Surveillance Camera Feature Quantity Masking for Tracking Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Surveillance cameras face challenges in accurately tracking targets due to overlapping objects within the search range, leading to erroneous changes in tracking targets.
Innovation Solution
The surveillance camera and information processing apparatus employ a feature quantity masking technique, using a control unit to set a search range and extract the tracking-target image by masking feature quantities of non-tracking objects, preventing incorrect target changes through artificial intelligence and user-designated masking areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the surveillance camera uses a search range to extract tracking target images, then the tracking coverage is improved, but false tracking of non-target objects occurs when other objects overlap within the search range
Solution Approach 1:
The patent divides the search range into multiple regions and processes feature quantities from different regions separately. By segmenting the feature extraction process, the system can identify and exclude non-target objects while maintaining comprehensive tracking coverage, thus resolving the contradiction between broad coverage and accurate target identification.
Solution Approach 2:
The patent applies different processing strategies to different local regions within the search range. High-confidence regions are processed with standard tracking, while low-confidence regions containing overlapping objects are processed with enhanced verification or excluded, improving overall tracking reliability without sacrificing coverage.
2Difficulty of detecting and measuring
If the surveillance camera extracts feature quantities from the entire search range, then the detection capability is improved, but the influence of non-tracking objects increases leading to incorrect target changes
Solution Approach 1:
The patent extracts and removes feature quantities from non-target objects before performing tracking decisions. By taking out the interfering feature quantities from overlapping objects, the system maintains high detection capability while preventing false tracking changes, thus resolving the contradiction between detection sensitivity and tracking stability.
Solution Approach 2:
The patent applies asymmetric weighting to different feature quantities based on their reliability. Feature quantities from regions with high target presence probability are given higher weight, while those from ambiguous regions are downweighted or excluded, allowing the system to utilize all available detection data while maintaining tracking stability.
Data Source
AI summary
A surveillance camera includes an imaging unit configured to output a video frame, and a control unit configured to set a search range in the video frame and to extract a tracking-target image by using a feature quantity of an image in the search range. In a case where an object image other than the tracking target is contained in the search range, the control unit masks a feature quantity of the object image and extracts the tracking-target image.