An aerial photography view angle planning method and device based on photogrammetry

Abstract

This invention provides an aerial photography viewpoint planning method based on superior view photogrammetry. The method includes: determining the UAV's viewpoint sampling space based on a 3D approximate model of the target scene; then performing Poisson disk sampling on the surface of the 3D approximate model to generate a set of sampling points and a pre-selected viewpoint set corresponding to the sampling point set; based on the sampling point set, performing reconstructibility modeling on the 3D approximate model to generate the reconstructibility of each sampling point; based on the pre-selected viewpoint set, performing viewpoint redundancy modeling to generate the redundancy of each pre-selected viewpoint; and performing optimization processing on the pre-selected viewpoint set based on the reconstructibility of the sampling points and the redundancy of the pre-selected viewpoints to generate the aerial photography viewpoint. This embodiment combines maximizing the reconstructibility of sampling points and minimizing viewpoint redundancy for the optimization processing of the pre-selected viewpoint set of the target scene. Therefore, while achieving automatic aerial photography planning for the target scene using a UAV, it also improves the accuracy of 3D reconstruction of the target scene.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to a method and apparatus for aerial perspective planning based on superior view photogrammetry. Background Technology

[0002] The quality and speed of most multi-view stereo matching (MVS) algorithms largely depend on the selection of input images. Reconstructing an object's surface requires sufficient views, but not all input images contribute equally to the reconstruction quality. Too many images lead to unnecessary processing time, and incorrect viewpoints or scale differences can introduce noise into structural motion and multi-view stereo matching, even degrading reconstruction quality. Therefore, viewpoint planning has a significant impact on the quality of multi-view stereo reconstruction, typically requiring images to be acquired from multiple different positions and directions. Modern techniques often employ a coarse-to-fine approach: first, image data is acquired using traditional methods such as initial flight or satellite imagery. A rough initial model (scene proxy) is then created based on the 3D geometry of the object to be reconstructed. This initial model can be existing 3D data with height information or a low-precision model obtained from the initial flight. Then, the optimal shooting viewpoint for dense view sampling is calculated, and the UAV aerial photography route for the tested scene is designed and optimized. Finally, images are acquired and a detailed 3D model is reconstructed.

[0003] Existing large-scale urban scene reconstruction often requires prior aerial photography data collection. Before collecting data for urban 3D reconstruction, drone operators need to plan flight paths for the target area. The core of drone flight path planning is predicting whether a set of viewpoints can effectively reconstruct the target area, and the quality of the planning directly determines the final model quality generated from the aerial images. However, the flight time of conventional quadcopter drones is usually limited to 25-30 minutes by battery capacity. Therefore, the main goal of drone trajectory planning is to capture the entire 3D scene better and faster, i.e., maximizing reconstruction quality while minimizing flight time. The viewpoint and path planning problem based on drones can be solved in two steps: ① without considering flight time and path quality, generate the optimal shooting viewpoint to maximize reconstruction quality; ② minimize the flight cost of passing through all viewpoints generated in the first step.

[0004] The first class of existing methods, such as the ETE method, constructs a set of adaptive observation planes for viewpoint selection after exploring and generating geometric proxies. It repeats data capture after identifying low-quality areas until the reconstruction quality reaches the desired level. However, observation planes generated based on region slopes may lead to gaps because they are created based on the average of the normals of patch points; in complex regions, the observation planes may not accurately cover all areas. The second class of existing methods evaluates the quality of the geometric point cloud and selects two types of new viewpoints in a constrained view sampling space. The first type of new viewpoint is used to improve incomplete or low-quality areas, and the second type of viewpoint is added to fully cover the entire scene. Since the sample points on the geometric proxies are uniformly distributed, there are many sample points with low quality scores at the edges of buildings or structures with more geometric information. The third class of existing methods finds the optimal position and orientation of each candidate camera and selects the optimal subset of camera positions as candidate viewpoints for the optimal flight path based on UAV constraints. This path is often obtained using heuristic solutions, such as reducing it to a Traveling Salesman Problem (TSP) to solve through all selected viewpoints. These methods separate path construction from view optimization, neglecting path continuity and quality, which can significantly increase the workload of on-site data acquisition. Furthermore, paths obtained in this way often contain many sharp turns; the necessary deceleration when the drone passes through these turns leads to decreased efficiency, while frequent acceleration consumes additional power. Summary of the Invention

[0005] To address the aforementioned problems in existing technologies, embodiments of the present invention provide an aerial perspective planning method, apparatus, and computer-readable medium based on superior view photogrammetry; this method can achieve high-precision full-coverage 3D reconstruction of a target scene at a relatively low cost, and improve the fidelity of the target scene model reconstruction.

[0006] According to a first aspect of the present invention, an aerial viewpoint planning method based on superior view photogrammetry is provided. The method includes: determining the viewpoint sampling space of a UAV based on a 3D schematic model of a target scene; performing Poisson disk sampling on the surface of the 3D schematic model to generate a set of sampling points and a set of pre-selected viewpoints corresponding to the set of sampling points; performing reconstructibility modeling on the 3D schematic model based on the set of sampling points to generate the reconstructibility of each sampling point; performing viewpoint redundancy modeling on the set of pre-selected viewpoints to generate the redundancy of each pre-selected viewpoint; and performing optimization processing on the set of pre-selected viewpoints based on the reconstructibility of the sampling points and the redundancy of the pre-selected viewpoints to generate an aerial viewpoint.

[0007] Optionally, the step of sampling the surface of the 3D schematic model using a Poisson disk to generate a set of sampling points and a set of pre-selected viewpoints corresponding to the set of sampling points includes: determining the sampling diameter of the Poisson disk based on the camera's field of view, the overlap rate between adjacent images, and a preset distance between viewpoints; sampling the surface of the 3D schematic model using a Poisson disk based on the sampling diameter to generate a set of sampling points; for any sampling point in the set of sampling points: obtaining the surface normal vector at the sampling point; moving the viewpoint a preset distance in the opposite direction of the surface normal vector to obtain a pre-selected viewpoint corresponding to the sampling point; and obtaining a set of pre-selected viewpoints corresponding to the set of sampling points based on the pre-selected viewpoints corresponding to each sampling point in the set of sampling points.

