Target Track Detection Method Based on RHT Idea in Active Sonar Image Process

Abstract

This paper proposes a target track detection method based on RHT idea in active sonar image process. The method takes advantage of speed limitation of underwater maneuvering target which means the target track distributed in local space, by means of statistical similarity analysis applied to image pixels lying on a line like curve, to make sure of pixels belonging to a line and then to accumulate in parameter space by marking the same line, with the threshold detection, it is efficiently to draw target tracks and maintain tracks' survival. This method can resolve the data gaps existed in different sonar echo ping images, and it needs not randomly sample pixels in whole image like RHT, and its realization doesn't purely rely on pixels' geometrical distribution property, i.e., more emphasis has been put on statistical similarity computation of multi-dimensional target characteristics. With these changes to standard RHT, it lowers down linearity demand in curve detection necessary in STH/RHT, and it also reduces track detection alarming rate and improve the accordance of same target track, as well as small computation burden that is favorable for noisy image track detection of large amount of pixels. This technique is very suitable for mono-stationary sonar/radar track detection in multi-ping images.