Multi-Instance Visual Tracking via Observer Motion Modelling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional visual tracking methods face challenges in accurately tracking moving objects, particularly when dealing with multiple instances and observer motion, as they often focus solely on target motion and struggle with error rates and accuracy in dynamic scenes.
Innovation Solution
A processor-implemented method for multi-instance visual tracking based on observer motion modeling, which involves extracting image frames, detecting bounding boxes, computing stereo correspondences and camera projection matrices, and using trifocal tensors and fundamental matrices to update and track objects across multiple frames, incorporating observer motion to improve tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional visual tracking methods focus solely on target motion, then the tracking process is simpler, but tracking accuracy deteriorates in dynamic scenes with observer motion
Solution Approach 1:
The patent segments the motion analysis into two independent components: target object motion and observer (camera) motion. By computing camera projection matrices separately for each frame using stereo correspondences, the method isolates observer motion effects, allowing the tracker to focus purely on target motion while compensating for camera movement through geometric transformations.
Solution Approach 2:
The patent introduces camera projection matrices and trifocal tensors as intermediary mathematical tools that mediate between raw image frames and tracking decisions. These intermediaries encode observer motion information, allowing the tracking algorithm to operate in a corrected coordinate system where target motion can be accurately measured despite camera movement.
2Reliability
If patch-based and filter-based methods are used for visual tracking, then implementation is straightforward, but error rate increases and accuracy decreases in multi-instance scenarios
Solution Approach 1:
The patent transitions from 2D patch-based tracking to 3D geometric tracking by introducing camera projection matrices and trifocal tensors. This dimensional elevation allows the system to reason about target positions in 3D space and project them accurately onto 2D image planes, resolving ambiguities that plague 2D patch-matching methods in multi-instance scenarios.
Solution Approach 2:
The patent changes the fundamental parameters of tracking from appearance-based patch similarity to geometry-based point correspondences. By using stereo correspondences and camera projection geometry as primary parameters, the method achieves more reliable tracking that is invariant to appearance changes, while the iterative refinement process maintains computational feasibility.
3Measurement precision
If iterative tracking with updated stereo correspondence and camera projection matrices is performed, then tracking accuracy improves, but computational cost increases
Solution Approach 1:
The patent performs preliminary computation of camera projection matrices and trifocal tensors using stereo correspondences before the actual tracking step. By pre-computing the geometric transformation parameters that encode observer motion, the iterative tracking loop only needs to apply these transformations and refine point correspondences, significantly reducing the computational burden of each iteration compared to re-computing all geometric parameters.
Data Source
AI summary
This disclosure relates generally to multi instance visual tracking based on observer motion modelling. Conventional state-of-the-art methods have major challenges in dealing with the error rate, accuracy of tracking the moving objects when observer motion is also considered. Embodiments of the present disclosure provide a method for multi instance visual tracking considering the motion of observer. The method tracks moving objects in an input video comprising a set of image frames, by detecting the moving objects in each image frame. The method detects the moving objects in each image frame by calculating trifocal tensors and fundamental matrix of the previous frames. The disclosed multi instance visual tracking can be used in remote surveillance problem for tracking moving objects.