2D model generation method and emergency aerial survey equipment
Through the image fusion processing of the front and rear camera modules, the problem of information missing in extreme environments of single-view aerial survey is solved, and a high-precision and high-completeness 2D model is generated to support disaster information extraction and disaster damage in disaster environments Evaluate.
Patent Information
- Application Number
- CN202411007433.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-12
AI Technical Summary
In extreme disaster environments, single-view drone aerial survey is prone to problems such as missing information and shadow occlusion, resulting in insufficient accuracy of the generated 2D model and affecting subsequent image analysis.
The front and rear camera modules are used to synchronize the image of the target area, and the image fusion processing is realized through feature matching and spatial registration, and a more comprehensive fusion image is generated to overcome the defects of a single-view angle.
It improves the reconstruction accuracy and completeness of the 2D model, reduces the risk of data loss caused by environmental interference, and achieves high-precision and high-reliability disaster information extraction and disaster loss assessment.
Smart Images

Figure CN118941667A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of emergency rescue equipment, and in particular to a 2D model generation method and emergency aerial survey equipment. Background Art
[0002] Natural disasters and emergencies pose a serious threat to people's lives, property and social and economic development. In disaster emergency response and rescue operations, timely and accurate acquisition of geographical information and damage in the disaster area is the key to making scientific decisions and taking effective actions. Traditional ground surveys and manual surveys can no longer meet the needs of current disaster management. It is urgent to develop fast, efficient and intelligent information acquisition and processing methods based on advanced remote sensing technology.
[0003] In related technologies, drone aerial survey technology is used to obtain high-resolution image data, and through algorithms such as image matching and correction, high-precision and high-detail 2D models can be generated. 2D models can provide intuitive and accurate disaster information for rescue personnel and command departments. By analyzing and interpreting 2D models, the scope, extent and distribution characteristics of disasters can be quickly assessed, and the damage to key objects such as affected buildings, roads, and farmland can be identified, providing a decision-making basis for the deployment of rescue forces and the transportation of materials.
[0004] However, disaster areas often contain a large amount of suspended particles such as smoke and dust, which can easily have a serious impact on the image quality of drone aerial surveys. In extreme cases, there are limitations on the deployment of drones in the same area. If a single drone is used for aerial surveys, the single-view aerial survey images are prone to information loss, shadow occlusion and other problems, resulting in insufficient accuracy of the generated 2D model, affecting subsequent image analysis. Summary of the invention
[0005] The present application provides a 2D model generation method and emergency aerial survey equipment for more accurately generating a 2D model when there is partial viewing angle occlusion.
[0006] In a first aspect, the present application provides a method for generating a 2D model, which is applied to emergency aerial survey equipment, the emergency aerial survey equipment including a front camera module and a rear camera module, the method including: in response to an acquisition instruction sent by a user terminal, obtaining a first image taken by the front camera module and a second image taken by the rear camera module; the first image includes preset feature points; flipping the second image, and extracting matching feature points in the second image at the same position as the preset feature points; based on the preset feature points and the matching feature points, fusing the second image and the first image to obtain a fused image; based on the fused image, generating a 2D model and sending it to the user terminal.
[0007] In the above embodiment, the emergency aerial survey equipment can automatically obtain images of different perspectives taken by the front and rear camera modules, and by flipping the rear image, extracting matching feature points, and then fusing it with the front image, a fused image with a more comprehensive perspective and richer information can be generated. The 2D model constructed on this basis can more completely reflect the geometric structure and texture details of the target scene.
[0008] In combination with some embodiments of the first aspect, in some embodiments, based on preset feature points and matching feature points, the second image and the first image are fused to obtain a fused image, specifically including: dividing the first image into a number of first sub-regions, each first sub-region including at least one preset feature point; dividing the second image into the same number of second sub-regions as the first image, each second sub-region including matching feature points at the same position as the preset feature points; based on the preset feature points of the first sub-region, feature matching is performed on the matching feature points of the second sub-region to obtain a mapping relationship between pixel points in the first sub-region and pixel points in the second sub-region; according to the mapping relationship, the pixel values of the first sub-region are mapped to obtain target fused pixel points of the second sub-region; and all target fused pixel points of the second sub-region are spliced to obtain a fused image.
[0009] In the above embodiment, the emergency aerial survey equipment introduces a sub-region division and feature matching mechanism in the image fusion process. By dividing the front and rear images into multiple corresponding sub-regions and performing local feature matching in each sub-region, the defect that the global registration algorithm is susceptible to local distortion can be overcome, and a more refined and reliable pixel-level fusion can be achieved. At the same time, by splicing the target fusion pixel points generated by each sub-region, the final fused image is obtained, which not only ensures global consistency but also takes into account local accuracy.
[0010] In combination with some embodiments of the first aspect, in some embodiments, the pixel values of the first sub-area are mapped according to a mapping relationship to obtain target fused pixel points of the second sub-area, specifically including: extracting the pixel value of each pixel point in the first sub-area, mapping it to the corresponding pixel point position of the second sub-area through a mapping relationship, to obtain an initial fused pixel point; performing smoothing correction on the initial fused pixel point to obtain a target fused pixel point; extracting independent pixels in the second sub-area that have no direct mapping relationship with the pixels of the first sub-area; and calculating and generating target fused pixel points of independent pixels through an interpolation algorithm based on the color values of the generated target fused pixel points in the neighborhood of the independent pixel points and combined with the spatial distance difference.
[0011] In the above embodiment, when generating target fused pixel points, the emergency aerial survey equipment not only considers the pixel value transfer under the direct mapping relationship, but also performs reasonable interpolation calculations on independent pixel points that lack direct mapping. First, by smoothing and correcting the initial fused pixel points, local noise is eliminated and the smoothness of the fused image is improved; secondly, the color information of the generated fused pixel points in the neighborhood of the independent pixel points is used, combined with the spatial distance difference for weighted interpolation, to fill the vacant area in a soft and natural way, and to maximize the integrity of the image.
[0012] In combination with some embodiments of the first aspect, in some embodiments, the second image is flipped, and matching feature points in the second image at the same position as the preset feature points are extracted, specifically including: acquiring posture parameters of the first image and the second image; the posture parameters include the three-axis rotation angle, focal length and resolution during shooting; based on the posture parameters, calculating the position offset and angle offset of the second image relative to the first image; according to the position offset, the second image is translationally corrected so that it coincides with the coordinate origin of the first image, and at the same time, according to the angle offset, the second image after translation is rotationally corrected so that it maintains the same angle direction as the first image; the second image after the translation correction and rotation correction is completed is scale-normalized so that it has the same resolution and size as the first image, to obtain a second target image; extracting the pixel position in the second target image that is the same as the two-dimensional coordinates of the preset feature point in the first image as the initial position point; taking the initial position point as the center, intercepting a local image block of a preset range size on the second target image; in the local image block, searching for image points whose similarity with the preset feature points of the first image is higher than a preset similarity threshold as the matching feature points.
[0013] In the above embodiment, the emergency aerial survey equipment makes full use of the posture parameters of the front and rear camera modules when extracting matching feature points. By obtaining key parameters such as the three-axis rotation angle, focal length, and resolution during shooting, the position and angle offset of the rear image relative to the front image can be accurately calculated, and the rear image can be accurately translated, rotated, and scaled normalized. After the rear image and the front image are aligned at the pixel level, the optimal matching feature points are extracted through local similarity search, which greatly improves the accuracy and robustness of feature matching.
[0014] In combination with some embodiments of the first aspect, in some embodiments, before the step of acquiring the first image taken by the front camera module and the second image taken by the rear camera module in response to the acquisition instruction sent by the user terminal, the method also includes: acquiring the current first posture information of the front camera module and the current second posture information of the rear camera module, and determining the first posture adjustment instruction and the second posture adjustment instruction based on the preset aerial survey posture; the preset aerial survey posture includes the three-axis rotation angle, the translation position and the lens parameters; sending the first posture adjustment instruction to the front camera module, and sending the second posture adjustment instruction to the rear camera module, so as to adjust the front camera module and the rear camera module respectively.
[0015] In the above embodiment, the emergency aerial survey equipment can automatically adjust the attitude parameters of the front and rear camera modules according to the preset aerial survey attitude to meet the optimal data collection conditions. By obtaining the current attitude information and comparing it with the preset aerial survey attitude, generating attitude adjustment instructions and feeding back to the camera module in real time, it can ensure that key factors such as shooting angle, field of view, and imaging quality are always in an ideal state. This active attitude control mechanism overcomes external interference such as drone flight vibration and wind speed changes, and ensures image quality from the source of the data.