[0008] Optionally, the step of performing reconstructibility modeling on the 3D approximate model based on the sampling point set to generate the reconstructibility of each sampling point includes: for any sampling point in the sampling point set: obtaining a pre-selected viewpoint and at least one observation viewpoint corresponding to the sampling point; determining a reconstruction quality model of the sampling point between the pre-selected viewpoint and the observation viewpoint; the observation viewpoint is used to indicate a viewpoint located outside the normal vector of the sampling point and capable of observing the sampling point; for any observation viewpoint among the at least one observation viewpoint: determining a reconstructibility metric of the sampling point between the pre-selected viewpoint and the observation viewpoint according to the reconstruction quality model; and determining the reconstructibility of the sampling point based on the reconstructibility metric corresponding to each of the at least one observation viewpoints.

[0009] Optionally, the step of performing viewpoint redundancy modeling based on the preselected viewpoint set to generate the redundancy of each preselected viewpoint includes: for any preselected viewpoint in the preselected viewpoint set: determining the sampling points that can be observed by the preselected viewpoint as first target sampling points, obtaining a plurality of first target sampling points; obtaining a plurality of reconstructibility based on the reconstructibility corresponding to each of the plurality of first target sampling points; and selecting the minimum reconstructibility value from the plurality of reconstructibility values ​​as the redundancy corresponding to the preselected viewpoint.

[0010] Optionally, the step of optimizing the set of pre-selected viewpoints based on the reconstructability of the sampling points and the redundancy of the pre-selected viewpoints to generate aerial viewing angles includes: obtaining the redundancy of each pre-selected viewpoint in the set of pre-selected viewpoints; when it is determined that the reconstructability of the sampling points corresponding to each pre-selected viewpoint in the set of pre-selected viewpoints meets a preset condition, performing redundancy minimization processing on the set of pre-selected viewpoints to obtain a set of candidate viewpoints; when it is determined that the number of candidate viewpoints in the set of candidate viewpoints remains unchanged, based on the set of pre-selected viewpoints, performing reconstructability maximization processing on the sampling points corresponding to each candidate viewpoint in the set of candidate viewpoints to generate a set of quasi-viewpoints; and determining each quasi-viewpoint in the set of quasi-viewpoints as an aerial viewing angle.

[0011] Optionally, when the reconstructability of the sampling points corresponding to each preselected viewpoint in the preselected viewpoint set meets a preset condition, the redundancy minimization process is performed on the preselected viewpoint set to obtain a candidate viewpoint set; this includes: obtaining the redundancy corresponding to each preselected viewpoint in the preselected viewpoint set, and sorting the preselected viewpoints in the preselected viewpoint set in descending order of redundancy to generate a sorting result; taking the first preselected viewpoint in the sorting result as the viewpoint to be deleted, and performing a deletion operation on the viewpoint to be deleted in the preselected viewpoint set to obtain... Candidate viewpoint set; obtain at least one second target sampling point associated with the viewpoint to be deleted, and update the reconstructability of the second target sampling point based on the deletion operation; determine whether the reconstructability of each second target sampling point in the at least one second target sampling point is greater than a preset threshold; if so, use the candidate viewpoint set as a pre-selected viewpoint set; if not, add the viewpoint to be deleted to the candidate viewpoint set; continue to perform the deletion operation on the second-ranked pre-selected viewpoint in the sorting result until the last pre-selected viewpoint, to obtain the candidate viewpoint set.

[0012] Optionally, when the number of candidate viewpoints in the candidate viewpoint set remains unchanged, the step of maximizing the reconstructability of the sampling points corresponding to each candidate viewpoint in the candidate viewpoint set based on the pre-selected viewpoint set to generate a quasi-viewpoint set includes: determining a set of deleted viewpoints based on the candidate viewpoint set and the pre-selected viewpoint set; for any candidate viewpoint in the candidate viewpoint set: determining the similarity between each deleted viewpoint in the deleted viewpoint set and the candidate viewpoint; based on the similarity, selecting deleted viewpoints with a similarity greater than a preset threshold from the deleted viewpoint set, and using the selected deleted viewpoints as related viewpoints of the candidate viewpoint to obtain a related viewpoint set; if there exists a related viewpoint in the related viewpoint set that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoint; then using this related viewpoint to replace the candidate viewpoint; if there is no related viewpoint in the related viewpoint set that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoint, then no processing is performed; performing the above operations for each candidate viewpoint in the candidate viewpoint set to generate a quasi-viewpoint set.

[0013] According to a second aspect of the present invention, an aerial photography perspective planning device based on superior view photogrammetry is also provided, comprising: a determination module, configured to determine the viewpoint sampling space of a UAV based on a 3D schematic model of a target scene; a sampling module, configured to perform Poisson disk sampling on the surface of the 3D schematic model to generate a set of sampling points and a set of pre-selected viewpoints corresponding to the set of sampling points; a reconstructibility modeling module, configured to perform reconstructibility modeling on the 3D schematic model based on the set of sampling points to generate the reconstructibility of each sampling point; a redundancy modeling module, configured to perform viewpoint redundancy modeling based on the set of pre-selected viewpoints to generate the redundancy of each pre-selected viewpoint; and an optimization processing module, configured to perform optimization processing on the set of pre-selected viewpoints based on the reconstructibility of the sampling points and the redundancy of the pre-selected viewpoints to generate an aerial photography perspective.

[0014] According to a third aspect of the present invention, an electronic device is also provided, the electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the method as described in the first aspect.

[0015] According to a fourth aspect of the present invention, a computer-readable medium is also provided, on which a computer program is stored, wherein the program, when executed by a processor, implements the method described in the first aspect.

