Binary Edge Descriptor for Real-Time Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current feature matching algorithms are inadequate for environments with insufficient corner features or rich textures, such as non-cluttered living rooms, where long straight edge features are prevalent, lacking robust and real-time description or matching methods.
Innovation Solution
The development of a binary edge descriptor system that captures edges in images by defining an offset patch of pixels based on grayscale gradient and contrast, enabling accurate edge description and matching across images in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If corner features and rich textures are used for feature matching, then matching accuracy is improved, but availability of features deteriorates in non-cluttered environments
Solution Approach 1:
The patent changes the type of feature being detected from corner features to edge features, and changes the descriptor type from traditional float descriptors to binary descriptors. This parameter change allows the system to work effectively in non-cluttered environments where corner features are scarce, while maintaining matching accuracy through the use of gradient and contrast information in the binary descriptor
Solution Approach 2:
The binary edge descriptor is segmented into multiple substrings, where each substring captures different aspects of the edge (gradient direction, gradient magnitude, contrast). This segmentation allows the descriptor to comprehensively characterize edge features using multiple independent binary comparisons, maintaining accuracy while using efficient binary operations
2Measurement precision
If traditional float descriptors are used for edge description, then description accuracy is improved, but computational speed deteriorates
Solution Approach 1:
The patent replaces the traditional float-based descriptor computation with a binary descriptor system. Instead of using continuous floating-point values that require complex arithmetic operations, the system uses binary comparisons (greater than, less than) that can be efficiently implemented using bitwise operations, dramatically improving computational speed while maintaining descriptive accuracy
Solution Approach 2:
The patent changes the data type parameter from floating-point to binary, and changes the comparison metric from Euclidean distance to Hamming distance. This parameter change enables the use of efficient bitwise operations and population count instructions for descriptor comparison, achieving real-time performance while maintaining matching accuracy
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Systems and methods are disclosed for describing and tracking edges within the field of view of one or more imaging devices. In one example, the present system defines a row of pixels taken across a width of the edge, and then determines a binary edge descriptor for the edge by comparing at least one of grayscale values and contrast of pixels within respective pixel pairs from the row of pixels, the result of the comparisons setting bits within the binary descriptor.