Multi-Object Tracking Using ROI Segmentation and HMM Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection and tracking methods struggle with detecting and tracking multiple aircraft at long range, especially when objects are not discernable in every image or video frame, and existing detectors fixate on a single object, leading to low detection probability and difficulty in tracking multiple objects.
Innovation Solution
A method and system that utilize a hidden Markov model (HMM) to detect and track multiple objects by initializing an object tracker, setting detection probability to zero within the tracked region, and using a region of interest to efficiently detect and track additional objects, allowing for navigation based on predicted trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional object detector uses a hidden Markov model to track a single object, then tracking accuracy for that object is improved, but the detector cannot detect more than one object at a time and switches back to low detection probability
Solution Approach 1:
The patent divides the detection space into multiple regions of interest (ROIs), each associated with a separate object tracker. When an object is detected, an ROI is created around it and assigned to a specific tracker. This segmentation allows multiple objects to be tracked simultaneously in different regions without interference, resolving the contradiction between single-object tracking accuracy and multi-object detection capability.
2Stability of the object's composition
If the detector fixates on a single object using HMM, then tracking stability is improved, but detection probability drops when the object moves away or noise occurs
Solution Approach 1:
The patent applies different detection strategies to different regions. Within each ROI assigned to a tracker, the system uses localized detection with adjusted parameters. When an object moves out of its assigned ROI or noise is detected, the system can switch to global detection mode. This local quality approach maintains tracking stability within ROIs while preserving the ability to detect objects that move away, resolving the contradiction between tracking stability and detection reliability.
3Adaptability or versatility
If the system searches the entire image for objects, then detection coverage is improved, but computational load increases
Solution Approach 1:
The patent segments the image into multiple regions of interest rather than searching the entire image. Each ROI is processed by a dedicated object tracker, significantly reducing the computational search space. This segmentation maintains detection coverage within relevant regions while dramatically lowering computational load compared to full-image search, resolving the contradiction between detection coverage and computational complexity.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method includes receiving a first image that is captured at a first time. The method also includes detecting a location of a first object in the first image. The method also includes determining a region of interest based at least partially upon the location of the first object in the first image. The method also includes receiving a second image that is captured at a second time. The method also includes identifying the region of interest in the second image. The method also includes detecting a location of a second object in a portion of the second image that is outside of the region of interest.