基于几何特征的声呐图像人造目标检测算法  

Man-made target detection algorithm of sonar image based-on geometric feature

作　　者：邹岗[1] 田晓东[1] 刘忠[1] 

机构地区：[1]海军工程大学电子工程学院,湖北武汉430033

出　　处：《舰船科学技术》2007年第6期177-179,183,共4页Ship Science and Technology

摘　　要：针对高频声呐图像目标检测问题,基于数学形态学算子,提出了一种新的快速目标检测算法。该算法通过定义多尺度多形状的形态结构元素,对各像素点进行形态开闭运算,并采用阈值评判方法实现目标检测。仿真结果表明,该算法具有实时性好、准确度高的特点,并且可有效地克服背景中岩石等自然景物的影响,实现人造目标的准确提取。Aiming at terms of target detection of the high-frequency sonar image, based on morphological operator,a novel algorithm used to target detection was presented. It defines frame units with multi-scale and multi-shape, and open operator and close operator are adapted to each pixel in sonar image. In the end, thresholding method is used to detect man-made target. Simulation results indicate that this method has advantages such as good real-time and high precision. Synchronously, the shape feature and size feature of target are both preserved. Influence of rocks in background of sonar image is decreased effectively.

关 键 词：声呐图像 目标检测 几何特征 形态学 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]