Point cloud enhancement method and device based on binocular reconstruction, equipment and storage medium
By using a binocular reconstruction method, the problem of point cloud sparsity in reflective or black surface areas is solved by correcting sparse point clouds with local homography matrix and epipolar constraints, thus achieving efficient 3D reconstruction results.
Patent Information
- Application Number
- CN202511077470.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-01
- Publication Date
- 2025-11-07
AI Technical Summary
When performing 3D scanning or reconstruction on reflective or black surface areas, existing technologies struggle to effectively acquire structured light information, leading to point cloud sparsity and making it impossible to achieve complete data acquisition while maintaining the original surface state.
A binocular reconstruction-based method is adopted, which acquires sparse point clouds and projects them onto a preset pixel plane. The method is then corrected using local homography matrix and epipolar constraints, and verified by average depth value, to achieve accurate reconstruction of the point clouds.
It effectively enhances the reconstruction accuracy and integrity of sparse point clouds, and improves the efficiency and accuracy of 3D reconstruction, especially in areas with high reflectivity or black surfaces.
Smart Images
Figure CN120912781A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] Embodiments of the present application relate to the technical field of three-dimensional scanning, and particularly relate to a point cloud enhancement method and device based on binocular reconstruction, equipment and a storage medium. BACKGROUND
[0002] In the field of structured light three-dimensional scanning or three-dimensional reconstruction, due to the diffuse reflection / light absorption effect of light, the structured light information of the surface of an object is sometimes difficult to be collected and acquired by a camera, resulting in that local point cloud information is difficult to be effectively established, so that the overall point cloud is sparse in the reflective or black surface area.
[0003] Specifically, the surface characteristics of an object fundamentally limit data collection, especially in the scanning process of high-reflective surfaces (such as polished metals) and pure black objects (such as rubber products), due to the directional scattering of light caused by specular reflection or the light absorption effect of the material, the projected structured light pattern is difficult to be effectively captured by the camera. The current mainstream solutions include adding a polarizer group (reducing 60% light flux), multi-exposure fusion (producing motion artifacts) or surface spraying matte powder (polluting the object), these methods cannot achieve complete data collection under the premise of maintaining the original surface state, resulting in significant sparseness of the reconstructed point cloud in the special surface area. SUMMARY
[0004] An object of embodiments of the present application is to provide a point cloud enhancement method and device based on binocular reconstruction, equipment and a storage medium, to solve the technical problem that the point cloud obtained is sparse when three-dimensional scanning or reconstruction is performed on a reflective or black surface area in the related art.
[0005] In a first aspect, embodiments of the present application provide a point cloud enhancement method based on binocular reconstruction, the method comprising:
[0006] obtaining a sparse point cloud separated from a global point cloud;
[0007] projecting the sparse point cloud to a preset left pixel plane and a preset right pixel plane respectively, to obtain left pixel coordinates of the sparse point cloud in the left pixel plane and right pixel coordinates of the sparse point cloud in the right pixel plane;
[0008] determining a bounding box region in the left pixel plane according to the distribution relationship of the left pixel coordinates in the left pixel plane, and determining a matching point pair set about the left pixel coordinates and the corresponding right pixel coordinates by traversing the pixel points in the bounding box region;
[0009] According to the matching point pair set, a local homography matrix and an average depth value are calculated, a matching point of a right pixel coordinate is generated by using the homography matrix, and the matching point is corrected by a epipolar constraint, three-dimensional reconstruction is performed on the corrected matching point pair by using a binocular vision principle, and a reconstructed point is obtained;
[0010] According to the average depth value, the reconstructed point is verified, and the reconstructed point meeting a verification condition is added to the global point cloud.
[0011] With reference to the first aspect, in a possible implementation manner, the bounding box region in the left pixel plane is determined according to the distribution relationship of the left pixel coordinates in the left pixel plane, and the bounding box region in the left pixel plane comprises:
[0012] The maximum and minimum values of the left pixel coordinates in the u and v directions in the left pixel plane are respectively calculated as u_min_l, v_min_l, u_max_l, and v_max_l.
