Enhancement method of stereoscopic vision system and stereoscopic vision system

Abstract

The invention provides a stereoscopic vision system enhancement method and a stereoscopic vision system, and the method comprises the following steps: S1, building a three-dimensional analysis model, and carrying out the building and simulation analysis of the three-dimensional analysis model according to an obtained image; s2, common scenes are classified, and common visual scenes are classified according to existing data; s3, a real-time database is constructed, and the real-time database which can be called at any time is constructed for the three-dimensional analysis model; s4, acquiring scene information; s5, intelligent auxiliary identification; s6, hole pixel filling is carried out; and S7, outputting a complete image. According to the method, the three-dimensional analysis model is constructed to carry out image three-dimensional analysis, the simulation of the three-dimensional model is completed according to common scene matching, the matching of the three-dimensional model and the image content is carried out according to the image content and material matching result when the image is acquired, and the image is filled through the three-dimensional simulation model, so that the filled image content is more effective, and the image filling efficiency is improved. And multi-angle filling and restoration of the same visual scene are facilitated.

Description

Technical Field

[0001] The present invention relates to a stereoscopic vision system enhancement method, in particular to a stereoscopic vision system enhancement method and a stereoscopic vision system, belonging to the field of stereoscopic vision technology. Background Art

[0002] Stereoscopic vision is the feeling of being able to distinguish the shapes of objects near and far when observing scenes with both eyes. Stereoscopic vision is an important topic in the field of computer vision. Its purpose is to reconstruct the three-dimensional geometric information of the scene. The research on stereoscopic vision has important application value. Its applications include autonomous navigation systems of mobile robots, aviation and remote sensing measurements, industrial automation systems, etc. There are many ways to obtain images used for stereoscopic vision research. There are large ranges of changes in time, viewpoint, and direction, which are directly affected by the application field. The research on stereoscopic vision is mainly concentrated in three application fields, namely the interpretation of aerial pictures in automatic mapping, the guidance and obstacle avoidance of autonomous vehicles, and the functional simulation of human stereoscopic vision. Different application fields involve different types of scenes;

[0003] Chinese patent application (CN201811629404.1) discloses a stereoscopic vision system enhancement method and a stereoscopic vision system. The application accurately distinguishes the hole pixels as background or foreground to accurately fill the hole pixels. However, the existing stereoscopic vision system enhancement method lacks the recognition of image content material and intelligent analysis of image occlusion surfaces, resulting in insufficient accuracy in filling some content and prone to errors when composing images from multiple perspectives.

[0004] In order to solve the above technical problems, a stereoscopic vision system enhancement method and a stereoscopic vision system are proposed. Summary of the invention

[0005] In view of this, the present invention provides a stereoscopic vision system enhancement method and a stereoscopic vision system to solve or alleviate the technical problems existing in the prior art and at least provide a beneficial option.

[0006] The technical solution of the present invention is implemented as follows: a method for enhancing a stereoscopic vision system comprises the following steps:

[0007] S1, three-dimensional analysis model, constructing a three-dimensional analysis model, and performing three-dimensional model construction and simulation analysis based on the acquired images;

[0008] S2, classify common scenes, classify common visual scenes based on existing data;

[0009] S3. Build a real-time database for the 3D analysis model that can be called at any time, and feed the analysis model with data in the database for deep learning;

[0010] S4, acquiring scene information, taking multiple shots of the same visual scene to acquire image data;

[0011] S5, intelligent auxiliary recognition, the three-dimensional analysis model performs three-dimensional image analysis based on the acquired image data;

[0012] S6, filling the hole pixels, filling the hole pixels of the image according to the analysis results of the three-dimensional analysis model;

[0013] S7, outputting the complete image, and outputting the filled complete image to a visual display terminal.

[0014] Further preferably, in S1, after constructing the three-dimensional analysis model, existing data is fed to perform deep learning of the three-dimensional analysis model, and visual information in common visual scenes is classified and fed to the three-dimensional analysis model, so that the three-dimensional analysis model can construct a three-dimensional simulation model based on the key information after obtaining the key information.

[0015] Further preferably, in S2, when classifying common scenes, common visual information in the scene is classified separately, and common collocation of visual information is matched with the visual scene.

[0016] Further preferably, in said S4, when capturing and acquiring image data, material recognition and matching is performed on the content of the image data, so that the three-dimensional analysis model can quickly construct a simulation model for image filling.

[0017] Further preferably, in S5, auxiliary analysis of the image scene is performed based on existing common combinations, and common visual information is matched with the image data.

[0018] Further preferably, in S7, the filled complete image can be fed into the three-dimensional analysis model as a valid case for deep learning, and the filled image can be recognized and classified and stored in the real-time database as data call.