[0016] In combination with some embodiments of the first aspect, in some embodiments, after the step of generating a 2D model based on the fused image and sending it to the user terminal, the method also includes: obtaining image annotation information sent by the user terminal; the image annotation information is generated by the user terminal based on the 2D model, including the category, outline and position of the target object marked on the 2D model; based on the image annotation information, the 2D model is semantically segmented to obtain a segmented 2D model; the segmented 2D model includes the target object after classification and annotation; according to the preset feature category, the category information and position information of multiple target objects after classification and annotation are counted to generate feature distribution information of the corresponding area of the 2D model; and the feature distribution information is sent to the user terminal.
[0017] In the above embodiment, the emergency aerial survey equipment can obtain and process interactive image annotation information from the user terminal. The user can conveniently annotate the key semantics of the target object such as category, outline, and position on the 2D model. The equipment automatically performs semantic segmentation on the 2D model based on the annotation information to achieve refined extraction and classification of the target object. Furthermore, by statistically analyzing the location distribution of each category of objects, intuitive and comprehensive object distribution information is generated.
[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of sending the feature distribution information to the user terminal, the method also includes: obtaining a damage assessment request sent by the user terminal; the damage assessment request includes the category of the feature to be assessed; according to the category of the feature to be assessed, extracting the corresponding feature to be assessed from the segmented 2D model; performing pixel-by-pixel detection on the outline of the feature to be assessed, counting the number of complete pixel points of the outline, and obtaining the integrity of the feature to be assessed; if the integrity of the feature to be assessed is lower than a preset integrity threshold, determining that the feature to be assessed is a damaged feature; determining the damage level of the feature to be assessed based on the proportion of missing pixels in the outline of the feature to be assessed; generating a damage assessment report for the feature to be assessed, and sending the damage assessment report to the user terminal; the damage assessment report includes the location, damage level and quantity statistical information of each damaged feature.
[0019] In the above embodiment, the emergency aerial survey equipment can automatically perform integrity analysis and damage level determination of key features according to the damage assessment request submitted by the user terminal. First, according to the category of the feature to be assessed specified by the user, the target feature in the segmented 2D model is accurately located and extracted; second, the feature contour is detected pixel by pixel, the completeness of the contour is quantitatively assessed, and the damaged feature is identified; finally, the specific level of damage is further determined based on the proportion of missing pixels. The assessment results are fed back to the user in the form of a damage assessment report, which intuitively presents detailed information such as the spatial location, damage status, and quantity statistics of key features.
[0020] In a second aspect, an embodiment of the present application provides an emergency aerial survey device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the emergency aerial survey device to execute the method described in the first aspect and any possible implementation method of the first aspect.
[0021] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on emergency aerial survey equipment, enables the emergency aerial survey equipment to execute the method described in the first aspect and any possible implementation of the first aspect.
[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on an emergency aerial survey device, enable the emergency aerial survey device to execute the method described in the first aspect and any possible implementation manner of the first aspect.
[0023] It can be understood that the emergency aerial survey equipment provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, and will not be repeated here.
[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. Since the front and rear camera modules are independently photographed and fused, multi-perspective image data of the target scene can be obtained, which effectively solves the problems of incomplete single-perspective information and missing texture in the existing technology, and thus realizes a more comprehensive two-dimensional reconstruction of the target scene.
[0025] 2. Due to the adoption of an image fusion strategy based on sub-region division and local feature matching, accurate registration and seamless stitching of the front and rear images can be achieved at the pixel level, effectively solving the problem in the existing technology that global registration is easily affected by local distortion and the fusion accuracy is not high, thereby achieving the generation of orthophotos with higher fidelity and more natural texture.
[0026] 3. Due to the use of a processing mechanism that combines user interactive image annotation with semantic segmentation, it is possible to accurately extract and classify key features in the 2D model and count their spatial distribution characteristics, effectively solving the problem that the existing model lacks feature semantics and is difficult to meet professional analysis needs, thereby achieving the refinement of disaster perception and decision-making command. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 It is a flow chart of a method for generating a 2D model in an embodiment of the present application; Figure 2 This is a schematic diagram of an application scenario of emergency aerial survey equipment in an embodiment of the present application; Figure 3 is another flowchart of the method for generating a 2D model in an embodiment of the present application; Figure 4 It is a schematic diagram of the structure of a physical device of the emergency aerial survey equipment in the embodiment of the present application. DETAILED DESCRIPTION
[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification of the present application, the singular expressions "one", "a kind of", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.
[0029] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.
[0030] For ease of understanding, the application scenarios of the embodiments of the present application are introduced below.
[0031] In a certain area, due to the influence of heavy rainfall for several days, the water levels of several rivers have risen sharply, and many villages and towns have been invaded by floods. The rescue command center urgently needs to grasp the geographical information and disaster distribution of the disaster-stricken areas in order to formulate a scientific rescue plan. The traditional manual on-site survey method is difficult to obtain comprehensive and accurate information in a short period of time, and cannot meet the timeliness requirements of emergency rescue. In order to quickly and efficiently obtain image data of the disaster area, the rescue department decided to use drone aerial survey technology to take aerial photos of key areas through airborne camera equipment, and use the obtained high-resolution image data to build a digital map model of the disaster area to provide a reference for subsequent rescue operations. In the process of drone aerial survey, due to the large-scale flood disaster, there is a lot of fog and suspended particles in some areas, coupled with low visibility under severe weather conditions, which seriously affects the quality of aerial survey images.
[0032] In the related art, the geographical information of the disaster area can be quickly acquired and the 2D model can be automatically generated by using drone aerial survey. The following describes the scenario of using the 2D model generation method in the related art.
[0033] In order to obtain the geographic information model of the disaster area as quickly as possible, the rescue headquarters sent a drone with high-definition camera equipment to the worst-hit areas for aerial survey. The drone hovered over the village, continuously photographing the disaster situation below through the onboard camera. After half an hour of flight and shooting, the drone returned to the take-off point and transmitted the collected image data to the data processing center of the ground headquarters.
[0034] Using the acquired image data and traditional 3D reconstruction algorithms, the technicians tried to construct a 2D geographic information model of the area. However, due to the large amount of smoke and shadows in the images collected by the drone, many key features of the objects were missing, which directly affected the accuracy of matching and correction. At the same time, due to the limitations of the aerial survey perspective, it was difficult to fully record the information of some narrow streets and building interiors. The resulting 2D model lacked details and had large blank areas, which was far from the actual shape of the objects.
[0035] After receiving the 2D model, the rescuers found that many important landmarks and roads could not be accurately located in the model, and the damage to some buildings could not be determined. Such a model cannot provide strong support for rescue decisions and actions, but may mislead rescue forces and delay precious rescue time. It can be seen that traditional single-view aerial survey methods are often inadequate in extreme disaster environments, and there is an urgent need to develop more intelligent and efficient surveying and mapping solutions.
[0036] The 2D model generation method in the embodiment of the present application is adopted, by using the front and rear camera modules to synchronously shoot the target area, and performing feature matching and spatial registration on the two sets of images, the defect repair and information completion of the single-view image are realized, which not only improves the accuracy and completeness of the 2D model reconstruction, but also greatly reduces the risk of data loss caused by environmental interference. The following introduces the scene using the 2D model generation method in the present application.
[0037] In view of the above problems, the rescue command decided to adopt the 2D model generation method proposed in this application. The new aerial survey drone is equipped with two camera modules, front and rear, which can simultaneously shoot the target area from different angles. At the same time, the data processing center introduced a 2D model reconstruction algorithm based on dual-view image fusion, which can make full use of the image information collected by the front and rear cameras to generate a detailed, complete and accurate 2D geographic information model.
[0038] The drone passed over the disaster area and took pictures of the disaster situation through the front and rear cameras simultaneously. Even if the same feature cannot be fully imaged from a certain perspective due to smoke or obstacles, key feature information can be captured from another perspective. After the data is transmitted back, the system automatically extracts the key feature points in the two sets of images, finds the same-name points under different perspectives through feature matching, and then realizes the spatial fusion and superposition of the two sets of images through the transformation and correction of the coordinate system. In this way, the missing or polluted areas under a single perspective can be supplemented and repaired with images from another perspective, greatly improving the success rate of model reconstruction. In the end, a high-resolution, high-completeness 2D geographic information model was successfully generated. On this model map, the distribution of features and scene details in the disaster-stricken area, collapsed houses, broken bridges, and flooded roads can all be clearly identified.
[0039] It can be seen that the 2D model generation method in the embodiment of the present application can not only realize the rapid and automatic generation of 2D models, but also effectively solve the problems of information occlusion and data missing in extreme disaster environments, thereby realizing high-precision and highly reliable disaster information extraction and disaster damage assessment needs.
[0040] For ease of understanding, the following describes the process of the method provided by this implementation in combination with the above scenario. Figure 1, which is a flow chart of the method for generating a 2D model in an embodiment of the present application.
[0041] S101: In response to a capture instruction sent by a user terminal, acquire a first image taken by a front camera module and a second image taken by a rear camera module.
