A three-dimensional depth reconstruction method, device, equipment and medium integrating binocular vision and Gaussian sputtering

Abstract

The present invention discloses a three-dimensional depth reconstruction method, device, equipment and medium integrating binocular vision and Gaussian sputtering, including: obtaining first video data corresponding to a target object, and obtaining a first Gaussian ellipse model based on the first video data; obtaining second video data corresponding to the target object, and obtaining a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and the first position of shooting the first video data and the second position of shooting the second video data satisfy the binocular vision theorem; performing similarity matching on the first Gaussian ellipse model and the second Gaussian ellipse model, determining the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity; processing a number of depth values ​​based on an interpolation method, and obtaining dense depth point cloud data corresponding to the target object. The present invention belongs to the field of three-dimensional detection. The present invention can achieve high-precision depth detection.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional detection, and in particular to a three-dimensional depth reconstruction method, device, equipment and medium integrating binocular vision and Gaussian sputtering. Background Art

[0002] In the field of computer vision, the acquisition of depth information is critical to many applications, such as augmented reality, virtual reality, and autonomous driving. Traditional depth acquisition methods, such as those based on structured light and lidar, have limitations such as high equipment costs and strict environmental requirements. Depth reconstruction methods based on binocular vision have attracted much attention because of their relatively low cost and ability to provide richer depth information.

[0003] However, existing binocular depth reconstruction methods often face problems such as poor depth accuracy and sparse point clouds when dealing with complex scenes. Gaussian Splatting, as an emerging 3D reconstruction technology, can better represent 3D scenes, but the depth information obtained by monocular Gaussian Splatting still has certain limitations. Therefore, how to combine the two to achieve more accurate depth detection is an urgent problem to be solved. Summary of the invention

[0004] The present invention solves the technical problem of inaccurate depth detection in the prior art and achieves the technical effect of high-precision depth detection by providing a three-dimensional depth reconstruction method, device, equipment and medium that integrates binocular vision and Gaussian sputtering.

[0005] In a first aspect, the present invention provides a three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering, the method comprising:

[0006] Acquire first video data corresponding to the target object, and obtain a first Gaussian ellipse model based on the first video data;

[0007] Acquire second video data corresponding to the target object, and obtain a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and a first position at which the first video data is shot and a second position at which the second video data is shot satisfy a binocular vision theorem;

[0008] Performing similarity matching on the first Gaussian ellipse model and the second Gaussian ellipse model to determine the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity;

[0009] Based on the interpolation method, several depth values ​​are processed to obtain dense depth point cloud data corresponding to the target object.

[0010] Furthermore, obtaining first video data corresponding to the target object and obtaining a first Gaussian ellipse model based on the first video data includes:

[0011] Extracting frames from the first video data to obtain a plurality of image frames;

[0012] Based on the preset deep learning network model, feature extraction is performed on the image frame to obtain several feature points;

[0013] Gaussian distribution parameters are determined according to a number of feature points, and a first Gaussian ellipse model is generated.

[0014] Furthermore, the first position for shooting the first video data and the second position for shooting the second video data satisfy the binocular vision theorem, including:

[0015] A baseline between the single camera at the first position and the single camera at the second position is greater than a preset distance; and,

[0016] The optical axis of the single camera at the first position is parallel to the optical axis of the single camera at the second position.

[0017] Further, similarity matching is performed on the first Gaussian ellipse model and the second Gaussian ellipse model to determine the disparity corresponding to the feature points in the target object, including:

[0018] Based on the feature similarity measurement method, the feature descriptor corresponding to each feature point in the first Gaussian ellipse model and the second Gaussian ellipse model is expressed;

[0019] Calculate the similarity between any feature descriptor in the first Gaussian ellipse model and any feature descriptor in the second Gaussian ellipse model;

[0020] The disparity is calculated based on the feature points corresponding to the pair of feature descriptors with the highest similarity.

[0021] Further, according to the disparity, determining a depth value corresponding to the disparity includes:

[0022]

[0023] in, is the depth value, is the focal length of a single camera, is the length of the baseline between the single camera at the first position and the single camera at the second position, For parallax.

