Human Posture Estimation Using Motion-Aware Heatmaps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 2D camera-based human posture estimation is vulnerable to dynamic movement and suffers from jittering due to image noise, lighting changes, and camera movement, leading to inaccurate posture estimation.
Innovation Solution
An apparatus and method that generate motion-aware heatmaps from continuous images, using a processor to create intersection heatmaps by averaging products of motion-aware heatmaps at different time points, and estimate human posture based on these heatmaps, incorporating motion vectors and weights to stabilize the estimation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If 2D camera-based posture estimation is used, then equipment cost is reduced, but measurement precision deteriorates due to vulnerability to dynamic movement and jittering
Solution Approach 1:
The system performs preliminary actions by generating motion-aware heatmaps from multiple consecutive images before final posture estimation. This includes predicting future and past posture heatmaps, calculating motion vectors, and creating motion-aware representations that compensate for dynamic movement effects before the actual posture measurement is made, thereby improving precision without changing the 2D camera equipment
Solution Approach 2:
Motion-aware heatmaps serve as an intermediary between the raw 2D camera images and the final posture estimation. The system introduces motion vectors and temporal context as intermediate representations that mediate the effect of dynamic movement and jittering, allowing the final posture measurement to be more accurate despite using inexpensive 2D cameras
2Measurement precision
If motion-aware heatmaps from multiple time points are integrated, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system segments the posture estimation process into distinct temporal components: current frame heatmaps, future frame predicted heatmaps, and past frame predicted heatmaps. Each segment is processed separately to generate motion-aware heatmaps, which are then integrated. This segmentation manages complexity by breaking down the temporal processing into manageable, modular steps rather than attempting to process all frames simultaneously
3Reliability
If continuous images are processed to generate motion-aware heatmaps, then reliability is improved, but loss of time increases due to processing multiple frames
Solution Approach 1:
The system uses periodic action by processing discrete consecutive image frames at regular time intervals rather than continuous streams. By selecting specific time points (current, future, and past frames) and processing them periodically, the system achieves reliable posture estimation while controlling processing time through the periodic sampling of frames rather than continuous analysis
Data Source
AI summary
The present invention relates to an apparatus and a method for estimating human posture. The method for estimating human posture comprises generating a plurality of motion-aware heatmaps for each joint based on a plurality of previously input images corresponding to continuous time, generating intersection heatmaps by considering motions between motion-aware heatmaps at different time points from among the plurality of motion-aware heatmaps and estimating a human posture based on the intersection heatmap.