[0013] The bounding box region corresponding to the left pixel plane of the left pixel coordinates is determined according to the u_min_l, v_min_l, u_max_l, and v_max_l.
[0014] With reference to the first aspect, in a possible implementation manner, the matching point pair set about the left pixel coordinates and corresponding right pixel coordinates is determined by traversing the pixel points in the bounding box region, and the matching point pair set comprises:
[0015] Each pixel point in the bounding box region is traversed, and a plurality of sampling points are selected according to a preset sampling interval.
[0016] A target sampling point is determined, and the target sampling point is any one of the plurality of sampling points.
[0017] The left pixel coordinates are searched in a preset neighborhood of the target sampling point, and the right pixel coordinates corresponding to the searched left pixel coordinates are determined.
[0018] The searched left pixel coordinates and the corresponding right pixel coordinates are stored in a preset matching point pair set.
[0019] With reference to the first aspect, in a possible implementation manner, after the matching point pair set about the left pixel coordinates and corresponding right pixel coordinates is determined, the method further comprises:
[0020] The number of point pairs in the matching point pair set is determined.
[0021] It is judged whether the number of point pairs is less than a preset threshold.
[0022] If yes, continue to traverse the pixel points in the bounding box region, increase the number of point pairs in the matching point pair set until the number of point pairs is not less than the preset threshold value;
[0023] If no, perform the step of calculating a local homography matrix and an average depth value according to the matching point pair set.
[0024] In combination with the first aspect, in a possible implementation manner, the generating a matching point of the right pixel coordinate by using the homography matrix and correcting the matching point by using the epipolar constraint comprises:
[0025] determining, by using the homography matrix, a pixel coordinate Pr of a right pixel plane corresponding to any one pixel point Pl in the bounding box region;
[0026] calculating an epipolar line corresponding to the point Pl in the right pixel plane, and projecting the pixel coordinate Pr to the epipolar line to obtain a corrected matching point.
[0027] In combination with the first aspect, in a possible implementation manner, the verifying the reconstructed point according to the average depth value and adding a reconstructed point meeting a verification condition to the global point cloud comprises:
[0028] determining a coordinate X of the reconstructed point;
[0029] determining a directional component of the coordinate X in a depth direction;
[0030] calculating an absolute value of a difference value between the directional component and the average depth value;
[0031] if the absolute value is less than a preset fixed value, adding the reconstructed point to the global point cloud.
[0032] In combination with the first aspect, in a possible implementation manner, the method projects the sparse point cloud to a preset left pixel plane and a preset right pixel plane respectively based on internal and external parameters of a left camera corresponding to the left pixel plane and internal and external parameters of a right camera corresponding to the right pixel plane.
[0033] In a second aspect, an embodiment of the present application provides a binocular reconstruction three-dimensional point cloud enhancement device based on homography estimation, comprising:
[0034] a data acquisition module configured to acquire a sparse point cloud separated from a global point cloud;
[0035] a point cloud projection module configured to project the sparse point cloud to a preset left pixel plane and a preset right pixel plane respectively, to obtain a left pixel coordinate of the sparse point cloud in the left pixel plane and a right pixel coordinate of the sparse point cloud in the right pixel plane;
[0036] a data matching module configured to determine a bounding box region in the left pixel plane according to a distribution relationship of the left pixel coordinates in the left pixel plane, and determine a set of matching point pairs about the left pixel coordinates and corresponding right pixel coordinates by traversing pixel points in the bounding box region;
[0037] a three-dimensional reconstruction module configured to calculate a local homography matrix and an average depth value according to the set of matching point pairs, generate matching points of right pixel coordinates by using the homography matrix, correct the matching points by using epipolar constraint, and perform three-dimensional reconstruction on the corrected matching point pairs by using binocular vision principle to obtain reconstructed points;
[0038] a data verification module configured to verify the reconstructed points according to the average depth value, and add reconstructed points meeting a verification condition to the global point cloud.
