An unmanned aerial vehicle evidence providing method based on intelligent segmentation of map patches and adaptive zoom control

Abstract

The application discloses a kind of unmanned plane evidence-giving methods based on intelligent segmentation and adaptive zoom control of map spot, it is related to unmanned plane aerial photography and map spot evidence-giving technical field.Optimization of evidence-giving perspective is as core, structured import vector map spot data, different scale map spot is implemented different processing: large map spot adopts minimum circumscribed rectangle iterative segmentation, small map spot starts adaptive zoom control algorithm, special shape map spot is handled by segmentation-zoom collaborative optimization mechanism, solve the problem of missing details and associated map spot in shooting.Shotting point according to the planned back calculation unmanned plane shooting position, automatically plan to generate optimal flight task queue, and unmanned plane executes breakpoint continuation and completes evidence-giving task independently.The achievement is uploaded to platform after being checked and encrypted and compliance self-checking guarantee anti-fake nature.The application realizes the closed loop mode of unmanned plane map spot evidence-giving whole process automation, establishes standardization operation paradigm, and provides efficient, accurate technical support for map spot evidence-giving.

Description

Technical Field

[0001] This invention relates to the field of drone aerial photography and image patch evidence technology, and more specifically to a drone evidence method based on intelligent image patch segmentation and adaptive zoom control. Background Technology

[0002] Land use change parcel verification is a survey procedure that uses remote sensing imagery and on-site photography to confirm the spatial location and land use attributes of land use change parcels. The resulting data is a crucial basis for spatial planning and natural resource management. Traditional verification methods involve manual on-site visits and data collection via mobile phones or drones, which is inefficient and lacks standardized, uniform photography. As land use change verification becomes more routine and refined, traditional drone aerial photography solutions are no longer sufficient to meet the demands for high efficiency and accuracy.

[0003] Natural resource monitoring patches face multiple technical bottlenecks in evidence collection due to their spatial dispersion, massive data scale, and heterogeneous morphological characteristics. Traditional manually operated drone-based evidence collection methods cannot dynamically adjust shooting parameters, making it difficult to adapt to the significant scale differences of mixed-morphology discrete patches. This results in blind spots or insufficient resolution in single shots, failing to meet the technical requirement that "evidence images must accurately cover changing areas." To address these technical bottlenecks, there is an urgent need to develop intelligent drone-based field evidence collection solutions, focusing on overcoming key technical challenges such as accurate coverage and standardized acquisition of multi-scale patches.

[0004] Therefore, proposing a UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control to address the difficulties in existing technologies is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control. It establishes a collaborative optimization operation mode combining a minimum bounding rectangle (MBR) alignment segmentation strategy and adaptive zoom control, solving the problems of associated patches caused by differences in patch area and the clarity of evidence photos in land use patch evidence collection. This invention not only achieves fully automated operation of the field evidence collection process but also establishes a standardized operation paradigm for UAV evidence collection from the technical source.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for providing evidence using unmanned aerial vehicles (UAVs) based on intelligent patch segmentation and adaptive zoom control includes the following steps: S1. Import Vector Patch Data: Import patch data containing GeoJSON and SHAPE format vectors and attribute information. Create a structured data import template based on patch attributes, automatically extract core field information of patches, complete batch data verification and format conversion, and generate a set of tasks to be processed. S2. Patch Data Analysis and Classification: Based on the task set to be processed, the patch data is classified and analyzed sequentially. First, the actual coverage area of ​​the camera frame is dynamically calculated according to the flight altitude parameters set by the UAV mission. The minimum bounding rectangle is calculated through the patch vector boundary. A multi-scale classification system combining patch geometry and the coverage capability of the UAV's shooting frame is constructed to classify patches into large patches, small patches, regular patches, and special-shaped patches. S3. Differentiated processing strategy for image patches: After classifying image patches according to the multi-scale classification system, a differentiated processing strategy is adopted. For large image patches, an intelligent image patch segmentation algorithm is introduced, and an iterative geometric segmentation is performed using the minimum bounding rectangle long side alignment strategy. For small image patches, an adaptive zoom control algorithm is triggered for shooting, and the image resolution is improved to the 0.2m / pixel standard through focal length compensation. For regular image patches, the standard shooting mode is used to directly cover them completely. For image patches with special shapes, an image patch segmentation-zoom collaborative optimization mechanism is activated. S4. Generation of UAV Evidence-Collecting Shooting Locations: Based on the image patch processing results, for each captured image patch or segmented sub-image patch, one vertical orthogonal aerial photography point and N oblique aerial photography points are planned according to the evidence-collecting specifications to construct a collection scheme that meets the requirements of multi-view evidence collection. At the same time, the spatial range and corresponding focusing parameters of each image patch or segmented unit are determined. Based on the photogrammetric collinearity equation, the captured image frame unit is inversely solved into UAV shooting spatial parameters, thereby determining the spatial coordinates of each shooting point. S5. Automated and secure execution of drone evidence collection: Based on the calculated spatial coordinates of each photo point, the system automatically plans and generates the globally optimal evidence collection flight route; the drone autonomously executes the flight mission along the optimal evidence collection flight route and automatically adjusts the gimbal to a preset angle for shooting at each waypoint; S6. Encrypted transmission and compliance self-check of evidence results: After completing the shooting task, the drone automatically packages the evidence photos and uploads them to the drone control cloud server. The drone generates and encrypts the evidence results, encapsulates them into a DB format result package that meets the requirements, performs a compliance self-check on the result package, and pushes the result package to the cloud platform after verification.

