Sequential ROI Detection for Person-Object Association in Crowds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional technologies face challenges in accurately tracking and associating persons with target objects in crowded environments, leading to detection omission and erroneous associations.
Innovation Solution
A detection program and method that analyze video data to identify regions of interest, track objects appearing from these regions, and determine the person responsible for moving the object out of the region of interest, even in congested conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a region of interest (ROI) is set to limit detection range for acquiring target objects, then detection precision for acquisition behavior is improved, but reliability of person-object association deteriorates in congested environments
Solution Approach 1:
The patent segments the detection process into multiple phases: first detecting object appearance from ROI in frame N, then determining person identifiability, and finally identifying the person responsible in frame N+1. This segmentation allows the system to maintain detection precision while addressing reliability issues through multi-step verification.
Solution Approach 2:
The patent performs preliminary detection of object appearance from the ROI in frame N before finalizing the person identification. By preliminarily identifying which objects appeared from the ROI, the system can then focus the person identification process in frame N+1 on the correct objects, preventing erroneous associations in congested environments.
2Device complexity
If conventional detection methods are used in congested environments, then processing simplicity is maintained, but detection accuracy deteriorates due to erroneous associations
Solution Approach 1:
The patent introduces dynamic adjustment based on congestion detection. When congestion is detected in frame N, the system dynamically switches to an alternative detection process that uses frame N+1 to identify the person responsible for moving objects from the ROI. This dynamic approach maintains accuracy without permanently increasing processing complexity.
Solution Approach 2:
The patent changes the detection parameters based on environmental conditions. By detecting congestion level as a parameter and adjusting the detection process accordingly (using different frames and criteria based on congestion state), the system maintains high accuracy while keeping the base processing simple.
3Productivity
If person tracking is performed in congested areas, then completeness of detection is improved, but measurement precision deteriorates due to detection omission
Solution Approach 1:
The patent extracts the critical information of object appearance from the ROI in frame N, separate from the complex person tracking process. By extracting this key information first, the system can then focus person identification efforts in frame N+1 on the correct objects, improving measurement precision while maintaining completeness.
Data Source
AI summary
A non-transitory computer-readable recording medium has stored therein a detection program that causes a computer to execute a process. The process includes acquiring a video in which a region of interest where an object is located is set, detecting an object that has appeared from the region by analyzing a first frame in the acquired video, in a case where the appearance of the object is detected, determining whether or not a person who has moved the object out of a range of the region is identifiable, in a case where it is determined that the person is not identifiable, identifying a person using the object by analyzing a second frame after the first frame, and registering the identified person as the person who has moved the object out of the range of the region in a storage unit in association with the object that has appeared from the region.