A three-dimensional model generation system and method

Abstract

The present invention discloses a three-dimensional model generation system and method, which relates to the field of model generation technology, including a data acquisition unit, wherein the data acquisition unit is used to collect image data of a gymnasium under conditions of no wind and sufficient light through a drone, and to perform quality assessment and overlap detection on the collected drone images, and to screen the collected drone images based on the quality assessment results and overlap detection results. The present invention proposes a three-dimensional model generation system and method, which collects images of a gymnasium through a set data acquisition unit, and performs quality assessment and overlap detection on the collected images, thereby ensuring the quality of the acquired images and providing support for subsequent three-dimensional modeling, and through a set data processing unit, based on the screened images, acquires point cloud data and removes irrelevant point clouds, and removes abnormal points in relevant point clouds.

Description

Technical Field

[0001] The present invention relates to the technical field of model generation, and in particular to a three-dimensional model generation system and method. Background Art

[0002] At present, the three-dimensional model of an object is mostly obtained by scanning the depth image of the real object, and then the depth image is processed by a three-dimensional modeling tool to construct a three-dimensional model of the real object. In the Chinese patent with application number 201710703847.X, "a three-dimensional model generation method and system are disclosed, the method includes obtaining two-dimensional position data of a two-dimensional element corresponding to a target real scene element and terrain data of the target real scene element; generating three-dimensional position data of a target three-dimensional model required for the target real scene element according to the two-dimensional position data and terrain data; triangulating the two-dimensional element corresponding to the target real scene element; and finally generating a target three-dimensional model corresponding to the target real scene element based on the three-dimensional position data and the triangulation result. It can be seen that the present invention provides a solution for automatically generating a three-dimensional model according to actual two-dimensional position data and terrain data, which has high data production efficiency and less model data volume. Moreover, since the present invention uses actual two-dimensional position data and terrain data as the data source of the three-dimensional model, the generated three-dimensional model has a high position matching degree with the real scene element, thereby improving the reference value of the three-dimensional electronic map.";

[0003] This prior art only solves the problem that the actual two-dimensional data collected manually and the data obtained from oblique photography are currently used as the data sources of the three-dimensional model. The two data standards are different, resulting in large errors in the constructed three-dimensional model, so that the generated three-dimensional model is difficult to match the positions of the real-life elements, affecting the reference value of the three-dimensional electronic map. It does not take into account that when three-dimensional modeling is performed based on the captured real-life images, the image quality should be evaluated and screened first, and when modeling is performed through point cloud data, irrelevant point cloud data should be removed, and when performing point cloud matching, the matching results should be verified for accuracy to ensure the authenticity of the three-dimensional model. Summary of the invention

[0004] The purpose of the present invention is to provide a three-dimensional model generation system and method to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solutions: a three-dimensional model generation system, comprising a data acquisition unit, the data acquisition unit is used to collect image data of a gymnasium under the condition of no wind and sufficient light through a drone, and to perform quality assessment and overlap detection on the collected drone images, and to screen the collected drone images based on the quality assessment results and the overlap detection results;

[0006] A data processing unit, the data processing unit is used to obtain point cloud data of each component of the gymnasium based on the filtered drone image, construct point clouds and three-dimensional models of each component of the gymnasium, and perform ground point cloud removal and point cloud segmentation operations on the point cloud;

[0007] A model generation unit, the model generation unit is used to match the point cloud of each component of the gymnasium with the point cloud of the gymnasium scene, and to build a matching accuracy evaluation system, and to verify the accuracy of the matching result through the evaluation indicators in the matching accuracy evaluation system, so as to ensure the accuracy of the matching between the point cloud of each component of the gymnasium and the point cloud of the gymnasium scene, thereby realizing the three-dimensional modeling of the gymnasium, making the three-dimensional model more realistic and complete;

[0008] The three-dimensional model management unit is used to manage the constructed three-dimensional model scene of the gymnasium and to manage the constructed three-dimensional model roaming of the gymnasium.