[0016] This invention provides an aerial photography viewpoint planning method based on superior view photogrammetry. The method includes: first, determining the viewpoint sampling space of a UAV based on a 3D approximate model of the target scene; then, performing Poisson disk sampling on the surface of the 3D approximate model to generate a set of sampling points and a pre-selected viewpoint set corresponding to the sampling point set; second, performing reconstructibility modeling on the 3D approximate model based on the sampling point set to generate the reconstructibility of each sampling point; then, performing viewpoint redundancy modeling based on the pre-selected viewpoint set to generate the redundancy of each pre-selected viewpoint; finally, performing optimization processing on the pre-selected viewpoint set based on the reconstructibility of the sampling points and the redundancy of the pre-selected viewpoints to generate the aerial photography viewpoint. This embodiment combines maximizing the reconstructibility of sampling points and minimizing viewpoint redundancy for the optimization processing of the pre-selected viewpoint set of the target scene, thereby improving the accuracy of 3D reconstruction of the target scene while achieving automatic aerial photography planning for the target scene. Attached Figure Description

[0017] The following sections will describe some specific embodiments of the invention in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0018] Figure 1This is a flowchart illustrating an aerial perspective planning method based on superior view photogrammetry, provided as an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of a process for generating aerial viewpoints based on preferred processing, according to an embodiment of the present invention.

[0020] Figure 3 This is a schematic diagram of an aerial perspective planning device based on superior view photogrammetry, provided as an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] like Figure 1 The diagram shown is a flowchart illustrating an aerial perspective planning method based on superior view photogrammetry according to an embodiment of the present invention.

[0023] A method for planning aerial viewpoints based on best-of-view photogrammetry includes at least the following steps:

[0024] S101, based on a 3D rough model of the target scene, determines the viewpoint sampling space of the UAV;

[0025] S102, perform Poisson disk sampling on the viewpoint sampling space to generate a set of sampling points and a set of pre-selected viewpoints corresponding to the set of sampling points;

[0026] S103, Based on the set of sampling points, perform reconstructability modeling on the 3D rough model and generate the reconstructability of each sampling point;

[0027] S104, Based on the pre-selected viewpoint set, perform viewpoint redundancy modeling and generate the redundancy of each pre-selected viewpoint;

[0028] S105, based on the reconstructability of the sampling points and the redundancy of the pre-selected viewpoints, the pre-selected viewpoint set is optimized to generate an aerial viewpoint.

[0029] In S101, in practical applications of scene-oriented 3D reconstruction, the 3D schematic model of the target scene is primarily used as input information for over-the-view photogrammetry. In other words, the 3D schematic model forms the basis for trajectory planning in over-the-view photogrammetry. The 3D schematic model indicates the basic geometric structure representation of the target scene or object. It includes geometric information of the observation site and is mainly used to determine the aerial viewpoint to plan the safe airspace for the UAV's flight path.

[0030] The 3D rough model was obtained through the following method:

[0031] 1) Obtain image data of the target scene using oblique photogrammetry and generate a 3D rough model;

[0032] 2) First, use two-dimensional planar information such as aerial orthophotos to plan the acquisition route of the target scene. Then, the UAV flies according to the acquisition route to collect image data of the target scene. Finally, a 3D rough model is generated based on the image data.

[0033] Based on the UAV's pose control capabilities, the level of detail in the 3D schematic model, and the complexity of the target scene, determine the safe distance d between the UAV and the target scene. min Based on the furthest shooting distance d from the drone to the target scene. max and the safe distance d between the drone and the target scene min Determine the viewpoint sampling space from the drone to the target scene; where the furthest shooting distance d max The maximum observable distance corresponding to a viewpoint is preset based on empirical values; d max The smaller the value, the higher the accuracy of the target scene model reconstruction.

[0034] After identifying the 3D rough model, a safe distance d is used. min Using a scale of d, the drone's no-fly zone is generated by expanding both horizontally and vertically along the target scene. This no-fly zone is used to prevent the drone from planning its trajectory, ensuring flight safety. After recognizing the 3D rough model, the furthest shooting distance d is used... max Using a scale of [scale value], dilation is performed along both the horizontal and vertical directions of the target scene to generate the drone's flight space. Based on the drone's flight space and no-fly zones, the drone's viewpoint sampling space is determined. This not only avoids collisions between the drone and buildings but also increases the drone's sampling frequency.

[0035] It should be noted that if the target scene has a large area, the target scene can be divided into smaller segments, and then the viewpoint sampling space of each segmented smaller scene can be determined.

[0036] In S102, Poisson disk sampling refers to dividing a 3D schematic model into several small local regions, configuring each local region accordingly, and finally generating the corresponding aerial viewpoint, thus achieving the goal of breaking down the whole into smaller parts. Compared with other sampling methods, Poisson disk sampling can better highlight the geometric features of the 3D schematic model surface.

[0037] The surface of the 3D schematic model is sampled using a Poisson disk according to the preset sampling diameter of the Poisson disk. In order to effectively control the density of sampling viewpoints, the sampling diameter of the Poisson disk is determined based on the camera's field of view, the overlap rate between adjacent images, and the preset distance between viewpoints; then, the surface of the 3D schematic model is sampled using the determined sampling diameter.

[0038] For example, the sampling diameter of the Poisson disk is determined based on the camera's field of view, the overlap rate between adjacent images, and a preset distance from the viewpoint; based on the sampling diameter, the surface of the 3D rough model is sampled using the Poisson disk to generate a set of sampling points; for any sampling point in the set of sampling points: the surface normal vector at the sampling point is obtained; the viewpoint is moved a preset distance in the opposite direction of the surface normal vector to obtain a pre-selected viewpoint corresponding to the sampling point; based on the pre-selected viewpoints corresponding to each sampling point in the set of sampling points, a set of pre-selected viewpoints corresponding to the set of sampling points is obtained.

[0039] For example: (1) the sampling diameter D of the Poisson disk disk It is calculated using the following formula (1):

[0040] D disk =d prj ·(1-r overlap );

[0041] d prj =2·d GSD ·tan(θ / 2); Equation (1);

[0042] Where θ is the camera's field of view, r overlap The overlap rate between adjacent images is a preset value obtained based on experience; d prj d represents the field of view projected by the camera onto the surface of the target scene model; GSD Indicates the preset distance of the viewpoint.

[0043] (2) Using the Poisson disk sampling method, according to the sampling diameter D obtained above... disk Poisson disk sampling is performed on the surface of the 3D approximate model to uniformly generate a set of sampling points S for observation.

[0044] S = {{s1,n1},{s2,n2},…,{s i ,n i}}; Equation (2);

[0045] in, N represents i sampling regions i The average normal vector, denoted as the sampling point s i The surface normal vector at that point.

