混合高斯模型和Radon变换用于声呐图像背景抑制  

A Method for Background Suppression of Sonar Image Using Gaussian Mixture Model and Radon Transform

作　　者：范威 朱代柱 张德泽 曾赛 FAN Wei;ZHU Dai-zhu;ZHANG De-ze;ZENG Sai(Key Laboratory of Underwater Acoustic Countermeasure Technology,Shanghai 201108,China)

机构地区：[1]水声对抗技术重点实验室,上海201108

出　　处：《水下无人系统学报》2018年第5期492-497,共6页Journal of Unmanned Undersea Systems

基　　金：国家自然科学基金(61601415);水声对抗技术重点实验室基础基金(2018JCJJ-1-04)

摘　　要：水下运动小目标的声呐图像背景随着空间和时间起伏变化,声呐图像的背景抑制处理是水下小目标检测的关键技术之一。文中采用混合高斯模型建立序列声呐图像背景的各个方位-距离单元幅度的统计模型,抑制声呐图像中成片亮区、亮带干扰背景。针对混合高斯模型"剩余"噪声问题,利用水下运动目标航迹的连续变化特征,通过Radon变换和Radon逆变换提取历程累积声呐图像的线特征,滤除混合高斯模型处理后的"剩余"噪声。对实测水下小目标直线运动和跨方位曲线运动2种情况下的多波束声呐数据进行了分析,结果表明,通过混合高斯模型和Radon变换滤波方法处理,可以实现复杂声呐背景的去除,获得"干净"的声呐背景,有利于声呐对水下运动目标的自主检测。The background of sonar image of small underwater moving target fluctuates with space and time.The background suppression is one of the key processes in sonar detection.In this paper,the Gaussian mixture model is used to model the amplitude variation of each azimuth-range resolution cell of sequential sonar image,so as to suppress the main background of highlight area and highlight band in the sonar image.In order to solve the problem of residual noise from the Gaussian mixture model,the continuous change characteristics of the track from an underwater moving target are considered,the line features of the history-accumulated sonar image are extracted,and the residual noise is filtered out by Radon transform and inverse Radon transform.The analysis of multi-beam sonar data from measurement under the conditions of small underwater target’s linear motion and cross-azimuth curvilinear motion shows that the complex sonar background can be removed and the clean sonar background can be obtained by using the Gaussian mixture model and the filtering method via Radon transform,which can be used in sonar detection of underwater moving targets.

关 键 词：水下运动目标 混合高斯模型 RADON变换 声呐背景抑制 

分 类 号：TJ6[兵器科学与技术—武器系统与运用工程] TB566[交通运输工程—水声工程]