Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Gamma-Based Statistical Modeling for Extended Target Detection in mmWave Automotive Radar

Vinay Kulkarni, V. V. Reddy

TL;DR

The paper tackles extended-target detection in mmWave automotive radar by moving from cell-wise CFAR to a Range-Doppler segment approach that preserves the target's scattering spread. It introduces a Gamma-distribution-based statistical model for RD segments, estimated via MLE and Gibbs sampling, and defines a skewness-based test statistic that depends only on the Gamma shape parameter to enable real-time binary decisions. An end-to-end pipeline slides RD segments across the map, centers peaks, and uses IoU to merge overlaps, achieving reduced redundancy and single-dwell operation. Empirical results on simulated and real data show strong detection performance with low false alarms compared to OS-CFAR, demonstrating practical viability for automotive radar systems. The work provides a statistically grounded, low-complexity framework that improves extended-target detection by exploiting scattering structure and robust moment-based discrimination.

Abstract

Millimeter-wave (mmWave) radar systems, owing to their large bandwidth, provide fine range resolution that enables the observation of multiple scatterers originating from a single automotive target, commonly referred to as an extended target. Conventional CFAR-based detection algorithms typically treat these scatterers as independent detections, thereby discarding the spatial scattering structure intrinsic to the target. To preserve this scattering spread, this paper proposes a Range-Doppler (RD) segment framework designed to encapsulate the typical scattering profile of an automobile. The statistical characterization of the segment is performed using Maximum Likelihood Estimation (MLE) and posterior density modeling based on the Gamma distribution, facilitated through Gibbs Markov Chain Monte Carlo (MCMC) sampling. A skewness-based test statistic, derived from the estimated statistical model, is introduced for binary hypothesis classification of extended targets. Additionally, the paper presents a detection pipeline that incorporates Intersection over Union (IoU) and segment centering based on peak response, optimized to work within a single dwell. Extensive evaluations using both simulated and real-world datasets demonstrate the effectiveness of the proposed approach, underscoring its suitability for automotive radar applications through improved detection accuracy.

Gamma-Based Statistical Modeling for Extended Target Detection in mmWave Automotive Radar

TL;DR

The paper tackles extended-target detection in mmWave automotive radar by moving from cell-wise CFAR to a Range-Doppler segment approach that preserves the target's scattering spread. It introduces a Gamma-distribution-based statistical model for RD segments, estimated via MLE and Gibbs sampling, and defines a skewness-based test statistic that depends only on the Gamma shape parameter to enable real-time binary decisions. An end-to-end pipeline slides RD segments across the map, centers peaks, and uses IoU to merge overlaps, achieving reduced redundancy and single-dwell operation. Empirical results on simulated and real data show strong detection performance with low false alarms compared to OS-CFAR, demonstrating practical viability for automotive radar systems. The work provides a statistically grounded, low-complexity framework that improves extended-target detection by exploiting scattering structure and robust moment-based discrimination.

Abstract

Millimeter-wave (mmWave) radar systems, owing to their large bandwidth, provide fine range resolution that enables the observation of multiple scatterers originating from a single automotive target, commonly referred to as an extended target. Conventional CFAR-based detection algorithms typically treat these scatterers as independent detections, thereby discarding the spatial scattering structure intrinsic to the target. To preserve this scattering spread, this paper proposes a Range-Doppler (RD) segment framework designed to encapsulate the typical scattering profile of an automobile. The statistical characterization of the segment is performed using Maximum Likelihood Estimation (MLE) and posterior density modeling based on the Gamma distribution, facilitated through Gibbs Markov Chain Monte Carlo (MCMC) sampling. A skewness-based test statistic, derived from the estimated statistical model, is introduced for binary hypothesis classification of extended targets. Additionally, the paper presents a detection pipeline that incorporates Intersection over Union (IoU) and segment centering based on peak response, optimized to work within a single dwell. Extensive evaluations using both simulated and real-world datasets demonstrate the effectiveness of the proposed approach, underscoring its suitability for automotive radar applications through improved detection accuracy.

Paper Structure

This paper contains 19 sections, 29 equations, 12 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Problem Statement
  3. Proposed Approach
  4. Binary Hypothesis Construction for an RD segment
  5. Noise and Target response distributions
  6. Test statistic
  7. Parameter estimation
  8. Likelihood Formulation for Gamma Mixture Model
  9. Bayesian inference using MCMC Gibbs sampling
  10. Statistical third-order moment function
  11. RSP detection pipeline
  12. Data Synthesis
  13. Results and Analysis
  14. Optimization results
  15. Performance analysis of skewness-based detection within the pipeline
  16. ...and 4 more sections

Figures (12)

  • Figure 1: Experimental data for the vehicular target is shown in (b), with its corresponding RD map in (a).
  • Figure 2: Illustration of localized RD segment in RD map.
  • Figure 3: Histograms for 10500 samples: (a) Alternate hypothesis, (b) Null hypothesis, (c) envelopes for the two hypothesis.
  • Figure 4: The proposed detection pipeline for RD segment-based detection.
  • Figure 5: MLE for 2 components of Gamma mixtures: (a) NLL optimization with $W_2$ colorbar ; Contour plots of the parameter surfaces for (b) $\text{Gamma}_1(W_1,\alpha_1,\beta_1)$ and (c) $\text{Gamma}_2(W_2,\alpha_2,\beta_2)$, shown with a shared NLL colorbar.
  • ...and 7 more figures