Person Tracking via Sampled Image Search for Travel Time Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking systems face challenges in automatically estimating travel time between areas covered by multiple cameras, requiring manual user intervention and incurring high computational costs, especially when tracking individuals across different camera views.
Innovation Solution
The system automatically detects and tracks individuals across cameras by extracting image features, storing them in a database, and conducting sample queries to calculate travel time without user operation, using image matching techniques with filtering mechanisms to ensure accuracy and reduce computational costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image matching is performed to all person images in newly input images, then tracking accuracy across cameras is improved, but computational cost increases significantly
Solution Approach 1:
The patent applies partial action by performing image matching only on sampled person images rather than all detected persons. The system selects representative samples from the database and compares them against newly input images, achieving acceptable tracking accuracy while significantly reducing computational resources required for cross-camera person re-identification.
Solution Approach 2:
The system performs preliminary actions by pre-processing and storing person images in a database with extracted features before actual tracking is needed. This allows the system to quickly retrieve and compare pre-processed images during real-time operation, reducing computational burden during critical tracking phases while maintaining accuracy.
2Measurement precision
If manual user operation is required to select tracking targets, then tracking precision is maintained, but ease of operation deteriorates and automation is reduced
Solution Approach 1:
The system implements self-service by automatically selecting sampling targets and performing image matching without requiring manual user intervention. The algorithm autonomously identifies representative persons from the database, performs comparisons with newly input images, and calculates travel times, thereby maintaining tracking precision while significantly improving ease of operation and enabling full automation.
Solution Approach 2:
The system uses feedback mechanisms to automatically adjust and improve tracking performance. By continuously comparing sampled images with newly input images and evaluating matching results, the system refines its selection criteria and matching algorithms, maintaining high tracking precision through automated learning and adaptation without user intervention.
3Productivity
If sampling is introduced to reduce computational cost, then productivity is improved, but measurement precision may deteriorate
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting sampling rates, similarity thresholds, and comparison criteria based on scene complexity, camera configurations, and performance requirements. This allows the system to optimize the balance between processing speed and tracking accuracy, improving productivity while maintaining sufficient measurement precision for practical applications.
Data Source
AI summary
In example implementations described herein, person images obtained by object detection and tracking are stored in an image database. Then the system automatically samples a person image as a query of image search for past images. Filtering is performed on the search results to extract a person images having a similarity higher than certain threshold. Travel time is calculated by sorting information (camera identifier, time) attached to person images on a timeline, and the system detects gap between periods that the person appears in frame.