Research On Location Of Fruit Leaf Diseases And Pests And Spray Uniformity Detection Technology Based On Binocular Vision

In order to ensure that the mechanized pesticide application under the high spindle planting mode does not cause drug waste,uniform pesticide application and precise pesticide application,the situation of pesticide spraying and the severity of pests and diseases are detected,and the spraying machinery is precisely controlled.The progress of binocular vision technology provides a good technical means for this kind of detection and control.In this paper,how to locate pests and diseases by binocular vision under mechanized spraying conditions and detect the uniformity of pesticide spraying after spraying pesticides were studied.This study will help to improve the automatic application in high spindle orchards,reduce labor intensity,improve the effective utilization of pesticides,reduce the occurrence of pests and diseases,and promote higher quality output of agricultural products.The automatic spraying method of the fruit tree pesticide is: the tractor pulls the medicine barrel,and there are three fans on the left and right sides,and the parallel binocular camera detects the pests and diseases and distinguishes the uniformity of the spraying pesticides.Based on the existing mechanical devices,contents of the study:1.A binocular vision pixel matching method without obvious feature regions is proposed.In the binocular matching of this topic,it is often necessary to match regions without obvious features.A binocular matching method for this type of regions is proposed.The basic idea is to use the relative distance and azimuth information between the abundant SIFT feature points around the region and the matching points in the target region to match the pixels in the region without obvious features.The results show that the method can accurately match the pixels of no obvious features regions.2.Binocular vision ranging is realized.In binocular ranging,the checkerboard calibration method is used to calibrate the binocular camera,obtain the camera's internal and external parameters and correct the distortion of the image.Then the corrected image is matching,and the pixel coordinates and parallax of the target point are obtained,which are brought into the ranging formula to obtain the final depth information.In this paper,the formula of ranging is deduced in detail,and the two methods of ranging are compared and analyzed.3.Select Mobile Net convolution neural network model with portability to detect and locate fruit leaf diseases and insect pests as well as spraying uniformity.By using Mobile Net model to train the data set captured and collated manually,the trained model is used to identify the pests and diseases of fruit leaves and the uniformity of spraying pesticides,and the recognition area is accurately positioned by matching and ranging methods.Compared with traditional image processing methods and existing neural network methods,this method can accurately locate fruit leaf diseases and insect pests and uneven areas of spraying pesticides.The research results in this paper,combined with the spraying machinery and equipment manufactured by other units of the project,can realize the automatic spraying of pesticides in new orchards,which has a certain practical significance and is helpful to promote the development of orchard automation.