[0046] (3) Sample point s i along the surface normal vector n i move in the opposite direction d GSD Distance, thereby determining the viewpoint v corresponding to the sampling point. i viewpoint v i The shooting direction and the normal vector n i The directions are opposite. Analysis is performed based on stereoscopic vision observations and observability constraints between adjacent sampling points to generate a corresponding viewpoint for each sampling point. This viewpoint v is based on the viewpoint v corresponding to each sampling point in the sampling point set S. i Generate a preselected viewpoint set V, where viewpoint v i ∈V by camera position and direction R i ∈SO(3) definition; distance d GSD This determines the overlap rate of images captured by the preselected viewpoint set V.

[0047] In S103, no restrictions are placed on the method of reconstructability modeling.

[0048] Based on sampling points and corresponding pre-selected viewpoints, quantitative methods are applied to pre-evaluate the accuracy and completeness of the target scene reconstruction results. To this end, a quantitative index, reconstructability, is proposed based on the theory of two-view geometric analysis.

[0049] For example, for any sampling point in the sampling point set: obtain a pre-selected viewpoint and at least one observation viewpoint corresponding to the sampling point; determine a reconstruction quality model of the sampling point between the pre-selected viewpoint and the observation viewpoint; the observation viewpoint is used to indicate a viewpoint located outside the normal vector of the sampling point and capable of observing the sampling point; for any one of the at least one observation viewpoints: determine a reconstructibility metric of the sampling point between the pre-selected viewpoint and the observation viewpoint according to the reconstruction quality model; determine the reconstructibility of the sampling point based on the reconstructibility metric corresponding to each of the at least one observation viewpoints.

[0050] For example: Reconstructability q(s,v) i ,v j ) is used to indicate the preselected viewpoint v corresponding to any sampling point s. i and observation viewpoint v j The reconstruction quality model under observation is shown in Equation (3) below.

[0051] q(s,v i ,v j )=w1(α)w2(d m w3(α)cos(θ) m )

[0052] Equation (3);

[0053] In the formula, α represents the two viewpoints (v i ,v j The angle between the directions of the lines connecting each sampling point s; It represents two viewpoints (v i ,v j The larger of the distances to the sampling point s; θ m =max(θ) i ,θ j ), which represents two viewpoints (v i ,v j The larger of the included angles, the viewpoint included angle is used to indicate the pre-selected viewpoint v. i Or observation viewpoint v j The angle between the sampling point vector and the sampling point normal vector n, respectively; w1, w2 and w3 are weight terms, defined as shown in equation (4) below.

[0054]

[0055] Where α1 represents the preselected viewpoint v i The corresponding preset angle, α3 represents the observation viewpoint v j The corresponding preset angle. k3 and k1 represent coefficients.

[0056] Each sampling point can be observed from multiple viewpoints. For any sampling point s: obtain the set of viewpoints V = {v...} that can observe sampling point s. i} i＝1,…M ; Obtain the pre-selected viewpoint and at least one observation viewpoint corresponding to the sampling point from the viewpoint set; For any observation viewpoint among the at least one observation viewpoint: Calculate the reconstructibility between the pre-selected viewpoint and the observation viewpoint based on equation (3); Then, sum the reconstructibility corresponding to each observation viewpoint among the at least one observation viewpoint to generate the reconstructibility h(s,V) of the sampling point. The specific calculation formula is shown in equation (5) below:

[0057]

[0058] Wherein, δ(s,v) i ) and δ(s,v j ) is the visibility function, representing the pre-selected viewpoint v i and observation viewpoint v jVisibility of sampling point s.

[0059] In S104, for example, for any preselected viewpoint in the preselected viewpoint set: the sampling points that can be observed by the preselected viewpoint are determined as first target sampling points, and a plurality of first target sampling points are obtained; based on the reconstructability corresponding to each of the plurality of first target sampling points, a plurality of reconstructability is obtained; the minimum reconstructability value is selected from the plurality of reconstructability as the redundancy corresponding to the preselected viewpoint.

[0060] For example: at viewpoint v i Several sampling points can be observed at a given location. The reconstructibility h(s,V) corresponding to each sampling point is obtained, resulting in several reconstructibility values. These reconstructibility values ​​are then sorted in chronological order. The minimum reconstructibility h(s,V) among the sorted values ​​is determined as the redundancy r(v) of that viewpoint, i.e.:

[0061] r(v,V)=min{h(s,V)|s∈S,δ(s,v)=1} Formula (6);

[0062] It should be noted that if at viewpoint v i If all observed sampling points have high reconstructability, then the viewpoint v i It may be redundant and requires optimization.

[0063] In S105, based on the reconstructability of sampling points and the redundancy of pre-selected viewpoints, the pre-selected viewpoint set is optimized using preset rules or models to generate aerial viewpoints.

[0064] The pre-selected viewpoint set generated by dense sampling based on a 3D approximate model can ensure sufficient coverage of the target scene, but it also introduces redundant viewpoints. Therefore, it cannot be directly used for trajectory planning in photogrammetry and requires viewpoint optimization. The optimal process involves finding a quasi-viewpoint set from the pre-selected viewpoint set. This quasi-viewpoint set is small enough to fully observe all sampling points of the target scene, maximizing the reconstructability of all sampling points while minimizing the redundancy of the quasi-viewpoint set. In particular, incomplete and low-quality areas of the scene require focused observation.

[0065] This embodiment is based on optimization processing, which can optimize the quasi-viewpoint set from the pre-selected viewpoint set, so that the sampling points can obtain sufficient observation with minimal viewpoint redundancy, that is, maximize the reconstructability of the model sampling points, while minimizing the redundancy of all viewpoints; thus, it can not only effectively eliminate viewpoints with low quality scores in target scenes with more geometric information, but also improve the restoration accuracy of model reconstruction.

[0066] like Figure 2The diagram shown is a flowchart illustrating the process of generating aerial viewpoints based on preferred processing according to an embodiment of the present invention.

