Multi-Object Tracking with Overlap-Aware Kalman Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking technologies face challenges in accurately tracking multiple detection targets due to overlapping images, leading to tracking errors and reduced accuracy, especially when using Kalman filters.
Innovation Solution
The object tracking device employs a processor that uses a Kalman filter for each detection target, allowing overlapping and linking of detection results, and manages observation values, Kalman filters, and unique identification information in a hierarchical manner to improve tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional tracking techniques are used for multiple detection targets, then the tracking process is simpler, but tracking accuracy deteriorates due to overlapping detection results causing misassociations
Solution Approach 1:
The patent applies segmentation by dividing the tracking process into distinct stages: detection result generation, overlap identification, and selective association. Each detection target is processed independently with its own Kalman filter instance, and overlapping regions are segmented and handled separately to prevent misassociations between multiple targets
Solution Approach 2:
The patent introduces an intermediary mechanism in the form of a data association module that mediates between multiple detection results and tracking targets. This intermediary layer resolves overlaps by evaluating measurement residuals and making intelligent association decisions, preventing direct conflicts between overlapping detection results
2Reliability
If multiple sensor data sources are integrated for tracking, then tracking reliability improves, but computational load increases
Solution Approach 1:
The patent merges multiple sensor data sources (camera, radar, LIDAR) into a unified tracking framework where detection results from different sensors are integrated. The Kalman filter combines measurements from multiple sources to produce more reliable tracking estimates, improving robustness through sensor fusion while managing computational complexity through efficient data association algorithms
3Adaptability or versatility
If detection results are allowed to overlap during tracking, then tracking of multiple targets becomes possible, but misassociations increase with conventional methods
Solution Approach 1:
The patent changes the parameters used for data association by incorporating measurement residuals and probability-based metrics into the association process. Instead of simple distance-based association that fails with overlapping targets, the system uses modified association parameters that account for target dynamics, measurement uncertainty, and spatial-temporal relationships to maintain high association accuracy in overlapping scenarios
Data Source
AI summary
Provided are an object tracking device and an object tracking method that allow multiple objects to be tracked with high accuracy. An object tracking device (20) includes an input interface (21), a processor (23), and an output interface (24). The input interface (21) is configured to acquire sensor data. The processor (23) is configured to detect multiple detection targets from the sensor data and perform tracking using a Kalman filter for each of the multiple detection targets. The output interface (24) is configured to output detection results of the detection targets. The processor (23) allows overlapping of detection results during the process of tracking the multiple detection targets.