Personal Item Reidentification Using Skeleton-Based Occlusion Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Incorrect personal item reidentification frequently occurs when a personal item detected from an image is unconditionally selected as an item to reidentify due to differences in robustness to occlusion between object detection and reidentification.
Innovation Solution
An image processing apparatus that includes an image acquiring unit, personal item detecting unit, skeleton information detecting unit, reidentification personal item selecting unit, and personal item reidentifying unit to determine whether personal items in different images are identical by using skeleton information to select non-obstructed personal items for reidentification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a personal item detected from an image is unconditionally selected as an item to reidentify, then the reidentification process can be performed without additional complexity, but the accuracy of personal item reidentification deteriorates due to occlusion by persons
Solution Approach 1:
The system performs preliminary detection of skeleton information representing persons before the reidentification process. This preliminary action allows the system to identify and exclude personal items that are occluded by persons, thereby improving reidentification accuracy without adding significant complexity to the overall process.
Solution Approach 2:
Skeleton information acts as an intermediary element that mediates between the detected personal items and the reidentification process. By introducing this intermediate representation of persons, the system can accurately determine which personal items should be excluded from reidentification due to occlusion, resolving the contradiction between simplicity and accuracy.
2Measurement precision
If skeleton information detection is added to filter obstructed personal items, then personal item reidentification accuracy is improved, but the device complexity increases due to additional detection components
Solution Approach 1:
The skeleton information detection component serves multiple functions: it identifies persons in the image, determines occlusion relationships, and provides filtering criteria for personal item selection. This multi-functionality reduces the need for separate specialized components, thereby limiting the increase in device complexity while achieving improved accuracy.
Solution Approach 2:
The system changes the parameter of personal item selection from unconditional inclusion to conditional exclusion based on occlusion detection. By modifying the selection criterion rather than adding complex filtering mechanisms, the system achieves improved accuracy with minimal increase in device complexity.
3Productivity
If all detected personal items are processed for reidentification, then the completeness of reidentification is maintained, but the frequency of incorrect reidentification results increases due to occluded items
Solution Approach 1:
The system extracts and excludes personal items that are occluded by persons from the reidentification process. By taking out these problematic items rather than processing them, the system maintains completeness of valid reidentification while eliminating sources of incorrect results, thus improving reliability without sacrificing productivity.
Solution Approach 2:
The system uses skeleton information as feedback to dynamically adjust which personal items should undergo reidentification. This feedback mechanism ensures that only non-occluded items are processed, maintaining high completeness while preventing incorrect results, thereby resolving the contradiction between productivity and reliability.
Data Source
AI summary
An image processing apparatus includes: an image accruing means acquiring a plurality of images captured at different times; a personal item detecting means detecting a personal item from each of the images; a skeleton information detecting means detecting skeleton information of a person from each of the images; a reidentification personal item selecting means selecting a personal item to reidentify by the detected personal item and skeleton information of the person; and a personal item reidentifying means determining whether or not a personal item to reidentify selected from one image of the plurality of images and a personal item to reidentify selected from another image of the images are identical personal items.