Study On The Moving Target Detection Based On Undersea Image Sequences

Strong emphasis has been put on ocean exploration in order to fulfill the needs of Ocean Economic and military Development. Computer vision technology is an important means with which ROV sensing the environment information, motion detection play a key role in environment perception, target location and tracking. Under this background, this paper did some research on undersea moving target detection based on image sequence.When sequence images was captured, due to the camera motion, the background changed accordingly, so we must compensate the motion before detection. Global motion compensation technology are discussed firstly, and then developed an compensation algorithm based on Weighted Least Square Estimation(WLSE) method which made the difference binary image after compensation only reserve the moving target.Based on the analysis of the classical optical flow estimation algorithm, we point that the LSE method was not robust when residuals was not subject to normal distribution ,so we put forward a improved edition in order to enhance the tolerance. This paper presents a optical flow calculation method based on multi-resolution which aim at the large-motion problem. In multi-scale framework, with the resolution reduce, multiplied velocity decreased in the small velocity conditions, constant illumination conditions was easily satisfied, which made the optical flow estimation more accurate. Finally, we use a clustering algorithm based on the optical flow field model to implement the moving target segmentation.This paper finally discussed the higher order statistical approach in moving target detection. HOS is a signal characteristic of effective extraction of non-Gaussian With a total volume of high-Gaussian noise, which provided a theoretical basis for extraction of motion template. On the other hand, associated Markov random model with moving target detection, we firstly established the Markov model of template image and construct its energy cost function, In order to optimize the energy function, Iterated Conditional Modes (ICM) is used. This method can effectively suppress the underwater environmental disturbance and detect the exact target edge.