Representative Point Estimation for Occluded Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In situations of object congestion, where objects and people overlap, leading to hidden joint points and representative points, existing posture estimation methods suffer from decreased accuracy due to occlusion.
Innovation Solution
An estimation device and method that includes feature point estimation, representative point candidate determination, and estimation, which detects and integrates feature points to accurately determine the representative point of an object even when some points are hidden, using a system comprising a learning device, storage device, and object detection device with neural network-based models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional posture estimation methods are used, then the estimation process is simple, but the estimation accuracy deteriorates under congestion situations where objects overlap and shield each other
Solution Approach 1:
The patent segments the representative point estimation into multiple independent steps: feature point estimation, representative point candidate determination, and representative point estimation. This segmentation allows each step to handle specific aspects of the estimation problem, improving overall accuracy while maintaining manageable system complexity through modular design
Solution Approach 2:
The patent performs preliminary feature point estimation and representative point candidate determination before final representative point estimation. This preliminary action prepares and pre-processes the data, filtering out invalid candidates and establishing reliable feature points beforehand, which improves the accuracy of the final estimation while organizing the complex processing into structured phases
2Reliability
If feature points are used for posture estimation, then the estimation can be performed, but the estimation accuracy decreases when feature points are hidden by occlusion
Solution Approach 1:
The patent introduces representative point candidates as intermediary elements between feature points and the final representative point. These candidates serve as mediators that can be derived from multiple feature points, allowing the system to infer the representative point location even when some feature points are occluded, thereby maintaining reliability under occlusion conditions
Solution Approach 2:
The patent creates multiple representative point candidates based on different feature point combinations and uses voting mechanisms to select the final representative point. This copying approach generates alternative estimates that can compensate for occluded features, improving reliability by providing redundant estimation paths
3Measurement precision
If multiple representative point candidates are determined and integrated, then the estimation accuracy improves, but the processing time increases
Solution Approach 1:
The patent determines multiple representative point candidates (excessive action) but then applies filtering and voting mechanisms to select only the most reliable ones (partial action). This approach generates enough candidates to ensure accuracy while avoiding processing all possible combinations, thereby balancing accuracy improvement with time efficiency
Data Source
AI summary
An estimation device 3X mainly includes a feature point estimation means 35X, a representative point candidate determination means 37X, and a representative point estimation means 38X. The feature point estimation means 35X estimates plural feature points relating to an object. The representative point candidate determination means 37X determines plural representative point candidates that are candidates of a representative point of the object based on the plural feature points. The representative point estimation means 38X estimates the representative point based on the plural representative point candidates.