Attribute-Change Monitoring for Suspicious Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing monitoring systems fail to detect suspicious objects when individuals disguise themselves or change their clothing or gait, as these changes are not inherently determined to be suspicious, and existing techniques only focus on specific actions like shoplifting.
Innovation Solution
A monitoring device that associates objects across time-series images, detects attribute changes in objects and accessories, and identifies suspicious objects based on these changes, using association units, attribute change detection units, and suspicious object detection units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If monitoring systems focus on detecting specific suspicious behaviors (e.g., shoplifting actions), then detection accuracy for known suspicious actions is improved, but the system cannot detect suspicious objects when individuals disguise themselves or change attributes
Solution Approach 1:
The monitoring system segments the detection process into two independent modules: behavior pattern recognition and attribute change detection. This segmentation allows the system to simultaneously maintain high accuracy for known suspicious behaviors while adding the capability to detect attribute changes such as clothing or mask changes, thereby resolving the contradiction between detection precision and adaptability against disguise
Solution Approach 2:
The system implements multi-functionality by integrating both behavior analysis and attribute change detection capabilities into a single monitoring framework. The attribute change detection module universally monitors any changes in physical attributes (clothing, masks, accessories) regardless of the specific context, enabling the system to adapt to various disguise scenarios while maintaining its original behavior detection functions
2Device complexity
If the system tracks only behavior patterns within the monitoring range, then processing complexity is reduced, but the system cannot detect when a person has left a carried object in a place not shown in the monitoring video
Solution Approach 1:
The system performs preliminary action by detecting and recording attribute changes (such as when a person picks up or drops an object) at the moment they occur within the monitoring range. This preliminary detection creates a record that can be later correlated with the person's movement轨迹, enabling the system to infer events that occurred outside the direct monitoring view without requiring continuous complex tracking of every movement
3Measurement precision
If the system monitors all attribute changes of objects, then detection capability for suspicious objects is improved, but false alarms increase because clothing changes are not inherently suspicious
Solution Approach 1:
The system applies local quality by differentiating the evaluation criteria for different types of attribute changes based on their contextual significance. Critical attributes such as face masks, hats, and outerwear receive higher monitoring priority and trigger alerts when changed, while less significant attributes are monitored with lower sensitivity. This localized quality control reduces false alarms while maintaining high detection capability for truly suspicious changes
Data Source
AI summary
A monitoring device and the like are provided which are capable of detecting an attribute change in a suspicious object that cannot be determined from the behavior of the object. An associating unit associates, among a plurality of objects detected from time-series image data, identical objects with one another. An attribute change detecting unit detects from the time-series image data a change in an attribute of at least one of the identical objects and an attendant item. A suspicious object detecting unit detects a suspicious object on the basis of the change in attribute.