Research And Application Of Binocular Stereo Vision Ship Ranging Technology Based On Feature Points

With the development of intelligent equipment and related technologies,more and more attention has been paid to the intelligent and unmanned technology of surface equipment.Water surface environment perception is one of the key directions of ship intelligent research.The acquisition of ship distance information is conducive to ship obstacle avoidance,unmanned ship autopilot,ship path planning and so on.Compared with radar equipment,the vision measurement scheme based on binocular ranging system with low price and high accuracy has received more and more attention.This paper designs and implements a set of surface ship binocular ranging system based on feature points.Through binocular calibration,stereo correction,feature point extraction,sparse stereo matching and other technologies,the ship distance measurement in water environment is realized.Aiming at the problem of binocular ranging in water environment,this paper mainly studies and improves the feature point extraction and sparse stereo matching.In this paper,the chessboard calibration method and MATLAB Calibration Toolbox are used to complete the calibration of the binocular ranging system,and the internal and external parameters of the binocular ranging system are obtained.The image is stereo corrected by Bouguet stereo correction algorithm.Aiming at the problem of feature point extraction in binocular ranging system,an improved feature point extraction scheme is proposed in this paper.Before feature points are extracted,median filtering algorithm is used to preprocess the image to reduce the influence of noise.By analyzing different feature points extraction algorithms,brisk algorithm is selected to extract feature points of ship targets in water environment.Aiming at the problem of more interference feature points and feature points aggregation in surface ship image,a feature point extraction scheme is designed.Firstly,some interference feature points in the fixed background are removed by using continuous images.Then,by using the image edge detection algorithm,the feature points in the range target contour region are retained and the invalid feature points in other regions are suppressed to reduce the number of invalid feature points.Finally,aiming at the problem of poor clustering and distribution of feature points,the improved adaptive non maximum suppression algorithm is used to improve the distribution of feature points.In order to improve the accuracy of sparse stereo matching,this paper divides sparse stereo matching into two stages: rough matching and fine matching.In rough matching stage,this paper improves GMS feature matching algorithm based on grid motion statistics using stereo matching constraints,and completes rough matching with RANSAC algorithm.In the precision matching stage,a disparity optimization scheme based on ZNCC algorithm is designed and the disparity information at sub-pixel level is obtained by curve fitting,which further improves the calculation accuracy of disparity.Finally,the ranging experiment of unmanned surface craft is carried out in this paper.The disparity is calculated by using the improved feature point extraction scheme and the improved sparse stereo matching algorithm,and the distance information of unmanned surface craft is calculated by the parameters of binocular distance measuring system.In this paper,GPS equipment is used to obtain the true distance between unmanned surface ship and camera.By comparing the measured value with the true value,the measurement accuracy and error of the binocular distance measuring system in this paper are analyzed.