Object Pose Estimation via Tracked and Identified Density Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object pose estimation methods, particularly in computer vision, face challenges in achieving accurate and continuous estimation, especially under conditions of high rotation, velocity, and abrupt illumination changes, leading to tracking drift and inaccurate classification.
Innovation Solution
An apparatus and method that combine tracked pose probability density from a tracking scheme with identified pose probability density from a training model, using Gabor wavelet transforms and random forest models, to estimate object pose effectively, with online updating of the training model to improve accuracy and adapt to dynamic environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If tracking-based method is used to estimate pose through pair matching of current and previous frames, then relatively accurate estimation over short time is achieved, but tracking drift occurs due to accumulated errors
Solution Approach 1:
The system uses feedback by comparing tracked pose results with identified pose results from training models. The identified pose serves as a reference to correct accumulated tracking errors, creating a closed-loop system that maintains long-term accuracy while preserving the speed benefits of tracking.
Solution Approach 2:
The system performs preliminary action by pre-training pose estimation models with labeled data before actual tracking. This pre-computed knowledge base enables rapid correction of tracking drift without requiring real-time complex computations, allowing the system to maintain accuracy while operating at tracking speeds.
2Reliability
If key frame is used to eliminate tracking drift, then tracking stability is improved, but reasonable selection and updating of key frame becomes difficult
Solution Approach 1:
The system implements self-service by automatically selecting and updating key frames based on accumulated tracking error thresholds. The algorithm autonomously determines when to invoke the training model for correction without requiring manual intervention or complex external control logic, simplifying the overall system architecture.
3Productivity
If training-based method is used for object pose estimation through classification or regression, then pose estimation is performed, but accurate estimate is not obtained due to rough classification or susceptibility to real environment
Solution Approach 1:
The system merges tracking-based estimation and training-based estimation into a unified framework. The tracking component provides fast, continuous pose estimates while the training model provides accurate reference poses. By combining these two approaches, the system achieves both high speed and high accuracy that neither method could achieve alone.
4Adaptability or versatility
If object rotation range or velocity increases due to abrupt change in illumination, then tracking becomes more challenging, but pose estimation effectiveness is reduced
Solution Approach 1:
The system implements dynamics by adaptively adjusting the frequency of training model invocation based on detected environmental changes. When illumination changes or rapid motion is detected, the system dynamically increases the frequency of reference pose calculations, ensuring accuracy is maintained under varying conditions without wasting computational resources during stable periods.
Data Source
AI summary
An apparatus and method for estimating a pose of an object are provided. The apparatus includes an object input unit configured to input an object in an object tracking unit and an object identifying unit, an object tracking unit configured to obtain a tracked pose probability density of the object based on a tracking scheme, an object identifying unit configured to obtain an identified pose probability density of the object based on a training model, and a combination unit configured to obtain an estimated pose probability density of the object using a combination of the tracked pose probability density and the identified pose probability density and to estimate a pose of the object based on the estimated pose probability density of the object. Through the combination, a cumulative error occurring in the object tracking may be corrected, resulting in more accurate object estimation.