[0019] A stereoscopic vision system comprises an image acquisition unit, a three-dimensional analysis unit and an image output unit;

[0020] The image acquisition unit includes a camera module and a material recognition module, the three-dimensional analysis unit includes a deep learning module, a three-dimensional analysis module, a three-dimensional simulation module, a data matching module, a pixel extraction module and a pixel filling module, and the image output unit includes a visualization module and a data storage module.

[0021] Further preferably, the camera module includes multiple camera devices in the same stereoscopic vision system, and the material recognition module performs material recognition analysis on the image content after the camera module acquires the image, and matches the image content material with the data in the database.

[0022] Further preferably, the three-dimensional analysis module performs three-dimensional analysis on the acquired image, the three-dimensional analysis module cooperates with the three-dimensional simulation module to simulate the three-dimensional model of the image, and performs material matching on the three-dimensional model according to the image content material, extracts the image content material pixels through the pixel extraction module, and cooperates with the data matching module to complete data matching, and fills the image hole pixels through the pixel filling module according to the matching results.

[0023] Further preferably, the visualization module is used to display the padded image and display the original image together as a comparison reference, and the data storage module is used to store various types of data.

[0024] The embodiment of the present invention has the following advantages due to the adoption of the above technical solution:

[0025] 1. The present invention performs three-dimensional analysis of images by constructing a three-dimensional analysis model, and completes the simulation of the three-dimensional model according to common scene combinations, matches the three-dimensional model with the image content according to the image content material matching result when the image is acquired, and fills the image through the three-dimensional simulation model, thereby making the content of the filled image more effective and facilitating multi-angle filling and restoration of the same visual scene;

[0026] Second, the present invention constructs a real-time database so that the data in the database can be used as a reference in real time during image analysis, and common scene combinations are classified to improve the efficiency of image classification and recognition.

[0027] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 A flowchart of a method for enhancing a stereoscopic vision system in the present invention;

[0030] Figure 2 This is a functional module diagram of a stereoscopic vision system in the present invention.

[0031] Attached figure numbers: 10, image acquisition unit; 11, camera module; 12, material recognition module; 20, three-dimensional analysis unit; 21, deep learning module; 22, three-dimensional analysis module; 23, three-dimensional simulation module; 24, data matching module; 25, pixel extraction module; 26, pixel filling module; 30, image output unit; 31, visualization module; 32, data storage module. DETAILED DESCRIPTION

[0032] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0033] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0034] like Figure 1 As shown, an embodiment of the present invention provides a method for enhancing a stereoscopic vision system, comprising the following steps:

[0035] S1, three-dimensional analysis model, constructing a three-dimensional analysis model, and performing three-dimensional model construction and simulation analysis based on the acquired images;

[0036] S2, classify common scenes, classify common visual scenes based on existing data;

[0037] S3. Build a real-time database for the 3D analysis model that can be called at any time, and feed the analysis model with data in the database for deep learning;

[0038] S4, acquiring scene information, taking multiple shots of the same visual scene to acquire image data;

[0039] S5, intelligent auxiliary recognition, the three-dimensional analysis model performs three-dimensional image analysis based on the acquired image data;

[0040] S6, filling the hole pixels, filling the hole pixels of the image according to the analysis results of the three-dimensional analysis model;

[0041] S7, outputting the complete image, and outputting the filled complete image to a visual display terminal.

[0042] In one embodiment, in S1, after building the three-dimensional analysis model, existing data is fed into the three-dimensional analysis model for deep learning, and visual information in common visual scenes is classified and fed into the three-dimensional analysis model, so that the three-dimensional analysis model can build a three-dimensional simulation model based on the key information after obtaining the key information.

[0043] In one embodiment, in S2, when classifying common scenes, common visual information in the scene is separately classified, and common collocations of visual information are matched with the visual scene.

[0044] In one embodiment, in S4, when capturing and acquiring image data, material recognition and matching is performed on the content of the image data so that the three-dimensional analysis model can quickly construct a simulation model for image filling.

[0045] In one embodiment, in S5, auxiliary analysis of the image scene is performed based on existing common combinations, and common visual information is matched with the image data.

[0046] In one embodiment, in S7, the filled complete image can be fed into the three-dimensional analysis model as a valid case for deep learning, and the filled image can be recognized and classified and stored in the real-time database as data call.

[0047] like Figure 2 As shown, the embodiment of the present invention further provides a stereoscopic vision system, including an image acquisition unit 10, a three-dimensional analysis unit 20 and an image output unit 30;

[0048] The image acquisition unit 10 includes a camera module 11 and a material recognition module 12. The camera module 11 includes a plurality of camera devices in the same stereoscopic vision system. The material recognition module 12 performs material recognition analysis on the image content after the camera module 11 acquires the image, and matches the image content material with the data in the database.

