Semi-Global Matching Disparity Derivation for Weak Texture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing range-finding methods using stereo cameras struggle to accurately detect disparity for objects with weak texture due to interference from objects with strong texture, leading to imprecise disparity detection.
Innovation Solution
The implementation of a semi-global matching (SGM) method that calculates synthesis costs by aggregating costs from surrounding pixels, reducing the impact of strong-texture objects on weak-texture objects, and using directional path costs to derive disparity values with enhanced precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cost aggregation method is used to derive disparity for objects with weak texture, then disparity accuracy for weak-texture objects is improved, but disparity detection precision deteriorates when strong-texture objects are present at far positions
Solution Approach 1:
The patent applies local quality by differentiating the treatment of pixels based on their textual properties. Pixels are classified as belonging to strong-texture objects or weak-texture objects, and different cost aggregation strategies are applied accordingly. For weak-texture objects, cost aggregation from surrounding pixels is performed to derive accurate disparity. For strong-texture objects, the patent uses a different approach that does not rely on aggregation, thereby preventing their disparity values from interfering with weak-texture object disparity detection.
Solution Approach 2:
The patent segments the image into regions corresponding to strong-texture objects and weak-texture objects. By identifying which pixels belong to which object type, the system can apply appropriate disparity derivation methods to each segment independently. This segmentation allows the cost aggregation method to be applied only where needed (for weak-texture objects) while avoiding the interference problem in strong-texture regions.
2Measurement precision
If semi-global matching method is used to calculate synthesis costs, then disparity detection precision is improved, but computational complexity increases
Solution Approach 1:
The patent applies local quality by implementing directional path costs that weigh nearby pixels more heavily than distant pixels. The cost function includes a weight component that decreases with distance, effectively limiting the influence radius of each pixel on its neighbors. This localized approach maintains the benefits of semi-global matching while reducing the computational burden compared to fully global methods.
Solution Approach 2:
The patent changes the parameters of the cost aggregation process by introducing directional path costs with distance-dependent weights. Instead of uniformly aggregating costs from all surrounding pixels, the system uses parameterized weight functions that adjust the contribution of each pixel based on its position and textual properties. This parameterization allows for optimized computational efficiency while maintaining high disparity detection precision.
Data Source
Figure 1
Figure 2A~2C
Figure 3A~4
AI summary
A disparity deriving apparatus (3) for deriving disparity of an object based on a reference image of the object captured at a first image capturing position and a comparison image of the object captured at a second image capturing position includes a calculator (310) to calculate costs between a first reference area in the reference image and each one of corresponding areas corresponding to the first reference area in a given region in the comparison image, and costs between a second reference area, surrounding the first reference area in the reference image, and each one of corresponding areas corresponding to the second reference area in a given region in the comparison image; a synthesizer (320) to synthesize the costs of the first reference area multiplied by a first weight, the costs calculated by the calculator (310), and the costs of the second reference area multiplied by a second weight smaller than the first weight, the cost calculated by the calculator (310), as synthesis costs; and a deriving unit (330) to derive a disparity value of an object captured on the first reference area based on the synthesis costs synthesized by the synthesizer (320).