Object Detection Using Global-Local Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer vision object detection methods, particularly local object detection algorithms, perform poorly in uncooperative environments with issues like resolution, frame rate, and illumination problems, and existing solutions to enhance image quality are not always effective.
Innovation Solution
A system that combines global and local object detection techniques, where global detection identifies objects and local detection is applied to specific regions of interest within those objects, providing feedback to correct errors in both global and local object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If local object detection algorithms are used to improve object recognition accuracy, then detection precision is improved, but reliability deteriorates in uncooperative environments with resolution, frame rate, illumination, or obscuration issues
Solution Approach 1:
The patent divides the detection process into two segments: global object detection that identifies major objects in entire images, and local object detection that analyzes specific regions of interest within those global objects. This segmentation allows the system to apply different detection strategies appropriate to each scale, improving overall reliability while maintaining precision.
Solution Approach 2:
The system performs global object detection as a preliminary step before applying local object detection. By first identifying global objects and their regions of interest, the system prepares the data structure and focuses subsequent detailed analysis only on relevant areas, improving efficiency and reliability in challenging environments.
2Measurement precision
If local object detection is applied to entire images to improve detection coverage, then detection completeness is improved, but processing time increases
Solution Approach 1:
The patent segments the image processing task by first performing global object detection on entire images to identify regions of interest, then applying computationally intensive local object detection only to those specific regions. This segmentation maintains detection completeness while dramatically reducing processing time compared to applying local detection to entire images.
Solution Approach 2:
The system applies local object detection partially, only to regions of interest within global objects rather than to entire images. This partial application of the detection algorithm maintains completeness for relevant objects while reducing overall processing time by avoiding redundant analysis of background and non-relevant areas.
Data Source
AI summary
The present invention therefore provides a system and method of object detection that employs both global object detection and local object detection. In particular, the present invention applies global object detection techniques to detect global objects, and then applies local object detection techniques on select portions of detected global objects to detect local objects.