Automated Video Analytics for Sports Pose Skeleton Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video analytics techniques require manual annotation and fail to provide dynamic annotations for video analysis, limiting their effectiveness in providing quick and accurate posture and movement analysis for athletes.
Innovation Solution
A system and method for automated video analytics that detects individuals in images or videos, generates posture skeletons, and outputs angular data, enabling quick analysis and comparison of movement techniques, with both cloud-based and local processing capabilities for real-time feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation is used for video analytics, then annotation accuracy can be ensured, but time consumption and labor requirements increase significantly
Solution Approach 1:
The system enables automated self-annotation by having the computer vision system automatically detect persons, generate posture skeletons, and create annotations without requiring manual human input for each frame, thus resolving the contradiction between annotation accuracy and time consumption
Solution Approach 2:
The system performs preliminary detection and skeleton generation in advance, creating reusable annotation data structures that can be quickly applied across multiple video frames, reducing the time required for detailed annotation while maintaining accuracy
2Measurement precision
If conventional video analytics are used, then detailed frame-by-frame analysis is possible, but dynamic annotations that move with video frames cannot be provided
Solution Approach 1:
The system generates dynamic posture skeletons that automatically update and move with each video frame, allowing annotations to track persons throughout the video sequence rather than remaining static, thus enabling both detailed analysis and dynamic visualization
Solution Approach 2:
The system continuously processes video frames in sequence, maintaining persistent person detections and updating posture skeletons frame-by-frame, which enables smooth dynamic annotations that follow persons throughout the entire video duration
3Productivity
If automated processing is implemented, then analysis speed increases, but complexity of the system increases
Solution Approach 1:
The system uses a unified computer vision pipeline that handles multiple tasks (person detection, pose estimation, skeleton generation, annotation creation) through integrated algorithms, reducing overall system complexity while maintaining high processing speed
Solution Approach 2:
The system replaces manual mechanical annotation processes with automated computer vision algorithms and machine learning models, significantly increasing analysis speed while the modular software architecture keeps system complexity manageable
Data Source
AI summary
A method for providing automated video analytics includes receiving, at a computing device, an image and detecting a person within the image. A pose of the person within the image is determined and a plurality of points are identified based on the pose. Each point is indicative of a location of a body part of the person. A posture skeleton is generated based on the plurality of points. The posture skeleton includes lines interconnecting at least some of the plurality of points. The image is outputted with the posture skeleton superimposed over the person.