Machine Vision Object Recognition With Modified Census Transform
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition methods in machine vision and image analysis require high equipment complexity and computational demand, making them costly and slow, and are limited by the need for constant backgrounds or predefined patterns.
Innovation Solution
A method using a modified census transform combined with a threshold setting for object recognition, allowing for a simplified binarized comparison that reduces computational effort and enables real-time recognition, using a correlation method with Hamming distance to determine object presence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object recognition methods are used, then recognition accuracy can be achieved, but equipment complexity and computational demand become high
Solution Approach 1:
The patent divides the image processing into distinct stages: preprocessing (brightness/contrast adjustment), feature extraction (edge detection, corner detection, Hough transform), and recognition. This segmentation allows each stage to be optimized independently, reducing overall system complexity while maintaining accuracy.
Solution Approach 2:
The patent extracts key features from images (edges, corners, lines) and uses them for recognition rather than processing the entire image data. This extraction approach reduces computational demand by focusing only on informative features while maintaining recognition accuracy.
2Measurement precision
If traditional object recognition methods are used, then recognition accuracy can be achieved, but processing speed becomes slow
Solution Approach 1:
The patent performs preliminary actions by preprocessing images (adjusting brightness and contrast) and extracting features before the actual recognition. This preparation reduces the computational load during recognition, enabling faster processing while maintaining accuracy.
Solution Approach 2:
The patent creates a simplified representation (copy) of the object using extracted features and Hough transform lines, rather than processing the complete image data. This copying approach enables rapid comparison and recognition while maintaining accuracy through feature fidelity.
3Adaptability or versatility
If traditional object recognition methods are used, then recognition can be performed, but constant background or predefined patterns are required
Solution Approach 1:
The patent creates a universal recognition system that can identify objects without requiring constant backgrounds or predefined patterns. The Hough transform and feature extraction methods work across various image conditions, making the system adaptable to different environments and object configurations.
Solution Approach 2:
The patent employs dynamic feature extraction and adaptive thresholding that can adjust to different image conditions. This dynamic approach allows the system to handle varying lighting, backgrounds, and object positions without requiring predefined patterns or constant background assumptions.
Data Source
AI summary
A method is provided for machine vision and image analysis for recognizing an object in an electronic image, which is captured with the aid of an optical sensor. A reference image of the object to be recognized is trained during a learning phase and compared with the image of the scene during a working phase, the pattern comparison between the object and the scene takes place with the aid of a modified census transform, using a determination of maximum and which must exceed a threshold value for a positive statement on a degree of correspondence.