Image Processing Subject Prioritization for Real-Time Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing shoplifting prevention systems face challenges in maintaining responsiveness when detecting a large number of subjects due to increased calculation loads, particularly on low-cost generic PCs, leading to potential degradation in real-time performance.
Innovation Solution
An image processing apparatus that includes a subject detection unit, independence detection unit, measurement unit, and determination unit to prioritize and limit image processing based on subject independence and residence time, allowing efficient processing even with a large number of subjects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If action analysis processing is performed on all detected subjects, then detection accuracy is improved, but processing time increases proportionally with the number of subjects
Solution Approach 1:
The patent segments the set of detected subjects into two categories: independent subjects (those alone or with others at distance) and non-independent subjects (those in close proximity). This segmentation allows the system to apply different processing strategies to different subject groups, avoiding uniform processing of all subjects and thereby reducing overall processing time while maintaining detection accuracy for suspicious behaviors.
Solution Approach 2:
The patent applies local quality by treating independent subjects and non-independent subjects differently based on their spatial relationships. Independent subjects receive full action analysis processing to detect suspicious behaviors, while non-independent subjects (likely companions or groups) receive reduced or no processing. This localized differentiation optimizes processing resources based on the specific characteristics of each subject group.
2Speed
If the system processes all detected subjects in real-time, then responsiveness is maintained, but calculation load increases significantly with more subjects
Solution Approach 1:
The patent applies partial action by performing action analysis processing only on a subset of subjects (independent subjects) rather than all detected subjects. This selective processing maintains responsiveness for potentially suspicious individuals while significantly reducing the overall calculation load, allowing the system to operate efficiently on low-cost generic PCs even when many subjects are present.
3Ease of manufacture
If the system uses a low-cost generic PC, then cost is reduced, but processing capability is limited when the number of subjects increases
Solution Approach 1:
The patent changes the parameter of processing scope by introducing an upper limit on the number of subjects subjected to action analysis processing. This parameter modification allows the system to maintain manageable calculation loads on low-cost hardware while still effectively detecting shoplifting suspects. The upper limit parameter enables the system to operate within the computational constraints of generic PCs.
Data Source
AI summary
An image processing apparatus comprising a subject detection unit that detects a plurality of subjects from a captured image; an independence detection unit that detects an independent subject from the detected subjects, wherein the independent subject includes a subject for which other subjects are not present in a predetermined range; a measurement unit that measures residence times for the detected subject; a determination unit that determines a degree of priority for the subject such that the degree of priority becomes higher for the subject that is independent and have relatively long residence times, based on detection results from the independence detection unit, and measurement results from the measurement unit; and an image processing unit that performs predetermined image processing on the subjects up to a predetermined upper limit for the number of subjects in order from the highest degree of priority.