Research On Pose Estimation Of Underwater Target Based On Vision

The accurate estimation of the underwater target pose has important engineering value for the large docking of underwater equipment and the manipulation of the end space of the underwater manipulator.The vision-based target pose estimation algorithm is more mature in aerospace equipment and terrestrial ground equipment,and has the characteristics of high positioning accuracy.Due to the particularity of the underwater environment,such as the scattering of light by water particles and the absorption and attenuation of light by water,it is difficult to extract the image feature contour acquired by the underwater camera system,making it difficult to obtain the ideal pose estimation value..In response to this problem,this paper has carried out the following work.Firstly,the mathematical model of underwater imaging is studied,and the mathematical model of undegraded image is derived based on Lambert-Beer's law.Then,the red dark channel transmissivity image in the undegraded image model is obtained by using the red dark channel prior theory,and the image is refined by the filtering algorithm to obtain the undegraded image with high contrast.Then,the CLAHE algorithm is used to further improve the contrast of the undegraded image,so that the image has higher contrast and clearer contour,which makes the extraction of the underwater target feature contour more accurate,and finally obtains the ideal pose estimation accuracy.Finally,based on the algorithm proposed in this paper,the pose estimation test of the target is carried out by using the image of the underwater target.The experimental results show that the proposed algorithm can enhance the underwater image well,and the obtained underwater target pose estimation accuracy is close to the land position estimation accuracy.