Visual Object Tracker Using Segmented Detection and Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current visual object tracking systems face challenges in achieving real-time performance due to missed detections and limited computing resources, particularly with deep neural network-based tracking-by-detection schemes, which can lead to fragmented tracklets and low frame rates.
Innovation Solution
The system employs a combination of outer and inner tracking mechanisms, where a deep neural network-based outer tracker is complemented by lightweight region-of-interest (ROI) trackers to enhance detection confidence and frame rates, with an image traffic throttling mechanism to adjust frame rates and ensure real-time processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep neural network-based tracking-by-detection schemes are used to improve detection accuracy, then detection precision is improved, but frame rate decreases and real-time performance deteriorates
Solution Approach 1:
The tracking system is divided into two independent modules: a detection module that operates at lower frame rates to identify objects, and a tracking module that operates at high frame rates to follow detected objects. This segmentation allows each module to be optimized for its specific function, resolving the contradiction between detection precision and frame rate.
Solution Approach 2:
A lightweight tracking module serves as an intermediary between the detection module and the final output. This intermediary takes sparse detection results and generates dense tracking trajectories at high frame rates, bridging the gap between low-rate detection and high-rate tracking requirements.
2Measurement precision
If deep neural network-based tracking-by-detection schemes are used to improve detection capability, then detection precision is improved, but computational resources are excessively consumed
Solution Approach 1:
The system segments computational tasks between a resource-intensive detection module that runs periodically and a computationally efficient tracking module that runs continuously. This division reduces overall computational resource consumption while maintaining detection capability.
Solution Approach 2:
The tracking module uses lightweight, computationally inexpensive algorithms that can be executed rapidly and discarded each frame, replacing the need for expensive deep neural network computations at every frame. This allows high frame rate operation with minimal resource consumption.
3Reliability
If detection frequency is increased to reduce missed detections, then detection reliability is improved, but frame rate decreases
Solution Approach 1:
The system separates detection and tracking functions, allowing detection to occur at lower frequencies while tracking maintains high frame rates. This segmentation ensures that missed detections are compensated by the continuous tracking process, maintaining both reliability and productivity.
Solution Approach 2:
The tracking module operates continuously at high frame rates, providing uninterrupted tracking even when detection occurs periodically. This continuous action ensures that detection reliability is maintained without sacrificing frame rate, as the tracker fills in gaps between detection events.
Data Source
AI summary
An approach is disclosed for real-time object tracking. The approach involves, for example, using a first object tracking mechanism to detect and associate one or more objects from frame to frame of a video. The approach also involves initiating one or more second object tracking mechanisms to track the one or more objects detected by the first object tracking mechanism from frame to frame of the video in parallel with the first object tracking mechanism. The approach further involves using a tracking output of the one or more second object tracking mechanisms in place of the first object tracking mechanism for a frame of the video based on determining that first object tracking mechanism has missed a detection of the object in the frame of the video.