[0007] Optionally, the specific details of dynamically calculating the actual coverage area of ​​the camera frame based on the flight altitude parameters set in S2 according to the UAV mission are as follows: Flight altitude parameters set according to the UAV mission Calculate the actual coverage area captured by the drone camera. : Single-span coverage area width :

[0008] Single-span ground coverage height :

[0009] Maximum coverage area per frame :

[0010] In the formula, To ensure the drone camera's field of view covers the width of the ground, , The actual width of the sensor. The actual height of the sensor. This is the actual focal length.

[0011] Optionally, the specific content of the intelligent patch segmentation algorithm in S3 is as follows: The specific content of the S3 image patch intelligent segmentation algorithm is as follows: Step 1: Generate the minimum bounding rectangle based on the target feature;

[0012] in, For the set of boundary points of the map patch, For convex hull calculation functions, To be the candidate bounding rectangle that can cover the patch, Rectangle area, This is the bounding rectangle with the smallest area obtained through optimization. Step 2, generating candidate segmentation boxes: at the four corner points of the smallest bounding rectangle Generate two types of drone frame bounding boxes with fixed horizontal and vertical dimensions as segmentation candidate boxes:

[0013] In the formula, These are the coordinates of the corner point. For corner index, To fix the width and height of the ground covered by the drone camera's image frame, [] represents the horizontal coverage area of ​​the rectangle. [] represents the vertical coverage area of ​​the rectangle; Step 3: Optimal Segmentation Box Decision: Select the best candidate box from the segmentation candidate boxes. Choose the candidate box that covers the largest effective area of ​​the patch as the optimal segmentation box region. The formula is as follows:

[0014] In the formula, For the candidate box set, This is a function for calculating the area of ​​a region. For the original target patch, Candidate boxes With the pattern The geometric intersection; Step 4: Update remaining polygons: Subtract the segmented regions from the remaining polygons. Iteratively update the remaining set of blobs using the following formula:

[0015] In the formula, For the first t The set of remaining graph spots at the next iteration In the first t The region segmented in the next iteration. For the updated version t +1 iterations of the remaining graph spot set; Repeat steps 1-3 for iterative segmentation until the length and width of the minimum bounding rectangle of the remaining patch are both smaller than the coverage size of the drone's frame. Step 5: Full Coverage of Remaining Image Areas: For the final remaining image area, maintaining the original frame parameters, select the frame position that can completely cover the remaining image area and has the highest overlap with the already photographed area as the final shooting frame, as shown in the following formula:

[0016] In the formula, To ensure the candidate boxes completely cover the remaining area. This represents the actual overlap area between the candidate bounding box and the original patch. The area is the candidate box area.

[0017] Optionally, the specific content of the adaptive zoom control algorithm in S3 is as follows: Step 1: Calculate the minimum bounding rectangle of the target patch and determine the direction of the major axis and the coordinates of the center point; Step 2: Gimbal attitude calibration. Control the rotation of the gimbal camera to make the long side of the camera frame parallel to the long axis of the MBR, and align the center point of the camera frame with the center point of the MBR. Step 3: Zoom parameter decision. Based on the gimbal attitude calibration results in Step 2, the theoretical zoom value is calculated using the focal length-field-of-view mapping model. The formula is as follows:

[0018] In the formula, The base sensor area for 1.0x zoom. To set the coverage factor, the default value is 0.8. The proportion of the target image patch's MBR area to the zoomed image area; Step 4: Control the camera lens motor to step to the selected zoom level, and automatically trigger shooting after the gimbal stabilizes.

