Research On Relative Pose Estimation For UUV Docking Based On Vision Guided Method

The demand and competition of marine resources is increasing as the continuous expansion of economic globalization and human activities.UUV(Unmanned Underwater Vehicle,UUV)is playing a huge role whether in the use of military or civilian.With the demand for the development of the national economy and national defense construction,better performance of UUV get more attention than before.The safe and recovery of UUV is the key factor in its ability of endurance and underwater combat capabilities.The technique of relative pose estimation is the key to UUV underwater close-end navigation and positioning.Vision,not only provide lavish information,but also accuracy of close positioning has been widely used on land.But the underwater environment is more complicated and often accompanied by severe noise,bringing much more difficulties for underwater vision applications.In this situation,underwater vision technology is more of a meaningful research.This paper mainly studies the following aspects in the background of UUV underwater recovery,noise underwater and the presence of the pseudo-light shooting conditions,also the non-ideal shooting conditions leads to low accuracy.First of all,by means of pretreatment of captured underwater light images provides conditions for the smooth progress of the following research work.By reasoning of underwater image degradation,using the scattering model based on the recovery of underwater images,verify the feasibility of the simulation method by simulation comparison test.Weaken the noise,while highlighting the image features to improve the quality of underwater target source image through a comparative analysis of Gaussian filtering and Wallis underwater light filtering target image smoothing and Enhancement act.Secondly,considering the target light underwater image feature extraction and pseudo problems of light,use a combination of Canny detection and Snake model to improve the accuracy of edge detection.For a special heart-shaped light source,use the quantum particle swarm optimization algorithm contour matching to improve real-time matching.For pseudo-light interference underwater,using Blob analysis and Bayesian decision to remove the false light,improve the efficiency of image processing,laying the foundation for subsequent pose estimation.Thirdly,according to the UUV recycling program located in the dock space to design monocular vision guidance system and complete the build of the target light source coordinate system,using the heart-shaped target sources to estimate the relative heading angle,and using the common light source estimated vertical,horizontal and vertical relative position.The use of virtual camera rotation method improved relative pose estimation algorithm and the accuracy and applicability of the method,particularly for the actual recovery over the non-ideal conditions of trim or heel may exist in UUV movement.