Design And Implementation Of Crop Disease And Pest Monitoring Aircraft Based On Vision

In the process of agricultural production,diseases and pests are important factors that affect crop yields.It is significant to detect and identify diseases and pests in time.In traditional agriculture,people usually rely on artificial judgment to identify diseases and pests.The accuracy of recognition is limited to practical experience and the rationality of sampling.Aiming at the problems of heavy workload,low efficiency,poor working environment and damag e to crop,this paper designs a crop diseases and pests monitoring aircraft based on vision.Firstly,the software and hardware designs of the aircraft are presented and the dynamic model of the aircraft is introduced.The flight control unit is designed with STM32F429 and u C/OS-III,the airborne vision detection system is implemented by Jetson TK1,Ubuntu and Open CV.In order to facilitate visualization operation,the ground monitoring software platform is developed with QT,Open CV and wireless communication technology.In addition,based on this aircraft,the monitoring strategy of diseases and pests is put forward and the optimal range of flying height for the aircraft is determined by experiments.Secondly,since there are few publicly published datasets of diseases and pests which is photographed by aircraft,the thesis self-builds datasets to meet the requirements of diseases and pests detection.The related technologies of visual detection of crop diseases and pests are discussed,including target detection classification,image acquisition,image preprocessing algorithms and image feature extraction.We study target detection algorithms of crop diseases and pests and analyze the principles of two classifiers which include cascade classifier and Support Vector Machine(SVM)classifier.Through experimental tests,the accuracies and efficiencies of the combination detection of two classifiers with different feature description operators are compared.According to the accuracies and efficiencies,the final visual detection scheme which uses the HOG feature and SVM classifier to monitor crop diseases and pests is determined.The effective algorithm is transplanted to airborne embedded platform.Finally,experimental results show that the aircraft can complete the task of monitoring the crop diseases and pests contained in self-built datasets.