[0009] Preferably, the data acquisition unit includes an image data acquisition module and an image quality assessment module. The image data acquisition module is used to collect image data of the gymnasium under windless and sufficient lighting conditions through a drone, and to collect images by oblique camera, and the preset heading and lateral overlap are both 75%, and the collected images are numbered according to the order of camera shooting. The image quality assessment module is used to perform quality assessment on the collected drone images, specifically to perform image quality assessment through the BRISQUE algorithm, and to construct an image rating scatter plot based on the labels of the collected images and the quality assessment results of the corresponding images.

[0010] Preferably, the data acquisition unit further includes an image overlap detection module and an image screening module. The image overlap detection module is used to perform overlap detection on the collected drone images. Specifically, based on the position information of the collected images, the drone aerial route map is simulated to determine the lateral image and the aerial image corresponding to each collected image, and the resolution of the collected images is adjusted. The collected images are convolved using a scaled Gaussian difference kernel to obtain a Gaussian difference scale space. The Gaussian difference scale space is used to locate the extreme point by comparing each sampling point with all its adjacent points, and the direction of the extreme point is calculated by a direction algorithm. The specific calculation formula is as follows:

[0011] θ(x,y)=tan -1 ((L(x,y+1)-L(x,y-1)) / (L(x+1,y)-L(x-1,y)))

[0012] In the formula, (x, y) represents the coordinates of the extreme point, L represents the scale of the extreme point, and by obtaining the relative positional relationship between different images, the mapping point of the extreme point on another image is obtained. Based on the positional relationship of the same extreme point on different images, the overlap of the pixels between the two images is recursively detected. The overlap of the two images is obtained according to the ratio of the overlapping pixels to the total pixels of the image. The image screening module is used to perform average processing on the quality of the collected image based on the image quality assessment result and the overlap detection result, on the premise that the overlap between the collected image and the corresponding lateral image is less than 50%, and the overlap with the corresponding heading image is less than 60%, and remove the image with the quality below the average to achieve the screening of the collected image.

[0013] Preferably, the data processing unit includes a point cloud data acquisition module and a primitive assembly module. The point cloud data acquisition module is used to obtain the point cloud data of each component of the gymnasium based on the screened drone images, and specifically obtain the point cloud data of each basic primitive of the gymnasium through the basic primitive division method, including gymnasium component point cloud data and gymnasium scene point cloud data. The primitive assembly module is used to perform basic transformation operations such as translation and rotation on each basic primitive point cloud, move each basic primitive to a corresponding position, form a complete gymnasium component grid model, and realize the assembly of primitives.

[0014] Preferably, the data processing unit also includes a point cloud processing module and a point cloud segmentation module. The point cloud processing module is used to perform ground point cloud removal operations on the point cloud, specifically to obtain the three-dimensional coordinates of all gymnasium scene point cloud data in the three-dimensional scene, obtain the initial point cloud, and copy the point cloud data to the output point cloud, calculate the height difference between each point in the initial point cloud and the corresponding point in the output point cloud, and compare it with a preset height difference threshold. If the calculation result is within the height difference threshold range, the point is classified as a ground point cloud, otherwise it is not classified as a ground point cloud, the ground point cloud is removed, and an index is created for the obtained ground point cloud, the index is inverted to obtain non-ground point cloud data, and discrete abnormal points are removed by a statistical filtering method, thereby removing the residual abnormal point cloud on the ground. The point cloud segmentation module is used to segment the gymnasium scene point cloud by a Euclidean clustering method, dividing it into individual unconnected point clouds.

[0015] Preferably, the model generation unit includes a point cloud registration module and a judgment criterion selection module. The point cloud registration module is used to match the gymnasium component point cloud with the gymnasium scene point cloud, and specifically adopts a three-dimensional model recognition method based on Hough voting to identify and match the gymnasium component point cloud and the gymnasium scene point cloud. The judgment criterion selection module is used to construct a matching accuracy evaluation system, and specifically selects the Euclidean distance, SHOT difference and average curvature difference between the gymnasium component point cloud and the gymnasium scene point cloud after matching as evaluation indicators in the matching accuracy evaluation system.