[0039] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, and the processor, when executing the one or more computer programs, causes the electronic device to implement the method in any one of the above first aspect.
[0040] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, the computer readable storage medium storing a computer program, the computer program including program instructions, and the program instructions, when executed by a processor, causing the processor to execute the method in any one of the above aspects.
[0041] The embodiments of the present application can achieve the following technical effects:
[0042] Based on the method proposed in the embodiments of the present application, when performing sparse point cloud enhancement, the right camera pixel coordinates corresponding to the left camera pixel coordinates are determined based on the principle of homography matrix, and the right camera pixel coordinates are further accurately searched according to the principle of epipolar constraint, and finally the left / right pixel coordinates are accurately reconstructed by using the principle of binocular vision, which can effectively enhance the discrete sparse part of the point cloud, and is more efficient. BRIEF DESCRIPTION OF DRAWINGS
[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.
[0044] Figure 1A flowchart of a point cloud enhancement method based on binocular reconstruction provided by an embodiment of the present application is shown in FIG. 1.
[0045] Figure 2 A point pair comparison diagram before and after enhancement provided by an embodiment of the present application is shown in FIG. 2.
[0046] Figure 3 A structural diagram of an electronic device provided by an embodiment of the present application is shown in FIG. 3. DETAILED DESCRIPTION
[0047] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0048] It should be noted that the various features in the embodiments of the present application can be combined with each other without conflict, and all fall within the scope of protection of the present application. In addition, although the functional modules are divided in the device schematic diagram, and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from the module division in the device or the order in the flowchart. Furthermore, the "first", "second", "third" and the like used in the present application do not limit the data and execution order, but only distinguish the same items or similar items with basically the same function and effect.
[0049] In the first aspect, referring to Figure 1 The present application provides a point cloud enhancement method based on binocular reconstruction, which comprises the following steps:
[0050] Step S10, obtaining a sparse point cloud separated from a global point cloud;
[0051] Step S20, projecting the sparse point cloud to a preset left pixel plane and a preset right pixel plane respectively, to obtain left pixel coordinates of the sparse point cloud in the left pixel plane and right pixel coordinates of the sparse point cloud in the right pixel plane;
[0052] Step S30, determining a bounding box region in the left pixel plane according to the distribution relationship of the left pixel coordinates in the left pixel plane, and determining a matching point pair set about the left pixel coordinates and the corresponding right pixel coordinates by traversing the pixel points in the bounding box region;
[0053] Step S40, according to the matching point pair set, a local homography matrix and an average depth value are calculated, the homography matrix is used to generate a matching point of a right pixel coordinate, and the matching point is corrected through an epipolar constraint, a three-dimensional reconstruction is performed on the corrected matching point pair through a binocular vision principle, and a reconstructed point is obtained;
[0054] Step S50, according to the average depth value, the reconstructed point is verified, and the reconstructed point meeting the verification condition is added to the global point cloud.
[0055] In the embodiment, first, a local sparse point cloud (point set) is extracted from an existing global point cloud, and the point cloud is used as a basis for point cloud enhancement in subsequent steps.
[0056] As a feasible implementation manner, the sparse point cloud is separated, and the specific process includes but is not limited to the following:
[0057] The points in a certain local cubic range are cropped according to a spatial region; the key feature points such as edges and corner points are reserved, and the redundant plane points are removed according to a feature screening; and the point cloud is simplified by using voxel grid downsampling and the like.
[0058] In the embodiment, the projection from a three-dimensional point to an image pixel is realized through camera intrinsic parameters and extrinsic parameters.
[0059] For example, assuming that a three-dimensional point in the sparse point cloud is P (X, Y, Z) (a world coordinate system or a camera coordinate system), and the projection coordinates (ul, vl) and (ur, vr) of the three-dimensional point in a left / right pixel plane, the calculation formula of the three-dimensional point can be as follows:
[0060]
[0061] Wherein, K is a camera intrinsic parameter matrix (including a focal length, a principal point coordinate and the like), and Z is a depth (a distance along an optical axis direction) of the point to the camera plane.