[0019] Optionally, the specific details of S4, which uses the photogrammetric collinearity equation to inversely calculate the shooting frame unit into the UAV shooting spatial parameters, are as follows: With fixed camera pitch, roll, and flight altitude, the focal length and yaw angle are determined by combining the geometric properties of the target imaging unit. The spatial coordinates of the point the UAV needs to reach to fully cover the current imaging unit are then calculated. , , ).

[0020] Optionally, the specific details of automatically planning and generating the globally optimal evidence-gathering flight route in S5 based on the calculated spatial coordinates of each photo-taking point are as follows: Define the task priority function:

[0021] In the formula, For the image The center coordinates, The coordinates of the drone's take-off and landing points. Let be the Euclidean distance between the map patch and the take-off and landing points. The spatial attenuation coefficient, is a natural constant and the base of the exponential function; Based on the task priority function, the task priority coefficient of the image patch to be verified is calculated; based on the task priority coefficient and the set of aerial photography parameters, a flight path with the shortest total flight distance is planned and generated.

[0022] In the formula, To find the waypoint arrangement that minimizes the total distance, m The total number of waypoints. The distance between waypoints is expressed in Euclidean form. For the first j The spatial coordinates of the waypoint to be visited. In order to be in The spatial coordinates of the next waypoint to be visited.

[0023] Optionally, S6 includes an anti-counterfeiting verification scheme for evidence patches, wherein, The specific content of the evidence patch anti-counterfeiting verification scheme is as follows: When the drone takes pictures, it automatically embeds a digital watermark containing shooting coordinates, shooting time and equipment information, generates a unique hash value in real time, generates a verification code through key signing and encrypts the result data, and the key verification information is synchronously uploaded to the blockchain for evidence storage; the result package is transmitted in standard DB format, and the receiving end ensures the authenticity and integrity of the photo by comparing and verifying the photo metadata, verifying the hash value and the validity of the verification code.

[0024] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control, the beneficial effects of which are: By synergistically optimizing the patch segmentation strategy and zoom control algorithm, a closed-loop evidence collection process is constructed, encompassing intelligent patch analysis and flight control, thereby improving the efficiency and accuracy of field evidence collection. Specifically, this invention employs a patch minimum bounding rectangle alignment segmentation strategy to reduce non-target shooting areas, effectively suppressing the problem of associated patches. Simultaneously, adaptive zoom control enhances the resolution of internal details within patches, achieving a dual breakthrough in coverage integrity and detail clarity. Furthermore, the fully automated operation paradigm for UAV field evidence collection constructed in this invention establishes a complete technical link from patch import to flight control to result verification, providing efficient and accurate intelligent support for land use patch evidence collection. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0026] Figure 1 A flowchart of an UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control provided by the present invention; Figure 2 A flowchart illustrating the map patch data analysis, classification, and map patch differentiation processing strategies provided by this invention; Figure 3 This is a schematic diagram of the intelligent patch segmentation algorithm provided by the present invention; Figure 4 This is a schematic diagram of the adaptive zoom control algorithm provided by the present invention; Figure 5 This is a schematic diagram of drone shooting location generation provided by the present invention; Figure 6 The flowchart of the method for encrypted return of evidence and compliance self-inspection provided by this invention. Detailed Implementation

[0027] 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.

[0028] See Figure 1 As shown, this invention discloses a UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control, comprising the following steps: S1. Import Vector Patch Data: Import patch data containing GeoJSON and SHAPE format vectors and attribute information. Create a structured data import template based on patch attributes, automatically extract core field information of patches, complete batch data verification and format conversion, and generate a set of tasks to be processed. See Figure 2 As shown, S2, patch data analysis and classification: Based on the task set to be processed, the patch data is classified and analyzed sequentially. First, the actual coverage area of ​​the camera frame is dynamically calculated according to the flight altitude parameters set by the UAV mission. The minimum bounding rectangle is calculated through the patch vector boundary. A multi-scale classification system combining the patch geometry and the coverage capability of the UAV's shooting frame is constructed, classifying the patches into large patches, small patches, regular patches, and special-shaped patches. Specifically, (1) Large patches: When the area of ​​a patch is greater than the actual coverage area of ​​the camera frame, or when the length of the minimum bounding rectangle of a patch is greater than the length that the camera frame can cover, the patch intelligent segmentation algorithm is triggered, and iterative geometric segmentation is performed using the minimum bounding rectangle (MBR) long side alignment strategy.