[0016] Preferably, the model generation unit further includes a registration determination module and an automatic modeling module, wherein the registration determination module is used to verify the accuracy of the matching result through the evaluation index in the matching accuracy evaluation system, wherein the Euclidean distance between the gymnasium component point cloud and the gymnasium scene point cloud after matching is specifically calculated by a distance search algorithm, and the specific calculation formula is as follows:

[0017]

[0018] In the formula, n represents the total number of gymnasium component point clouds, i represents the i-th point in the gymnasium component point cloud, (x i ,y i ,z i ) represents the coordinates of the i-th point in the stadium component point cloud, (x i ',y i ',z i ') represents the coordinates of the nearest point in the stadium scene point cloud, and the calculation results are collected, wherein the SHOT difference between the stadium component point cloud and the stadium scene point cloud after matching is specifically calculated by the difference algorithm, and the specific calculation formula is as follows:

[0019]

[0020] Where k represents the kth dimension of the i-th point in the gymnasium component point cloud, m represents the total number of dimensions of the i-th point in the gymnasium component point cloud, and s ik represents the k-th SHOT feature value of the i-th point in the stadium component point cloud, s ik ' represents the k-th SHOT feature value of the nearest point in the stadium scene point cloud, and the calculation results are collected, wherein the average curvature difference between the stadium component point cloud and the stadium scene point cloud after matching is specifically calculated by the curvature difference algorithm, and the specific calculation formula is as follows:

[0021]

[0022] In the formula, H i represents the average curvature of the i-th point in the stadium component point cloud, H i' represents the average curvature of the nearest point in the stadium scene point cloud, and collects the calculation results. Based on the collected calculation results of the three evaluation indicators, they are compared with the preset thresholds of the corresponding evaluation indicators respectively. When the calculation results of the three evaluation indicators are not greater than the corresponding evaluation indicator thresholds, the match is judged to be successful, and when the match fails, re-select points and re-judge. The automatic modeling module is used to perform three-dimensional modeling of the stadium based on the successfully registered stadium component point cloud and the stadium scene point cloud.

[0023] Preferably, the three-dimensional model management unit includes a model scene management module and a model roaming management module. The model scene management module is used to manage the constructed three-dimensional model scene of the gymnasium through a scene management method based on octree results, specifically by dividing the three-dimensional model scene of the gymnasium into multiple small areas, so that each small area contains a certain number of gymnasium component models, and each small area is stored in each node in the octree structure respectively; the model roaming management module is used to manage the constructed three-dimensional model roaming of the gymnasium, specifically by setting the moving path of each gymnasium component model in the three-dimensional model of the gymnasium, thereby realizing the movement and perspective transformation of each gymnasium component model through the combined operations of translation, rotation and scaling matrix on each gymnasium component model.

[0024] A three-dimensional model generation method comprises the following steps:

[0025] S1: The image data acquisition module is used to collect image data of the stadium under the condition of no wind and sufficient sunlight, the image quality assessment module is used to assess the quality of the collected drone images, the image overlap detection module is used to perform overlap detection on the collected drone images, and the image screening module is used to screen the collected images based on the image quality assessment results and overlap detection results;

[0026] S2: The point cloud data acquisition module is used to obtain the point cloud data of each component of the stadium based on the filtered drone images. The primitive assembly module is used to perform basic transformation operations such as translation and rotation on each basic primitive point cloud, and each basic primitive is moved to the corresponding position. The point cloud processing module is used to remove the ground point cloud from the point cloud. The point cloud segmentation module is used to segment the stadium scene point cloud into individual unconnected point clouds.

[0027] S3: Match the gymnasium component point cloud with the gymnasium scene point cloud through the point cloud registration module, build a matching accuracy evaluation system through the judgment standard selection module, verify the accuracy of the matching result based on the evaluation indicators in the matching accuracy evaluation system through the registration judgment module, perform three-dimensional modeling of the gymnasium through the automatic modeling module, manage the constructed three-dimensional model scene of the gymnasium through the model scene management module, and manage the constructed three-dimensional model roaming of the gymnasium through the model roaming management module.

