Object Tracking via Temporal Frame Buffering for Low Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection and tracking technologies face challenges in achieving low latency and high accuracy, particularly in real-time applications like video conferencing, where object detection delays can lead to inaccurate tracking results.
Innovation Solution
An object tracking method and apparatus that temporarily store continuous image frames during object detection, allowing for subsequent tracking based on the detection results, combining high accuracy from object detection with efficient tracking to meet low latency requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection is performed first to ensure high accuracy, then tracking precision is improved, but processing time increases causing latency
Solution Approach 1:
The patent performs object detection in advance on a subset of frames (e.g., every Nth frame) to obtain detection results beforehand. These pre-computed detection results are then used for tracking in subsequent frames, eliminating the need to wait for complete detection processing before initiating tracking operations.
Solution Approach 2:
The patent divides the video stream into multiple frames and selectively processes only certain frames for object detection (e.g., frames at regular intervals). This segmentation allows the system to perform detection less frequently while maintaining accurate tracking information across all frames through the temporal continuity provided by the detection results.
2Measurement precision
If object detection is performed on every frame to ensure accuracy, then tracking accuracy is improved, but processing speed decreases
Solution Approach 1:
Object detection is performed in advance on selected frames to obtain detection results before tracking needs to occur. This preliminary detection allows the tracking process to proceed at full speed using pre-computed information, rather than waiting for frame-by-frame detection completion.
Solution Approach 2:
The patent segments the detection process by applying object detection only to a subset of frames (e.g., every Nth frame) rather than all frames. This selective detection maintains tracking accuracy while significantly reducing the total processing load and improving overall processing speed.
3Measurement precision
If frame buffer size is increased to store more continuous frames, then tracking accuracy is improved, but memory consumption increases
Solution Approach 1:
Instead of buffering all continuous frames, the patent selectively stores only specific frames (e.g., every Nth frame) in the frame buffer. This segmentation approach reduces the total number of frames stored in memory while maintaining sufficient information for accurate tracking through the temporal relationships between the stored frames.
Data Source
AI summary
An object tracking method and an object tracking apparatus, which are adapted for a low latency application, are provided. In the method, an object detection is performed on one of continuous image frames. The objection detection is configured to identify a target. The continuous image frames are temporarily stored. An objection tracking is performed on the temporarily stored continuous image frames according to a result of the object detection. The objection tracking is configured to associate the target in one of the continuous image frames with the target in another of the continuous image frames. Accordingly, the accuracy of object tracking may be improved, and the requirement for low latency may be satisfied.