Security Inspection Face Tracking for Accurate Person-Bag Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing security inspection systems face challenges in capturing high-quality facial images due to limited capture angles and lack of information interaction between multiple trays used by the same person, leading to missed or false associations during person-bag association.
Innovation Solution
A method that tracks the movement trajectory of a target person, continuously acquires facial images, evaluates their quality, and updates the candidate image to the highest quality image, binding it to the trajectory, while using tray records and identifiers to ensure accurate person-bag association.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a camera captures a face at a single angle when the tray is placed at a specific position, then the detection process is simple and quick, but the capture angle is limited and the quality of the captured facial image is poor
Solution Approach 1:
The patent implements continuous tracking of the target person's movement trajectory and continuously captures facial images throughout the process, rather than using a fixed single-angle capture. The system dynamically adjusts capture based on the person's movement, maintaining multiple capture angles and extending the capture time window, thereby improving facial image quality without requiring complex dynamic adjustment mechanisms
Solution Approach 2:
The patent adds the time dimension to the capture process by continuously capturing images throughout the person's movement trajectory, rather than capturing at a single moment. This transforms the capture from a static single-angle snapshot to a dynamic multi-angle sequence over time, improving image quality while keeping the camera system relatively simple
2Reliability
If face detection and recognition is performed for each tray separately without information interaction, then the processing of each tray is independent and simple, but the time consumption increases and the association accuracy decreases
Solution Approach 1:
The patent merges the detection processes of multiple trays by implementing information interaction between them. When a person uses multiple trays, the system combines facial images and trajectory information across different trays, performing unified recognition and association. This reduces redundant detection operations and improves both association accuracy and time efficiency
Solution Approach 2:
The patent maintains continuous tracking and detection across the entire process of a person using multiple trays, rather than performing discrete independent detections. The system continuously monitors the person's movement trajectory and maintains association information throughout, ensuring no useful detection opportunities are lost while reducing overall processing time through information reuse
3Measurement precision
If the detection ends when the tray is pushed into the conveyor line, then the detection process is short and efficient, but the time for capturing the face is too short to ensure high-quality images
Solution Approach 1:
The patent performs preliminary face capture and recognition operations before the tray is pushed into the conveyor line. By starting detection when the person enters the specific region and continuing through the tray placement process, the system ensures sufficient capture time and multiple capture opportunities before the final conveyor insertion, improving image quality without extending the overall process unnecessarily
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The present disclosure provides a security inspection method, including: tracking (S420) a movement trajectory of a target person and continuously acquiring a current image of a face of the target person when the target person is located in a specific region, where the movement trajectory includes a relative position information between the target person and at least one candidate position in the specific region; evaluating (S430) a quality of the current image and a quality of a candidate image after each acquisition of the current image, where the candidate image is a facial image of the target person with a highest quality after the target person enters the specific region; and updating (S460) the candidate image to be the current image and binding the candidate image to the movement trajectory in a case that the quality of the current image is higher than the quality of the candidate image. The present disclosure further provides a security inspection apparatus, a device, a storage medium, and a program product.