A New Multiple Hypothesis Tracker Using Validation Gate with Motion Direction Constraint

Jinping Sun, Ziwei Wang, Qing Li2020 · DOI

TL;DR

This paper introduces a novel Multiple Hypothesis Tracker (MHT) called MHT-MDC that improves multi-target tracking in dense and cluttered environments. By incorporating a motion direction constraint into the validation gate, it significantly reduces false measurements, leading to more accurate and efficient tracking compared to traditional MHT methods.

What Problem Does This Paper Address?

Multi-target tracking in environments with dense and heterogeneous clutter presents a significant challenge. Clutter, which consists of false measurements not originating from actual targets, can overwhelm traditional tracking algorithms. Specifically, the Multiple Hypothesis Tracker (MHT), while theoretically optimal, suffers from a rapid growth in the number of potential measurement-to-track association hypotheses as clutter increases. This leads to a substantial computational burden and a decrease in data association accuracy and overall tracking performance. Existing validation gate techniques, while helpful, struggle to balance the need to capture true target measurements with the need to exclude clutter, especially when the uncertainty in target measurements (innovation covariance) is high.

Methods

The core innovation of this paper is the Multiple Hypothesis Tracker with Motion Direction Constraint (MHT-MDC). It builds upon the track-oriented MHT (TOMHT) framework. The key enhancements are:

  • Motion Direction Constraint (MDC) Gate: Instead of relying solely on traditional elliptical validation gates, MHT-MDC introduces an MDC gate. This gate incorporates prior target maneuvering information to constrain the validation region. It considers the predicted target motion direction and only accepts measurements that align with this direction within a specified threshold ($\theta_G$). This effectively reduces the volume of the validation gate, especially when the innovation covariance is large, thereby diminishing the number of false measurements.
  • Adaptive Clutter Density Estimation: For heterogeneous clutter scenarios, MHT-MDC employs a Conditional Mean Estimator of Clutter Density (CMECD). This method adaptively estimates the clutter density within the MDC gate, allowing for more accurate calculation of track hypothesis score functions.
  • Improved Score Function: The paper leverages the adaptively estimated clutter density to compute the track score function. This is crucial in dense clutter environments where a constant clutter density assumption can lead to inaccurate scoring and poor data association.
  • Approximate MDC Gate Volume Calculation: Since the MDC gate's volume doesn't have a simple analytical solution, the paper proposes an approximation method based on truncated cone volumes, which is shown to be effective.

Datasets

Not discussed in the paper. The experiments were conducted using simulated scenarios.

Key Insights and Ideas

  • Prior Motion Information is Powerful: The paper demonstrates that incorporating prior knowledge about target motion (even general assumptions about low maneuverability) can dramatically improve tracking performance in cluttered environments.
  • Constraining the Gate is Key: By reducing the volume of the validation gate using motion direction constraints, MHT-MDC effectively filters out a significant portion of clutter, which is the primary source of errors in dense scenarios.
  • Adaptivity is Crucial for Heterogeneous Clutter: The CMECD for clutter density estimation is vital for handling varying clutter distributions, preventing performance degradation that would occur with a static clutter model.
  • MHT-MDC Balances Accuracy and Efficiency: The proposed method not only improves tracking accuracy but also reduces computational complexity by pruning false tracks earlier.

Key Results

The MHT-MDC algorithm was compared against the traditional MHT algorithm in two simulated scenarios with dense and heterogeneous clutter. The results consistently showed superior performance for MHT-MDC:

  • Scenario 1: MHT-MDC achieved a significantly lower fragmentary ratio ($R_{FT}$) (0.11 vs. 0.357), a much lower miscorrelation ratio ($R_{MC}$) (0.017 vs. 0.074), and a higher correct correlation ratio ($R_{CC}$) (0.972 vs. 0.917) compared to MHT. It also had a lower average processing time ($T_H$) (0.084 s vs. 0.134 s).
  • Scenario 2: Similar improvements were observed, with $R_{FT}$ dropping from 0.731 to 0.086, $R_{MC}$ from 0.206 to 0.049, and $R_{CC}$ increasing from 0.612 to 0.93. The average OSPA distances were also significantly lower for MHT-MDC.
  • Qualitative Results: Figures show that MHT-MDC produces cleaner tracks with fewer false tracks and less mutual interference.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the MDC gate volume calculation is an approximation.
  • The performance of the MDC gate is dependent on the accuracy of the prior target maneuvering information and the chosen motion direction threshold ($\theta_G$).
  • Future Work:
    • The authors suggest combining MHT-MDC with other prior information-aided tracking algorithms.
  • Incorporating estimated target trajectory information (e.g., continuous-time trajectory information, estimated target velocity, and location) could further enhance performance.

Summary updated

BibTeX
@misc{sun2020new,
  author = {Jinping Sun and Ziwei Wang and Qing Li},
  title = {{A New Multiple Hypothesis Tracker Using Validation Gate with Motion Direction Constraint}},
  year = {2020},
  doi = {10.3390/s20174816},
}
Figure