Brassica Napus L. Pest Detection Based On Image Processing Technology

Brassica napus L.,which has a large planting area and high grain yield,is a very important oil crop in China.But pests can seriously affect its growth and reduce grain yield.Precision pesticide spraying technology can effectively prevent and cure pests and diseases.The traditional crop pest detection method is time-consuming and laborious,and the accuracy is poor,which is difficult to meet the current requirements for efficient and rapid detection.In this paper,from the perspective of image segmentation,image feature extraction and design of classification model,the detection method of insect pest degree in Brassica napus L.leaves image based on image processing technology is studied.The paper focuses on the following:(1)Because the pictures collected in outdoor are greatly affected by illumination and external environment,this paper proposes a segmentation method based on color space model,which converts color images into several space models and finds the color component of each space model,in which the blade region can be most prominently selected.And the appropriate components are selected and superposed to obtain the grayscale image,which can recognition the blade region obviously.Thereby the grayscale image is performed by the threshold segmentation method.The method is compared with the segmentation method using Otsu algorithm,and the segmentation effect and segmentation error rate of two method are analyzed.The results show that the segmentation method based on color space model has good segmentation effect,the error rate is <0.024,and the average error rate is 0.00611.Finally,the segmented image is binarized and denoised,and a binary image retaining only the blade region is obtained.(2)According to the detection of insect pests in Brassica napus L.leaves,eight feature parameters are extracted: wormhole area,leaf occupancy,eccentricity and circularity before wormhole filling,leaf occupancy and eccentricity and the circularity after wormhole filling,integrity.Based on the MATLAB GUI platform,a eigenvalue extraction system of leaves image of Brassica napus L.is designed,and all the blade images are processed by the system.On the basis of the eigenvalues of all the blade images,the wormhole area,the difference in occupancy,and the integrity,which is effect to detect the degree of pests,are obtained after the selection and combination of feature.(3)There are many sample data,the characteristics are related to each other,and it is a multi-classification problem in paper.Considering these factors,a BP neural network classification model for image pest degree of Brassica napus L.is designed,in which count of the hidden layer is 1,the number of its neurons is 8.The experimental results show that the mean square error of the model on the training sample reaches to 0.0357,and the mean square of classification error is 0.177905,and the classification correct rate reaches to 97.142857% for the test sample.