[0029] (2) Small patch: When the area of ​​the patch is less than 0.80 × the actual coverage area of ​​the camera frame, and the length of the smallest bounding rectangle of the patch is less than the length that the camera can cover, and the width of the smallest bounding rectangle of the patch is less than the width that the camera can cover, the adaptive zoom mechanism is activated to improve the image resolution to 0.2m / pixel standard through focal length compensation.

[0030] (3) Regular patch: The area of ​​the patch is within the range of 0.80~1.0 × the actual coverage area of ​​the camera frame, and the longest side of the smallest bounding rectangle of the patch is... The camera's coverage length, the width of the smallest bounding rectangle of the patch. When the camera can cover a patch of width, use the standard shooting mode to directly cover it completely.

[0031] (4) Special-shaped patches: For unconventional patch shapes such as L-shaped / C-shaped patches, when the ratio of the minimum bounding rectangle of the patch to the actual area of ​​the patch is less than 40%, even if the full coverage condition of the frame is met, the special shape is prone to generating associated patches. Therefore, intelligent segmentation is still required. Non-rectangular sub-regions are generated by decomposing the convex hull before shooting. If the sub-patterns after segmentation meet the small patch standard, then shooting needs to be optimized in conjunction with the zoom strategy.

[0032] S3. Differentiated processing strategy for image patches: After classifying image patches according to the multi-scale classification system, a differentiated processing strategy is adopted. For large image patches, an intelligent image patch segmentation algorithm is introduced, and an iterative geometric segmentation is performed using the minimum bounding rectangle long side alignment strategy. For small image patches, an adaptive zoom control algorithm is triggered for shooting, and the image resolution is improved to the 0.2m / pixel standard through focal length compensation. For regular image patches, the standard shooting mode is used to directly cover them completely. For image patches with special shapes, an image patch segmentation-zoom collaborative optimization mechanism is activated. For details, see Figure 3 As shown, the intelligent patch segmentation algorithm employs an iterative minimum bounding rectangle (MBR) segmentation strategy for large patches requiring segmentation. This involves iteratively executing a "alignment segmentation - region subtraction - bounding rectangle update" process to progressively segment the large patches. In each iteration, the algorithm prioritizes aligning the longest side of the current patch region with the camera frame to extract the largest coverage sub-region, dynamically adjusting the segmentation boundary to reduce invalid shots. The remaining portion undergoes secondary segmentation by recalculating the MBR. When the remaining area is insufficient for a single complete shot, the algorithm maintains the original frame parameters to generate the optimal coverage frame, ensuring complete coverage while maintaining necessary overlap with already captured areas, ultimately achieving complete patch segmentation without omissions.

[0033] Adaptive zoom control algorithm: See Figure 4 As shown, an adaptive zoom control shooting method is implemented for small image patches. The minimum bounding rectangle (MBR) of the image patch is calculated, and the camera automatically rotates so that the long side of the image frame is parallel to the long side of the MBR, and the center point of the image frame coincides with the center point of the rectangle. The optimal focal length is selected by calculating the ratio of the image patch area to the camera's image area. The focal length is dynamically adjusted while standardized shooting parameters are fixed to ensure that the main body of the image patch is accurately covered in each shot, improving detail clarity.

[0034] Image patch segmentation-zoom collaborative optimization mechanism: For special-shaped images, even if the entire image can be covered by a single frame, to eliminate potential related image patch problems, an intelligent image patch decomposition algorithm is first used to divide the original image patch into multiple regular sub-regions. After decomposition, the size attributes of each sub-region are dynamically determined: if the sub-region meets the small image patch standard, an adaptive zoom mechanism is triggered to improve resolution; if it meets the standard for regular image patches, the standard shooting mode is used to directly cover the entire image. Through the above image patch segmentation-zoom collaborative optimization mechanism, efficient and accurate coverage of special-shaped images is achieved.