[0049] The three-dimensional analysis unit 20 includes a deep learning module 21, a three-dimensional analysis module 22 for performing three-dimensional analysis on the acquired image, a three-dimensional simulation module 23, a data matching module 24, a pixel extraction module 25 and a pixel filling module 26;

[0050] The image output unit 30 includes a visualization module 31 for displaying the padded image and a data storage module 32 for storing various types of data.

[0051] In one embodiment, the three-dimensional analysis module 22 cooperates with the three-dimensional simulation module 23 to simulate the three-dimensional model of the image, and matches the material of the three-dimensional model according to the image content material, extracts the image content material pixels through the pixel extraction module 25, and cooperates with the data matching module 24 to complete the data matching, and fills the image hole pixels through the pixel filling module 26 according to the matching results.

[0052] In one embodiment, when displaying an image, the visualization module 31 displays the original image as a reference.

[0053] When the present invention is working: the visual scene is photographed by the camera module 11, and after the photographing, the content and material of the acquired image are identified by the material recognition module 12, and the corresponding content of the image is matched with the data in the data storage module 32, the three-dimensional analysis unit 20 matches the image content with the common scene, and builds a three-dimensional simulation model according to the material recognition result, so that the material of the three-dimensional simulation model matches the image content, and the pixels of the image are filled according to the three-dimensional simulation model. During the pixel filling process, the corresponding pixel data is extracted from the image, so as to improve the restoration degree, and after the restoration, the original image and the filled image are displayed together through the visualization module 31.

[0054] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for enhancing a stereoscopic vision system, characterized in that: The following steps are involved: S1, three-dimensional analysis model, constructing a three-dimensional analysis model, and performing three-dimensional model construction and simulation analysis based on the acquired images; S2, classify common scenes, classify common visual scenes based on existing data; S3. Build a real-time database for the 3D analysis model that can be called at any time, and feed the analysis model with data in the database for deep learning; S4, acquiring scene information, taking multiple shots of the same visual scene to acquire image data; S5, intelligent auxiliary recognition, the three-dimensional analysis model performs three-dimensional image analysis based on the acquired image data; S6, filling the hole pixels, filling the hole pixels of the image according to the analysis results of the three-dimensional analysis model; S7, outputting the complete image, and outputting the filled complete image to a visual display terminal.

2. The method for enhancing a stereoscopic vision system according to claim 1, characterized in that: In S1, after the three-dimensional analysis model is constructed, existing data is fed to perform deep learning of the three-dimensional analysis model, and visual information in common visual scenes is classified and fed to the three-dimensional analysis model, so that the three-dimensional analysis model can construct a three-dimensional simulation model based on the key information after obtaining the key information.

3. The method for enhancing a stereoscopic vision system according to claim 1, characterized in that: In S2, when classifying common scenes, common visual information in the scene is separately classified, and common collocation of visual information is matched with the visual scene.

4. The method for enhancing a stereoscopic vision system according to claim 1, characterized in that: In the above S4, when the image data is captured, material recognition and matching is performed on the content of the image data, so that the three-dimensional analysis model can quickly build a simulation model for image filling.

5. The method for enhancing a stereoscopic vision system according to claim 1, characterized in that: In S5, auxiliary analysis of the image scene is performed based on existing common combinations, and common visual information is matched with the image data.

6. The method for enhancing a stereoscopic vision system according to claim 1, characterized in that: In the S7, the filled complete image can be fed into the three-dimensional analysis model as a valid case for deep learning, and the filled image can be recognized and classified and stored in the real-time database as data call.

7. A stereoscopic vision system, characterized in that: It comprises an image acquisition unit (10), a three-dimensional analysis unit (20) and an image output unit (30); The image acquisition unit (10) comprises a camera module (11) and a material recognition module (12); the three-dimensional analysis unit (20) comprises a deep learning module (21), a three-dimensional analysis module (22), a three-dimensional simulation module (23), a data matching module (24), a pixel extraction module (25) and a pixel filling module (26); and the image output unit (30) comprises a visualization module (31) and a data storage module (32).

8. A stereoscopic vision system according to claim 7, characterized in that: The camera module (11) includes a plurality of camera devices in the same stereoscopic vision system, and the material recognition module (12) performs material recognition analysis on image content after the camera module (11) acquires an image, and matches the image content material with data in a database.

9. A stereoscopic vision system according to claim 7, characterized in that: The three-dimensional analysis module (22) performs three-dimensional analysis on the acquired image, and cooperates with the three-dimensional simulation module (23) to simulate a three-dimensional model of the image, and performs material matching on the three-dimensional model according to the image content material, extracts the image content material pixels through the pixel extraction module (25), and cooperates with the data matching module (24) to complete data matching, and completes the filling of image hole pixels through the pixel filling module (26) according to the matching result.

10. A stereoscopic vision system according to claim 7, characterized in that: The visualization module (31) is used to display the filled image and to display the original image as a reference, and the data storage module (32) is used to store various types of data.

Images