Research on subsea small target recognition based on multiscale feature matching
A multi-scale matching sonar image recognition method is proposed to address the characteristics of rever-beration background,blurred target edges,and the presence of sound shadows in unmanned underwater vehicles.Normalize the sonar image to obtain a grayscale image of the sonar echo intensity.Gaussian low-pass filtering was applied to enhance the contour features of the strong echo area.Adaptive threshold method was adapted to segment images and form relatively independent connected domains.Finally,a multi-scale matching model for target recognition in sonar images is proposed based on the multi-scale characteristics such as size,echo intensity,and sound shadow intensity of typical cylindrical targets in sonar images.The method was applied on different sonar images,and the results showed that the multi-scale matching sonar image recognition method has good recognition effect on typical small targets on the seabed.This providing an effect-ive method for automatic target recognition of underwater robots.
subsea small targetbottom reverberationacoustic shadowmultiscale matching
NETL
NSTL
万方数据