[0042] The user terminal refers to a mobile device that communicates and controls the emergency aerial survey equipment, such as a mobile phone or tablet computer, and is used to send task instructions to the emergency aerial survey equipment and receive aerial survey results. The acquisition instruction refers to a control instruction issued by the user terminal to the emergency aerial survey equipment to start collecting image data.
[0043] The front camera module and the rear camera module refer to two high-definition cameras installed on the front and rear of the emergency aerial survey equipment, which are used to synchronously shoot the target area from different perspectives. The first image refers to the image of the target area obtained by the front camera module, and the second image refers to the image of the target area obtained by the rear camera module. The first image contains preset feature points, and the preset feature points can represent a set of obvious image features pre-calibrated in the first image, such as corner points, edges, etc., for subsequent image registration and fusion processing. In some embodiments, the preset feature points can ignore the features and be defined as the corner points of the preset grid of the image. The grid size of the preset grid can be set according to the accuracy.
[0044] like Figure 2 The figure is a schematic diagram of an application scenario of the emergency aerial survey equipment in the embodiment of the present application. Figure 2 Reference numeral 201 is the body of the emergency aerial survey equipment, i.e., a drone, which flies from left to right. The emergency aerial survey equipment is provided with a rear camera module 202 and a front camera module 203 . Figure 2 In the figure, 204 and 205 are respectively the areas A and B to be tested on the ground, and both areas to be tested are blocked by smoke. It can be seen from the figure that since the shooting angles of the rear camera module 202 and the front camera module 203 are opposite, a more comprehensive geographic image can be scanned.
[0045] In step S101, when the emergency aerial survey equipment performs the disaster information collection task, it receives the collection instruction through the communication link with the user terminal, and then starts to perform the image data collection operation. Specifically, after receiving the collection instruction, the emergency aerial survey equipment controls the front camera module and the rear camera module to simultaneously shoot the target area, the front camera module shoots to obtain the first image, and the rear camera module shoots to obtain the second image. In the process of shooting the first image, the emergency aerial survey equipment also needs to extract and mark a set of preset feature points in the image for subsequent image registration.
[0046] S102: Flip the second image, and extract matching feature points in the second image at the same position as the preset feature points.
[0047] Image flipping refers to mirroring the second image in the horizontal or vertical direction to make it consistent with the shooting angle of the first image. Matching feature points refer to feature points in the flipped second image that correspond one-to-one with the preset feature points of the first image in terms of spatial position and scale, and are used for subsequent image fusion processing.
[0048] After acquiring the first image and the second image, due to the difference in the shooting angles of the two camera modules, directly aligning and fusing the two images will result in large errors. Therefore, the second image needs to be flipped first to keep it consistent with the direction of the first image. Specifically, the emergency aerial survey equipment first calculates the direction and angle that the second image needs to be flipped based on the installation position and posture relationship of the two camera modules; then uses image processing algorithms, such as affine transformation, perspective transformation, etc., to perform a flipping operation on the second image. After the flipping is completed, the matching feature points corresponding to the preset feature points of the first image are extracted from the second image to provide key point information for subsequent image fusion.
[0049] In some embodiments, matching feature points can be extracted in the following manner: Optionally, after the second image is flipped, the emergency aerial survey equipment maps the two-dimensional coordinates of the preset feature points to the coordinate system of the second image, and captures a small local image block on the second image with the mapping position as the center; then, using a feature matching algorithm, the local image block is searched for image points with the highest similarity to the preset feature points, and marked as matching feature points. Optionally, the emergency aerial survey equipment first uses a feature point extraction algorithm to extract a set of image feature points on the flipped second image; then, the spatial position relationship between these feature points and the preset feature points of the first image is calculated, and for feature point pairs with the closest distance and scale, the feature points on the second image are marked as matching feature points. It is understandable that other methods can also be used to extract matching feature points on the flipped second image, such as using a deep learning algorithm for feature matching, etc., which are not limited here.
[0050] S103: Based on the preset feature points and the matching feature points, the second image and the first image are fused to obtain a fused image.
[0051] Image fusion processing refers to the use of image registration and stitching algorithms to superimpose the first image and the second image to generate a target image containing complete information from two perspectives. A fused image refers to a new image obtained through image fusion processing, which contains the comprehensive feature information of the first image and the second image.
[0052] After completing the flipping of the second image and extracting the matching feature points, the emergency aerial survey equipment needs to further use the feature point information to align and fuse the two images. Specifically, the emergency aerial survey equipment first calculates the coordinate transformation relationship between the first image and the second image based on the preset feature points and matching feature points, using homography matrix estimation or other parameter solving methods; then uses the transformation relationship to geometrically correct the two images so that they are accurately aligned in space. Based on the alignment, the two images are stitched together using image stitching algorithms, such as weighted averaging, pyramid fusion, etc., to generate a fused image containing complete target area information.
[0053] In some embodiments, image fusion processing can be achieved in the following ways: Optionally, the emergency aerial survey equipment first estimates the homography matrix based on the preset feature points and matching feature points using the RANSAC algorithm, and uses the matrix to perform a projection transformation on the second image to align it with the first image; then the two images are divided into blocks, and pixel-level fusion is performed in each image block using a weighted average algorithm, and the stitching seams are processed using an image pyramid technique; finally, all image blocks are stitched together to obtain the final fused image. Optionally, the emergency aerial survey equipment first constructs a Laplacian pyramid for the first image and the second image respectively, and then performs image registration and fusion processing layer by layer; in each layer, the registration parameters are first estimated using the preset feature points and matching feature points, and then the pyramid image is transformed using the registration parameters, and then the image is fused using the weighted average method; finally, the fused image is reconstructed through the Laplacian pyramid. It is understandable that other methods can also be used to fuse the first image and the second image, such as using an end-to-end image fusion model based on deep learning, etc., which is not limited here.
[0054] S104: Generate a 2D model based on the fused image and send it to the user terminal.
[0055] Among them, the 2D model refers to the conversion of the fused image into a vector map model in two-dimensional space coordinates through computer vision algorithms. The model contains rich information such as the contours, locations, and attributes of the objects in the target area.
[0056] After obtaining the fused image, the emergency aerial survey equipment needs to further generate a 2D vector map model suitable for analysis and display based on user needs using image interpretation and feature extraction technology. Specifically, the emergency aerial survey equipment first pre-processes the fused image, such as image enhancement and noise removal, to improve image quality; then uses the image segmentation algorithm to divide the image into several semantic areas and extract the contour features of each area; then, according to the pre-established feature classification system, the semantic attributes of each area are annotated to form 2D map elements containing information such as feature categories and vector boundaries; finally, all map elements are organized into a 2D map data format that meets industry standards and sent to the user terminal through a wireless communication network for users to view and apply.
[0057] In some embodiments, a 2D model can be generated in the following manner: Optionally, the emergency aerial survey equipment first performs geometric correction and orthorectification processing on the fused image to eliminate distortion and projection errors; then uses a semantic segmentation model based on deep learning to perform pixel-level classification on the corrected image to obtain a segmentation result containing land object categories such as buildings, roads, and vegetation; then uses the Douglas-Peucker algorithm to vectorize the segmentation result and extract the vector outline of each land object; finally, according to the map drawing specifications, attribute annotations are added to each vector element, and it is saved in a universal 2D map data format and transmitted to the user terminal via a wireless communication network.
[0058] In the above embodiment, the basic principle and implementation process of generating a 2D model using dual-view image fusion technology are mainly described. In practical applications, the 2D model generated by this solution can be further combined with intelligent algorithms such as manual annotation and semantic segmentation to achieve automatic classification and attribute extraction of damaged objects, providing rich data support for post-disaster assessment and rescue decision-making. The following is a supplement to the scenario of this embodiment.
[0059] After the rescue work came to an end, post-disaster assessment and reconstruction work began immediately. The high-precision 2D model generated by this solution has become an important basis for disaster damage statistics and assessment. Local government departments manually marked the categories and contours of various damaged buildings, farmlands, roads and other target objects on the model and uploaded them to the rescue command platform. The platform automatically performs semantic segmentation on the annotated 2D model, classifies each pixel in the model according to the preset land object classification system, and finally generates a structured disaster-affected land object distribution database.
[0060] Based on this database, the emergency management department conducted a comprehensive damage assessment of the disaster area. The system performs integrity detection on each marked damaged feature, counts the percentage of pixels with complete outlines, and gives a corresponding damage rating based on the material and structural characteristics of the feature itself. For key assessment objects such as buildings, the system can also determine the direction of house collapse and force characteristics based on the missing pixels, providing a reference for the subsequent formulation of repair and reinforcement plans. All assessment results are fed back to government departments at all levels in the form of damage assessment reports to guide post-disaster reconstruction work.
[0061] At the same time, enterprises and the public can also query the geographical distribution information and disaster damage in the disaster area through relevant service platforms. Insurance companies review and settle property insurance claims in the disaster area based on this, and volunteer groups can also carry out targeted donation and assistance activities based on the disaster heat map. Through mobile phone applications, the affected people can intuitively browse the damaged images and assessment results of their own houses, which greatly improves the transparency and convenience of post-disaster government services. The digital disaster management system constructed by this solution realizes the sharing of disaster information by multiple parties and improves the scientificity and orderliness of post-disaster reconstruction work.
