Study On The Method Of Movement State Sensing Of Small Underwater Vehicle Based On Optical Vision

With the in-depth exploration of the ocean and the continuous development of technology,the need for improving the intelligence of underwater robots is more and more urgent.During underwater operations,sensing the state of movement is a necessary condition for the robot to navigate and control its own motion,and it is also an important indicator of the robot's intelligence.Therefore,the research on the estimating method of movement state has a practical significance for small underwater vehicles.The main work of this article is as follows:(1)Data expansion methods such as cropping are used to solve the problem of data imbalance in existing data set.According to the speed and accuracy,the YOLOv3 is selected and retrained to achieve the classification and positioning of marine organisms.Aiming at the problem that the recall rate is not high,the soft-NMS is introduced to improve the recall rate of marine organism.(2)By comparing six underwater image enhancement methods,the histogram equalization method is selected,which greatly enhances the image in a short time.Aiming at inaccurate matching points during feature point matching,the RANSAC is introduced to effectively reduce inaccurate matching.An object comparison strategy is proposed,which comprehensively uses the number of object matching points,object categories and object positions to accurately determine the possibility that objects in different images are the same object in reality.(3)A method for sensing the movement state of small underwater vehicles based on key object libraries is proposed.The method performs 3D positioning on targets,and then obtains the movement state based on the 3D information.Then box plot analysis and five-point cubic filtering are used to obtain the final movement state.A key object library with distributed reliability is proposed to solve the problem of instability of movement state sensing by relying on two adjacent frames and effectively improve the stability of movement state sensing.(4)The movement state sensing experimental platform is set up to carry out experimental research on the movement state of small underwater vehicles.Design different speeds,different terrain movement state sensing experiments and comparison experiments with an accelerometer for movement state sensing,verify the feasibility,adaptability,and superiority of the movement state sensing method proposed in this paper.