Skeleton-Based Running Detection for Accurate Speed Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing monitoring systems struggle to accurately calculate the movement speed of individuals due to changes in background, camera installation location, or variations in the size of the detected person, making it difficult to enforce movement rules effectively.
Innovation Solution
A running detection system that utilizes skeleton information from video analysis to track joint points and calculate movement speed, correcting for size changes and camera placement variations, and provides notifications for non-compliant movements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If background subtraction method is used to detect person size and shape, then movement speed can be calculated, but accuracy deteriorates when background changes, camera location changes, or person size varies
Solution Approach 1:
The patent replaces the mechanical/image-processing-based background subtraction method with an AI-based skeleton detection system. The skeleton detector uses deep learning to directly detect human skeleton information from images, eliminating the need for background subtraction and making the detection accurate regardless of background changes, camera locations, or person size variations.
Solution Approach 2:
The patent changes the detection parameters from relying on person size and shape (which vary with distance and background) to using skeleton joint positions and movements. By detecting the positions and movements of specific skeleton joints (such as ankle joints), the system achieves consistent and accurate movement speed calculation across different conditions.
2Measurement precision
If skeleton information is used to detect movement, then measurement accuracy improves across different conditions, but device complexity increases
Solution Approach 1:
The patent extracts only the essential skeleton information needed for movement detection, specifically focusing on key joint positions (such as ankle joints) rather than processing entire person images or complex background data. This extraction approach maintains high measurement accuracy while reducing computational complexity compared to comprehensive image analysis.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
To provide a running detection system, a running detection method, and a running detection program capable of accurately calculating the movement speed of a person even in a case where a monitoring region, an installation location of a camera, or the like is different or in a case where the size of the person changes. The running detection system includes a video acquirer 330, a skeleton detector 410, and a speed calculator 340. The video acquirer 330 acquires a video obtained by imaging a predetermined region including a mobile object. The skeleton detector 410 detects skeleton information of the mobile object included in the video. The speed calculator 340 calculates information on the movement speed of the mobile object on the basis of the skeleton information of the mobile object.