[0028] Compared with the prior art, the beneficial effects of the present invention include at least: the present invention proposes a three-dimensional model generation system and method, through the set data acquisition unit, the gymnasium is imaged, and the quality of the acquired images is evaluated and overlapped, so as to ensure the quality of the acquired images and provide support for subsequent three-dimensional modeling, through the set data processing unit, based on the screened images, point cloud data is acquired and irrelevant point clouds are removed, and abnormal points in relevant point clouds are removed, through the set model generation unit, the component point cloud is matched with the scene point cloud, and the matching result is verified for accuracy from three aspects, so as to ensure the accuracy of the matching between the component point cloud and the scene point cloud, so that the three-dimensional model obtained by subsequent modeling is more realistic and complete, and the constructed three-dimensional model scene and roaming are managed through the set three-dimensional model management unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the overall process of an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of the data acquisition unit flow in an embodiment of the present invention;

[0031] Figure 3 This is a schematic diagram of the data processing unit flow in an embodiment of the present invention;

[0032] Figure 4 A schematic diagram of a flow chart of a model generation unit according to an embodiment of the present invention;

[0033] Figure 5 The figure is a flow chart of a three-dimensional model management unit according to an embodiment of the present invention.

[0034] In the figure: 100, data acquisition unit; 101, image data acquisition module; 102, image quality assessment module; 103, image overlap detection module; 104, image screening module; 200, data processing unit; 201, point cloud data acquisition module; 202, primitive assembly module; 203, point cloud processing module; 204, point cloud segmentation module; 300, model generation unit; 301, point cloud registration module; 302, judgment standard selection module; 303, registration judgment module; 304, automatic modeling module; 400, three-dimensional model management unit; 401, model scene management module; 402, model roaming management module. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] See also Figure 1-5 , the present invention provides a technical solution: a three-dimensional model generation system, comprising a data acquisition unit 100, the data acquisition unit 100 is used to collect image data of a gymnasium under conditions of no wind and sufficient light through a drone, and to perform quality assessment and overlap detection on the collected drone images, and to screen the collected drone images based on the quality assessment results and the overlap detection results;

[0037] The data processing unit 200 is used to obtain point cloud data of each component of the gymnasium based on the filtered drone image, construct point clouds and three-dimensional models of each component of the gymnasium, and perform ground point cloud removal and point cloud segmentation operations on the point cloud;

[0038] The model generation unit 300 is used to match the point cloud of each component of the gymnasium with the point cloud of the gymnasium scene, and to build a matching accuracy evaluation system, and to verify the accuracy of the matching result through the evaluation index in the matching accuracy evaluation system, so as to ensure the accuracy of the matching between the point cloud of each component of the gymnasium and the point cloud of the gymnasium scene, thereby realizing the three-dimensional modeling of the gymnasium, making the three-dimensional model more realistic and complete;

[0039] The three-dimensional model management unit 400 is used to manage the constructed three-dimensional model scene of the gymnasium and to manage the constructed three-dimensional model roaming of the gymnasium.

[0040] The data acquisition unit 100 includes an image data acquisition module 101 and an image quality assessment module 102. The image data acquisition module 101 is used to collect image data of the gymnasium under the condition of no wind and sufficient light through a drone, and to collect images by using an oblique camera, and the overlap between the heading and the lateral direction is preset to be 75%, and the collected images are numbered according to the order of shooting. The image quality assessment module 102 is used to perform quality assessment on the collected drone images, specifically, to perform image quality assessment through a BRISQUE algorithm, and to construct an image rating scatter plot based on the number of the collected images and the quality assessment results of the corresponding images;

[0041] The data acquisition unit 100 also includes an image overlap detection module 103 and an image screening module 104. The image overlap detection module 103 is used to perform overlap detection on the collected drone images. Specifically, based on the position information of the collected images, the drone aerial route map is simulated to determine the lateral image and the aerial image corresponding to each collected image, and the resolution of the collected images is adjusted. The collected images are convolved using a scaled Gaussian difference kernel to obtain a Gaussian difference scale space. The Gaussian difference scale space is used to locate the extreme point by comparing each sampling point with all its adjacent points, and the direction of the extreme point is calculated by a direction algorithm. The specific calculation formula is as follows:

[0042] θ(x,y)=tan -1 ((L(x,y+1)-L(x,y-1)) / (L(x+1,y)-L(x-1,y)))

[0043] In the formula, (x, y) represents the coordinates of the extreme point, L represents the scale of the extreme point, and by obtaining the relative positional relationship between different images, the mapping point of the extreme point on another image is obtained. Based on the positional relationship of the same extreme point on different images, the overlap of the pixels between the two images is recursively detected. The overlap of the two images is obtained according to the ratio of the overlapping pixels to the total pixels of the images. The image screening module 104 is used to perform average processing on the quality of the collected images based on the image quality assessment results and the overlap detection results, on the premise that the overlap between the collected images and the corresponding lateral images is less than 50%, and the overlap with the corresponding heading images is less than 60%, and removes images with lower than average quality, so as to screen the collected images.

