Research On Key Technology Of Vision Measurement For Manipulator Of Unmanned Underwater Vehicle

China has a long maritime border and a vast territorial sea.In order to protect the rights and interests of the territorial sea,Chinese researchers have been committed to the exploration of new marine equipment and made major breakthroughs.This thesis will rely on the national key research and development plan "ship borne unmanned underwater vehicle retraction and release system",research on the link of binocular camera installed on the manipulator to guide the manipulator to complete the recovery of the submersible.Firstly,The basic theory of binocular camera imaging and calibration is introduced.Starting from the linear and nonlinear model of pinhole imaging and the transformation relationship between coordinate systems involved in the imaging process,the general mathematical model of binocular camera is derived.The common camera calibration methods and principles are briefly described,and MATLAB is selected for calibration.The re projection error of calibration result is 0.13 pixel,which indicates that the calibration result is good.Secondly,he defogging algorithm,color based shape feature extraction and feature tracking algorithm are analyzed in detail.This thesis analyzes the dark channel defogging algorithm,changes the minimum filter to corrosion,and verifies through experiments that the improved algorithm can save running time;uses the combination of color features and shape features to improve the stability of feature extraction,uses Hough transform to fit the shape features of the lifting ring,and verifies through experiments;compares several tracking methods,and selects the best one The KCF algorithm is adopted,and it is found that it can track accurately when it moves slowly through experiments,which shows that it can meet the requirements of slow-moving when the unmanned underwater vehicle recovers.Then,This thesis describes the stereo matching of binocular camera and the solution of target pose,and uses v-rep simulation to verify the algorithm.This thesis introduces the types and main principles of the mainstream binocular stereo matching algorithms,and proposes a stereo matching method based on epipolar constraint.The center point data of the ring is extracted from the epipolar corrected image,and the three-dimensional coordinates are obtained by stereo matching.At the same time,a method of calculating the rotation by using the minimum circumscribed rectangle of the ring in the left camera image is proposed According to the relationship between the side length and the rotation angle of the minimum circumscribed rectangle,the rotation of the ring around each axis of the left camera coordinate is calculated by coordinate transformation;the algorithm is verified by v-rep simulation,and the results show that the average error rate of coordinate measurement is 0.32%,and the average error rate of rotation measurement is 0.40%,which shows that the algorithm is feasible.Finally,according to the actual application scenarios,the platform is built and different experiments are designed.The experimental results show that the working range of binocular camera is [263.7mm,1590mm];The average rotation error of the ring around the left camera coordinate axis is 2.1 °,The average error of target coordinate measurement is 4.14 mm,the average error rate is 3.1%,the average feature detection rate is 95.3%,and the success rate of manipulator grasping is 92.31%.The experiment shows that it meets the requirements.