[0035] S4. Generation of UAV Evidence-Collecting Shooting Locations: Based on the image patch processing results, for each captured image patch or segmented sub-image patch, one vertical orthogonal aerial photography point and N oblique aerial photography points (N≥1) are planned according to the evidence-collecting specifications to construct a collection scheme that meets the requirements of multi-view evidence collection. At the same time, the spatial range and corresponding focusing parameters of each image patch or segmented unit are determined. Based on the photogrammetric collinearity equation, the captured image frame unit is inversely solved into UAV shooting spatial parameters, thereby determining the spatial coordinates of each shooting point. For details, see Figure 5 As shown, in accordance with the evidentiary requirements, one vertical orthophoto point and N oblique aerial photography points (N≥1) are planned, as follows: (a) Orthophoto point: located directly above the center point of the target, with the gimbal tilt angle vertically downward (90 degrees) to ensure that a complete vertical projection image of the target area is obtained; (b) Inclined aerial survey points: These points are deployed around the target at different azimuth angles (selecting directions with an azimuth difference ≥ 120 degrees), with a gimbal pitch angle of approximately 45 degrees, to provide multi-angle evidence of the actual land use. For each unit to be photographed, the following aerial survey parameters are planned:

[0036] In the formula, The waypoint coordinates (longitude, latitude) For the shooting angle of the gimbal, including the pitch angle. Roll angle And yaw angle κ. Among them, the pitch angle for orthographic shooting is 90 degrees vertically downward, the pitch angle for oblique shooting is 45 degrees, the roll angle is generally 0, and the yaw angle can be automatically adjusted by the gimbal.

[0037] S5. Automated and secure execution of drone evidence collection: Based on the calculated spatial coordinates of each photo point, the system automatically plans and generates the globally optimal evidence collection flight route; the drone autonomously executes the flight mission along the optimal evidence collection flight route and automatically adjusts the gimbal to a preset angle for shooting at each waypoint; Specifically, the drone relies on a high-precision RTK positioning system to autonomously perform evidence-gathering aerial photography tasks along a preset route, automatically adjusting the gimbal to a preset angle at each waypoint to complete the shooting. It employs a breakpoint-resume flight strategy, triggering autonomous return to recharge when the battery level falls below a safe threshold. After recharging, it resumes flight based on precise positioning to the point of interruption and continues the mission. Upon completion of the evidence-gathering task, it automatically returns to base and simultaneously uploads a mission report containing aerial photography parameters and evidence-gathering results.

[0038] S6. Encrypted transmission and compliance self-check of evidence results: After completing the shooting task, the drone automatically packages the evidence photos and uploads them to the drone control cloud server. The drone generates and encrypts the evidence results, encapsulates them into a DB format result package that meets the requirements, performs a compliance self-check on the result package, and pushes the result package to the cloud platform after verification.

[0039] Furthermore, the specific details of S2's dynamic calculation of the camera's actual coverage area based on the flight altitude parameters set for the UAV mission are as follows: Flight altitude parameters set according to the UAV mission Calculate the actual coverage area captured by the drone camera. : Single-span coverage area width :

[0040] Single-span ground coverage height :

[0041] Maximum coverage area per frame :

[0042] In the formula, To ensure the drone camera's field of view covers the width of the ground, , The actual width of the sensor. The actual height of the sensor. This is the actual focal length.

[0043] Furthermore, the specific details of the intelligent segmentation algorithm for image patches in S3 are as follows: Step 1: Generate the minimum bounding rectangle based on the target feature;

[0044] in, For the set of boundary points of the map patch, For convex hull calculation functions, To be the candidate bounding rectangle that can cover the patch, Rectangle area, This is the bounding rectangle with the smallest area obtained through optimization. Step 2, generating candidate segmentation boxes: at the four corner points of the smallest bounding rectangle Generate two types of drone frame bounding boxes with fixed horizontal and vertical dimensions as segmentation candidate boxes:

[0045] In the formula, These are the coordinates of the corner point. For corner index, To fix the width and height of the ground covered by the drone camera's image frame, [] represents the horizontal coverage area of ​​the rectangle. [] represents the vertical coverage area of ​​the rectangle; Step 3: Optimal Segmentation Box Decision: Select the best candidate box from the segmentation candidate boxes. Choose the candidate box that covers the largest effective area of ​​the patch as the optimal segmentation box region. The formula is as follows:

[0046] In the formula, For the candidate box set, This is a function for calculating the area of ​​a region. For the original target patch, Candidate boxes With the pattern The geometric intersection; Step 4: Update remaining polygons: Subtract the segmented regions from the remaining polygons. Iteratively update the remaining set of blobs using the following formula:

[0047] In the formula, For the first t The set of remaining graph spots at the next iteration In the first t The region segmented in the next iteration. For the updated version t +1 iterations of the remaining graph spot set; Repeat steps 1-3 for iterative segmentation until the length and width of the minimum bounding rectangle of the remaining patch are both smaller than the coverage size of the drone's frame. Step 5: Full Coverage of Remaining Image Areas: For the final remaining image area, maintaining the original frame parameters, select the frame position that can completely cover the remaining image area and has the highest overlap with the already photographed area as the final shooting frame, as shown in the following formula:

[0048] In the formula, To ensure the candidate boxes completely cover the remaining area. This represents the actual overlap area between the candidate bounding box and the original patch. The area is the candidate box area.

