A weak target detection algorithm based on decision fusion of multi-dimensional sea echo features
DUAN Guoxing
WANG Yunhua
ZHANG Yanmin
LYU Letian
Abstract：The fusion detection using the multi-dimensional features of sea surface scattered echoes is one of the effective ways to solve weak target detection problem.Based on the differences of X-band ra-dar sea surface echo features such as the time domain,frequency domain and time-frequency domain in target detection performance,in order to improve the target detection accuracy and achieve the seconda-ry detection of sea surface targets by using the base detector and fusion detector,this paper proposed an open target detection algorithm framework for the multi-dimensional feature decision fusion of sea sur-face echoes.Firstly,this paper converted the maritime target detection problem into a feature-based bi-nary classification problem and utilized the feature quantities in the time domain,frequency domain and time-frequency domain of the radar echo data for target detection based on the K-Nearest Neighbor(KNN)algorithm,respectively.A multi-dimensional decision probability space was constructed and analyzed based on the multiple detection probabilities of different echo sample data.Then,the KNN was optimized by calculating the target and sea clutter classification thresholds with different false a-larm probabilities(Pfa)in order to achieve controllable false alarm of maritime target detection.Fur-ther,the optimized KNN was used as a fusion detector to obtain a multi-feature decision fusion detec-tion algorithm.Finally,the experiments on McMaster's IPIX radar real measurement dataset showed that the detection algorithm in this paper could obtain a detection probability of 0.87 when the signal to clutter ratio(SCR)was 0 dB and the Pfa was 10-3.The detection performance was better than other feature-based detection algorithms.It could better satisfy the target detection requirements under the radar short-time observation and low SCR sea conditions.
Keywords：sea clutterweak small targetsK Nearest Neighbor(K-NN)multi-dimensional decision probability spacedecision fusion detection
Publication Date：2025-12-30
Online Publishing Date：2026-05-22(First online date of this platform, not the publication date of the document)
Pages：9( 157-165 )