[0062] After combining the above scenarios, the following is a more detailed description of the process of the method provided by this implementation. Figure 3 , is another flow chart of the method for generating a 2D model in an embodiment of the present application.
[0063] S301: In response to a capture instruction sent by a user terminal, acquire a first image taken by a front camera module and a second image taken by a rear camera module.
[0064] Referring to step S101 , the emergency aerial survey equipment acquires a first image and a second image.
[0065] In some embodiments, the emergency aerial survey equipment will obtain the current first posture information of the front camera module and the current second posture information of the rear camera module, and determine the first posture adjustment instruction and the second posture adjustment instruction based on the preset aerial survey posture; the preset aerial survey posture includes three-axis rotation angle, translation position and lens parameters; the first posture adjustment instruction is sent to the front camera module, and the second posture adjustment instruction is sent to the rear camera module to adjust the front camera module and the rear camera module respectively.
[0066] Among them, attitude information refers to a set of parameters that describe the position and direction of an object in three-dimensional space, usually including three-axis (i.e. pitch angle, roll angle and yaw angle) rotation angle, translation position (i.e. x, y, z coordinates) and lens parameters (such as focal length, field of view, etc.). The attitude adjustment instruction represents a control instruction to adjust the attitude of the camera module according to the difference between the preset aerial survey attitude and the current attitude, so as to make the attitude of the camera module meet the aerial survey requirements. For example, if the preset aerial survey attitude requires the camera lens to remain horizontal and downward, and the current attitude deviates from this requirement, it is necessary to control the rotation of the camera's pan / tilt motor through the attitude adjustment instruction to restore it to the preset horizontal orientation.
[0067] Specifically, after obtaining the current posture information of the front and rear camera modules, the emergency aerial survey equipment needs to compare it with the preset aerial survey posture, determine the posture difference between the two, and generate corresponding posture adjustment instructions accordingly to control the movement of the camera module so that it meets the aerial survey requirements. The emergency aerial survey equipment first reads the current posture parameters such as the three-axis rotation angle and translation position from the camera module's gyroscope, accelerometer and other sensors; then compares these parameters with the corresponding parameters in the preset aerial survey posture one by one, calculates their difference, and obtains the posture deviation; according to the size and direction of the posture deviation, the aerial survey equipment determines the specific amount of rotation and translation adjustment of the camera module, and converts it into the control instruction format of the pan-tilt motor; finally, the generated posture adjustment instructions are sent to the front and rear camera modules respectively, driving them to perform the corresponding posture adjustment actions according to the instruction requirements until the postures of the two camera modules meet the preset aerial survey posture requirements.
[0068] In some embodiments, the camera module attitude adjustment can be achieved in the following two ways: Optionally, the emergency aerial survey equipment first estimates the current flight attitude of the UAV based on the onboard GPS and IMU data, and then infers the current attitude of the camera module based on the fixed installation position relationship between the UAV and the camera module; compares the current attitude of the camera module with the preset aerial survey attitude, calculates the attitude deviation, and inputs the deviation data into the PID controller to generate a continuous control quantity; converts the control quantity into a PWM wave to drive the servo motor and linear motor of the camera module gimbal to move, thereby achieving tracking control of the camera module attitude. Optionally, the emergency aerial survey equipment first aligns the three-dimensional model of the camera module in the preset aerial survey posture with the three-dimensional model of the camera module in the current flight state to obtain the transformation matrix between the two sets of models; extracts the Euler angles that the camera module needs to rotate and the displacement vector that needs to be translated according to the transformation matrix, and converts them into the number of pulses of the rotary motor and the linear motor; controls the two rotational degrees of freedom and one translational degree of freedom of the gimbal through the motor drive module, so that the position and posture of the camera module gradually approaches the preset aerial survey posture; repeats the above process until the two sets of models overlap and the posture of the camera module is consistent with the preset value. It can be understood that other methods can also be used to achieve the adjustment of the posture of the camera module, such as machine vision servo control, which is not limited here.
[0069] S302: Flip the second image, and extract matching feature points in the second image at the same position as the preset feature points.
[0070] Referring to step S102 , the emergency aerial survey equipment extracts matching feature points in the second image.
[0071] In some embodiments, the emergency aerial survey equipment will obtain the attitude parameters of the first image and the second image; the attitude parameters include the three-axis rotation angle, focal length and resolution during shooting; based on the attitude parameters, the position offset and angle offset of the second image relative to the first image are calculated; according to the position offset, the second image is translated and corrected so that it coincides with the coordinate origin of the first image, and at the same time, according to the angle offset, the translated second image is rotated and corrected so that it maintains the same angular direction as the first image; the second image after the translation correction and rotation correction is completed is scaled and normalized so that it has the same resolution and size as the first image, and a second target image is obtained; the pixel position in the second target image that is the same as the two-dimensional coordinates of the preset feature point in the first image is extracted as the initial position point; with the initial position point as the center, a local image block of a preset range size is intercepted on the second target image; within the local image block, image points whose similarity with the preset feature points of the first image is higher than a preset similarity threshold are searched as matching feature points.
[0072] Among them, the attitude parameters refer to a set of parameters that describe the state of image shooting, including the three-axis rotation angle, focal length and resolution of the camera at the moment of shooting. Among them, the three-axis rotation angle (such as pitch angle, roll angle, yaw angle) represents the angular attitude of the camera optical axis relative to the reference coordinate system, the focal length represents the distance from the projection center to the imaging plane, and the resolution represents the number of pixels contained in a unit physical length on the imaging plane. The attitude parameters are an important basis for multi-view image registration. The position offset and angle offset respectively represent the degree of difference between the two images in spatial position and angle direction, and are usually expressed in the form of vectors or matrices. For example, the position offset can be expressed by a two-dimensional translation vector [dx, dy], and the angle offset can be expressed by a three-dimensional rotation vector [rx, ry, rz] or a rotation matrix. Translation correction and rotation correction refer to the process of applying translation and rotation transformations to the image to align it with the reference image. Scale normalization refers to the process of scaling the image to make its resolution and size consistent with the reference image. The local image block represents a small image area centered on the feature point, and the preset feature point refers to the significant feature point pre-extracted in the reference image, such as corner points, edge points, etc.
[0073] Specifically, after obtaining the two images to be registered and their corresponding attitude parameters, the emergency aerial survey equipment needs to achieve accurate image registration through a series of attitude registration and feature matching operations. The emergency aerial survey equipment first extracts the attitude parameters such as the three-axis rotation angle, focal length and resolution of the camera when the two images were taken; then uses attitude parameters, such as the rotation matrix and focal length ratio, to calculate the position offset and angle offset of the second image relative to the first image; based on the calculated position offset, the second image is translated so that its coordinate origin is aligned with the first image; then, based on the angle offset, the translated second image is rotated so that its angle is consistent with that of the first image; the rotated second image is scaled so that its resolution and size are the same as those of the first image. , obtain the second target image after registration; pre-extract a set of significant feature points and their two-dimensional pixel coordinates in the first image, and use the same pixel coordinates as the initial position points in the second target image; take each initial position point as the center, intercept a local image block in the second target image; search for the point with the highest similarity to the preset feature point of the first image and higher than the threshold in the local image block, as the matching point of the preset feature point in the second target image; finally, use a set of matched feature point pairs, through the least squares and other optimization algorithms, solve the accurate homography transformation matrix between the two images, and complete the image registration process.
[0074] In some embodiments, image posture registration and feature matching can be achieved in the following two ways: Optionally, the emergency aerial survey equipment first extracts features from the two images to be registered to obtain a set of scale-invariant feature transform (SIFT) feature points; the random sampling consensus (RANSAC) algorithm is used to roughly estimate the homography transformation matrix of the two images based on SIFT feature matching to complete the initial registration; on this basis, the emergency aerial survey equipment optimizes the homography matrix using posture parameters, obtains an accurate homography transformation matrix by minimizing the reprojection error, and performs a registration transformation on the second image to make it geometrically completely aligned with the first image; finally, pre-extract significant feature points such as corner points in the first image, and search for matching points with the highest similarity near the corresponding positions of the second registered image to generate a set of accurately matched feature point pairs. Optionally, the emergency aerial survey equipment first uses the attitude parameters to roughly align the two images; then uses algorithms such as regularized cross-correlation to calculate the pixel-by-pixel similarity scores of the two roughly registered images, and extracts the top N pixels with the highest scores and their coordinates in the first image as preset feature points; in the second image, a local image block is captured with the preset feature point coordinates as the center, and a matching point with the highest similarity to the preset feature point is searched in the local range; all preset feature points and matching points constitute control point pairs, which are substituted into the homography transformation model, and the optimal transformation matrix is solved by the iterative nearest point algorithm, and the second image is registered and transformed to obtain a precisely registered image. It is understandable that other methods can also be used to achieve image registration, such as end-to-end registration based on deep learning, which is not limited here.
