Target Detection for RD Images of HFSWR Based on CNN-ELM Model
Maokai Wu, Jiong Niu, Ling Zhang, Q. M. Jonathan Wu, Chenlu Shi, Jingzhi Zhang
Abstract
High-frequency surface wave radar (HFSWR) can effectively detect ship targets. However, ship target signals are often affected by strong clutter and complex interference. In this paper, we propose an HFSWR target detection algorithm based on a two-stage cascade detector combined with a convolutional neural network-extreme learning machine (CNN-ELM) model. In the first stage, an extremum detector (ED) is used to obtain suspicious target regions (STRs) in the range-Doppler (RD) spectrum image. In the second stage, a CNN-ELM model is employed. The features of the STRs are learned by a lightweight convolutional neural network (LW-CNN) and fast classification is performed by an extreme learning machine (ELM). The experiments show that the proposed algorithm can achieve better performance in the measured RD image than the traditional detection algorithms.