Real-Time Object Tracking With Delayed Server Detection Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality applications face challenges in achieving real-time object tracking due to trade-offs between computational complexity, memory requirements, and accuracy, particularly when object detection is performed on the client or server sides, leading to delayed and inaccurate tracking.
Innovation Solution
A client-server architecture is employed where object detection is performed remotely by an OD server, and object tracking is done locally on the client using asynchronous model synchronization and location updates, allowing for real-time object tracking by correcting outdated detection information with delayed updates from the server.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If object detection is performed on the client side using lightweight algorithms, then real-time performance is improved, but detection accuracy deteriorates
Solution Approach 1:
The system segments the object detection and tracking functions across different devices: the server performs accurate object detection on uploaded frames, while the client performs real-time tracking using lightweight algorithms. This division allows each component to optimize for its specific requirements without compromising overall system performance.
Solution Approach 2:
The patent introduces an intermediary mechanism where detection results from the server are transmitted to the client, and tracking information is exchanged between client and server. This intermediary communication enables the client to leverage server-based accuracy while maintaining real-time local tracking capabilities.
2Measurement precision
If object detection is performed on the server side using powerful algorithms, then detection accuracy is improved, but real-time performance deteriorates due to communication delays
Solution Approach 1:
The system performs preliminary object detection on uploaded video frames before they are fully processed for display. By detecting objects in advance on the server and transmitting the results to the client, the system prepares accurate detection data ahead of time, reducing delays during real-time tracking and display operations.
Solution Approach 2:
The patent implements dynamic adjustment of tracking parameters and model updates based on received detection information. The client adapts its tracking behavior according to the timing and accuracy of server responses, allowing flexible real-time performance despite varying communication conditions.
3Measurement precision
If accurate object detection models with large memory requirements are deployed, then detection accuracy is improved, but device complexity and memory requirements increase
Solution Approach 1:
The patent extracts the computationally intensive object detection function from the client device and relocates it to the server. This extraction removes the memory burden of large detection models from client devices while maintaining accurate detection capabilities through server-based processing.
Solution Approach 2:
Instead of deploying full detection models on client devices, the system uses simplified tracking models that copy and adapt detection results from the server. The client maintains a reduced model that leverages server-generated detection data, achieving accurate tracking with minimal local memory requirements.
Data Source
Figure 1~5
Figure 2
Figure 3~4
AI summary
An object tracking, in particular adapted for real-time augmented reality applications, involves determining a location of an object (20) in a current frame (10) of a video stream (15), at a point in time following output of a preceding frame (11) of the video stream (15) but preceding output of the current frame (10), by starting from a location of the object (20) determined by an object-detection server (5) for a previous frame (12) of the video stream (15) and recursively track the location of the object (20) in frames (11) of the video stream (15) following the previous frame (12) up to the current frame (10) and recursively update a model of the object (20). Accurate objection detection from an object-detection server (5) can thereby be used even if the object was detected in a past frame (12) of the video stream (15) that has already been visualized.