross-spectral anomaly pairing optimization and anomaly type output are performed as follows: After all the ultraviolet-infrared candidate combination relationships have been checked in one round, if no replacement has occurred in the round of checks, the second selected combination relationship set is determined as the valid corresponding relationship set under the component number; if a replacement has occurred in the round of checks, the next round of checks on all ultraviolet-infrared candidate combination relationships is carried out again according to the above sorting order, until no replacement occurs in a certain round of checks. Anomaly types include transspectral anomalies, separated composite anomalies, ultraviolet-only anomalies, and infrared-only anomalies. For any set of ultraviolet-infrared correspondences in the effective correspondence set, if the boundary envelope overlap rate of the set of ultraviolet-infrared correspondences is greater than zero, then the second ultraviolet anomaly region and the second infrared anomaly region corresponding to the set of ultraviolet-infrared correspondences are determined to be transspectral anomalies. If the boundary envelope overlap rate of the set of ultraviolet-infrared correspondences is zero, then the second ultraviolet anomalous region and the second infrared anomalous region corresponding to the set of ultraviolet-infrared correspondences are jointly determined as separate composite anomalous regions within the same target component region. For the second ultraviolet anomaly region that does not enter the valid correspondence set, it is determined to be an ultraviolet anomaly only; For the second infrared anomaly region that does not enter the valid correspondence set, it is determined to be an infrared anomaly only; When outputting the results, the visible light image is used as the display base map. The boundaries of the second ultraviolet anomaly region and the second infrared anomaly region are superimposed on the corresponding locations of the target component regions to be measured. The corresponding component number, anomaly type, and cross-spectral anomaly center offset value are also output. At the same time, the component number and component center coordinates of each target component region to be measured at the current acquisition time, the coordinates of the second ultraviolet anomaly center at the current acquisition time, and the coordinates of the second infrared anomaly center at the current acquisition time are stored as the anomaly center storage results of the previous acquisition time for the next acquisition time.

8. A multi-spectral imaging detection system for unmanned aerial vehicles (UAVs) based on multiple spectral information, comprising an ultraviolet imaging unit, a thermal infrared imaging unit, a visible light imaging unit, a ranging unit, an attitude acquisition unit, a control and processing unit, and a storage unit: The ultraviolet imaging unit is used to acquire ultraviolet images of the target under test, so as to extract the coordinate values ​​of ultraviolet anomalous regions and the center of ultraviolet anomalous regions; The thermal infrared imaging unit is used to acquire infrared images of the target under test in order to extract the coordinates of infrared anomaly regions and the center of infrared anomalies. The visible light imaging unit is used to acquire visible light images of the target under test in order to identify the component area of ​​the target under test and generate component numbers; The ranging unit is used to measure the distance between the UAV and the target. The attitude acquisition unit is used to acquire the gimbal pitch angle value at the time of acquisition. The control and processing unit is used to synchronously process the data output by each unit and complete the extraction of abnormal regions, correction of the abnormal center position, cross-spectral pairing optimization, and output of detection results. The storage unit is used to store the abnormal central storage results from the previous acquisition time as well as the intermediate data required for current processing.

Images