[0075] S303: Divide the first image into a plurality of first sub-regions, each of which includes at least one preset feature point.
[0076] The first image refers to the target area image captured by the front camera module of the emergency aerial survey equipment. The first sub-region refers to a number of local regions obtained by dividing the first image according to certain rules, and each sub-region contains part of the image content. The preset feature points refer to a set of obvious image features extracted and marked during the shooting of the first image, such as corner points, edges, etc., which are used for subsequent image registration and fusion processing.
[0077] In order to improve the accuracy and efficiency of image registration and fusion, after obtaining the first image, the emergency aerial survey equipment needs to process it in blocks. Specifically, the emergency aerial survey equipment first determines the appropriate block size and number based on the resolution of the first image and the distribution of preset feature points; then uses image segmentation algorithms, such as grid-based segmentation, graph-cut-based segmentation, etc., to divide the first image into several non-overlapping first sub-areas; during the segmentation process, the emergency aerial survey equipment needs to ensure that each first sub-area contains at least one preset feature point to provide sufficient key point information for subsequent feature matching.
[0078] In some embodiments, the first image segmentation can be implemented in the following manner: Optionally, the emergency aerial survey equipment first divides the first image evenly into rectangular grids of a fixed size, and counts the number of preset feature points in each grid; then the grid size is adaptively adjusted to merge adjacent grids with fewer preset feature points than a threshold until all grids meet the requirements; finally, each grid is used as a first sub-region. Optionally, the emergency aerial survey equipment first constructs a triangulation of the first image using the positions of the preset feature points; then, based on each triangle, a regional growing algorithm is used to expand outward and merge adjacent triangles until a sufficient number of preset feature points are contained in the growing region; finally, each growing region is used as a first sub-region. It is understandable that other methods may also be used to segment the first image, such as using a semantic segmentation algorithm to extract different terrain areas as first sub-regions, etc., which are not limited here.
[0079] S304: Divide the second image into second sub-regions of the same number as that of the first image, each second sub-region including matching feature points at the same position as the preset feature points.
[0080] The second image refers to the image of the target area captured by the rear camera module of the emergency aerial survey equipment and flipped. The second sub-area represents a number of local areas obtained by dividing the second image according to the same rules as the first image. The matching feature point refers to the feature point extracted from the second image that corresponds one-to-one with the preset feature point of the first image in spatial position, which is used for image registration.
[0081] While dividing the first image into blocks, the emergency aerial survey equipment also needs to perform corresponding block processing on the second image to ensure the spatial consistency of the two groups of sub-regions. Specifically, the emergency aerial survey equipment first refers to the block scheme of the first image to determine the block size and number of the second image so that it corresponds one-to-one with the first sub-region of the first image; then, using the same segmentation algorithm and block parameters as the first image, the second image is divided into the same number of second sub-regions; during the division process, the emergency aerial survey equipment needs to ensure that each second sub-region contains at least one matching feature point, and the matching feature point corresponds to the preset feature point in the first sub-region in spatial position.
[0082] In some embodiments, the second image segmentation can be implemented in the following manner: Optionally, the emergency aerial survey equipment first maps the segmentation result of the first image to the coordinate system of the second image to obtain the initial segmentation scheme of the second image; then, using the correspondence between the preset feature points and the matching feature points, the initial segmentation scheme is geometrically corrected to make each second sub-region accurately correspond to the first sub-region in space; finally, it is checked whether each second sub-region contains matching feature points, and the sub-regions that do not meet the requirements are fine-tuned to ensure one-to-one correspondence. Optionally, the emergency aerial survey equipment first counts the spatial distribution of matching feature points in the second image and compares it with the preset feature point distribution of the first image; then, referring to the segmentation scheme of the first image, an equal area division is performed on the matching feature point distribution map of the second image to obtain the initial second sub-region; finally, the region growing algorithm is used to optimize the initial sub-region so that the number and position of matching feature points in each second sub-region are consistent with the first sub-region. It can be understood that other methods can also be used to segment the second image, as long as the number and spatial position of the second sub-region and the first sub-region are guaranteed to correspond one-to-one, which is not limited here.
[0083] S305 . Based on the preset feature points of the first sub-region, feature matching is performed on the matching feature points of the second sub-region to obtain a mapping relationship between the pixel points in the first sub-region and the pixel points in the second sub-region.
[0084] Among them, feature matching refers to the process of establishing a spatial correspondence between two sets of feature points, which is often used in tasks such as image registration and 3D reconstruction. Pixels represent the smallest sampling unit in a digital image, and each pixel contains information such as color and brightness. The mapping relationship refers to the geometric transformation relationship between corresponding pixels in two images, which is often represented by mathematical models such as homography matrices and affine transformations.
[0085] After completing the block division of the two images, the emergency aerial survey equipment needs to further establish the pixel-level correspondence between the first sub-region and the second sub-region to prepare for the subsequent image fusion. Specifically, the emergency aerial survey equipment first extracts the preset feature points in each first sub-region and the matching feature points in the second sub-region, and constructs a feature descriptor; then uses feature matching algorithms, such as SIFT, SURF, etc., to calculate the similarity between the two sets of feature points and establish the optimal feature point correspondence; on this basis, the emergency aerial survey equipment uses the matching feature point pairs to estimate the geometric transformation matrix between the first sub-region and the second sub-region, and uses the matrix to coordinate map all pixel points in the first sub-region, obtain their corresponding positions in the second sub-region, and establish a pixel-level mapping relationship.
[0086] In some embodiments, feature matching and mapping relationship estimation can be achieved in the following ways: Optionally, the emergency aerial survey equipment first uses the SIFT feature extraction algorithm to extract scale-invariant key points in each first sub-region and the second sub-region; then uses the SIFT feature descriptor to calculate the 128-dimensional vector of the key points, and uses the Euclidean distance to measure the similarity between the two groups of key points; for the key point pairs with the highest similarity, the RANSAC algorithm is used to estimate the homography matrix, and the matrix is used to transform the coordinates of all pixels in the first sub-region to establish a pixel-level mapping relationship. Optionally, the emergency aerial survey equipment first uses the Harris corner detection algorithm to extract corner features in the two sub-regions; then uses the normalized cross-correlation algorithm to calculate the similarity of the two groups of corner features, and for the corner pairs with similarities greater than the threshold, the least squares method is used to estimate the affine transformation matrix; finally, the matrix is used to map the coordinates of the pixels in the first sub-region to obtain their corresponding positions in the second sub-region. It can be understood that other methods can also be used to achieve feature matching and mapping relationship estimation between the two sub-regions, such as using a feature matching network based on deep learning, which is not limited here.
[0087] S306: Map the pixel values of the first sub-region according to the mapping relationship to obtain target fused pixel points of the second sub-region.
[0088] The pixel value refers to the numerical representation of the color, brightness and other attributes of each pixel in the image, and is usually represented by the component value of color space such as RGB and HSV. The target fused pixel represents a new pixel obtained from the first sub-region through the mapping relationship and used to fill the second sub-region. The pixel integrates the image information of the two sub-regions.
[0089] After establishing the pixel-level mapping relationship between the two sub-areas, the emergency aerial survey equipment needs to further use the mapping relationship to fuse the images of the two sub-areas. Specifically, the emergency aerial survey equipment first traverses each pixel point in the first sub-area, and uses the pixel mapping matrix to calculate the corresponding coordinates of the pixel point in the second sub-area; then, at the corresponding position of the second sub-area, a small neighborhood centered on the coordinate is extracted, and the fusion weight of the pixel points in the neighborhood is calculated using an interpolation algorithm; finally, the pixel value of the pixel point in the first sub-area is weighted averaged with the neighborhood to obtain the target fused pixel point at that position, which is used to update the image information in the second sub-area.
[0090] In some embodiments, pixel value mapping and fusion can be achieved in the following manner: Optionally, the emergency aerial survey equipment first calculates the floating-point coordinates of each pixel in the first sub-region in the second sub-region according to the mapping relationship between the first sub-region and the second sub-region; then, with the floating-point coordinates as the center, a 3x3 neighborhood window is extracted on the second sub-region, and the fusion weight of each pixel in the neighborhood is calculated using a bilinear interpolation algorithm; finally, the RGB value of the pixel in the first sub-region is weighted averaged with the neighborhood to obtain the RGB value of the target fused pixel, and fill it into the corresponding position of the second sub-region. Optionally, the emergency aerial survey equipment first adjusts the brightness and contrast of the images of the first sub-region and the second sub-region respectively to make them have similar appearances; then, using the pixel mapping relationship, the floating-point coordinates of the first sub-region corresponding to each integer coordinate in the second sub-region are calculated, and the pixel value is sampled on the first sub-region using a bicubic interpolation algorithm; finally, the sampled pixel values of the first sub-region are used as the target fused pixel points and directly filled into the integer coordinate position of the second sub-region. It can be understood that other methods can also be used to achieve pixel value mapping and fusion, such as using an image fusion algorithm based on optimization, which is not limited here.
