HIGH-RESOLUTION RADAR GROUND CLUTTER GENERATION BY DENOISING DIFFUSION PROBABILITY MODEL

Environmental clutter significantly impacts the performance of High-Resolution Range Profiles (HRRP) in detection and recognition. Hence, it's crucial to model and simulate radar clutter for effective radar system design. However, traditional methods are limited to fixed probability distributions, struggle to accurately replicate clutter in complex real-world scenarios. In this paper, we propose to generate clutter based on the Denoising Diffusion Probabilistic Model (DDPM). DDPM can learn the distribution of training data through its probabilistic modelling method, solving the limitations brought by traditional methods. In addition, in order to evaluate the amplitude characteristics of generated clutter, we propose to compare the fitting results of specific physical probability distributions of generated clutter with those of real clutter, as well as evaluate the distribution distance with real clutter. Experimental results on measured clutter data show that the clutter synthesized by DDPM can well match the distribution of the measured clutter. And it is able to demonstrate good performance in different scenarios.