[0049] Specifically, while ensuring the candidate boxes Completely cover the remaining area Under the premise of selecting the original complete image patch The shooting frame with the largest overlapping area is selected to retain enough overlapping area to ensure the consistency of image stitching.

[0050] Furthermore, the specific details of the adaptive zoom control algorithm in S3 are as follows: Step 1: Calculate the minimum bounding rectangle of the target patch and determine the direction of the major axis and the coordinates of the center point; Step 2: Gimbal attitude calibration. Control the rotation of the gimbal camera to make the long side of the camera frame parallel to the long axis of the MBR, and align the center point of the camera frame with the center point of the MBR. Step 3: Zoom parameter decision. Based on the gimbal attitude calibration results in Step 2, the theoretical zoom value is calculated using the focal length-field-of-view mapping model. The formula is as follows:

[0051] In the formula, The base sensor area for 1.0x zoom. To set the coverage factor, the default value is 0.8. The proportion of the target image patch's MBR area to the zoomed image area; Specifically, the actual zoom value is dynamically optimized within the camera's optical zoom range [1.0x, 6.0x], generating a discrete sequence with a step size of 0.1x. Choose the minimum effective zoom magnification that satisfies the constraints: .

[0052] Step 4: Control the camera lens motor to step to the selected zoom level, and automatically trigger shooting after the gimbal stabilizes.

[0053] Furthermore, in S4, based on the photogrammetric collinearity equation, the specific details of inversely solving the image frame unit into the spatial parameters for UAV shooting are as follows: With fixed camera pitch, roll, and flight altitude, the focal length and yaw angle are determined by combining the geometric properties of the target imaging unit. The spatial coordinates of the point the UAV needs to reach to fully cover the current imaging unit are then calculated. , , ).

[0054] Specifically, the calculation of waypoint coordinates for UAV photography includes: based on the collinearity equation principle in photogrammetry, accurately calculating the spatial parameters of the UAV from the image area of ​​the patch. The basic principle is described as follows:

[0055] In the formula: , The image coordinates of the image point on the image plane; The focal length used when taking a picture with a camera; , , The object space coordinates of the camera station; , , These are the object space coordinates of the object point; , , The nine direction cosines are composed of the three exterior azimuth elements of the image.

[0056] Camera pitch angle, roll angle, and flight altitude Under fixed conditions, the focal length is uniquely determined based on the geometric properties of the target imaging unit. With the yaw angle κ, the collinearity equation is simplified to a function of the UAV's planar coordinates. and This is a deterministic mathematical problem. By solving this equation, the coordinates of the spatial points that the drone needs to reach to fully cover the imaging unit are output. , , ).

[0057] Furthermore, the specific details of S5's automatic planning and generation of the globally optimal evidence-gathering flight route based on the calculated spatial coordinates of each photo-taking point are as follows: Define the task priority function:

[0058] In the formula, For the image The center coordinates, The coordinates of the drone's take-off and landing points. Let be the Euclidean distance between the map patch and the take-off and landing points. The spatial attenuation coefficient, is a natural constant, approximately equal to 2.71828, and is the base of the exponential function; Based on the task priority function, the task priority coefficient of the image patch to be verified is calculated; based on the task priority coefficient and the set of aerial photography parameters, a flight path with the shortest total flight distance is planned and generated.

[0059] In the formula, To find the waypoint arrangement that minimizes the total distance, m The total number of waypoints. The distance between waypoints is expressed in Euclidean form. For the first j The spatial coordinates of the waypoint to be visited. In order to be in The spatial coordinates of the next waypoint to be visited.

[0060] Specifically, the above formula generates the globally optimal evidence-gathering flight path. The optimal evidence-gathering flight path ensures that the UAV visits all target waypoints efficiently and in an orderly manner, and automatically adjusts the gimbal to a preset angle at each waypoint to complete the shooting.

