Object Tracking via Location Prediction and CCTV Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object tracking systems require significant computing power to identify and track objects across multiple CCTV videos, making the process inefficient.
Innovation Solution
An object tracking system that utilizes a terminal capable of location tracking, incorporating modules for cross-CCTV and current-CCTV detection, basic image detection, and object tracking, which collects location history and current location to efficiently detect and track objects by determining overlapping CCTV areas and moving directions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning technology is used to extract, classify and re-identify objects from video, then object recognition accuracy is improved, but computing power requirements increase significantly
Solution Approach 1:
The system performs preliminary actions by collecting terminal location history and predicting future locations before object detection is needed. The server determines which CCTVs are likely to capture the terminal based on location prediction, pre-filtering the search space. This preliminary location-based filtering reduces the computational burden during actual object detection and re-identification, while maintaining high accuracy by focusing resources on relevant video streams only.
2Reliability
If object detection is performed on all CCTV videos to track objects across multiple cameras, then tracking coverage is improved, but processing time increases
Solution Approach 1:
The server performs preliminary location prediction to determine which CCTVs are likely to capture the terminal before detection is needed. This advance determination allows the system to focus detection resources only on relevant CCTVs rather than processing all videos, maintaining comprehensive tracking coverage while significantly reducing processing time by eliminating unnecessary camera inspections.
Solution Approach 2:
The system segments the CCTV network into relevant and irrelevant groups based on terminal location prediction. Only CCTVs that are predicted to capture the terminal are selected for object detection, creating a segmented processing approach that maintains tracking coverage across the network while reducing overall processing time by working with a subset of cameras.
3Productivity
If location information from terminals is used to predict movement and determine relevant CCTVs, then detection efficiency is improved, but system complexity increases
Solution Approach 1:
The server acts as an intermediary that receives location information from terminals and translates it into predictions about which CCTVs will capture the terminal. This intermediary function simplifies the overall system by centralizing the complex prediction logic in one component, allowing terminal devices to remain simple while achieving efficient detection through the server's coordination.
Data Source
AI summary
Provided are an object tracking system and an object tracking method. The object tracking system includes: a terminal identifier and reference time providing module configured to receive an identifier of a terminal and a reference time for tracking an object corresponding to the identifier; a cross-CCTV detection module configured to detect a cross-CCTV for the terminal by using a CCTV installation information and a location of the terminal before the reference time; a basic image detection module configured to detect an object repeatedly appearing in the cross-CCTV as a basic image; a current-CCTV detection module configured to detect a current-CCTV currently recording the terminal by detecting a location and a moving direction of the terminal after the reference time; an object detection module configured to detect an object appearing in the current-CCTV based on the location and the moving direction of the terminal after the reference time; and an object tracking module configured to track an object corresponding to the identifier by determining whether the detected object from the current-CCTV and the basic image.