Research On System Of Binocular Vision Weed Locating Based On Corner Feature

In the traditional agriculture,the way of using herbicide is the well-distributed spraying.The method not only increase the cost of herbicide but also damage the quality of field and pollute air.For this,"fine agriculture",this concept is advanced at home and abroad.According to the weed's distributing aera to spray herbicide,the using of herbicide is evidently reduced.Now the main research aims at recognition of weed.The insuffiation system of weed cover the area of crop and weed by many spray nozzles,but the precondition is to measure position of weed.Therefore,technique premise of achieving fine agricultrue is measure position of weed by the binocular vision system.To obtain the position of weed,basing on the binocular vision system,the paper study the weed orientation technology in the static state.The main contents are as follows:For chessboard comer detection,the author propose a new algorithm for chessboard comer detection based on USAN according to sysmmetrical geometry structure of USAN area,the theory of algorithm is simple and it can detect comer not only quickly,but also the invariability of revolving transformation.for SUSAN comer detection,The authors point out the limitations of SUSAN corner detector in detecting X,Y,T style comer,describe an improved SUSAN corner detector alogrithm based on connectivity.The experimental result indicate that the improved alogrithm can detect X,Y,T style comer effectively and accurately,it is also robust to image noise.In addition,the paper propose to check its validity according to cross ratio invariability in perspective projection.Camera calibration method of the binoular vision system is studied in this paper. Aiming at the distrortion of camera lens,one rank radial model is builded.According to cross ratio inbariability,distrortion coefficient is obtained.Images gathered by the camera are corrected through the coefficient,and then a method baseing on the plane template is using to obtain inside and outside of camera.the camera calibration method is very simple and has a high precision.for the comer macthing,correlation coefficient is using to make first match and compute foundation matrix through points obtained in the process of first matching,then,according to fundamental matrix to optimize last step reslut.on the based of inside and outside parameters of camera obtained by calibration and corresponding point's matching result,position of weed can be obtained.The paper study the weed orientation technique based on binocular vision system,which will provide the theoretical support,The results of research are definite instruction to development of the automatic killed-weed system based on machine vision.