[0024] Furthermore, several depth values ​​are processed based on the interpolation method to obtain dense depth point cloud data corresponding to the target object, including:

[0025] Processing several depth values ​​based on linear interpolation or cubic spline interpolation;

[0026] Convert a number of depth values ​​and the two-dimensional coordinates corresponding to each depth value into three-dimensional coordinates;

[0027] Combine several three-dimensional coordinates to obtain dense depth point cloud data.

[0028] Furthermore, the linear interpolation method includes:

[0029]

[0030] in, for The depth value of the point, for The depth value of the point, for The depth value of the point, for Point The coordinate value in the direction, for Point The coordinate value in the direction, for Point Coordinate value in direction.

[0031] In a second aspect, the present invention provides a three-dimensional depth reconstruction device integrating binocular vision and Gaussian sputtering, the device comprising:

[0032] A first acquisition module, used to acquire first video data corresponding to the target object, and obtain a first Gaussian ellipse model based on the first video data;

[0033] a second acquisition module, configured to acquire second video data corresponding to the target object, and obtain a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and a first position at which the first video data is shot and a second position at which the second video data is shot satisfy a binocular vision theorem;

[0034] A depth value determination module, used to perform similarity matching between the first Gaussian ellipse model and the second Gaussian ellipse model, and determine the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity;

[0035] The point cloud module is used to process several depth values ​​based on the interpolation method to obtain dense depth point cloud data corresponding to the target object.

[0036] In a third aspect, the present invention provides an electronic device, comprising:

[0037] processor;

[0038] a memory for storing processor-executable instructions;

[0039] Wherein, the processor is configured to execute to implement a three-dimensional depth reconstruction method that integrates binocular vision and Gaussian sputtering as provided in the first aspect.

[0040] In a fourth aspect, the present invention provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement a three-dimensional depth reconstruction method that integrates binocular vision and Gaussian sputtering as provided in the first aspect.

[0041] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0042] The present invention generates the first and second Gaussian ellipse models by performing frame extraction and feature extraction on videos shot at different positions. The model not only contains the spatial position information of the feature points, but also includes rich local feature descriptions such as color, size (variance), etc., which enhances the accuracy and robustness of the matching. And due to the existence of multi-view information, the effect of binocular vision can be simulated even with a single camera, thereby improving the accuracy of depth estimation. The calculated depth value is processed by an interpolation algorithm (such as linear interpolation or cubic spline interpolation), the gaps between feature points are filled, and dense depth point cloud data is generated. Not only the density of the point cloud is improved, but also the three-dimensional reconstruction result is smoother and more continuous. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A schematic diagram of a flow chart of a three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering provided by the present invention;

[0045] Figure 2 A schematic diagram of the structure of a three-dimensional depth reconstruction device that integrates binocular vision and Gaussian sputtering provided by the present invention;

[0046] Figure 3 The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0047] The embodiment of the present invention solves the technical problem of inaccurate depth detection in the prior art by providing a three-dimensional depth reconstruction method that integrates binocular vision and Gaussian sputtering.

[0048] The technical solution of the present invention is to solve the above technical problems, and the overall idea is as follows:

[0049] A three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering, the method comprising: obtaining first video data corresponding to a target object, and obtaining a first Gaussian ellipse model based on the first video data; obtaining second video data corresponding to the target object, and obtaining a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and a first position for shooting the first video data and a second position for shooting the second video data satisfy the binocular vision theorem; performing similarity matching on the first Gaussian ellipse model and the second Gaussian ellipse model to determine the disparity corresponding to feature points in the target object and the depth value corresponding to each disparity; processing a number of depth values ​​based on an interpolation method to obtain dense depth point cloud data corresponding to the target object.

[0050] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0051] First of all, the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0052] The present invention provides Figure 1 A three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering is shown, comprising steps S11-S14:

[0053] Step S11, acquiring first video data corresponding to the target object, and obtaining a first Gaussian ellipse model based on the first video data.

[0054] Specifically, the method includes: extracting frames of the first video data to obtain a plurality of image frames; extracting features of the image frames based on a preset deep learning network model to obtain a plurality of feature points; determining Gaussian distribution parameters according to the plurality of feature points, and generating a first Gaussian ellipse model.

