Object Recognition Device Using Frequency-Based Voting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition systems face increased calculation complexity and decreased performance as the number of image patterns and object classes grows, making them inefficient for real-time applications.
Innovation Solution
An object recognition device that uses a function table to relate class and position information with image pattern probabilities, calculates pattern appearance frequencies, selects matching image patterns, extracts classes with high frequency, and votes position information for recognition, reducing calculation and improving performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of image patterns and object classes is increased to improve recognition coverage, then the recognition capability is improved, but the calculation amount increases and recognition performance deteriorates
Solution Approach 1:
The patent divides the object recognition process into multiple stages: initial pattern matching, frequency calculation, and selective voting. By segmenting the recognition tasks and processing them in stages, the system can handle multiple object classes and patterns without overwhelming computational resources at any single stage, thus maintaining recognition performance while improving recognition capability
Solution Approach 2:
The patent changes the parameter of pattern frequency by calculating and utilizing the frequency of appearance for each image pattern. This parameter change allows the system to prioritize more common patterns during voting, reducing the computational burden on rare patterns while maintaining comprehensive recognition capability across multiple object classes
2Adaptability or versatility
If multiple recognition devices for 2-class recognition are combined to recognize all objects, then the recognition coverage is improved, but the processing time and device complexity increase
Solution Approach 1:
The patent merges multiple pattern matching operations into a unified voting framework. Instead of using separate recognition devices for different object classes, the system combines all image patterns from multiple classes into a single pattern library and uses a unified voting mechanism to recognize all objects, thereby reducing device complexity while maintaining comprehensive recognition coverage
Solution Approach 2:
The voting unit serves as a universal component that can handle recognition for all object classes simultaneously. Rather than having dedicated recognition devices for each class, the voting mechanism is designed to be multi-functional, accepting votes from patterns of any class and producing unified recognition results, thus reducing overall system complexity
3Measurement precision
If the number of image patterns is increased to improve recognition accuracy, then the recognition accuracy is improved, but the calculation amount for recognizing the object increases
Solution Approach 1:
The patent applies partial action by performing complete pattern matching for all image patterns during the frequency calculation stage, but then using only the frequency information for selective voting. This allows the system to benefit from comprehensive pattern matching for accuracy while reducing the calculation amount in the voting stage by prioritizing patterns based on their frequency of appearance
Data Source
AI summary
A learning unit generates a function table indicating the relationship between the class number and position information of an object and the probability of appearance of the object for each small area image pattern of a code book, calculates a sharing matrix indicating the commonality of a feature amount between the classes, makes a tree diagram in which the classes with a similar feature amount are clustered, and calculates the weight of each node in the tree diagram for each small area image pattern. The recognition processing unit compares image data captured by a camera with the code book, selects the closest small area image pattern, extracts the class related to the node with the smallest weight among the nodes with a weight equal to or greater than a threshold value, and votes the position information of the small area image pattern for the class, thereby recognizing the object.