[0067] Based on the reconstructability of sampling points and the redundancy of pre-selected viewpoints, the pre-selected viewpoint set is optimized to generate aerial viewpoints; this includes at least the following steps:

[0068] S201, Obtain the redundancy of each pre-selected viewpoint in the pre-selected viewpoint set;

[0069] S202, when it is determined that the reconstructability of the sampling points corresponding to each preselected viewpoint in the preselected viewpoint set meets the preset conditions, the redundancy minimization process is performed on the preselected viewpoint set to obtain the candidate viewpoint set;

[0070] S203, when the number of candidate viewpoints in the candidate viewpoint set remains unchanged, based on the pre-selected viewpoint set, the reconstructability of the sampling points corresponding to each candidate viewpoint in the candidate viewpoint set is maximized to generate a quasi-viewpoint set;

[0071] S204, which sets each vantage point as the aerial viewpoint in the vantage point set.

[0072] In S201, for each preselected viewpoint in the preselected viewpoint set: the redundancy of each preselected viewpoint is determined based on formula (6).

[0073] In S202, for example, the redundancy of each pre-selected viewpoint in the pre-selected viewpoint set is obtained, and the pre-selected viewpoints in the pre-selected viewpoint set are sorted in descending order of redundancy to generate a sorting result; the pre-selected viewpoint ranked first in the sorting result is taken as the viewpoint to be deleted, and a deletion operation is performed on the viewpoint to be deleted in the pre-selected viewpoint set to obtain a candidate viewpoint set; at least one second target sampling point associated with the viewpoint to be deleted is obtained, and the reconstructability of the second target sampling point is updated based on the deletion operation; it is determined whether the reconstructability of each second target sampling point in the at least one second target sampling point is greater than a preset threshold; if yes, the candidate viewpoint set is taken as the pre-selected viewpoint set; if no, the viewpoint to be deleted is added to the candidate viewpoint set; the deletion operation is continued on the second-ranked pre-selected viewpoint in the sorting result until the last pre-selected viewpoint, to obtain a candidate viewpoint set.

[0074] For example:

[0075] Step 1: Determine the redundancy r(v) of each preselected viewpoint in the preselected viewpoint set V, and sort the preselected viewpoints in the preselected viewpoint set in descending order of redundancy to generate the sorting result;

[0076] Step 2: Select the pre-selected viewpoint v with the highest redundancy in the sorting results. iAs the viewpoints to be deleted, the deletion operation is performed on the viewpoints to be deleted in the pre-selected viewpoint set to obtain the candidate viewpoint set W;

[0077] Step 3: Obtain and pre-select viewpoint v i At least one associated second target sampling point, wherein the second target sampling point is used to indicate at the preselected viewpoint v i The observable sampling points; and the reconstructability of each second target sampling point is updated based on equation (5);

[0078] Step 4: Determine whether the reconstructability of each second target sampling point after the update is greater than the threshold t. h If so, return to step 2 and process the second target sampling viewpoint v according to the sorting result. i+1 If not, then the viewpoint to be deleted, v i Add it back to the candidate viewpoint set and return to step 2 to process the second target sampling viewpoint v according to the sorting result. i+1 .

[0079] In S203, for example, based on the candidate viewpoint set and the pre-selected viewpoint set, a set of viewpoints to be deleted is determined; for any candidate viewpoint in the candidate viewpoint set: the similarity between each deleted viewpoint in the deleted viewpoint set and the candidate viewpoint is determined; based on the similarity, deleted viewpoints with a similarity greater than a preset threshold are selected from the deleted viewpoint set, and the selected deleted viewpoints are used as related viewpoints of the candidate viewpoint to obtain a set of related viewpoints; if there is a related viewpoint in the set of related viewpoints that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoint; then the candidate viewpoint is replaced by the related viewpoint; if there is no related viewpoint in the set of related viewpoints that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoint, no processing is performed; the above operations are performed for each candidate viewpoint in the candidate viewpoint set to generate a quasi-viewpoint set.

[0080] For example:

[0081] Step 1: Determine candidate viewpoints v in the candidate viewpoint set i The similarity between the viewpoint and each deleted viewpoint in the deleted viewpoint set. The similarity calculation formula is as shown in Equation (7), which is used to find the nearest candidate viewpoint v from the deleted viewpoint set. i All deleted viewpoints.

[0082]

[0083] Among them, o i ·o j It is the dot product that measures the similarity between viewpoint directions; This represents the distance between the centers of the two cameras; ε is a tiny constant used to avoid division by zero.

[0084] Step 2: Find the corresponding candidate viewpoint v in the candidate viewpoint set W from the deleted viewpoint set V. i Similar related viewpoint set Ω(v i );

[0085] Step 3: From the relevant viewpoint set Ω(v i The query in ) makes candidate viewpoints v i The relevant viewpoint that maximizes the reconstructability H(S,U) of the corresponding sampling points

[0086] Step 4: If the query result representation exists, then the relevant viewpoints will be... Replace v i And for the next candidate viewpoint v in the candidate viewpoint set. i+1 Continue with step 1; if the query result does not exist, do not process it, and proceed to the next candidate viewpoint v in the candidate viewpoint set. i+1 Continue with step 1.

[0087] This embodiment combines minimizing viewpoint redundancy with maximizing reconstructability. By optimizing the aerial photography perspective, it can achieve the goal of maximizing model reconstruction accuracy with the fewest images. This not only realizes automatic aerial photography planning for UAVs, but also reduces aerial photography costs.

[0088] It should be noted that the viewpoint mentioned in each embodiment of the present invention refers to the image acquisition viewpoint when the UAV performs aerial photography of the target scene.

[0089] The following section provides a detailed description of an aerial perspective planning method based on superior view photogrammetry, as provided in this embodiment, using specific application scenarios as examples.

[0090] A method for planning aerial viewpoints based on best-of-view photogrammetry includes at least the following steps:

[0091] S1, based on a 3D rough model of the target scene, determines the viewpoint sampling space of the UAV.

[0092] S2, based on the camera's field of view, the overlap rate between adjacent images, and the preset distance to the viewpoint, determine the sampling diameter of the Poisson disk; based on the sampling diameter, perform Poisson disk sampling on the surface of the 3D rough model to generate a set of sampling points; for any sampling point in the set of sampling points: obtain the surface normal vector at the sampling point; move the viewpoint a preset distance in the opposite direction of the surface normal vector to obtain the pre-selected viewpoint corresponding to the sampling point; based on the pre-selected viewpoint corresponding to each sampling point in the set of sampling points, obtain the pre-selected viewpoint set corresponding to the set of sampling points.

