Image Recognition Using Whole-Body Features for Obscured Passenger ID
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current airport systems face challenges in efficiently identifying and monitoring passengers and their belongings due to insufficient data quality, reliance on face features, and high false negative rates, especially when faces are obscured or not aligned with the camera view, and existing systems are labor-intensive and costly.
Innovation Solution
A system using artificial intelligence and machine learning to identify entities based on a set of unique characteristics, such as dents, stickers, or unusual shapes, without requiring face landmarks or article identifiers, leveraging cameras and machine learning models to track and recognize passengers and their belongings across various angles and alignments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If biometric face detection systems rely solely on face features to identify passengers, then identification can be performed when faces are clearly visible, but the system fails when faces are obscured or not aligned with the camera view, resulting in high false negative rates
Solution Approach 1:
The system changes the detection parameters from face-specific features to whole-body representation features. By detecting and tracking the entire body silhouette and characteristics rather than relying on face features alone, the system maintains identification capability even when faces are obscured or misaligned with camera views, thereby improving reliability without sacrificing detection precision
Solution Approach 2:
The system transitions from two-dimensional face feature detection to three-dimensional whole-body representation analysis. By capturing and analyzing the complete body shape, posture, and spatial characteristics across multiple dimensions, the system overcomes the limitation of face-only detection and achieves reliable identification under various viewing conditions
2Area of stationary object
If CCTV footage is used to monitor passengers, then broad coverage is achieved, but passengers can be obscured by others who are closer to the camera, reducing detection effectiveness
Solution Approach 1:
The system employs a multi-functional detection approach that can identify passengers through various body parts and characteristics, not just faces. By using whole-body representation that includes clothing patterns, body shape, posture, and movement characteristics, the system maintains detection precision across wide surveillance areas even when passengers are partially obscured
Solution Approach 2:
The system uses body silhouette and clothing characteristics as intermediary features for identification when direct face detection is not possible. These intermediary features serve as alternative identifiers that maintain detection precision even when the primary identification target (face) is obscured by other passengers
3Adaptability or versatility
If algorithms use whole body representation to identify passengers, then identification can occur without face alignment, but the ratio of false negative results remains high
Solution Approach 1:
The system performs preliminary action by capturing and storing multiple whole-body representation features for each passenger during periods when clear detection is possible. These pre-captured body characteristics, clothing patterns, and movement signatures are stored as reference data, enabling reliable identification later even when detection conditions are suboptimal, thus improving reliability while maintaining detection flexibility
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is provided an image processing system and method for identifying a user. The system comprises a processor configured to identify a first user in an image, determine a plurality of characteristic vectors associated with the first user, compare the characteristic vectors associated with the first user with a plurality of predetermined characteristic vectors associated with a plurality of users including the first user, and identify the first user based on the comparison.