[0061] Further, see Figure 6 As shown, S6 includes an anti-counterfeiting verification scheme for evidence patches, wherein, The specific content of the evidence patch anti-counterfeiting verification scheme is as follows: When the drone takes pictures, it automatically embeds a digital watermark containing shooting coordinates, shooting time and equipment information, generates a unique hash value in real time, generates a verification code through key signing and encrypts the result data, and the key verification information is synchronously uploaded to the blockchain for evidence storage; the result package is transmitted in standard DB format, and the receiving end ensures the authenticity and integrity of the photo by comparing and verifying the photo metadata, verifying the hash value and the validity of the verification code.

[0062] Specifically, if the verification passes, the result is valid; if it fails, the data and result are recorded for future reference. Once the verification is successful, the results are pushed to the platform.

[0063] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0064] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for providing evidence using unmanned aerial vehicles (UAVs) based on intelligent image segmentation and adaptive zoom control, characterized in that, Includes the following steps: S1. Import Vector Patch Data: Import patch data containing GeoJSON and SHAPE format vectors and attribute information. Create a structured data import template based on patch attributes, automatically extract core field information of patches, complete batch data verification and format conversion, and generate a set of tasks to be processed. S2. Patch Data Analysis and Classification: Based on the task set to be processed, the patch data is classified and analyzed sequentially. First, the actual coverage area of ​​the camera frame is dynamically calculated according to the flight altitude parameters set by the UAV mission. The minimum bounding rectangle is calculated through the patch vector boundary. A multi-scale classification system combining patch geometry and the coverage capability of the UAV's shooting frame is constructed to classify patches into large patches, small patches, regular patches, and special-shaped patches. S3. Differentiated processing strategy for image patches: After classifying image patches according to the multi-scale classification system, a differentiated processing strategy is adopted. For large image patches, an intelligent image patch segmentation algorithm is introduced, and an iterative geometric segmentation is performed using the minimum bounding rectangle long side alignment strategy. For small image patches, an adaptive zoom control algorithm is triggered for shooting, and the image resolution is improved to the 0.2m / pixel standard through focal length compensation. For regular image patches, the standard shooting mode is used to directly cover them completely. For image patches with special shapes, an image patch segmentation-zoom collaborative optimization mechanism is activated. S4. Generation of UAV Evidence-Collecting Shooting Locations: Based on the image patch processing results, for each captured image patch or segmented sub-image patch, one vertical orthogonal aerial photography point and N oblique aerial photography points are planned according to the evidence-collecting specifications to construct a collection scheme that meets the requirements of multi-view evidence collection. At the same time, the spatial range and corresponding focusing parameters of each image patch or segmented unit are determined. Based on the photogrammetric collinearity equation, the captured image frame unit is inversely solved into UAV shooting spatial parameters, thereby determining the spatial coordinates of each shooting point. S5. Automated and secure execution of drone evidence collection: Based on the calculated spatial coordinates of each photo point, the system automatically plans and generates the globally optimal evidence collection flight route; the drone autonomously executes the flight mission along the optimal evidence collection flight route and automatically adjusts the gimbal to a preset angle for shooting at each waypoint; S6. Encrypted transmission and compliance self-check of evidence results: After completing the shooting task, the drone automatically packages the evidence photos and uploads them to the drone control cloud server. The drone generates and encrypts the evidence results, encapsulates them into a DB format result package that meets the requirements, performs a compliance self-check on the result package, and pushes the result package to the cloud platform after verification.

2. The UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control according to claim 1, characterized in that, The specific details of S2's dynamic calculation of the actual coverage area of ​​the camera frame based on the flight altitude parameters set for the UAV mission are as follows: Flight altitude parameters set according to the UAV mission Calculate the actual coverage area captured by the drone camera. : Single-span coverage area width : Single-span ground coverage height : Maximum coverage area per frame : In the formula, To ensure the drone camera's field of view covers the width of the ground, , The actual width of the sensor. The actual height of the sensor. This is the actual focal length.