[0062] In the embodiment, according to the distribution of the left pixel coordinates, a minimum rectangle capable of covering all the left pixel points is calculated, the bounding box region is obtained, the whole image (usually 1920*1080 pixels) is avoided to be traversed, and the calculation amount is reduced. Meanwhile, in the bounding box region, corresponding right pixel points are found for each left pixel point through pixel correlation or feature matching, and the matching point pair is formed.
[0063] In the embodiment, the matching precision is optimized through a local geometric constraint, and a three-dimensional coordinate (reconstructed point) is inversely deduced through a binocular vision principle, wherein the local homography matrix is used to describe a local mapping relationship (which is suitable for an approximately planar scene) of the left pixel plane and the right pixel plane in the bounding box region.
[0064] It is easy to understand that in the binocular system, the corresponding right pixel point of the left pixel point must be on its epipolar line (epipolar constraint), so that the epipolar lines of the left and right images can be adjusted to be horizontal through correction (such as using a stereo correction matrix), and the search range of the right pixel point is reduced from two dimensions to one dimension (the same row), which greatly improves the matching accuracy.
[0065] As shown in FIG. 1, the left image and the right image are obtained by a binocular system, and the left image and the right image are used as input images. Figure 2 , Figure 2 As shown in FIG. 1, the left image and the right image are obtained by a binocular system, and the left image and the right image are used as input images.
[0066] After the reconstruction point is obtained through step S40, the average depth value thereof is verified to ensure that the three-dimensional point obtained by binocular reconstruction is reliable, accurate, and consistent with the current scene depth information, and the reconstruction point is included in the final global point cloud model only after meeting these conditions.
[0067] The average depth value is obtained in step S40, and is used to indicate the average value of the depth of all matching points for calculation in the local region (bounding box region) determined in step S30 (in addition, a typical depth value in the bounding box region can also be used instead of the average depth value for calculation).
[0068] The embodiment provides a process for verifying the average depth value, for reference.
[0069] First, spatial consistency is checked. Although the reconstruction point generated in step S40 is derived from the same local region, due to image noise, matching errors, and subtle changes in the scene, the actual depth (Z coordinate) of some reconstruction points may deviate significantly from the average depth value of the region.
[0070] A tolerance range is set. The system defines an acceptable tolerance depth range based on the calculated average depth value. For example, the tolerance range can be the average depth value ± a fixed depth threshold (such as ± 0.5 meters); or the average depth value ± the average depth value × a percentage (such as ± 10%), that is, by dynamically setting the tolerance range, the variation amplitude of the depth value itself is fused (for example, the absolute error allowed for objects close to the camera should be small, and the absolute error allowed for objects far from the camera can be large).
[0071] Then, for each reconstruction point generated by S40, it is checked whether the depth coordinate falls within the above set tolerance range.
[0072] If the depth value of a reconstruction point is within the set tolerance range (for example, within [average depth - threshold, average depth + threshold]), the reconstruction point is considered to be reasonable, reliable, and consistent with the depth information of the current scene region.
[0073] If the depth value of a reconstructed point is far beyond the set tolerance range (for example, too deep or too shallow than the average depth value), it is considered that the reconstructed point is an outlier or an error reconstructed point caused by error matching, noise, occlusion or other abnormalities.
[0074] Further, the method further comprises:
[0075] respectively obtaining maximum and minimum values of the left pixel coordinates in u and v directions of the left pixel plane, u_min_l, v_min_l, u_max_l, v_max_l; and determining the bounding box region of the left pixel plane corresponding to the left pixel coordinates according to the u_min_l, v_min_l, u_max_l, v_max_l.
