3D Model Generation with Depth-Guided Similar Point Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing three-dimensional model generation techniques face challenges in improving accuracy and reducing processing time, particularly in the search for similar points between images, which can lead to inaccuracies and prolonged processing times.
Innovation Solution
A method that determines a search range in three-dimensional space based on map information generated by camera calibration, limiting the search for similar points to a specific range using map information, and employing interpolation techniques to accurately estimate three-dimensional positions, thereby enhancing accuracy and reducing processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the search for similar points is performed over the entire epipolar line without restriction, then the search accuracy may be maintained, but the processing time increases significantly
Solution Approach 1:
The patent divides the epipolar line into multiple search regions based on the depth map information. Instead of searching the entire epipolar line, the search is segmented into specific regions where the similar point is most likely to be found, thereby reducing processing time while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary depth estimation using the depth map before conducting the similar point search. This preliminary action provides prior information about the likely location of the similar point, allowing the search to be focused on a restricted region rather than the entire epipolar line.
2Productivity
If the search range is restricted to improve processing speed, then the processing time decreases, but the search accuracy may deteriorate
Solution Approach 1:
The patent dynamically adjusts the search range parameters based on the depth map information. The search range is set to extend from the projected point to the estimated depth position, optimizing the balance between search speed and accuracy by adapting the search parameters to the specific scene geometry.
Solution Approach 2:
The depth map serves as an intermediary that bridges the gap between speed and accuracy. It provides intermediate depth information that guides the search process, enabling the system to achieve both fast processing and high accuracy by using the depth map as a mediator to inform the search strategy.
3Measurement precision
If map information is obtained through camera calibration to guide the search, then the search precision improves, but the device complexity increases
Solution Approach 1:
The depth map, initially obtained through camera calibration, serves multiple functions: it guides the search range determination, provides depth information for three-dimensional model generation, and assists in identifying similar points. This multi-functionality reduces the need for separate mechanisms, thereby limiting the increase in device complexity.
Solution Approach 2:
The system uses its own calibration data (the depth map) to guide the search process. The depth map, which is already part of the system's operational data, is repurposed to inform the similar point search, allowing the system to serve itself and reducing the need for additional external devices or complex mechanisms.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A three-dimensional model generation method includes: obtaining map information generated by camera calibration executed by controlling one or more cameras to shoot a subject from a plurality of viewpoints, the map information including three-dimensional points each indicating a position on the subject in a three-dimensional space (S101); obtaining a first image from a first view point and a second image from a second viewpoint (S101); determining a search range in the three-dimensional space, based on the map information, the search range including a first three-dimensional point on the subject, the first three-dimensional point corresponding to a first point in the first image (S104); searching for a similar point that is similar to the first point, in a range in the second image which corresponds to the search range (S105); and generating a three-dimensional model using a search result in the searching (S106).