Research On Image Recognition Method Of Rape Insect Pests Based On YOLO

Rape is an important economic crop in China because of its wide planting area,high yield and high economic value.However,in the process of rape planting,frequent insect pests have a serious threat to the yield of rape in China,which directly affects the national economy.Therefore,in order to ensure the quality of rape and the development of national economy,the pest control of rape is of great significance.The traditional identification and diagnosis of rape pests mainly rely on artificial identification.This method is time-consuming and laborintensive,low efficiency and high rate of misjudgment,which is difficult to meet the actual needs.In view of the shortcomings of traditional methods,this paper aims at the actual needs of rape pest detection and recognition,and takes ten common rape pests,such as vegetable stink bug and cabbage butterfly,as the research object,puts forward a rape pest image recognition method based on Yolo,which mainly carries out the following work:(1)A typical pest data set of rape was constructed by field collection and related literature collection.In order to solve the bias of the result caused by the small sample size of the data set,data enhancement method is used to expand the data set,and finally the data set containing6948 images is obtained.(2)In view of the shortcomings of traditional methods of rape pest identification,this paper proposes a rape pest identification method based on Yolo v3.The experiment results show that the recognition rate of this method is 94.13%,which is better than the traditional algorithm and other single-stage target detection algorithm,but the recognition accuracy of small-scale pest is not high.(3)In order to improve the accuracy of small-scale pest identification,reduce the missed detection rate,and enhance the accuracy of the algorithm,this paper improves the proposed rape pest identification method.By improving the multi-scale feature extraction network architecture,the target’s feature extraction ability is improved to achieve The sharing of highlevel and low-level feature information is also proposed;at the same time,a new loss function Dlo U is proposed to make the features extracted by the model more comprehensive and effectively improve the ability to detect small-scale pests.Finally,a comparative experiment was carried out on the collected rape pest image data sets.The experimental results verified that the proposed rape pest image recognition algorithm based on improved YOLO v3 had higher recognition accuracy,reaching 96.08%,which not only solved the traditional rape pest pests The detection method is not robust and accurate,and can accurately detect and identify small-scale rape insect pests.