| As a big agricultural country,the development of agriculture plays an extremely important role in social progress.At present,in the process of agricultural production,plant diseases and insect pests are the main reason for crop yield reduction.Pesticide spraying is the main means to deal with plant diseases and insect pests.A large number of pesticides will cause environmental pollution and affect the safety of food.Accurate application of pesticides is an effective way to solve pesticide abuse.First of all,we need to make an accurate judgment on the degree and types of diseases and insect pests.The traditional detection methods of crop diseases and insect pests are time-consuming and laborious,which are difficult to meet the requirements of rapid and accurate modern agriculture.In this paper,using image processing technology,based on FPGA and MATLAB,from the perspective of image processing and image feature extraction,through the machine learning method of support vector machine,the detection method of pest degree and recognition of leaf image of Brassica napus was studied.This paper mainly carries out the following research work:(1)Based on DE10-Nano development board,it is equipped with Linux operating system.Build cross compiling environment,TFTP server,NFS server,u-boot,Linux kernel and root file system through Linux.The opencv library was transplanted to complete the leaf image acquisition of Brassica napus.(2)In view of the acquired image of the leaves of Brassica napus,the histogram statistics and equalization are completed based on FPGA image processing through Quartus Prime 18.0 design circuit,and the image segmentation of the target area of the leaves image of Brassica napus is completed based on MATLAB 2016 a,and the pretreatment of the leaves image of Brassica napus is preliminarily completed.(3)In order to study the pest degree and recognition of the leaf image of Brassica napus,this paper selects four characteristic parameters: the number of wormholes,the total area of wormholes,the roundness and the integrity of the leaf after eating.These four characteristic parameters are used to detect the effectiveness of pest degree.Through comparative analysis,it is found that with the change of pest degree,the total area of wormholes and the integrity of leaves change greatly,while the roundness change is not obvious.It can be seen that the total area of wormholes and the integrity of leaves can be used to detect the degree of insect damage.28 features,such as color feature,color texture feature and gray texture feature,are extracted as the input feature vector of SVM to identify the pest types.(4)The method of machine learning support vector machine(mlsvm)is adopted.28 feature vectors,such as color feature,color texture feature and gray texture feature,are used as input feature vectors of SVM,and a SVM classification model is established.The SVM classification of RBF radial basis kernel function is finally determined by selecting samples to be tested with different kernel function types for classification and recognitionThe average recognition accuracy of this method is 91%. |
