基于几何特征的声呐图像人造目标检测算法

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.

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

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.