[0093] S3, for any sampling point in the sampling point set: obtain a pre-selected viewpoint and at least one observation viewpoint corresponding to the sampling point; determine the reconstruction quality model of the sampling point between the pre-selected viewpoint and the observation viewpoint; the observation viewpoint is used to indicate a viewpoint located outside the normal vector of the sampling point and capable of observing the sampling point; for any one of the at least one observation viewpoints: determine the reconstructibility metric of the sampling point between the pre-selected viewpoint and the observation viewpoint according to the reconstruction quality model; determine the reconstructibility of the sampling point based on the reconstructibility metric corresponding to each of the at least one observation viewpoints.

[0094] S4, for any preselected viewpoint in the preselected viewpoint set: determine the sampling points that can be observed by the preselected viewpoint as the first target sampling points, and obtain a plurality of first target sampling points; based on the reconstructability corresponding to each of the plurality of first target sampling points, obtain a plurality of reconstructability; select the minimum reconstructability value from the plurality of reconstructability as the redundancy corresponding to the preselected viewpoint.

[0095] S5, obtain the redundancy of each preselected viewpoint in the preselected viewpoint set; obtain the redundancy corresponding to each preselected viewpoint in the preselected viewpoint set, and sort the preselected viewpoints in the preselected viewpoint set in descending order of redundancy to generate a sorting result; take the first preselected viewpoint in the sorting result as the viewpoint to be deleted, and perform a deletion operation on the viewpoint to be deleted in the preselected viewpoint set to obtain a candidate viewpoint set; obtain at least one second target sampling point associated with the viewpoint to be deleted, and update the reconstructability of the second target sampling point based on the deletion operation; determine whether the reconstructability of each second target sampling point in the at least one second target sampling point is greater than a preset threshold; if yes, then take the candidate viewpoint set as the preselected viewpoint set; if no, then add the viewpoint to be deleted to the candidate viewpoint set; continue to perform a deletion operation on the second-ranked preselected viewpoint in the sorting result until the last preselected viewpoint to obtain a candidate viewpoint set.

[0096] S6. Based on the candidate viewpoint set and the pre-selected viewpoint set, determine the deleted viewpoint set; for any candidate viewpoint in the candidate viewpoint set: determine the similarity between each deleted viewpoint in the deleted viewpoint set and the candidate viewpoint; based on the similarity, select deleted viewpoints from the deleted viewpoint set whose similarity is greater than a preset threshold, and use the selected deleted viewpoints as related viewpoints of the candidate viewpoint to obtain a related viewpoint set; if there is a related viewpoint in the related viewpoint set that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoint; then use the related viewpoint to replace the candidate viewpoint; if there is no related viewpoint in the related viewpoint set that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoint, then do not perform any processing; perform the above operations for each candidate viewpoint in the candidate viewpoint set to generate a quasi-viewpoint set.

[0097] This embodiment proposes an aerial viewpoint planning method based on superior view photogrammetry. This method acquires a box-shaped 3D coarse model of the target scene and uses it as the basis for planning, constraining the generation and selection of a pre-selected viewpoint set. Then, based on a criterion combining minimizing viewpoint redundancy and maximizing reconstructability, the sampled point set is evaluated and optimized to select the optimal UAV aerial viewpoint that adapts to the geometric structure of the target scene and meets the 3D reconstruction requirements of the target scene. This achieves automatic UAV aerial photography planning and maximizes model reconstruction accuracy with the fewest possible images.

[0098] like Figure 3 The diagram shown is a structural schematic of an aerial perspective planning device based on superior view photogrammetry provided in an embodiment of the present invention.

[0099] A camera view planning device based on superior view photogrammetry, the device 300 includes: a determination module 301, used to determine the viewpoint sampling space of a UAV based on a 3D schematic model of a target scene; a sampling module 302, used to perform Poisson disk sampling on the surface of the 3D schematic model to generate a set of sampling points and a set of pre-selected viewpoints corresponding to the set of sampling points; a reconstructibility modeling module 303, used to perform reconstructibility modeling on the 3D schematic model based on the set of sampling points to generate the reconstructibility of each sampling point; a redundancy modeling module 304, used to perform viewpoint redundancy modeling based on the set of pre-selected viewpoints to generate the redundancy of each pre-selected viewpoint; and an optimization processing module 305, used to perform optimization processing on the set of pre-selected viewpoints based on the reconstructibility of the sampling points and the redundancy of the pre-selected viewpoints to generate the camera view.

[0100] In a preferred embodiment of this example, the sampling module includes: a determining unit, used to determine the sampling diameter of the Poisson disk based on the camera's field of view, the overlap rate between adjacent images, and a preset distance from the viewpoint; a sampling unit, used to perform Poisson disk sampling on the surface of the 3D approximate model based on the sampling diameter to generate a set of sampling points; an acquiring unit, used for any sampling point in the set of sampling points to: acquire the surface normal vector at the sampling point; move the viewpoint a preset distance in the opposite direction of the surface normal vector to obtain a pre-selected viewpoint corresponding to the sampling point; and an acquiring unit, used to obtain a set of pre-selected viewpoints corresponding to the set of sampling points based on the pre-selected viewpoints corresponding to each sampling point in the set of sampling points.

[0101] In a preferred embodiment of this example, the reconstructibility modeling module is further configured to, for any sampling point in the sampling point set: acquire a pre-selected viewpoint and at least one observation viewpoint corresponding to the sampling point; determine a reconstruction quality model of the sampling point between the pre-selected viewpoint and the observation viewpoint; the observation viewpoint is used to indicate a viewpoint located outside the normal vector of the sampling point and capable of observing the sampling point; for any one of the at least one observation viewpoints: determine a reconstructibility metric of the sampling point between the pre-selected viewpoint and the observation viewpoint based on the reconstruction quality model; and determine the reconstructibility of the sampling point based on the reconstructibility metric corresponding to each of the at least one observation viewpoints.

[0102] In a preferred embodiment of this example, the redundancy modeling module includes: a determining unit, configured to, for any preselected viewpoint in the preselected viewpoint set, determine the sampling points observable by the preselected viewpoint as first target sampling points, thereby obtaining a plurality of first target sampling points; an obtaining unit, configured to, based on the reconstructability corresponding to each of the plurality of first target sampling points, obtain a plurality of reconstructability; and a selecting unit, configured to, select the minimum reconstructability value from the plurality of reconstructability values ​​as the redundancy corresponding to the preselected viewpoint.

