Image Region Bridging for Occlusion-Robust Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image detection technologies fail to accurately integrate object regions divided by wide-range occlusion, leading to detection as multiple separated pieces.
Innovation Solution
An image processing device and method that analyze luminance value transitions between object regions, integrate them if determined to be part of the same object, and correct the detection result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If object regions are corrected by color correction only in areas adjacent to detection target objects, then processing complexity is reduced, but detection accuracy deteriorates when occlusion occurs in wide range
Solution Approach 1:
The patent divides the predetermined region into multiple small regions, and further identifies occlusion small regions within them. This segmentation allows targeted processing of only the necessary areas (occlusion regions) rather than the entire image or large portions of it, thus maintaining low processing complexity while improving detection accuracy in occluded areas.
Solution Approach 2:
The patent applies color correction selectively only to occlusion small regions that are adjacent to detection target objects, rather than uniformly correcting the entire predetermined region. This localized approach reduces processing complexity while maintaining detection accuracy by focusing computational resources only where occlusion affects target detection.
2Use of energy by moving object
If only adjacent small regions are corrected, then processing load is reduced, but objects divided by occlusion are detected as multiple separated pieces
Solution Approach 1:
The patent uses luminance value transition analysis as feedback to determine whether two object regions should be integrated. By analyzing the luminance values along the trajectory connecting end points of object regions, the system can identify occlusion patterns and make informed decisions about integration, thus improving detection reliability without excessive processing load.
Solution Approach 2:
The patent performs preliminary analysis of luminance value transitions between object regions before final integration decisions are made. This preliminary action allows the system to identify potential occlusion cases and prepare integration strategies in advance, ensuring reliable object detection while managing processing load efficiently.
3Measurement precision
If luminance value analysis is performed between all object regions, then integration accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the analysis to focus only on pairs of object regions that are spatially close and potentially connected by occlusion. By calculating trajectories only between relevant region pairs rather than all possible combinations, the system maintains high integration accuracy while significantly reducing processing time through selective analysis.
Solution Approach 2:
The patent performs luminance value analysis only for object region pairs that meet specific spatial and contextual criteria, rather than analyzing all possible pairs. This partial action approach applies the computationally intensive luminance analysis only where necessary, maintaining integration accuracy for critical cases while minimizing overall processing time.
Data Source
Figure 1~2
Figure 3
Figure 4A~4D
AI summary
An image processing device includes an input unit to receive an input image, a feature amount extraction unit to extract a feature amount from the input image, an object region detection unit to detect an object region in the image through use of the feature amount, an object end-point estimation unit to estimate coordinates of an end point of an object in the object region, an object bridging unit to calculate a trajectory complementing a space between end points in two object regions on the input image, a same-object determination unit to determine, from luminance value transition on the trajectory, whether the two object regions are included in the same object, an object region correction unit to correct the object region output by the object region detection unit based on a determination result output by the same-object determination unit, and an output unit to output the corrected object region.