[0076] Further, the method further comprises:
[0077] traversing each pixel point in the bounding box region, selecting a plurality of sampling points according to a preset sampling interval; determining a target sampling point, the target sampling point being any one of the plurality of sampling points; searching for the left pixel coordinates in a preset neighborhood of the target sampling point, and determining the right pixel coordinates corresponding to the searched left pixel coordinates; and storing the searched left pixel coordinates and the corresponding right pixel coordinates into a preset matching point pair set.
[0078] Further, after the matching point pair set corresponding to the left pixel coordinates and the corresponding right pixel coordinates is determined, the method further comprises:
[0079] determining the number of point pairs in the matching point pair set; judging whether the number of point pairs is less than a preset threshold; if yes, continuing to traverse the pixel points in the bounding box region, increasing the number of point pairs in the matching point pair set, until the number of point pairs is not less than the preset threshold; and if no, performing the step of calculating the local homography matrix and the average depth value according to the matching point pair set.
[0080] Further, the method further comprises:
[0081] determining, by using the homography matrix, the pixel coordinates Pr of any one pixel point Pl in the bounding box region corresponding to the right pixel plane; calculating the epipolar line corresponding to the point Pl in the right pixel plane, projecting the pixel coordinates Pr to the epipolar line to obtain the corrected matching point.
[0082] Further, the above-mentioned verifying the reconstructed point according to the average depth value, and adding the reconstructed point meeting the verification condition to the global point cloud, comprises:
[0083] determining a coordinate X of the reconstructed point; determining a directional component of the coordinate X in a depth direction; calculating an absolute value of a difference between the directional component and the average depth value; and adding the reconstructed point to the global point cloud if the absolute value is less than a preset fixed value.
[0084] Further, the above-mentioned method projects the sparse point cloud to a preset left pixel plane and a preset right pixel plane based on internal and external parameters of a left camera corresponding to the left pixel plane and internal and external parameters of a right camera corresponding to the right pixel plane.
[0085] Based on the above-mentioned embodiments, the application proposes a three-dimensional point cloud enhancement method based on homography estimation, which adopts the principle of homography matrix to find the right camera pixel coordinates corresponding to the left camera pixel coordinates, simultaneously adopts the principle of epipolar constraint to further accurately search the right camera pixel coordinates, and finally utilizes the principle of binocular vision to accurately reconstruct the left / right pixel coordinates, thereby realizing efficient point cloud enhancement.
[0086] In a second aspect, the application embodiments further propose a point cloud enhancement device based on binocular reconstruction, comprising:
[0087] a data acquisition module, configured to acquire a sparse point cloud separated from a global point cloud;
[0088] a point cloud projection module, configured to project the sparse point cloud to a preset left pixel plane and a preset right pixel plane, to obtain left pixel coordinates of the sparse point cloud in the left pixel plane and right pixel coordinates of the sparse point cloud in the right pixel plane;
[0089] a data matching module, configured to determine a bounding box region in the left pixel plane according to a distribution relationship of the left pixel coordinates in the left pixel plane, and determine a matching point pair set about the left pixel coordinates and corresponding right pixel coordinates by traversing pixel points in the bounding box region;
[0090] a three-dimensional reconstruction module, configured to calculate a local homography matrix and an average depth value according to the matching point pair set, generate a matching point of the right pixel coordinates by using the homography matrix, correct the matching point by epipolar constraint, and perform three-dimensional reconstruction on the corrected matching point pair by using the principle of binocular vision, to obtain a reconstructed point;
[0091] a data verification module, configured to verify the reconstructed point according to the average depth value, and add the reconstructed point meeting a verification condition to the global point cloud.
[0092] It should be noted that the point cloud enhancement device based on binocular reconstruction provided in the above embodiment can execute the point cloud enhancement method based on binocular reconstruction provided in the embodiment, and has the function modules and advantages corresponding to the execution method. Technical details not described in detail in the device embodiment can be referred to the point cloud enhancement method based on binocular reconstruction provided in the embodiment.