[0091] In some embodiments, the emergency aerial survey equipment extracts the pixel value of each pixel in the first sub-area, maps it to the corresponding pixel position in the second sub-area through a mapping relationship, and obtains an initial fused pixel; performs smoothing correction on the initial fused pixel to obtain a target fused pixel; extracts independent pixels in the second sub-area that have no direct mapping relationship with the pixels in the first sub-area; and calculates the target fused pixel of the independent pixel through an interpolation algorithm based on the color value of the generated target fused pixel in the neighborhood of the independent pixel and the difference in spatial distance.
[0092] Among them, the pixel value refers to the color, brightness, transparency and other values corresponding to each pixel on the image. Common pixel value representation methods include RGB, RGBA, HSV, etc. The mapping relationship represents the corresponding relationship between the corresponding pixels in the two images, which can usually be represented by a two-dimensional coordinate transformation matrix. The initial fused pixel point and the target fused pixel point represent the intermediate pixel point and the final pixel point generated in the image fusion process, respectively. Smoothing correction refers to a pixel value correction method that reduces defects such as distortion or cracks caused by direct mapping while ensuring a smooth transition of pixel values. Independent pixel points represent pixels that have no direct corresponding points in the reference image, and their pixel values need to be generated through interpolation and other methods. Interpolation algorithm is a mathematical method that estimates unknown pixel values through known pixel values. Common ones include nearest neighbor interpolation, bilinear interpolation, bicubic interpolation, etc.
[0093] Specifically, after obtaining the two registered sub-images to be fused, the emergency aerial survey equipment needs to further generate a seamless stitching image containing information of the two sub-regions through pixel-level fusion. The emergency aerial survey equipment first extracts the pixel value and coordinates of each pixel in the reference image (such as the first sub-region image); then, according to the mapping relationship between the two images (such as the homography matrix), the pixel value of the reference image is mapped to the corresponding coordinates in the target image (such as the second sub-region image) to obtain the initial fused pixel; due to the influence of factors such as image registration error, the value of the initial fused pixel may be discontinuous with the neighborhood pixel value, so it is necessary to perform a smooth correction to make the pixel value smoothly transition to obtain the target fused pixel; for the area in the target image where there is no direct mapping of the reference image pixel, that is, the independent pixel, the emergency aerial survey equipment needs to estimate its pixel value by interpolation; the common idea is to take the independent pixel as the center, extract the target fused pixel generated in its neighborhood, and calculate the target fused pixel value of the independent pixel based on their numerical value and the difference in spatial distance with the independent pixel, using interpolation algorithms such as weighted average and spline surface fitting; finally, all the target fused pixels and interpolation pixels are combined together to obtain a complete target fused image, so as to achieve seamless stitching of the two sub-regions.
[0094] In some embodiments, sub-image pixel-level fusion can be achieved in the following two ways: Optionally, the emergency aerial survey equipment first forward maps the two sub-images based on the mapping relationship, that is, maps the pixel values of the source image to the target image coordinate system, and uses the nearest neighbor interpolation to determine the initial values of the target image pixels; then the initial fused image is morphologically processed, and the holes generated in the mapping process are filled by dilation and erosion operations; on this basis, the emergency aerial survey equipment uses the Laplace pyramid fusion algorithm to achieve a natural transition of information in different frequency bands of the fused image through multi-scale decomposition and reconstruction; for independent pixels, first use the iterative closest point algorithm (ICP) to estimate their optimal mapping position in the source image, and then use bicubic interpolation to calculate their fused pixel values. Optionally, the emergency aerial survey equipment first performs wavelet transform on the two images to be fused to obtain their respective high-frequency and low-frequency sub-band coefficients; then fuses the coefficients on the low-frequency sub-band using the weighted average method, and selects the coefficients on the high-frequency sub-band using the maximum value method; inverse transforms the fused wavelet coefficients to obtain a frequency domain fused image; finally, optimizes the frequency domain fusion result using the graph cut algorithm, finds the optimal segmentation path in the fused image through the minimum cut model, and replaces the pixels on both sides of the segmentation path with the pixel values of the adjacent area to eliminate the splicing marks and generate the final fused image. It is understandable that other methods can also be used to achieve sub-image fusion, such as end-to-end fusion based on deep learning, which is not limited here.
[0095] S307: splice the target fusion pixel points of all the second sub-areas to obtain a fused image.
[0096] Image stitching refers to the process of seamlessly stitching multiple local images into a complete image, which is often used in tasks such as panoramic image synthesis and satellite image mosaicking. A fused image refers to a new image obtained by image stitching, which combines the complete information of the original two perspectives.
[0097] After completing pixel fusion for all second sub-regions, the emergency aerial survey equipment needs to further seamlessly stitch these fused sub-regions to obtain the fusion result under the perspective of the complete second image. Specifically, the emergency aerial survey equipment first performs geometric correction on all fused second sub-regions to eliminate the slight deformation between sub-regions; then uses image stitching algorithms, such as multi-band fusion, pyramid fusion, etc., to find the best stitching path in the overlapping area of adjacent sub-regions, and adjusts the brightness and color to make the stitching boundary naturally smooth; finally, all second sub-regions are arranged in the order of the stitching path to obtain a complete fused image, which is completely consistent with the second image in perspective and resolution, but incorporates the texture detail information of the first image.
[0098] S308: Generate a 2D model based on the fused image and send it to the user terminal.
[0099] Referring to step S104, the emergency aerial survey equipment generates and sends a 2D model.
[0100] S309: Obtain image annotation information sent by the user terminal.
[0101] Image annotation information refers to the information that users mark on the target objects of interest on a two-dimensional map or image, usually including the category of the target object (such as buildings, roads, trees, etc.), outline (such as the boundary polygon of the object) and location (such as the geographic coordinates of the object). 2D model refers to a two-dimensional map or image, which can be a satellite image, aerial image or other geographic information product. User terminal refers to an electronic device with human-computer interaction capabilities, such as a smartphone, tablet computer, PC, etc. Users can browse and annotate 2D models through the terminal device.
[0102] In emergency aerial survey tasks, in order to obtain more accurate and detailed target information, the system usually allows users to further annotate and supplement the two-dimensional map or image based on the three-dimensional model. Specifically, the user terminal first generates a two-dimensional orthophoto, i.e., a 2D model, based on the three-dimensional model; then the user can browse the area of interest on the 2D model and use tools such as points, lines, and polygons to annotate the target object; the user can enter the category attributes of the object while annotating, such as annotating a polygon as "residential residence" and a broken line as "road", etc.; the user can also automatically generate the boundary polygon of the object by dragging the outline of the annotated object; the object information annotated by the user will be automatically associated with the 2D model coordinate system to obtain the geographic location information of the object; after the annotation is completed, the user terminal will upload the image annotation information to the emergency aerial survey system; after the system server obtains the annotation information, it needs to associate it with the three-dimensional model to update the recognition and positioning results of the object.
[0103] In some embodiments, user image annotation can be achieved in the following two ways: Optionally, the emergency aerial survey system first orthophoto projects the three-dimensional model onto a two-dimensional plane to generate an orthophoto image; then the system slices the orthophoto image and transmits it to the user terminal for display through a WebGIS web page; the user uses a vector drawing tool on the web page to mark the target object and inputs attribute information such as the object category and description; the vector data drawn by the user is displayed superimposed on the orthophoto image and can be edited and modified; after the annotation is completed, the web page associates the vector annotation information with the image coordinates to generate an annotation file, which is then transmitted back to the system server through the Web service interface. Optionally, the emergency aerial survey system first performs semantic segmentation on the three-dimensional model, extracts land features such as buildings, roads, and green spaces, and generates a vectorized two-dimensional map; the system distributes the vector map data to the user terminal through a mobile APP; the user loads the vector map in the APP and can translate and zoom the map; when the user finds a target object of interest, he can use his finger to draw around the outline of the object on the screen to generate a closed polygon; the user can also add object categories and attribute labels through voice input and other methods; after completing the annotation of an object, the APP associates the annotated polygon with the spatial reference of the vector map, obtains the geographic coordinates of the polygon, and uploads the complete annotation information to the system server in the GIS spatial data format. It is understandable that other methods can also be used to implement user image annotation, such as using augmented reality technology for real-scene annotation, which is not limited here.
[0104] S310 , performing semantic segmentation on the 2D model based on the image annotation information to obtain a segmented 2D model.
[0105] Semantic segmentation refers to dividing an image into several regions with semantic meanings and assigning specific category labels to each region. It is often used in tasks such as scene understanding and target extraction. The segmented 2D model represents a new structured representation based on the original 2D model by incorporating semantic segmentation information. Each pixel in the model contains not only coordinates and colors, but also the category label of the target object to which it belongs.