[0103] In a preferred embodiment of this example, the preferred processing module includes: an acquisition unit, configured to acquire the redundancy of each of the preselected viewpoints in the preselected viewpoint set; a redundancy processing unit, configured to perform redundancy minimization processing on the preselected viewpoint set to obtain a candidate viewpoint set when the reconstructibility of the sampling points corresponding to each of the preselected viewpoints in the preselected viewpoint set meets a preset condition; a reconstructibility processing unit, configured to perform reconstructibility maximization processing on the sampling points corresponding to each of the candidate viewpoints in the candidate viewpoint set based on the preselected viewpoint set when the number of candidate viewpoints in the candidate viewpoint set remains unchanged, to generate a quasi-viewpoint set; and a determination unit, configured to determine each quasi-viewpoint in the quasi-viewpoint set as an aerial photography viewpoint.

[0104] In a preferred embodiment of this example, the redundancy processing unit includes: a sorting subunit, used to obtain the redundancy corresponding to each pre-selected viewpoint in the pre-selected viewpoint set, and sort the pre-selected viewpoints in the pre-selected viewpoint set in descending order of redundancy to generate a sorting result; a deletion subunit, used to take the first pre-selected viewpoint in the sorting result as the viewpoint to be deleted, and perform a deletion operation on the viewpoint to be deleted in the pre-selected viewpoint set to obtain a candidate viewpoint set; an updating subunit, used to obtain at least one second target sampling point associated with the viewpoint to be deleted, and update the reconstructability of the second target sampling point based on the deletion operation; and a judging subunit, used to judge whether the reconstructability of each second target sampling point in the at least one second target sampling point is greater than a preset threshold; if yes, then the candidate viewpoint set is taken as the pre-selected viewpoint set; if no, then the viewpoint to be deleted is added to the candidate viewpoint set; and the deletion operation is continued on the second pre-selected viewpoint in the sorting result until the last pre-selected viewpoint, to obtain a candidate viewpoint set.

[0105] In a preferred embodiment of this example, the reconstructability processing unit includes: a determination subunit, configured to determine a set of deleted viewpoints based on the candidate viewpoint set and the pre-selected viewpoint set; a similarity selection subunit, configured to, for any candidate viewpoint in the candidate viewpoint set: determine the similarity between each deleted viewpoint in the deleted viewpoint set and the candidate viewpoint; based on the similarity, select deleted viewpoints from the deleted viewpoint set whose similarity is greater than a preset threshold, and use the selected deleted viewpoints as related viewpoints of the candidate viewpoint, thereby obtaining a related viewpoint set; if there exists a related viewpoint in the related viewpoint set that maximizes the reconstructability of the sample points corresponding to the candidate viewpoint; then, use the related viewpoint to replace the candidate viewpoint; if there is no related viewpoint in the related viewpoint set that maximizes the reconstructability of the sample points corresponding to the candidate viewpoint, then no processing is performed; and a generation subunit, configured to perform the above operations for each candidate viewpoint in the candidate viewpoint set to generate a quasi-viewpoint set.

[0106] The above-described apparatus can execute an aerial perspective planning method based on best-of-view photogrammetry provided in an embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing an aerial perspective planning method based on best-of-view photogrammetry. Technical details not described in detail in this embodiment can be found in an aerial perspective planning method based on best-of-view photogrammetry provided in an embodiment of the present invention.

[0107] The present invention also provides an electronic device, comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the aerial perspective planning method based on superior view photogrammetry described in the present invention.

[0108] In addition to the methods and apparatus described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0109] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0110] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps of the methods according to the following embodiments of this application as described in the "Exemplary Methods" section above.

[0111] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0112] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0113] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0114] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0115] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0116] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

[0117] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0118] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0119] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for planning aerial viewpoints based on best-of-view photogrammetry, characterized in that, include: Based on the 3D rough model of the target scene, the viewpoint sampling space of the UAV is determined; The sampling diameter of the Poisson disk is determined based on the camera's field of view, the overlap rate between adjacent images, and the preset distance from the viewpoint. Based on the sampling diameter, Poisson disk sampling is performed on the surface of the 3D rough model to generate a set of sampling points; for any sampling point in the set of sampling points: the surface normal vector at the sampling point is obtained; the viewpoint is moved a preset distance in the opposite direction of the surface normal vector to obtain the pre-selected viewpoint corresponding to the sampling point; Based on the pre-selected viewpoint corresponding to each sampling point in the sampling point set, a pre-selected viewpoint set corresponding to the sampling point set is obtained; For any sampling point in the sampling point set: obtain a pre-selected viewpoint and at least one observation viewpoint corresponding to the sampling point; determine the reconstruction quality model of the sampling point between the pre-selected viewpoint and the observation viewpoint; the observation viewpoint is used to indicate a viewpoint located outside the normal vector of the sampling point and capable of observing the sampling point; for any one of the at least one observation viewpoints: determine the reconstructibility metric of the sampling point between the pre-selected viewpoint and the observation viewpoint according to the reconstruction quality model; determine the reconstructibility of the sampling point based on the reconstructibility metric corresponding to each of the at least one observation viewpoints; Reconstructing the quality model The specific formula is shown in equation (3); Equation (3); In the formula, Represented as two viewpoints to sampling points respectively The directional angle between the lines connecting them; It represents two viewpoints. to sampling points respectively The larger value of the distance; It represents two viewpoints. The larger of the included angles, the viewpoint included angle is used to indicate the pre-selected viewpoint. or observation viewpoint The sampling point vector and the sampling point normal vector respectively The included angle; , and These are all weighted terms; for any pre-selected viewpoint in the pre-selected viewpoint set: the sampling points that can be observed by the pre-selected viewpoint are determined as first target sampling points, and several first target sampling points are obtained; based on the reconstructability corresponding to each of the several first target sampling points, several reconstructability values ​​are obtained; the minimum reconstructability value is selected from the several reconstructability values ​​as the redundancy corresponding to the pre-selected viewpoint; Obtain the redundancy of each pre-selected viewpoint in the pre-selected viewpoint set; when the reconstructability of the sampling points corresponding to each pre-selected viewpoint in the pre-selected viewpoint set meets a preset condition, perform redundancy minimization processing on the pre-selected viewpoint set to obtain a candidate viewpoint set; based on the candidate viewpoint set and the pre-selected viewpoint set, determine the set of viewpoints to be deleted; for any candidate viewpoint in the candidate viewpoint set: determine the similarity between each deleted viewpoint in the deleted viewpoint set and the candidate viewpoints; based on the similarity, select viewpoints from the deleted viewpoint set whose similarity is greater than a preset threshold. The deleted viewpoints are selected as related viewpoints for the candidate viewpoints, resulting in a set of related viewpoints. If there exists a related viewpoint in the set that maximizes the reconstructability of the sample points corresponding to the candidate viewpoint, then that related viewpoint replaces the candidate viewpoint. If there is no related viewpoint in the set that maximizes the reconstructability of the sample points corresponding to the candidate viewpoint, then no processing is performed. The above operation is performed on each candidate viewpoint in the set of candidate viewpoints to generate a set of quasi-viewpoints. Each quasi-viewpoint in the set of quasi-viewpoints is determined as the aerial photography viewpoint.

