Cotton Identification And Positioning System Based On Binocular Vision

Cotton is an important economic crop in China and an essential material which is beneficial to the people's livelihood.However,the backward of picking technology restricts yield of cotton.Aiming at the disadvantages of low efficiency,poor effect and high cost of artificial cotton picking.And hand-held and backpack cotton picker also face with problem of short fiber and pure low degree,so there is a great significance to research intelligent cotton picker.Extraction of spatial location is the key of intelligent cotton cotton pickerOn the basis of researching the existing cotton planting technology and picking technology at home and abroad.First of all,building a physical model based on model of southern Akesu area "Xinluzao 50" cotton planting and establish experimental model of cotton planting mode.Setting up four degrees of freedom experimental platform with three moves and one rotation freedom to simulate working conditions of cotton picker operation in cotton field.Designing and debugging orthogonal test with four factors and four levels considering of various factors affecting the camera.Using USB binocular camera module for shooting and evaluate the definition of the image.Acquiring the best parameters of the camera installation including best high H is 95 cm,best speed V is 100mm/s,best angle theta is 30 degrees and best illumination L is 22 thousand Luxor.By using of binocular camera to shoot 7*9 target image obtain original image data.Using the Zhangzhengyou calibration method for binocular camera calibration to acquire the internal and external parameters of the binocular camera.Then,the binocular camera is installed in the best position and the experiment model of cotton was shot at the best running speed in uniform motion to obtain the original image data of cotton.Using Matlab software to pre-process two corresponding images including gray scale,image enhancement and binarization processing.Then,segmenting two valued image and extracting feature points based on SIFT to obtain 156 feature points from left image and 144 feature points from right image,and describing the feature points.Finally,matching the left and right image feature points based on SIFT.Array feature points of each region which can be successfully matched according to the order from small to large.The characteristic points of the top three are selected and the disparity of the three points are calculated.The center of gravity of the cotton is calculated by the three points.Reconstructing three dimensional of cotton based on the theory of parallel binocular vision.Extracting three-dimensional space position of the center of cotton and measuring three-dimensional space position of the center of cotton.Comparing with the three-dimensional space position and the actual measurement,we find the mean deviations in the three directions X,Y,Z are 1.8727 mm,0.5686 mm and 2.4793 mm,the error is small and the experiment is successful.In this paper,the three-dimensional space position of cotton is designed and experienced from the aspects of modeling,orthogonal experiment,image processing and 3D reconstruction.Combination of theoretical model and experimental model,comparing of extracted three dimensional spatial position of cotton and measured three dimensional space position of cotton to provide a solution for the design of the three-dimensional space position of cotton(the front-end of intelligent cotton picker).