Image Retrieval System Using Adaptive Tracking and Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image retrieval systems on mobile devices face challenges in efficiently tracking and detecting objects within video streams, particularly in scenarios where motion-based tracking is inadequate, leading to suboptimal user experience and resource-intensive computations.
Innovation Solution
The implementation of a method that combines motion-based tracking and object detection using a discriminator and search engine to identify digital images depicting selected objects arranged in specific geometrical arrangements, optimizing computations by selectively performing object detection based on tracking confidence values and utilizing metadata generation for improved object tracking and retrieval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If motion-based tracking is used to track objects in video streams, then tracking speed is improved, but tracking reliability deteriorates when objects are stationary or motion is insufficient
Solution Approach 1:
The system dynamically switches between motion-based tracking and object detection modes based on the tracking confidence value. When motion-based tracking confidence is high, it continues using motion-based tracking for speed. When confidence drops below a threshold, it switches to object detection for reliability, creating a dynamic adaptive system that resolves the contradiction between speed and reliability.
2Reliability
If object detection is performed continuously to improve tracking reliability, then tracking reliability is improved, but computational resources are excessively consumed
Solution Approach 1:
Instead of performing object detection continuously (excessive action), the system performs object detection partially - only when the tracking confidence value drops below a threshold. This partial action maintains sufficient tracking reliability while avoiding the excessive computational resource consumption of continuous object detection.
Solution Approach 2:
The system uses periodic action by switching between motion-based tracking and object detection based on confidence thresholds. Object detection is performed periodically when needed rather than continuously, reducing computational resource consumption while maintaining tracking reliability when necessary.
3Use of energy by moving object
If object detection is selectively performed based on tracking confidence values, then computational resources are reduced, but tracking reliability may deteriorate when detection is skipped
Solution Approach 1:
The system uses feedback by continuously monitoring the tracking confidence value and adjusting the tracking approach accordingly. When confidence drops below a threshold, feedback triggers object detection to restore reliability. This feedback mechanism ensures tracking reliability is maintained when needed while avoiding unnecessary object detection when confidence is high.
4Measurement precision
If metadata generation is used to improve object tracking accuracy, then tracking precision is improved, but device complexity increases
Solution Approach 1:
The system performs preliminary action by generating metadata in advance that describes geometrical arrangements of selected objects. This pre-generated metadata is then used to improve tracking precision during video stream processing, avoiding the need for complex real-time analysis while maintaining high tracking accuracy.
Data Source
AI summary
A method of image retrieval includes obtaining information identifying a plurality of selected objects and selecting one among a plurality of candidate geometrical arrangements. This method also includes, by at least one processor, and in response to the selecting, identifying at least one digital image, among a plurality of digital images, that depicts the plurality of selected objects arranged according to the selected candidate geometrical arrangement.