2. The method according to claim 1, characterized in that, When the reconstructability of the sampling points corresponding to each preselected viewpoint in the preselected viewpoint set meets a preset condition, the redundancy minimization process is performed on the preselected viewpoint set to obtain a candidate viewpoint set; including: Obtain the redundancy of each preselected viewpoint in the preselected viewpoint set, and sort the preselected viewpoints in the preselected viewpoint set in descending order of redundancy to generate a sorting result; The first pre-selected viewpoint in the sorting result is taken as the viewpoint to be deleted, and the deletion operation is performed on the viewpoint to be deleted in the pre-selected viewpoint set to obtain the candidate viewpoint set; Obtain at least one second target sampling point associated with the viewpoint to be deleted, and update the reconstructability of the second target sampling point based on the deletion operation; Determine whether the reconstructability of each of the at least one second target sampling points is greater than a preset threshold; if yes, then use the candidate viewpoint set as a pre-selected viewpoint set; if no, then add the viewpoint to be deleted to the candidate viewpoint set; continue to perform the deletion operation on the second-ranked pre-selected viewpoint in the sorting result until the last pre-selected viewpoint, to obtain the candidate viewpoint set.

3. A device for aerial perspective planning based on best-of-view photogrammetry, characterized in that, include: The determination module is used to determine the viewpoint sampling space of the UAV based on the 3D rough model of the target scene; The sampling module is used to determine the sampling diameter of the Poisson disk based on the camera's field of view, the overlap rate between adjacent images, and the preset distance from the viewpoint. Based on the sampling diameter, Poisson disk sampling is performed on the surface of the 3D rough model to generate a set of sampling points; for any sampling point in the set of sampling points: the surface normal vector at the sampling point is obtained; the viewpoint is moved a preset distance in the opposite direction of the surface normal vector to obtain the pre-selected viewpoint corresponding to the sampling point; Based on the pre-selected viewpoint corresponding to each sampling point in the sampling point set, a pre-selected viewpoint set corresponding to the sampling point set is obtained; The reconstructibility modeling module is configured to, for any sampling point in the sampling point set: acquire a pre-selected viewpoint and at least one observation viewpoint corresponding to the sampling point; determine a reconstruction quality model of the sampling point between the pre-selected viewpoint and the observation viewpoint; the observation viewpoint is used to indicate a viewpoint located outside the normal vector of the sampling point and capable of observing the sampling point; for any one of the at least one observation viewpoints: determine a reconstructibility metric of the sampling point between the pre-selected viewpoint and the observation viewpoint based on the reconstruction quality model; determine the reconstructibility of the sampling point based on the reconstructibility metric corresponding to each of the at least one observation viewpoints; and define the reconstruction quality model. The specific formula is shown in equation (3); Equation (3); In the formula, Represented as two viewpoints to sampling points respectively The directional angle between the lines connecting them; It represents two viewpoints. to sampling points respectively The larger value of the distance; It represents two viewpoints. The larger of the included angles, the viewpoint included angle is used to indicate the pre-selected viewpoint. or observation viewpoint The sampling point vector and the sampling point normal vector respectively The included angle; , and These are all weighted items; A redundancy modeling module is used for any pre-selected viewpoint in the pre-selected viewpoint set to: determine the sampling points that can be observed by the pre-selected viewpoint as first target sampling points, and obtain a plurality of first target sampling points; based on the reconstructability corresponding to each of the plurality of first target sampling points, obtain a plurality of reconstructability; and select the minimum reconstructability value from the plurality of reconstructability as the redundancy corresponding to the pre-selected viewpoint. The optimization processing module obtains the redundancy of each pre-selected viewpoint in the pre-selected viewpoint set; when the reconstructability of the sampling points corresponding to each pre-selected viewpoint in the pre-selected viewpoint set meets a preset condition, it performs redundancy minimization processing on the pre-selected viewpoint set to obtain a candidate viewpoint set; based on the candidate viewpoint set and the pre-selected viewpoint set, it determines the deleted viewpoint set; for any candidate viewpoint in the candidate viewpoint set: it determines the similarity between each deleted viewpoint in the deleted viewpoint set and the candidate viewpoint; based on the similarity, it selects viewpoints from the deleted viewpoint set with similarity greater than a preset value. The threshold is used to delete viewpoints, and the selected deleted viewpoints are used as related viewpoints of the candidate viewpoints to obtain a related viewpoint set; if there is a related viewpoint in the related viewpoint set that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoints, then the candidate viewpoints are replaced by this related viewpoint; if there is no related viewpoint in the related viewpoint set that maximizes the reconstructability of the sampling points corresponding to the candidate viewpoints, then no processing is performed; the above operation is performed for each candidate viewpoint in the candidate viewpoint set to generate a quasi-viewpoint set; each quasi-viewpoint in the quasi-viewpoint set is determined as the aerial photography viewpoint.

4. An electronic device, the electronic device comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method as described in any one of claims 1-2.

5. A computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any one of claims 1-2.

Images