Abstract: Early-warning and recognition of torpedoes, which needs a comprehensive application of different kinds of recognition methods, is a complex process. The key point of the process is to extract the features of torpedo radiated noise. Considering the difference between torpedoes and non-torpedoes in the spectrum of radiated noise, the researches on image edge detection and feature extraction have been done. The characteristic quantity of spectrum of vehicle radiated noise is definition. A method, based on image edge detection and feature extraction, is put forward to get the characteristic quantity of torpedoes and applied to researches of recognition of torpedoes. The experimental results show that the method, which can extract the feature of spectrum well, can provide a dependable characteristic to recognition and early-warning of torpedoes.

Key words: edge detection, spectrum of radiated noise, feature extraction, recognition

摘要： 鱼雷的识别预警是一个复杂的特征信息融合过程，其关键是提取鱼雷目标的特征成分。鱼雷目标与非鱼雷目标辐射噪声声谱图的存在差异性，针对声谱图特征提取方法开展研究，通过定义目标辐射噪声的声谱图直纹特征量，提出了一种基于图像边缘检测理论的鱼雷目标辐射噪声声谱图特征提取方法，并将该方法应用于鱼雷目标分类识别研究。实测信号分析表明，该方法可以有效提取鱼雷目标辐射噪声的声谱图特征，将鱼雷目标与非鱼雷目标进行区分，为鱼雷目标的综合识别预警提供一种可靠的特征信息参考。

关键词: 边缘检测, 声谱图, 特征提取, 分类识别