3. The UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control according to claim 1, characterized in that, The specific content of the S3 image patch intelligent segmentation algorithm is as follows: Step 1: Generate the minimum bounding rectangle based on the target feature; in, For the set of boundary points of the map patch, For convex hull calculation functions, To be the candidate bounding rectangle that can cover the patch, Rectangle area, This is the bounding rectangle with the smallest area obtained through optimization. Step 2, generating candidate segmentation boxes: at the four corner points of the smallest bounding rectangle Generate two types of drone frame bounding boxes with fixed horizontal and vertical dimensions as segmentation candidate boxes: In the formula, These are the coordinates of the corner point. For corner index, To fix the width and height of the ground covered by the drone camera's image frame, [] represents the horizontal coverage area of ​​the rectangle. [] represents the vertical coverage area of ​​the rectangle; Step 3: Optimal Segmentation Box Decision: Select the best candidate box from the segmentation candidate boxes. Choose the candidate box that covers the largest effective area of ​​the patch as the optimal segmentation box region. The formula is as follows: In the formula, For the candidate box set, This is a function for calculating the area of ​​a region. For the original target patch, Candidate boxes With the pattern The geometric intersection; Step 4: Update remaining polygons: Subtract the segmented regions from the remaining polygons. Iteratively update the remaining set of blobs using the following formula: In the formula, For the first t The set of remaining graph spots at the next iteration In the first t The region segmented in the next iteration. For the updated version t +1 iterations of the remaining graph spot set; Repeat steps 1-3 for iterative segmentation until the length and width of the minimum bounding rectangle of the remaining patch are both smaller than the coverage size of the drone's frame. Step 5: Full Coverage of Remaining Image Areas: For the final remaining image area, maintaining the original frame parameters, select the frame position that can completely cover the remaining image area and has the highest overlap with the already photographed area as the final shooting frame, as shown in the following formula: In the formula, To ensure the candidate boxes completely cover the remaining area. This represents the actual overlap area between the candidate bounding box and the original patch. The area is the candidate box area.

4. The UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control according to claim 1, characterized in that, The specific details of the adaptive zoom control algorithm in S3 are as follows: Step 1: Calculate the minimum bounding rectangle of the target patch and determine the direction of the major axis and the coordinates of the center point; Step 2: Gimbal attitude calibration. Control the rotation of the gimbal camera to make the long side of the camera frame parallel to the long axis of the MBR, and align the center point of the camera frame with the center point of the MBR. Step 3: Zoom parameter decision. Based on the gimbal attitude calibration results in Step 2, the theoretical zoom value is calculated using the focal length-field-of-view mapping model. The formula is as follows: In the formula, The base sensor area for 1.0x zoom. To set the coverage factor, the default value is 0.

8. The proportion of the target image patch's MBR area to the zoomed image area; Step 4: Control the camera lens motor to step to the selected zoom level, and automatically trigger shooting after the gimbal stabilizes.

5. The UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control according to claim 1, characterized in that, The specific details of how S4 inversely calculates the spatial parameters of UAV photography based on the collinearity equation of photogrammetry are as follows: With fixed camera pitch, roll, and flight altitude, the focal length and yaw angle are determined by combining the geometric properties of the target imaging unit. The spatial coordinates of the point the UAV needs to reach to fully cover the current imaging unit are then calculated. , , ).

6. The UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control according to claim 1, characterized in that, The specific details of S5's automatic planning and generation of the globally optimal evidence-gathering flight route based on the calculated spatial coordinates of each photo-taking point are as follows: Define the task priority function: In the formula, For the image The center coordinates, The coordinates of the drone's take-off and landing points. Let be the Euclidean distance between the map patch and the take-off and landing points. The spatial attenuation coefficient, is a natural constant and the base of the exponential function; Based on the task priority function, the task priority coefficient of the image patch to be verified is calculated; based on the task priority coefficient and the set of aerial photography parameters, a flight path with the shortest total flight distance is planned and generated. In the formula, To find the waypoint arrangement that minimizes the total distance, m The total number of waypoints. The distance between waypoints is expressed in Euclidean form. For the first j The spatial coordinates of the waypoint to be visited. In order to be in The spatial coordinates of the next waypoint to be visited.

7. The UAV evidence collection method based on intelligent patch segmentation and adaptive zoom control according to claim 1, characterized in that, S6 includes an anti-counterfeiting verification scheme for evidence-based image patches, in which... The specific content of the evidence patch anti-counterfeiting verification scheme is as follows: When the drone takes pictures, it automatically embeds a digital watermark containing shooting coordinates, shooting time and equipment information, generates a unique hash value in real time, generates a verification code through key signing and encrypts the result data, and the key verification information is synchronously uploaded to the blockchain for evidence storage; the result package is transmitted in standard DB format, and the receiving end ensures the authenticity and integrity of the photo by comparing and verifying the photo metadata, verifying the hash value and the validity of the verification code.

Images