Adaptive Multi-Modal Object Identification Under Face and Pose Variations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object identification systems face challenges in accurately identifying objects when face features are invisible, significantly changed, or affected by head pose, and body identification is compromised by noisy backgrounds or large body poses, leading to reduced performance.
Innovation Solution
A system that integrates face and body features using deep convolutional neural networks, adaptively learns and updates databases with new features, and fuses identification results to provide robust and accurate object recognition, even when individual features are unreliable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If face identification is used for object recognition, then identification accuracy is improved when face features are visible, but identification performance deteriorates when face features are invisible, significantly changed, or affected by head pose
Solution Approach 1:
The patent combines face identification and body identification into a unified object recognition system. When face features are unavailable or unreliable, the system automatically switches to body identification, ensuring continuous and reliable object recognition across various conditions.
Solution Approach 2:
The system dynamically changes identification parameters by selecting different identification modes (face or body) based on the availability and quality of features. This allows the system to adapt to varying conditions and maintain high identification accuracy.
2Reliability
If body identification is used for object recognition, then identification capability is maintained when face features are unavailable, but identification accuracy deteriorates due to noisy backgrounds or large body poses
Solution Approach 1:
The patent integrates face identification and body identification results through result fusion. This combination allows the system to leverage the strengths of both identification methods, maintaining high accuracy even when body identification alone would be compromised by noisy backgrounds or large poses.
Solution Approach 2:
The system uses feedback from face identification results to guide body identification. When face identification is successful, it provides feedback that helps refine body identification, improving overall accuracy despite challenging body conditions.
3Reliability
If multiple identification methods are integrated, then identification reliability is improved under varying conditions, but system complexity increases
Solution Approach 1:
The patent creates a universal identification system that can perform both face and body identification using a single integrated framework. This multi-functional approach improves reliability across different conditions while avoiding the need for separate, independent identification systems.
Solution Approach 2:
The system dynamically adjusts its operation by automatically selecting and switching between face identification and body identification based on real-time conditions. This dynamic behavior maintains simplicity in operation while achieving high reliability through adaptive method selection.
Data Source
AI summary
Disclosed herein are systems, methods, and devices for using adaptive learning to identify objects. An object-identifying device performs a first object identification based on one or more features of a first modality of an object retrieved from an image frame including the object and a first database including first modality identification features. A second object identification is performed based on one or more features of a second modality of the object retrieved from the image frame and a second database including second modality identification features. The second database is updated by adaptively learning a new second modality identification feature according to a first identification result of the first object identification. The second object identification is trained with the updated second database and determines a final identification result by integrating a first identification result of the first object identification and a second identification result of the second object identification.