[0044] The data processing unit 200 includes a point cloud data acquisition module 201 and a primitive assembly module 202. The point cloud data acquisition module 201 is used to obtain the point cloud data of each component of the gymnasium based on the filtered drone image, and specifically obtain the point cloud data of each basic primitive of the gymnasium through the basic primitive division method, including the gymnasium component point cloud data and the gymnasium scene point cloud data. The primitive assembly module 202 is used to perform basic transformation operations such as translation and rotation on each basic primitive point cloud, move each basic primitive to a corresponding position, form a completed gymnasium component grid model, and realize the assembly of primitives;

[0045] The data processing unit 200 also includes a point cloud processing module 203 and a point cloud segmentation module 204. The point cloud processing module 203 is used to perform ground point cloud removal operations on the point cloud, specifically to obtain the three-dimensional coordinates of all gymnasium scene point cloud data in the three-dimensional scene, obtain the initial point cloud, and copy the point cloud data to the output point cloud, calculate the height difference between each point in the initial point cloud and the corresponding point in the output point cloud, and compare it with a preset height difference threshold. If the calculation result is within the height difference threshold range, the point is classified as a ground point cloud, otherwise it is not classified as a ground point cloud, the ground point cloud is removed, and an index is created for the obtained ground point cloud, the index is inverted to obtain non-ground point cloud data, and discrete abnormal points are removed by a statistical filtering method, thereby removing the residual abnormal point cloud on the ground. The point cloud segmentation module 204 is used to segment the gymnasium scene point cloud by a Euclidean clustering method, and divide it into individual unconnected point clouds;

[0046] The model generation unit 300 includes a point cloud registration module 301 and a judgment standard selection module 302. The point cloud registration module 301 is used to match the gymnasium component point cloud with the gymnasium scene point cloud, and specifically adopts a three-dimensional model recognition method based on Hough voting to identify and match the gymnasium component point cloud with the gymnasium scene point cloud. The judgment standard selection module 302 is used to build a matching accuracy evaluation system, and specifically selects the Euclidean distance, SHOT difference and average curvature difference between the gymnasium component point cloud and the gymnasium scene point cloud after matching as evaluation indicators in the matching accuracy evaluation system.

[0047] The model generation unit 300 also includes a registration determination module 303 and an automatic modeling module 304. The registration determination module 303 is used to verify the accuracy of the matching result through the evaluation index in the matching accuracy evaluation system. The Euclidean distance between the gymnasium component point cloud and the gymnasium scene point cloud after matching is specifically calculated by a distance search algorithm, and its specific calculation formula is as follows:

[0048]

[0049] In the formula, n represents the total number of gymnasium component point clouds, i represents the i-th point in the gymnasium component point cloud, (x i ,y i ,z i ) represents the coordinates of the i-th point in the stadium component point cloud, (x i ',y i ',z i ') represents the coordinates of the nearest point in the stadium scene point cloud, and the calculation results are collected, wherein the SHOT difference between the stadium component point cloud and the stadium scene point cloud after matching is specifically calculated by the difference algorithm, and the specific calculation formula is as follows:

[0050]

[0051] Where k represents the kth dimension of the i-th point in the gymnasium component point cloud, m represents the total number of dimensions of the i-th point in the gymnasium component point cloud, and s ik represents the k-th SHOT feature value of the i-th point in the stadium component point cloud, s ik ' represents the k-th SHOT feature value of the nearest point in the stadium scene point cloud, and the calculation results are collected, wherein the average curvature difference between the stadium component point cloud and the stadium scene point cloud after matching is specifically calculated by the curvature difference algorithm, and the specific calculation formula is as follows:

[0052]

[0053] In the formula, H i represents the average curvature of the i-th point in the stadium component point cloud, H i ' represents the average curvature of the nearest point in the stadium scene point cloud, and collects the calculation results, based on the collected calculation results of the three evaluation indicators, compares them with the preset thresholds of the corresponding evaluation indicators respectively, and judges that the match is successful when the calculation results of the three evaluation indicators are not greater than the corresponding evaluation indicator thresholds, and reselects points and re-judges when the match fails. The automatic modeling module 304 is used to perform three-dimensional modeling of the stadium based on the successfully registered stadium component point cloud and the stadium scene point cloud;

[0054] The three-dimensional model management unit 400 includes a model scene management module 401 and a model roaming management module 402. The model scene management module 401 is used to manage the constructed three-dimensional model scene of the gymnasium through a scene management method based on the octree result, specifically, the three-dimensional model scene of the gymnasium is divided into multiple small areas, each small area contains a certain number of gymnasium component models, and each small area is stored in each node in the octree structure. The model roaming management module 402 is used to manage the constructed three-dimensional model roaming of the gymnasium, specifically, the moving path of each gymnasium component model is set in the three-dimensional model of the gymnasium, so as to realize the movement and perspective transformation of each gymnasium component model by performing a combination of translation, rotation and scaling matrix operations on each gymnasium component model;

[0055] A three-dimensional model generation method comprises the following steps:

[0056] S1: The image data acquisition module 101 collects image data of the gymnasium under the condition of no wind and sufficient sunlight, the image quality assessment module 102 assesses the quality of the collected drone images, the image overlap detection module 103 performs overlap detection on the collected drone images, and the image screening module 104 screens the collected images based on the image quality assessment result and the overlap detection result;

[0057] S2: The point cloud data acquisition module 201 obtains the point cloud data of each component of the gymnasium based on the filtered drone image, and the primitive assembly module 202 performs basic transformation operations such as translation and rotation on each basic primitive point cloud to move each basic primitive to the corresponding position, and the point cloud processing module 203 performs ground point cloud removal on the point cloud, and the point cloud segmentation module 204 segments the gymnasium scene point cloud into individual unconnected point clouds;

[0058] S3: Match the gymnasium component point cloud with the gymnasium scene point cloud through the point cloud registration module 301, build a matching accuracy evaluation system through the judgment standard selection module 302, verify the accuracy of the matching result based on the evaluation indicators in the matching accuracy evaluation system through the registration judgment module 303, three-dimensionally model the gymnasium through the automatic modeling module 304, manage the constructed three-dimensional model scene of the gymnasium through the model scene management module 401, and manage the constructed three-dimensional model roaming of the gymnasium through the model roaming management module 402.

[0059] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0060] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A three-dimensional model generation system, characterized in that: The method comprises a data acquisition unit (100), wherein the data acquisition unit (100) is used to acquire image data of a gymnasium under windless and sufficient light conditions by using a drone, and to perform quality assessment and overlap detection on the acquired drone images, and to screen the acquired drone images based on the quality assessment results and overlap detection results; The data acquisition unit (100) further comprises an image overlap detection module (103) and an image screening module (104). The image overlap detection module (103) is used to perform overlap detection on the collected drone images. Specifically, based on the position information of the collected images, the drone aerial route map is simulated to determine the lateral image and the aerial image corresponding to each collected image, and the resolution of the collected images is adjusted. The collected images are convolved using a scaled Gaussian difference kernel to obtain a Gaussian difference scale space. The Gaussian difference scale space is used to locate the extreme point by comparing each sampling point with all its adjacent points, and the direction of the extreme point is calculated by a direction algorithm. and obtaining the mapping point of the extreme point on another image by obtaining the relative position relationship between different images, recursively detecting the overlap of the pixels between the two images based on the position relationship of the same extreme point on different images, and obtaining the overlap of the two images based on the ratio of the overlapping pixels to the total pixels of the images. The image screening module (104) is used to perform average processing on the quality of the collected images based on the image quality assessment results and the overlap detection results, provided that the overlap between the collected images and the corresponding lateral images is less than 50 percent and the overlap with the corresponding heading images is less than 60 percent, and remove images with quality below the average, thereby screening the collected images; A data processing unit (200), the data processing unit (200) being used to obtain point cloud data of each component of the gymnasium based on the filtered drone image, construct point clouds and three-dimensional models of each component of the gymnasium, and perform ground point cloud removal and point cloud segmentation operations on the point cloud; A model generation unit (300), the model generation unit (300) is used to match the point cloud of each component of the gymnasium with the point cloud of the gymnasium scene, and to construct a matching accuracy evaluation system, and to verify the accuracy of the matching result through the evaluation index in the matching accuracy evaluation system, so as to ensure the accuracy of the matching between the point cloud of each component of the gymnasium and the point cloud of the gymnasium scene, thereby realizing three-dimensional modeling of the gymnasium, making the three-dimensional model more realistic and complete; A three-dimensional model management unit (400) is used to manage the constructed three-dimensional model scene of the gymnasium and to manage the constructed three-dimensional model roaming of the gymnasium.

