Object Tracking with Adaptive Process Noise for Non-Linear Motion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking systems using linear motion models struggle with non-linear object motion, leading to loss of track or identity switches, particularly in complex scenes with noisy detections and multiple objects, due to inadequate process noise adjustment.
Innovation Solution
A dual-tracking approach is employed, combining a traditional tracking-by-detection tracker with a motion tracker, where the spatial overlap between their outputs is used to dynamically adjust the process noise of the linear motion model, allowing for higher or lower deviations based on tracking performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If process noise is increased to allow greater deviation from linear motion model, then ability to track non-linear motion is improved, but risk of erroneous associations and identity switches increases
Solution Approach 1:
The process noise is made dynamic rather than static. The system automatically adjusts the process noise level based on the detected motion characteristics. When non-linear motion is detected, process noise is increased to accommodate the deviation. When linear motion is detected, process noise is decreased to maintain tracking stability and reduce erroneous associations.
Solution Approach 2:
The system implements feedback by continuously monitoring the motion characteristics of tracked objects and using this information to adjust the process noise. The motion characteristics serve as feedback signals that trigger appropriate adjustments to the process noise level, creating a closed-loop control system that adapts to changing motion patterns.
2Stability of the object's composition
If process noise is decreased to maintain stable tracking, then tracking stability is improved, but ability to accommodate non-linear motion deteriorates
Solution Approach 1:
The process noise is made dynamic rather than static. The system automatically adjusts the process noise level based on the detected motion characteristics. When non-linear motion is detected, process noise is increased to accommodate the deviation. When linear motion is detected, process noise is decreased to maintain tracking stability and reduce erroneous associations.
Solution Approach 2:
The system changes the process noise parameter based on detected motion characteristics. By monitoring whether objects move linearly or non-linearly, the system adjusts the process noise parameter accordingly, transitioning between stable tracking mode and adaptive tracking mode to match the actual motion patterns.
3Device complexity
If linear motion model is used for object tracking, then computational simplicity is maintained, but accuracy in tracking non-linear motion deteriorates
Solution Approach 1:
The system changes the process noise parameter based on detected motion characteristics. By monitoring whether objects move linearly or non-linearly, the system adjusts the process noise parameter accordingly, transitioning between stable tracking mode and adaptive tracking mode to match the actual motion patterns.
Data Source
Figure 1~3
Figure 2
Figure 4
AI summary
There is provided a method for tracking an object in a sequence of image frames. A first tracker is used to determine a track of an object in a sequence of image frames by using a linear motion model associated with a process noise. A second tracker is used to determine a track of motion in the sequence of image frames. A spatial overlap in the image frames between the track of the object and the corresponding track of motion is monitored over time. The process noise used by the first tracker is adjusted to increase the uncertainty of the linear motion model as the spatial overlap decreases and decrease the uncertainty of the linear motion model as the spatial overlap increases.