Hai-Long Su

Millimetre‐wave road traffic surveillance radars are used for vehicle detection and tracking. Under rainy conditions, rain attenuation and rain backscatter severely degrade the vehicle detection performance of radars. In this paper, based on a big database of rain clutter collected by two radars at different rainfall levels, rain clutter modelling and the assessment of the vehicle detection ability of radars under rainy conditions are investigated. As the first contribution, the clutter‐to‐noise ratio (CNR) and signal‐to‐clutter‐noise ratio (SCNR) of the road traffic surveillance radar under rainy conditions are derived as functions of radial distance, based on the rain attenuation and backscatter models. The derived formulas highly accord with the measured CNR change of radars under rainy conditions, which can be used to evaluate radar performance and optimise the radar operating mode. As the second contribution, a range‐varying compound‐Gaussian model (CGM) with a clutter map cell partition is introduced to model rain clutter with range‐varying statistics, and a best‐type selection method of amplitude distributions is proposed. Based on the analysis of a big database of rain clutter at different rainfall levels, the range‐varying CGMs with gamma and lognormal textures are recommended to model rain clutter of millimetre‐wave road traffic surveillance radars.

Radar high-resolution range profiles (HRRPs) of ships are important in ship classification and recognition. Sparse recovery algorithms are a major tool for acquiring HRRPs from radar returns. Statistical models of ship HRRPs and sea clutter form the foundation to develop effective and efficient algorithms. In this article, ship HRRPs are modeled using biparametric lognormal distributions with heavy tails and high sparsity. Sea clutter is modeled using a compound-Gaussian model with inverse Gaussian texture (CGIG) distributions. Based on the two models, a fast sparse recovery algorithm, named the Amplitude–Position Bi-iterative Sparse Recovery Algorithm, is proposed to estimate ship HRRPs. In addition to sparsity along range cells, ship HRRPs exhibit nongrid structure, and ship scatterers are frequently not located at the centers of range cells, resulting in microposition offsets. The range-oversampled model can handle nongrid structures but requires excessive computational resources. In this context, a ship HRRP is represented by a complex amplitude vector and a real position vector. The bi-iterative algorithm is designed to alternatively optimize the two vectors. When the latter is held constant, the former is optimized using the sparse recovery through iterative minimization algorithm based on the lognormal ship HRRP model and the CGIG sea clutter model. When the former is held constant, the latter is optimized using the quasi-Newton algorithm. Simulation and measured data are employed to examine the proposed bi-iterative algorithm. The experiments demonstrate that it provides better estimates of ship HRRPs in shorter CPU time compared to the existing algorithms.

In complex scenes, it is a huge challenge to accurately detect motion-blurred, tiny, and dense objects in the thermal infrared images. To solve this problem, robust thermal infrared vehicle and pedestrian detection method is proposed in this paper. An important weight parameter β is first proposed to reconstruct the loss function of the feature selective anchor-free (FSAF) module in its online feature selection process, and the FSAF module is optimized to enhance the detection performance of motion-blurred objects. The proposal of parameter β provides an effective solution to the challenge of motion-blurred object detection. Then, the optimized anchor-free branches of the FSAF module are plugged into the YOLOv3 single-shot detector and work jointly with the anchor-based branches of the YOLOv3 detector in both training and inference, which efficiently improves the detection precision of the detector for tiny and dense objects. Experimental results show that the method proposed is superior to other typical thermal infrared vehicle and pedestrian detection algorithms due to 72.2% mean average precision (mAP).