[0055] A single camera can be used to shoot the target object to obtain the first video data. After obtaining the first video data, the first video data can be frame extracted. Frame extraction refers to extracting a series of static image frames from the video. The frames can be evenly distributed on the entire video timeline, or can be selected according to specific conditions (such as scene changes or key frames). Video files can be read using video processing tools (such as FFmpeg, OpenCV, etc.). Image frames are extracted at certain intervals (for example, one frame per second or one frame every N frames). The extracted image frames are saved as picture files (such as JPEG, PNG format) for subsequent processing.

[0056] After extracting the image frame, the next step is to use a deep learning model for feature extraction. The preset deep learning network model can be a convolutional neural network (CNN), or it can be a model such as ResNet, VGG, MobileNet, etc., or a model specifically used for feature extraction such as SuperPoint, LF-Net, etc.

[0057] Based on the preset deep learning network model, feature points (such as edges, textures or corners, etc.) in each image frame can be extracted.

[0058] For each feature point, it is assumed that the color and brightness distribution around it can be approximately described by a Gaussian distribution. The corresponding variance and covariance matrices are determined based on the color and geometric characteristics in the neighborhood of the feature point.

[0059] Each feature point corresponds to a Gaussian ellipse, which is defined by the parameters determined above. Specifically, the center position of the Gaussian ellipse corresponds to the position of the feature point, and its size and direction are determined by the variance and covariance matrices. The Gaussian ellipses corresponding to all feature points are combined to obtain the first Gaussian ellipse model.

[0060] It should be noted that it is necessary to simultaneously integrate and optimize several feature points corresponding to different image frames to ensure that the obtained first Gaussian ellipse model accurately presents the shape and position information of the three-dimensional scene objects.

[0061] Step S12, obtaining second video data corresponding to the target object, and obtaining a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both captured by a single camera, and the first position for capturing the first video data and the second position for capturing the second video data satisfy the binocular vision theorem.

[0062] Specifically, the method includes: extracting frames from the second video data to obtain a number of image frames; extracting features from the image frames based on a preset deep learning network model to obtain a number of feature points; determining Gaussian distribution parameters based on the number of feature points, and generating a second Gaussian ellipse model.

[0063] In addition, the first position for shooting the first video data and the second position for shooting the second video data satisfy the binocular vision theorem, including: the baseline between the single camera at the first position and the single camera at the second position is greater than a preset distance; and the optical axis of the single camera at the first position is parallel to the optical axis of the single camera at the second position.

[0064] The baseline distance refers to the horizontal distance between the optical centers of two single cameras. An appropriate baseline distance is the basis for accurately calculating the parallax (i.e., the horizontal displacement of the same object in two images). If the baseline is too short, the parallax of the object under different viewing angles will be very small, resulting in inaccurate depth estimation; on the contrary, if the baseline is too long, although the parallax can be increased, it may exceed the effective field of view of the camera or cause occlusion problems. The preset distance can be determined according to the actual situation.

[0065] The optical axes of the two cameras should be kept parallel or nearly parallel as much as possible, which helps to simplify the subsequent image processing steps and improve the accuracy of the disparity map. When the optical axes are parallel, it can be assumed that the images captured by the two cameras only have horizontal displacement changes (i.e., disparity) and no vertical displacement. In this way, for each pixel, only the horizontal matching needs to be considered, which greatly reduces the difficulty of matching.

[0066] Step S13: performing similarity matching on the first Gaussian ellipse model and the second Gaussian ellipse model to determine the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity.

[0067] The first Gaussian ellipse model and the second Gaussian ellipse model are matched for similarity to determine the disparity corresponding to the feature points in the target object, including: describing the feature descriptors corresponding to the feature points in the first Gaussian ellipse model and the second Gaussian ellipse model based on a feature similarity measurement method; performing similarity calculation on any feature descriptor in the first Gaussian ellipse model and any feature descriptor in the second Gaussian ellipse model; and performing disparity calculation on the feature points corresponding to the pair of feature descriptors with the highest similarity.

[0068] It can be understood that each Gaussian ellipse in the first Gaussian ellipse model represents a feature point. The feature descriptor of the feature point can be expressed based on SIFT or SURF. The feature points corresponding to a pair of feature descriptors with the highest similarity indicate that they are the same feature point in the target object.