[0106] After obtaining the image annotation information provided by the user, the emergency aerial survey equipment needs to further use this prior knowledge to perform semantic segmentation on the 2D model, so as to realize the automatic extraction and classification of the target of interest. Specifically, the emergency aerial survey equipment first parses the image annotation information into a computer-readable standard format, such as XML, JSON, etc.; then maps the annotation polygon to the 2D model coordinate system, and samples a set of training samples in the annotation area, each sample consisting of its coordinates and category labels; on this basis, the emergency aerial survey equipment uses semantic segmentation algorithms, such as fully convolutional neural networks, conditional random fields, etc., to learn the texture and shape features of the annotation area, and generalize the learned segmenter to the entire 2D model, predicting the most likely target category label for each pixel; finally, the aerial survey equipment fuses the pixel-level category label with the original 2D model to obtain a segmented structured model for subsequent analysis.
[0107] In some embodiments, semantic segmentation of 2D models can be achieved in the following ways: Optionally, the emergency aerial survey equipment first uses user annotations to establish a small training data set, and uses data enhancement to expand the sample size; then builds a U-Net semantic segmentation network, inputs the 2D model RGB image, and outputs a target probability map for each pixel; in the training phase, the labeled data is used for end-to-end supervised learning, and the network weights are tuned with the validation set; in the testing phase, the trained network is used to predict the category of the entire 2D model pixel by pixel, and post-processes with a probability threshold to obtain a segmentation result; finally, the segmentation result is used as a new channel and superimposed on the original 2D model to obtain a segmentation model with a category label. Optionally, the emergency aerial survey equipment first over-segments the 2D model to obtain a set of superpixels; then uses the user annotations to calculate the category histogram of the superpixels, and uses the graph cut algorithm to cluster the superpixels to obtain a set of candidate regions; on this basis, the emergency aerial survey equipment uses conditional random fields to model the candidate regions, and comprehensively utilizes regional appearance features and spatial adjacency to construct an energy function; obtains the category label of each candidate region by minimizing the energy function, thereby achieving pixel-level segmentation; finally, maps the regional category label back to the original pixel and reconstructs it in the 2D model to obtain the segmented model. It is understandable that other methods can also be used to achieve 2D model semantic segmentation, such as converting the 2D model to three-dimensional space for segmentation, etc., which is not limited here.
[0108] S311 , according to the preset ground object categories, the category information and location information of the multiple target objects after classification and labeling are counted to generate ground object distribution information of the area corresponding to the 2D model.
[0109] The preset feature categories refer to the target types that need to be focused on in specific application scenarios, such as buildings, roads, and waters in emergency rescue. Feature distribution information represents the spatial distribution of different feature categories in an area, usually described by indicators such as proportion, density, and location.
[0110] After completing the semantic segmentation of the 2D model, the emergency aerial survey equipment needs to further count and analyze the composition of the objects in the model to provide users with regional overview information. Specifically, the emergency aerial survey equipment first determines the key object categories in the scene and maps them to the category label set of the 2D segmentation model; then it traverses all the target objects in the segmentation model and counts the number of targets, average size, proportion of the total area, and other indicators of each key category; at the same time, the emergency aerial survey equipment also needs to calculate the center coordinates, minimum circumscribed rectangle, and other location information of each target object; finally, the aerial survey equipment integrates the statistical information and location information of each category to generate an easy-to-read and easy-to-understand object distribution report to describe the characteristics of the object composition in the area.
[0111] In some embodiments, feature distribution information can be generated in the following manner: Optionally, the emergency aerial survey equipment first manually defines a feature category system and maps it to the category labels used by the 2D segmentation model; then uses the connected domain analysis algorithm to extract the position, area, circumscribed rectangle and other geometric properties of each pixel block in the segmentation model; on this basis, the aerial survey equipment classifies the pixel blocks according to the category labels, and separately counts the number, total area, average area, etc. of each type of feature; at the same time, the coordinate mean of the pixel blocks of each type of feature is calculated to obtain the distribution center of this type of feature; finally, the statistical indicators and distribution center coordinates of each type of feature are tabulated and embedded in the regional map to intuitively present the quantity, scale and spatial distribution of each feature. Optionally, the emergency aerial survey equipment first converts the segmentation model into a vector format compatible with the geographic information system, such as Shapefile; then uses GIS spatial analysis tools, such as ArcGIS, to process the vector elements: statistics are performed on the element attribute table according to preset categories, and the number and area proportion of each type of element are calculated; the centroid of each element is extracted using the element conversion tool, and the distribution density of each type of element is calculated using the kernel density analysis tool; based on the statistical results, the aerial survey equipment uses GIS mapping tools, such as ArcMap, to design and produce thematic maps, organically combining statistical charts, density classification maps, element annotations, and other contents to obtain an intuitive map of the distribution of land features. It is understandable that other methods can also be used to generate land feature distribution information, such as using the spatial database and query language of the geographic information system, which is not limited here.
[0112] S312: Send the feature distribution information to the user terminal.
[0113] Specifically, the emergency aerial survey equipment sends the feature distribution information to the user terminal through the wireless communication network for the user to view and analyze. The feature distribution information can be presented in various forms such as charts, texts, and maps, so that users can quickly understand the overall situation and features of the target area. Users can browse the feature distribution information interactively, such as clicking on the target on the map to display its category and area, or filtering for features of a specific category.
[0114] At this point, the emergency aerial survey equipment uses the front and rear cameras on the drone to obtain dual-view images of the target area; through image registration, mapping, stitching and other processing, a high-resolution two-dimensional model covering the entire area is generated; and the model is semantically segmented and the ground feature statistics are combined with user annotations, and finally intuitive and rich regional ground feature distribution information is provided to users. This series of steps enables emergency aerial surveys to quickly and efficiently obtain key ground feature information, providing strong support for emergency command, rescue deployment and other tasks. At the same time, users can interact with the aerial survey system through mobile terminals and dynamically specify targets that need to be focused on and analyzed, making the acquisition of ground feature information more flexible and personalized.
[0115] In some embodiments, the emergency aerial survey equipment will obtain a damage assessment request sent by a user terminal; the damage assessment request includes the category of the object to be assessed; according to the category of the object to be assessed, the corresponding object to be assessed is extracted from the segmented 2D model; the outline of the object to be assessed is detected pixel by pixel, and the number of complete pixel points of the outline is counted to obtain the integrity of the object to be assessed; if the integrity of the object to be assessed is lower than a preset integrity threshold, the object to be assessed is determined to be a damaged object; according to the proportion of missing pixels in the outline of the object to be assessed, the damage level of the object to be assessed is determined; a damage assessment report of the object to be assessed is generated, and the damage assessment report is sent to the user terminal; the damage assessment report includes the location, damage level and quantity statistical information of each damaged object.
[0116] Among them, the damage assessment request refers to a request initiated by the user to evaluate and analyze the damage of the target object, which usually needs to include the category information of the object to be evaluated. Object category usually refers to the summary and division of the semantic attributes of the object, such as buildings, roads, green spaces, etc. The contour refers to the shape feature of the object boundary, which can be represented by a closed curve or polygon. The completeness refers to the degree of completeness of the object's contour, that is, the degree to which the contour curve is closed and continuous, which can be measured by indicators such as the ratio of the number of complete contour pixels to the total number of pixels. The damage level refers to the severity of the damage to the object, which can usually be divided into multiple levels such as intact, slightly damaged, moderately damaged, and severely damaged. The damage assessment report refers to the result file generated after the damage to the object is assessed, usually including the location, category, damage level, quantity statistics and other information of the damaged object.
[0117] Specifically, after receiving the damage assessment request sent by the user, the emergency aerial survey equipment needs to perform damage detection and assessment on the target objects in the specified area, and generate a damage assessment report to feedback to the user. The emergency aerial survey equipment first obtains the category information of the object to be assessed contained in the request; then, according to the object category, all the objects of this category are extracted from the segmented two-dimensional vector model; for each object, the emergency aerial survey equipment extracts its contour coordinate sequence, and counts the complete continuity of the pixels on the contour, and calculates the integrity index; compares the integrity index with the preset integrity threshold, and if it is lower than the threshold, the object is determined to be a damaged object; according to the proportion of the number of missing pixels in the damaged object contour to the total number of pixels, it is divided into corresponding damage levels; after damage detection and assessment of all objects to be assessed, the emergency aerial survey equipment generates a complete damage assessment report, including the geographical location, category, damage level and number statistics of damaged objects of each level of each damaged object; finally, the damage assessment report is sent to the requesting user terminal through the network to complete the complete damage assessment process.
