Person Tracking via Face and Appearance Feature Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing surveillance camera systems face challenges in accurately tracking suspicious persons due to varying installation environments and imaging conditions, leading to potential misidentification and increased processing loads, as they rely solely on feature extraction and transmission across multiple cameras.
Innovation Solution
A person tracking system that includes face detection cameras, appearance feature detection cameras, a collation server, and an instruction device, which collates face images, transmits appearance feature information, and instructs cameras to detect and transmit relevant data, improving tracking accuracy by utilizing both face and appearance features while managing processing loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all surveillance cameras extract and transmit feature data of suspicious persons, then tracking coverage is improved, but processing load and system complexity increase significantly
Solution Approach 1:
The surveillance camera system is divided into two functional segments: first detection cameras equipped with both face and appearance feature extraction capabilities, and second detection cameras equipped only with appearance feature extraction capabilities. This segmentation allows the system to distribute processing loads appropriately while maintaining comprehensive tracking coverage through coordinated operation of different camera types.
Solution Approach 2:
The first detection cameras are designed with multi-functionality, capable of performing both face feature extraction and appearance feature extraction. This universal capability allows them to serve as primary detection points that can independently provide comprehensive feature data, reducing the need for all cameras to have full functionality and thereby lowering overall system complexity.
2Measurement precision
If face images are transmitted and collated across multiple cameras, then identification accuracy is improved, but data transmission volume and processing time increase
Solution Approach 1:
The system extracts and transmits only the essential feature data needed for identification rather than transmitting complete images or all possible features. Face feature data and appearance feature data are extracted separately and transmitted in a condensed format, reducing data transmission volume while maintaining identification accuracy through the collation of these key features.
3Reliability
If multiple cameras with full detection capabilities are deployed, then tracking reliability is improved, but cost and device complexity increase
Solution Approach 1:
Different cameras are assigned different detection capabilities based on their local requirements and positions in the surveillance network. First detection cameras at critical locations have both face and appearance detection capabilities, while second detection cameras at other locations have only appearance detection capabilities. This local quality differentiation ensures tracking reliability is maintained where needed while reducing overall system complexity and cost.
Data Source
AI summary
In surveillance camera system (10), face detection is performed with Cam-A or Cam-F and in a case where there is a match with the face image of a specific person as a result of collation of face images, appearance feature information is transmitted from tracking client (30) to other Cam-B to Cam-E grouped in association with Cam-A or Cam-F. Upon detecting the appearance feature information, the other Cam-B to Cam-E transmit the person discovery information to a tracking client (30).