Foreground Area Motion Detection for Abandoned Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing surveillance technologies face challenges in accurately distinguishing between moving objects and abandoned objects, leading to potential misdetctions of theft or false alarms, due to difficulties in setting appropriate stay time thresholds and correctly identifying motionless objects as stolen or not.
Innovation Solution
An information processing apparatus that determines whether a foreground area represents a moving object by comparing it with a corresponding area from a previous image and using history information to confirm the object's movement status, thereby reducing false alarms and improving detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a simple difference operation between reference image and comparison image is performed to detect motionless objects, then detection speed is improved, but detection precision deteriorates due to inability to distinguish moving objects that stopped temporarily from abandoned objects
Solution Approach 1:
The system performs preliminary classification by detecting moving objects in real-time and tracking their trajectories before they stop. When an object stops, the system checks its motion history to determine if it was previously moving, thereby distinguishing abandoned objects from temporarily stopped objects. This preliminary tracking action enables accurate classification without requiring complex analysis at the moment of stopping.
2Measurement precision
If stay time threshold is set to distinguish abandoned objects from moving objects, then detection precision is improved, but reliability deteriorates due to difficulty in setting appropriate threshold values
Solution Approach 1:
The system performs preliminary tracking of object trajectories and calculates motion characteristics (speed, direction, acceleration) before the object stops. This preliminary analysis provides rich feature information that enables accurate classification without relying solely on stay time thresholds, thereby maintaining high reliability regardless of threshold settings.
Solution Approach 2:
Instead of relying on a single stay time parameter, the system utilizes multiple motion parameters including trajectory length, average speed, acceleration patterns, and stopping position. By changing from a single-parameter threshold approach to a multi-parameter analysis approach, the system achieves high detection precision without compromising reliability.
3Measurement precision
If image processing is performed to detect theft by comparing captured images, then detection precision is improved, but device complexity increases due to need for multiple image comparisons and analysis
Solution Approach 1:
The system performs preliminary processing by extracting only moving object regions from captured images and tracking their trajectories. When an object stops, the system checks its motion history and compares its trajectory pattern against predefined patterns for abandoned objects. This preliminary extraction and tracking approach simplifies the overall processing complexity while maintaining high detection precision.
Data Source
AI summary
A determination unit determines whether or not a first foreground area extracted from a first captured image represents a moving object. A warning unit outputs warning when the first foreground area is determined to be not representing the moving object. The first foreground area is determined to be not representing the moving object when the object represented by the first foreground area is not moving. First, the determination unit determines whether or not the object represented by the first foreground area is moving by comparing the first foreground area with a second foreground area extracted from a second captured image generated before the first captured image. Then, the determination unit determines whether or not the object is moving at time before the first captured image is generated based on information of the object at time before the first captured image is generated.