Optical Stereo Matching for Real-Time Depth Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stereo matching methods face difficulties in performing real-time calculations due to high computational loads, making them unsuitable for applications in actual robots and industrial settings.
Innovation Solution
A system and method that perform census transform on images taken by cameras using hardware, including a coordinate creating module, census transform module, delay XOR calculation module, and stereo matching module, to extract depth information in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo matching is performed using traditional methods (SAD or census) on general-purpose microprocessors, then depth information can be extracted, but the computational load is enormous and real-time processing is not achieved
Solution Approach 1:
The patent replaces the mechanical computation system (general-purpose microprocessor) with an optical computing system using light planes. The census transform and XOR operations are performed optically through light plane projections and intersections, eliminating the need for sequential digital computation and enabling parallel processing at the speed of light.
Solution Approach 2:
The patent introduces a temporal dimension by using light planes that propagate through space-time. The census transform is performed by projecting light planes from multiple temporal moments, and the XOR operation is achieved through the spatial intersection of these light planes, effectively using the fourth dimension (time) to solve computational problems.
2Measurement precision
If census transform and XOR operations are performed to achieve accurate stereo matching, then depth information can be obtained, but the quantity of operations increases enormously
Solution Approach 1:
The patent replaces complex digital computation (census transform and XOR operations) with simple optical phenomena. The census transform is achieved through light plane projections that naturally encode pixel intensity relationships, and the XOR operation is performed through optical interference and intersection of light planes, eliminating the need for complex logical operations.
Solution Approach 2:
The optical system performs the computation automatically through the physical properties of light. The light planes self-organize and intersect in space-time to produce the census transform and XOR results without requiring external control or complex processing algorithms, making the system self-sufficient and eliminating computational complexity.
3Adaptability or versatility
If general-purpose microprocessors are used to carry out stereo matching operations, then flexibility and adaptability are maintained, but the system bears enormous operational load and cannot operate in real-time
Solution Approach 1:
The patent replaces the flexible but slow general-purpose microprocessor with a dedicated optical computing system. The optical system is designed specifically for stereo matching operations, providing both the adaptability needed for different scenarios and the high-speed processing capability required for real-time operation.
Solution Approach 2:
The patent divides the stereo matching process into distinct optical stages: light plane generation, census transform through temporal projections, XOR operation through spatial intersections, and depth extraction. Each stage is handled by specialized optical components, allowing the system to maintain flexibility while achieving high-speed parallel processing.
Data Source
AI summary
Disclosed are a system and a method extensible for performing, in real-time, stereo snatching for calculating depth images with a result of searching for points of similarity by using images taken with two cameras. The system includes a coordinate creating module, a census transform module, a delay XOR calculation module, a stereo matching module, and a control module. Accordingly, by using the system extensible for performing stereo matching, depth information of corrected images can be acquired in real-time without using computer systems or software programs for special purposes. Furthermore, since the system extensible for performing stereo matching can be simply realized by hardware, the system and the method of the present invention can be easily applied to actual intellectual-type robots, industrial settings, etc.