[0093] Further, please refer to Figure 3 , Figure 3 is a structural schematic diagram of an electronic device 200 provided in an embodiment of the present application. The electronic device 200 includes one or more processors 51 and a memory 52. The memory 52 is connected to the one or more processors 51, for example, connected to the processor 51 through a bus.
[0094] The processor 51 is configured to support the electronic device 500 to execute the corresponding functions in the methods in the method embodiments described above. The processor 51 can be a central processing unit (CPU), a network processor (NP), a hardware chip or any combination thereof. The hardware chip described above can be an application specific integrated circuit (ASIC), a programmable logic device (PLD) or a combination thereof. The PLD described above can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL) or any combination thereof.
[0095] The memory 52 is used to store program codes and the like. The memory can include a volatile memory (VM), such as a random access memory (RAM); the memory can also include a non-volatile memory (NVM), such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD); the memory 52 can also include a combination of the above types of memories.
[0096] The memory 52 can be configured to store non-volatile software programs, non-volatile computer-executable programs and modules, such as program instructions / modules corresponding to the point cloud enhancement method based on binocular reconstruction in the embodiments of the present application. The processor 51 performs various functional applications and data processing of the point cloud enhancement method based on binocular reconstruction and the point cloud enhancement apparatus based on binocular reconstruction by running the non-volatile software programs, instructions and modules stored in the memory 52, i.e., realizes the functions of each module or unit of the point cloud enhancement method based on binocular reconstruction and the point cloud enhancement apparatus based on binocular reconstruction provided in the above method embodiments.
[0097] The memory 52 can include a program storage area and a data storage area, wherein the program storage area can store an operating system and application programs required by at least one function. The data storage area can store data created according to the use of the point cloud enhancement apparatus based on binocular reconstruction, etc. In some embodiments, the memory 52 can optionally include a memory remotely arranged with respect to the processor 51, and these remote memories can be connected to the point cloud enhancement apparatus based on binocular reconstruction through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.
[0098] One or more modules are stored in the memory 52, and when executed by the one or more processors 51, perform the point cloud enhancement method based on binocular reconstruction in any of the above method embodiments, e.g., perform the method steps described in the above method embodiments, and realize the functions of the modules described in the above apparatus embodiments.
[0099] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the computer program includes program instructions, which, when executed by a computer, cause the computer to perform the method of the foregoing embodiments.
[0100] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by a computer program instructing related hardware, and the program of all or part of the processes can be stored in a computer readable storage medium, and the program, when executed, can include the processes of the above method embodiments. The computer readable storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM) or a random access memory (RAM), etc.
[0101] The above only describes the preferred embodiments of the present application, and of course cannot limit the scope of the rights of the present application, so equivalent changes made according to the claims of the present application are still within the scope of the present application.
Claims
1. A binocular reconstruction three-dimensional point cloud enhancement method based on homography estimation, characterized in that, The method comprises: acquiring a sparse point cloud separated from a global point cloud; projecting the sparse point cloud to a preset left pixel plane and a preset right pixel plane respectively, to obtain left pixel coordinates of the sparse point cloud in the left pixel plane and right pixel coordinates of the sparse point cloud in the right pixel plane; determining a bounding box region in the left pixel plane according to a distribution relationship of the left pixel coordinates in the left pixel plane, and determining a matching point pair set about the left pixel coordinates and corresponding right pixel coordinates by traversing pixel points in the bounding box region; calculating a local homography matrix and an average depth value according to the matching point pair set, generating a matching point of the right pixel coordinates by using the homography matrix, correcting the matching point by a epipolar constraint, and performing three-dimensional reconstruction on the corrected matching point pair by using a binocular vision principle to obtain a reconstructed point; verifying the reconstructed point according to the average depth value, and adding a reconstructed point meeting a verification condition to the global point cloud.
2. The method of binocular reconstruction of a 3D point cloud based on homography estimation according to claim 1, characterized in that, The method further comprises: respectively obtaining maximum values and minimum values of the left pixel coordinates in u and v directions of the left pixel plane: u_min_l, v_min_l, u_max_l, v_max_l; determining a bounding box region corresponding to the left pixel plane of the left pixel coordinates according to the u_min_l, v_min_l, u_max_l, v_max_l.