[0069] After matching the feature points in the first Gaussian ellipse model and the second Gaussian ellipse model, several disparities can be obtained. Disparity is the horizontal displacement distance of the same object (or the same feature point) in two images.

[0070] According to the disparity, a depth value corresponding to the disparity is determined, including:

[0071]

[0072] in, is the depth value, is the focal length of a single camera, is the length of the baseline between the single camera at the first position and the single camera at the second position, For parallax.

[0073] Step S14, processing a plurality of depth values ​​based on an interpolation method to obtain dense depth point cloud data corresponding to the target object.

[0074] Specifically, it includes: processing a number of depth values ​​based on linear interpolation or cubic spline interpolation; converting a number of depth values ​​and the two-dimensional coordinates corresponding to each depth value into three-dimensional coordinates; and combining a number of three-dimensional coordinates to obtain dense depth point cloud data.

[0075] Linear interpolation methods, including:

[0076]

[0077] in, for The depth value of the point, for The depth value of the point, for The depth value of the point, for Point The coordinate value in the direction, for Point The coordinate value in the direction, for Point Coordinate value in direction.

[0078] In addition, several depth values ​​can be processed by cubic spline interpolation, which can better fit the curve and generate smoother results. Cubic spline interpolation approximates the relationship between data points through piecewise polynomial functions, ensuring that the interpolation result is not only continuous, but also has continuous first-order and second-order derivatives. Specifically, for a given sequence of depth values, a piecewise cubic polynomial function is constructed so that in each interval, the function and its first-order and second-order derivatives are continuous. Using the constructed spline function, interpolation is performed between known depth values ​​to generate a dense distribution of depth values.

[0079] In summary, the present invention provides a three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering, the method comprising: obtaining first video data corresponding to the target object, and obtaining a first Gaussian ellipse model based on the first video data; obtaining second video data corresponding to the target object, and obtaining a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and the first position of shooting the first video data and the second position of shooting the second video data satisfy the binocular vision theorem; performing similarity matching on the first Gaussian ellipse model and the second Gaussian ellipse model, determining the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity; processing a number of depth values ​​based on the interpolation method, and obtaining dense depth point cloud data corresponding to the target object. The present invention generates the first and second Gaussian ellipse models by performing frame extraction and feature extraction on videos shot at different positions. The model not only contains the spatial position information of the feature points, but also includes rich local feature descriptions such as color, size (variance), etc., which enhances the accuracy and robustness of the matching. And due to the existence of multi-view information, the effect of binocular vision can be simulated even with a single camera, thereby improving the accuracy of depth estimation. Interpolation algorithms (such as linear interpolation or cubic spline interpolation) are used to process the calculated depth values, fill the gaps between feature points, and generate dense depth point cloud data. This not only increases the density of the point cloud, but also makes the 3D reconstruction results smoother and more continuous.

[0080] Based on the same inventive concept, the present invention also provides Figure 2 A three-dimensional depth reconstruction device integrating binocular vision and Gaussian sputtering is shown, the device comprising:

[0081] A first acquisition module 21, configured to acquire first video data corresponding to the target object, and obtain a first Gaussian ellipse model based on the first video data;

[0082] A second acquisition module 22 is used to acquire second video data corresponding to the target object, and obtain a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and a first position for shooting the first video data and a second position for shooting the second video data satisfy a binocular vision theorem;

[0083] A depth value determination module 23, used for performing similarity matching between the first Gaussian ellipse model and the second Gaussian ellipse model, and determining the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity;

[0084] The point cloud module 24 is used to process a plurality of depth values ​​based on an interpolation method to obtain dense depth point cloud data corresponding to the target object.

[0085] Based on the same inventive concept, the present invention also provides Figure 3An electronic device as shown includes:

[0086] Processor 31;

[0087] A memory 32 for storing instructions executable by the processor 31;

[0088] The processor 31 is configured to execute to implement a three-dimensional depth reconstruction method that integrates binocular vision and Gaussian sputtering as provided above.

[0089] Based on the same inventive concept, the present invention also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by the processor 31 of the electronic device, the electronic device can execute a three-dimensional depth reconstruction method that integrates binocular vision and Gaussian sputtering as provided above.