2. A three-dimensional model generation system according to claim 1, characterized in that: The data acquisition unit (100) comprises an image data acquisition module (101) and an image quality assessment module (102). The image data acquisition module (101) is used to collect image data of a gymnasium under windless and sufficient light conditions by using a drone, and to collect images by using an oblique camera method, and the overlap between the heading and the lateral direction is preset to be 75 percent, and the collected images are numbered according to the order of the camera shooting. The image quality assessment module (102) is used to perform quality assessment on the collected drone images, specifically, to perform image quality assessment by using a BRISQUE algorithm, and to construct an image rating scatter plot based on the number of the collected images and the quality assessment results of the corresponding images.

3. A three-dimensional model generation system according to claim 1, characterized in that: The data processing unit (200) comprises a point cloud data acquisition module (201) and a graphic element assembly module (202). The point cloud data acquisition module (201) is used to acquire point cloud data of each component of the gymnasium based on the filtered drone image, and specifically obtain point cloud data of each basic graphic element of the gymnasium through a basic graphic element division method, including gymnasium component point cloud data and gymnasium scene point cloud data. The graphic element assembly module (202) is used to perform basic transformation operations such as translation and rotation on each basic graphic element point cloud, move each basic graphic element to a corresponding position, form a completed gymnasium component grid model, and realize the assembly of graphic elements.

4. A three-dimensional model generation system according to claim 1, characterized in that: The data processing unit (200) further comprises a point cloud processing module (203) and a point cloud segmentation module (204). The point cloud processing module (203) is used to perform a ground point cloud removal operation on the point cloud, specifically obtaining the three-dimensional coordinates of all stadium scene point cloud data in the three-dimensional scene to obtain an initial point cloud, and copying the point cloud data to the output point cloud, calculating the height difference between each point in the initial point cloud and the corresponding point in the output point cloud, and comparing it with a preset height difference threshold. If the calculation result is within the height difference threshold range, the point is classified as a ground point cloud, otherwise it is not classified as a ground point cloud, the ground point cloud is removed, and an index is created for the obtained ground point cloud, the index is inverted to obtain non-ground point cloud data, and discrete abnormal points are removed by a statistical filtering method, thereby achieving the removal of abnormal point clouds remaining on the ground. The point cloud segmentation module (204) is used to segment the stadium scene point cloud by a Euclidean clustering method to segment it into individual unconnected point clouds.

5. A three-dimensional model generation system according to claim 1, characterized in that: The model generation unit (300) comprises a point cloud registration module (301) and a judgment standard selection module (302). The point cloud registration module (301) is used to match the gymnasium component point cloud with the gymnasium scene point cloud, specifically using a three-dimensional model recognition method based on Hough voting to identify and match the gymnasium component point cloud with the gymnasium scene point cloud. The judgment standard selection module (302) is used to construct a matching accuracy evaluation system, specifically selecting the Euclidean distance, SHOT difference and average curvature difference between the gymnasium component point cloud and the gymnasium scene point cloud after matching as evaluation indicators in the matching accuracy evaluation system.