3. The method of binocular reconstruction of a 3D point cloud enhancement based on homography estimation according to claim 1, characterized in that, The method further comprises: traversing each pixel point in the bounding box region, and selecting a plurality of sampling points according to a preset sampling interval; determining a target sampling point, the target sampling point being any one of the plurality of sampling points; finding the left pixel coordinates in a preset neighborhood of the target sampling point, and determining right pixel coordinates corresponding to the found left pixel coordinates; storing the found left pixel coordinates and the corresponding right pixel coordinates to a preset matching point pair set.
4. The method of binocular reconstruction of a 3D point cloud enhancement based on homography estimation according to claim 1, characterized in that, After determining the matching point pair set about the left pixel coordinates and the corresponding right pixel coordinates, the method further comprises: determining a number of point pairs in the matching point pair set; judging whether the number of point pairs is less than a preset threshold value; if yes, continuing to traverse the pixel points in the bounding box region, increasing the number of point pairs in the matching point pair set, until the number of point pairs is not less than the preset threshold value; if no, performing a step of calculating a local homography matrix and an average depth value according to the matching point pair set.
5. The method of binocular reconstruction of a 3D point cloud enhancement based on homography estimation according to claim 1, characterized in that, The method further comprises: determining a pixel coordinate Pr corresponding to the right pixel plane of any one pixel point Pl in the bounding box region by using the homography matrix; calculating an epipolar line corresponding to the point Pl in the right pixel plane, projecting the pixel coordinate Pr to the epipolar line to obtain a corrected matching point.
6. The method of binocular reconstruction of a 3D point cloud enhancement based on homography estimation according to claim 1, characterized in that, The verifying the reconstructed point according to the average depth value comprises: determining a coordinate X of the reconstructed point; determining a directional component of the coordinate X in a depth direction; calculating an absolute value of a difference between the directional component and the average depth value; if the absolute value is less than a preset fixed value, adding the reconstructed point to the global point cloud.
7. The method of binocular reconstruction of a 3D point cloud enhancement based on homography estimation according to claim 1, characterized in that, The method projects the sparse point cloud to a preset left pixel plane and a preset right pixel plane based on internal and external parameters of a left camera corresponding to the left pixel plane and internal and external parameters of a right camera corresponding to the right pixel plane.
8. A device for enhancing a 3D point cloud reconstructed from binoculars based on homography estimation, characterized in that, The device comprises: a data acquisition module configured to acquire a sparse point cloud separated from a global point cloud; a point cloud projection module configured to project the sparse point cloud to a preset left pixel plane and a preset right pixel plane, to obtain left pixel coordinates of the sparse point cloud in the left pixel plane and right pixel coordinates of the sparse point cloud in the right pixel plane; a data matching module configured to determine a bounding box region in the left pixel plane according to a distribution relationship of the left pixel coordinates in the left pixel plane, and to determine a matching point pair set about the left pixel coordinates and corresponding right pixel coordinates by traversing pixel points in the bounding box region; a three-dimensional reconstruction module configured to calculate a local homography matrix and an average depth value according to the matching point pair set, to generate a matching point of the right pixel coordinates by using the homography matrix, to correct the matching point by epipolar constraint, and to perform three-dimensional reconstruction on the corrected matching point pair by using a binocular vision principle to obtain a reconstructed point; a data verification module configured to verify the reconstructed point according to the average depth value, and to add a reconstructed point meeting a verification condition to the global point cloud.
9. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory is connected to the processor, the processor is configured to execute one or more computer programs stored in the memory, and the processor is configured to enable the electronic device to implement the method according to any one of claims 1-7 when executing the one or more computer programs.
10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, the computer program comprises program instructions, and the program instructions enable the processor to execute the method according to any one of claims 1-7 when the program instructions are executed by the processor.