[0090] Since the electronic device introduced in this embodiment is an electronic device used to implement the information processing method in the embodiment of the present invention, based on the information processing method introduced in the embodiment of the present invention, a person skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present invention is not described in detail here. As long as the electronic device used by a person skilled in the art to implement the information processing method in the embodiment of the present invention, it belongs to the scope of protection of the present invention.

[0091] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0093] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0095] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0096] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering, characterized in that: The method comprises: Acquire first video data corresponding to the target object, and obtain a first Gaussian ellipse model based on the first video data; Acquire second video data corresponding to the target object, and obtain a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and a first position at which the first video data is shot and a second position at which the second video data is shot satisfy a binocular vision theorem; Performing similarity matching on the first Gaussian ellipse model and the second Gaussian ellipse model to determine the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity; Processing a plurality of depth values ​​based on an interpolation method to obtain dense depth point cloud data corresponding to the target object; Wherein, obtaining first video data corresponding to the target object and obtaining a first Gaussian ellipse model based on the first video data includes: Extracting frames from the first video data to obtain a plurality of image frames; Based on the preset deep learning network model, feature extraction is performed on the image frame to obtain several feature points; Gaussian distribution parameters are determined according to a number of feature points, and the first Gaussian ellipse model is generated.

2. A three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering as claimed in claim 1, characterized in that: The first position where the first video data is shot and the second position where the second video data is shot satisfy the binocular vision theorem, including: A baseline between the single camera at the first position and the single camera at the second position is greater than a preset distance; and, The optical axis of the single camera at the first position is parallel to the optical axis of the single camera at the second position.

3. The three-dimensional depth reconstruction method of integrating binocular vision and Gaussian sputtering as claimed in claim 1, characterized in that: Performing similarity matching on the first Gaussian ellipse model and the second Gaussian ellipse model to determine the disparity corresponding to the feature points in the target object, include: Describing the feature descriptors corresponding to the feature points in the first Gaussian ellipse model and the second Gaussian ellipse model based on a feature similarity measurement method; Performing similarity calculation on any feature descriptor in the first Gaussian ellipse model and any feature descriptor in the second Gaussian ellipse model respectively; The disparity is calculated based on the feature points corresponding to the pair of feature descriptors with the highest similarity.

4. A three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering as claimed in claim 3, characterized in that: According to the disparity, a depth value corresponding to the disparity is determined, including: in, is the depth value, is the focal length of a single camera, is the length of the baseline between the single camera at the first position and the single camera at the second position, For parallax.

5. The three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering as claimed in claim 1, characterized in that: Processing a number of depth values ​​based on the interpolation method to obtain dense depth point cloud data corresponding to the target object includes: Processing several depth values ​​based on linear interpolation or cubic spline interpolation; Convert a number of depth values ​​and the two-dimensional coordinates corresponding to each depth value into three-dimensional coordinates; Combine several three-dimensional coordinates to obtain dense depth point cloud data.

6. A three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering as claimed in claim 5, characterized in that: Linear interpolation methods, including: in, for The depth value of the point, for The depth value of the point, for The depth value of the point, for Point The coordinate value in the direction, for Point The coordinate value in the direction, for Point Coordinate value in direction.

7. A three-dimensional depth reconstruction device integrating binocular vision and Gaussian sputtering, characterized in that: A three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering applied to any one of claims 1 to 6, the device comprising: A first acquisition module, used to acquire first video data corresponding to the target object, and obtain a first Gaussian ellipse model based on the first video data; a second acquisition module, configured to acquire second video data corresponding to the target object, and obtain a second Gaussian ellipse model based on the second video data, wherein the first video data and the second video data are both obtained by shooting with a single camera, and a first position at which the first video data is shot and a second position at which the second video data is shot satisfy a binocular vision theorem; A depth value determination module, used for performing similarity matching between the first Gaussian ellipse model and the second Gaussian ellipse model to determine the disparity corresponding to the feature points in the target object and the depth value corresponding to each disparity; The point cloud module is used to process a plurality of depth values ​​based on an interpolation method to obtain dense depth point cloud data corresponding to the target object.

8. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement a three-dimensional depth reconstruction method integrating binocular vision and Gaussian sputtering as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement a three-dimensional depth reconstruction method that integrates binocular vision and Gaussian sputtering as described in any one of claims 1 to 6.

Images