Video Target Tracking with Motion-Constrained Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video target tracking methods struggle with accurately recognizing targets in videos when there are significant appearance differences between frames due to changes in the target's appearance information, leading to inaccurate prediction results.
Innovation Solution
A method that combines appearance and motion information to adjust the parameters of an image segmentation model using constraint information, including absolute positive and negative sample pixels, to enhance the accuracy of target object segmentation and tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an image segmentation model is adjusted using only appearance information from the first image frame, then the model can be adapted to the target object, but the recognition accuracy deteriorates when large appearance differences occur in subsequent frames
Solution Approach 1:
The patent combines appearance information and motion information to form constraint information for adjusting the image segmentation model. Appearance information comes from the first image frame while motion information is derived from optical flow between the first and second image frames. By merging these two types of information, the model can adapt to appearance changes while maintaining recognition accuracy through motion constraints.
Solution Approach 2:
The patent changes the parameters used for model adjustment from solely appearance-based features to a composite of appearance and motion features. The constraint information includes both appearance constraints from the target object in the first frame and motion constraints from the optical flow, allowing the model parameters to be adjusted in a way that accounts for both static and dynamic characteristics of the target.
2Reliability
If the model parameters are adjusted based on the first image frame, then the model becomes specific to that frame's target, but error propagation increases when the target appearance changes significantly
Solution Approach 1:
The patent introduces a feedback mechanism where the adjusted image segmentation model from the first frame is applied to the second image frame, and the results are used to further refine the model. The constraint information is continuously updated by combining appearance information from the current frame with motion information from optical flow, creating a feedback loop that corrects errors and adapts to appearance changes.
Solution Approach 2:
The patent performs preliminary adjustment of the image segmentation model using constraint information from the first image frame before applying it to subsequent frames. This preliminary action prepares the model to handle appearance variations, and the constraint information is further refined using motion information to prevent error propagation when significant appearance changes occur.
3Ease of operation
If only appearance information is used for target tracking, then the method is simple to implement, but tracking accuracy deteriorates under large appearance variations
Solution Approach 1:
The patent introduces motion information from optical flow as an intermediary element that bridges appearance information and tracking accuracy. The optical flow provides motion constraints that mediate between the appearance-based segmentation and the actual target position, improving tracking accuracy under appearance variations without significantly complicating the implementation, as optical flow computation is a standard technique.
Data Source
Figure 1a
Figure 1b(a)~1b(d)
Figure 2~3
AI summary
A video target tracking method and apparatus, a computer device and a storage medium are provided. The method comprises: acquiring a local detection map corresponding to a target image frame in a video to be detected (201); acquiring a relative motion saliency map corresponding to the target image frame (202); determining, according to the local detection map and the relative motion saliency map, constraint information corresponding to the target image frame (203); adjusting parameters of an image segmentation model by means of the constraint information, to obtain an adjusted image segmentation model (204); and extracting a target object from the target image frame by means of the adjusted image segmentation model (205).