6. A three-dimensional model generation system according to claim 1, characterized in that: The model generation unit (300) further comprises a registration determination module (303) and an automatic modeling module (304), wherein the registration determination module (303) is used to verify the accuracy of the matching result by using the evaluation index in the matching accuracy evaluation system, wherein the Euclidean distance between the gymnasium component point cloud and the gymnasium scene point cloud after matching is specifically calculated by a distance search algorithm, and the calculation results are collected, wherein the SHOT difference between the gymnasium component point cloud and the gymnasium scene point cloud after matching is specifically calculated by a difference algorithm, and the calculation results are collected, wherein the average curvature difference between the gymnasium component point cloud and the gymnasium scene point cloud after matching is specifically calculated by a curvature difference algorithm, and the calculation results are collected, based on the collected calculation results of the three evaluation indexes, they are respectively compared with the preset threshold values ​​of the corresponding evaluation indexes, and when the calculation results of the three evaluation indexes are not greater than the corresponding evaluation index threshold values, it is judged that the matching is successful, and when the matching fails, re-selection of points and re-judgment are performed, and the automatic modeling module (304) is used to perform three-dimensional modeling of the gymnasium based on the successfully registered gymnasium component point cloud and the gymnasium scene point cloud.

7. A three-dimensional model generation system according to claim 1, characterized in that: The three-dimensional model management unit (400) comprises a model scene management module (401) and a model roaming management module (402). The model scene management module (401) is used to manage the constructed three-dimensional model scene of the gymnasium through a scene management method based on an octree result, specifically dividing the three-dimensional model scene of the gymnasium into a plurality of small areas, so that each small area contains a certain number of gymnasium component models, and storing each small area in each node in the octree structure. The model roaming management module (402) is used to manage the constructed three-dimensional model roaming of the gymnasium, specifically setting a moving path for each gymnasium component model in the three-dimensional model of the gymnasium, thereby realizing the movement and perspective transformation of each gymnasium component model by performing a combination of translation, rotation and scaling matrix operations on each gymnasium component model.

8. A three-dimensional model generation method, characterized in that The following steps are involved: S1: The image data acquisition module (101) is used to collect image data of the gymnasium under the condition of no wind and sufficient sunlight, the image quality assessment module (102) is used to assess the quality of the collected drone images, and the image overlap detection module (103) is used to perform overlap detection on the collected drone images. Specifically, based on the position information of the collected images, the drone aerial route map is simulated to determine the lateral image and the aerial image corresponding to each collected image, and the resolution of the collected images is adjusted. The collected images are convolved using a scaled Gaussian difference kernel to obtain a Gaussian difference scale space, and the extreme value points are located in the Gaussian difference scale space by comparing each sampling point with all its adjacent points. The direction of the extreme point is calculated by a direction algorithm, and the mapping point of the extreme point on another image is obtained by obtaining the relative position relationship between different images. Based on the position relationship of the same extreme point on different images, the overlap of the pixels between the two images is recursively detected. The overlap of the two images is obtained based on the ratio of the overlapping pixels to the total pixels of the images. The image screening module (104) performs average processing on the quality of the collected images based on the image quality evaluation results and the overlap detection results, provided that the overlap of the collected images with the corresponding lateral images is less than 50 percent and the overlap with the corresponding heading images is less than 60 percent, and removes images with quality below the average, thereby implementing screening of the collected images. S2: acquiring point cloud data of each component of the gymnasium based on the filtered drone image through the point cloud data acquisition module (201), performing basic transformation operations such as translation and rotation on each basic primitive point cloud through the primitive assembly module (202), moving each basic primitive to a corresponding position, performing ground point cloud removal operations on the point cloud through the point cloud processing module (203), and segmenting the gymnasium scene point cloud through the point cloud segmentation module (204), segmenting it into individual unconnected point clouds; S3: Match the gymnasium component point cloud with the gymnasium scene point cloud through the point cloud registration module (301), build a matching accuracy evaluation system through the judgment standard selection module (302), verify the accuracy of the matching result based on the evaluation indicators in the matching accuracy evaluation system through the registration judgment module (303), three-dimensionally model the gymnasium through the automatic modeling module (304), manage the constructed three-dimensional model scene of the gymnasium through the model scene management module (401), and manage the constructed three-dimensional model roaming of the gymnasium through the model roaming management module (402).

Images