Object Tracking via 3D Feature Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking methods in videos face challenges in accurately capturing the location and dynamic features of target objects due to limitations in processing 3D feature blocks and integrating spatial and temporal information effectively.
Innovation Solution
A method and system that extract 3D feature blocks from videos, decompose them into 2D spatial and spatial-temporal feature maps, estimate object location, determine speed and acceleration, and calibrate the location using CNNs, enhancing tracking accuracy by processing spatial and temporal information separately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D feature blocks are processed directly to extract location and dynamic features, then tracking information can be obtained, but the accuracy of location estimation and dynamic feature detection is insufficient
Solution Approach 1:
The patent segments the 3D feature block processing into two distinct 2D feature maps: a spatial feature map for location estimation and a spatial-temporal feature map for dynamic feature detection. This segmentation allows each map to specialize in specific features, improving both location estimation accuracy and tracking reliability by preventing feature interference and enabling optimized processing for each feature type.
2Device complexity
If spatial and temporal information are integrated in a single feature map, then comprehensive object information is obtained, but the processing complexity and difficulty of detecting location and dynamic features increase
Solution Approach 1:
The patent divides the integrated spatial-temporal information into separate 2D feature maps - one dedicated to spatial information and another to spatial-temporal information. This segmentation reduces processing complexity by allowing each map to be handled independently with specialized algorithms, while also making feature detection easier by isolating spatial and temporal characteristics into distinct processing streams.
Solution Approach 2:
The patent transforms the 3D feature block into 2D feature maps, changing the dimensional representation. This dimensionality reduction simplifies the processing complexity while maintaining the essential spatial and temporal information needed for accurate feature detection, making the system more computationally efficient without sacrificing detection capability.
3Device complexity
If only location information is estimated without dynamic features, then the tracking system is simpler, but the accuracy of motion prediction and tracking deteriorates
Solution Approach 1:
The patent creates separate processing pathways: one for spatial feature maps that estimate location and another for spatial-temporal feature maps that detect dynamic features like speed and acceleration. This segmentation allows the system to maintain relatively simple individual processing modules while achieving high tracking accuracy through the combination of location and dynamic feature information.
Data Source
AI summary
The application relates to a method and system for tracking a target object in a video. The method includes: extracting, from the video, a 3-dimension (3D) feature block containing the target object; decomposing the extracted 3D feature block into a 2-dimension (2D) spatial feature map containing spatial information of the target object and a 2D spatial-temporal feature map containing spatial-temporal information of the target object; estimating, in the 2D spatial feature map, a location of the target object; determining, in the 2D spatial-temporal feature map, a speed and an acceleration of the target object; calibrating the estimated location of the target object according to the determined speed and acceleration; and tracking the target object in the video according to the calibrated location.