Multi-Task AI Object Tracking Through Occlusion and Orientation Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking technologies face challenges in accurately tracking multiple objects due to occlusions and changes in object orientation, leading to difficulties in maintaining identification and location accuracy.
Innovation Solution
An electronic device employs a multi-task artificial intelligence model with a backbone network, detection head, identification head, and body orientation head to extract feature maps, location information, identification features, and orientation angles, enabling precise tracking of objects by comparing these features with previously tracked objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing object tracking technology is used, then basic tracking functionality is achieved, but tracking accuracy deteriorates due to occlusions and orientation changes
Solution Approach 1:
The tracking system is divided into multiple specialized components: a detection head for location detection, an identification head for feature extraction, and a body orientation head for orientation estimation. This segmentation allows each component to specialize in one aspect, improving overall tracking accuracy and reliability under occlusions and orientation changes.
Solution Approach 2:
The system extracts and utilizes body orientation angles as an additional parameter to maintain reliable identification. By incorporating orientation information alongside location and identification features, the system can distinguish objects even when partially occluded or viewed from different angles, thereby maintaining identification reliability under varying conditions.
2Measurement precision
If multi-task AI model is employed, then tracking accuracy is improved, but device complexity increases
Solution Approach 1:
A single multi-task AI model is designed to perform multiple functions simultaneously: object detection, identification feature extraction, and body orientation estimation. This universal model processes input images through shared backbone networks and specialized heads, achieving accurate tracking without requiring separate systems for each function, thus managing complexity while improving precision.
3Measurement precision
If location and identification features are extracted separately, then processing efficiency is reduced, but extraction precision is improved
Solution Approach 1:
The detection head, identification head, and body orientation head are merged into a single integrated AI model that processes images simultaneously. By combining these extraction functions in one model rather than processing them separately, the system achieves both high extraction precision for location and identification features while maintaining efficient processing throughput.
Data Source
AI summary
A method of tracking at least one object, includes: obtaining an image; extracting, from the image, a feature map for performing a plurality of tasks related to object tracking; extracting, by using the extracted feature map, location information indicating a location of the at least one object, an identification feature for identifying the at least one object, and a body orientation angle at which a body of the at least one object is oriented; and tracking the at least one object, based on the location information, the identification feature, and the body orientation angle of the at least one object.