[0118] In some embodiments, the detection and assessment of damage to objects can be achieved in the following two ways: Optionally, the emergency aerial survey equipment first performs semantic segmentation on the image of the area to be assessed, and extracts all objects of the category to be assessed; then, for each object, a reference contour template in its complete state is extracted, and it is matched with the actual contour after segmentation, and the completeness is calculated by comparing the contour pixel by pixel; based on the comparison result of the completeness and the threshold, it is determined whether the object is damaged, and the damage level is divided according to the proportion of the number of missing pixels in the contour; on this basis, the emergency aerial survey equipment counts the number of damaged objects of each level, extracts their geographic location coordinates, and generates a damage assessment report; if the assessment request contains priority information, the damage assessment results are sorted according to the priority and then the report is generated. Optionally, the emergency aerial survey equipment first performs a planar segmentation on the point cloud of the area to be assessed to obtain a three-dimensional model of the object; associates the three-dimensional model with the object category, and extracts the contour features of each type of object; uses a three-dimensional contour matching algorithm to compare the current contour with the complete reference contour, and calculates the integrity of the object through the voxel integrity metric; determines the damage to the object according to the set integrity threshold and corresponds to each damage level; the emergency aerial survey equipment further counts the number and location information of objects at each damage level, generates a three-dimensional visual damage assessment report model, and finally renders it into a two-dimensional report and sends it to the user terminal. It is understandable that other methods can also be used to implement object damage assessment, such as multi-task damage detection based on deep learning, which is not limited here.
[0119] In the embodiment of the present application, due to the use of the aerial survey method of the front and rear dual cameras working together, the image information of different perspectives is fully utilized, and the automatic repair and optimization of the data level is realized through the image fusion algorithm, so the 2D model finally generated has the characteristics of both high resolution and high integrity. It effectively solves the problem that the traditional single-perspective aerial survey is easily affected by environmental interference and incomplete data, and thus realizes the stable surveying and accurate modeling requirements in complex disaster environments, providing reliable information guarantee for emergency command and rescue operations in disaster areas.
[0120] The following describes the emergency aerial survey equipment in the embodiment of the present invention from the perspective of hardware processing. Figure 4 , which is a schematic diagram of a physical device structure of emergency aerial survey equipment in an embodiment of the present application.
[0121] It should be noted that Figure 4 The structure of the emergency aerial survey equipment shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.
[0122] like Figure 4 As shown, the emergency aerial survey equipment includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 to the random access memory (RAM) 403, such as executing the method described in the above embodiment. Various programs and data required for system operation are also stored in the RAM 403. The CPU 401, ROM 402 and RAM 403 are connected to each other through a bus 404. The input / output (I / O) interface 405 is also connected to the bus 404.
[0123] The following components are connected to the I / O interface 405: an input section 406 including an audio input device, a button switch, etc.; an output section 407 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed so that a computer program read therefrom is installed into the storage section 408 as needed.
[0124] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 409, and / or installed from a removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, various functions defined in the present invention are performed.
[0125] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.
[0126] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.
[0127] Specifically, the emergency aerial survey equipment of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the method for generating a 2D model provided in the above embodiment is implemented.
[0128] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the emergency aerial survey equipment described in the above embodiment; or may exist independently without being assembled into the emergency aerial survey equipment. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the emergency aerial survey equipment, the emergency aerial survey equipment implements the 2D model generation method provided in the above embodiment.
[0129] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.
[0130] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.
[0131] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.
Claims
1. A method for generating a 2D model, characterized in that: Applied to emergency aerial survey equipment, the emergency aerial survey equipment includes a front camera module and a rear camera module, and the method includes: In response to a capture instruction sent by a user terminal, a first image taken by the front camera module and a second image taken by the rear camera module are acquired; the first image includes preset feature points; Flipping the second image, and extracting matching feature points in the second image at the same position as the preset feature points; Based on the preset feature points and the matching feature points, the second image and the first image are fused to obtain a fused image; Based on the fused image, a 2D model is generated and sent to the user terminal.
2. The method according to claim 1, characterized in that The fusing the second image and the first image based on the preset feature points and the matching feature points to obtain a fused image specifically includes: Dividing the first image into a plurality of first sub-regions, each of the first sub-regions including at least one preset feature point; Dividing the second image into second sub-regions of the same number as that of the first image, each of the second sub-regions including matching feature points at the same position as the preset feature points; Based on the preset feature points of the first sub-region, feature matching is performed on the matching feature points of the second sub-region to obtain a mapping relationship between pixel points in the first sub-region and pixel points in the second sub-region; According to the mapping relationship, the pixel values of the first sub-area are mapped to obtain target fused pixel points of the second sub-area; All target fusion pixels of the second sub-areas are spliced to obtain a fused image.
3. The method according to claim 2, characterized in that Mapping the pixel values of the first sub-region according to the mapping relationship to obtain the target fused pixel points of the second sub-region specifically includes: Extracting the pixel value of each pixel in the first sub-region, and mapping it to the pixel position corresponding to the second sub-region through the mapping relationship, to obtain an initial fused pixel; Performing smoothing correction on the initial fused pixel points to obtain target fused pixel points; Extracting independent pixel points in the second sub-region that have no direct mapping relationship with pixel points in the first sub-region; The target fused pixel point of the independent pixel point is generated by calculating the target fused pixel point of the independent pixel point through an interpolation algorithm according to the color value of the generated target fused pixel point in the neighborhood of the independent pixel point in combination with the spatial distance difference.
4. The method according to claim 1, characterized in that: The flipping of the second image and extracting matching feature points in the second image at the same position as the preset feature points specifically includes: Acquire posture parameters of the first image and the second image; the posture parameters include three-axis rotation angle, focal length and resolution during shooting; Based on the posture parameters, calculating the position offset and the angle offset of the second image relative to the first image; According to the position offset, the second image is subjected to translation correction so that it coincides with the coordinate origin of the first image, and according to the angle offset, the second image is subjected to rotation correction so that it maintains the same angle direction as the first image; Performing scale normalization processing on the second image after the translation correction and the rotation correction is completed so that the second image has the same resolution and size as the first image, thereby obtaining a second target image; Extracting a pixel position in the second target image having the same two-dimensional coordinates as the preset feature point in the first image as an initial position point; Taking the initial position point as the center, intercepting a local image block of a preset range size on the second target image; In the local image block, image points whose similarity with the preset feature points of the first image is higher than a preset similarity threshold are searched as matching feature points.
5. The method according to claim 1, characterized in that Before the step of acquiring the first image taken by the front camera module and the second image taken by the rear camera module in response to the acquisition instruction sent by the user terminal, the method further includes: Acquire the current first posture information of the front camera module and the current second posture information of the rear camera module, and determine the first posture adjustment instruction and the second posture adjustment instruction based on the preset aerial survey posture; the preset aerial survey posture includes three-axis rotation angle, translation position and lens parameters; The first posture adjustment instruction is sent to the front camera module, and the second posture adjustment instruction is sent to the rear camera module, so as to adjust the front camera module and the rear camera module respectively.
6. The method according to claim 1, characterized in that After the step of generating a 2D model based on the fused image and sending the 2D model to the user terminal, the method further includes: Acquire image annotation information sent by the user terminal; the image annotation information is generated by the user terminal based on the 2D model, including the category, outline and position of the target object annotated on the 2D model; Based on the image annotation information, the 2D model is semantically segmented to obtain a segmented 2D model; the segmented 2D model includes the target object after classification and annotation; According to the preset ground object categories, the category information and location information of the multiple target objects after classification and labeling are counted to generate ground object distribution information of the area corresponding to the 2D model; The ground feature distribution information is sent to the user terminal.
7. The method according to claim 6, characterized in that After the step of sending the feature distribution information to the user terminal, the method further includes: Acquire a damage assessment request sent by the user terminal; the damage assessment request includes a category of the object to be assessed; According to the category of the to-be-evaluated object, extract the corresponding to-be-evaluated object from the segmented 2D model; Detect the contour of the object to be evaluated pixel by pixel, count the number of complete pixel points of the contour, and obtain the completeness of the object to be evaluated; If the integrity of the object to be evaluated is lower than a preset integrity threshold, determining that the object to be evaluated is a damaged object; Determining the damage level of the object to be evaluated according to the proportion of missing pixels in the outline of the object to be evaluated; Generate a damage assessment report for the object to be assessed, and send the damage assessment report to the user terminal; the damage assessment report includes the location, damage level and quantity statistics of each damaged object.
8. An emergency aerial survey equipment, characterized in that: The emergency aerial survey equipment includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the emergency aerial survey equipment to execute the method described in any one of claims 1-7.
9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on the emergency aerial survey equipment, the emergency aerial survey equipment is caused to execute the method according to any one of claims 1 to 7.
10. A computer program product, characterized in that When the computer program product is run on the emergency aerial survey equipment, the emergency aerial survey equipment is enabled to execute the method according to any one of claims 1 to 7.
Citation Information
Cited By
Dual-camera combined sea surface aerial photography method and device
CN119946448A
