Design And Implementation Of Intelligent Investigation And Warning System Of Rice Planthoppers Based On Mobile Terminal

Through timely and accurate investigation and early warning of the population dynamics of rice planthoppers,it is conducive to better and effective control of planthoppers.Through timely and accurate investigation and early warning of the population dynamics of rice planthoppers,it is conducive to better and effective control of planthoppers.Currently,it is a hotspot of intelligent investigation and research on rice planthoppers to collect images of rice planthoppers with the help of hand-held image acquisition equipment,and to automatically detect and count rice planthoppers.Although the problem of the staff bending over to take pictures is solved by using a hand-held image acquisition instrument to collect the images of the rice planthoppers,it is difficult to control the shooting distance between the lens and the rice straw.The pictures taken in this way lead to a certain difference in the proportion of the same rice planthoppers in different images,which leads to the problem of missed detection of rice planthoppers targets and poor generalization ability of the model.In order to improve the precision and recall rate of rice planthoppers detection,this paper designs and implements a mobile terminal-based intelligent investigation and early warning system for rice planthoppers.The main research contents and results are as follows:(1)The research method of the detection model of rice planthoppers based on the deep learning target detection algorithm is proposed.YOLO-V4,YOLO-V5 in the one-stage object detection algorithm and Cascade R-CNN in the two-stage algorithm are trained on the same rice planthoppers image dataset.The experimental results show that under the same test set,Cascade R-CNN in the two-stage detection algorithm has a better detection effect on rice planthoppers,and the model has a higher precision and recall rate;The accuracy of the original model is improved by changing the backbone network of the Cascade R-CNN model,where the precision and recall of the Cascade R-CNN-Res Ne Xt-50 model are 75.3% and 62.3%,respectively.(2)Image preprocessing methods with different image segmentation strategies are designed to improve the model detection effect.Since it is difficult to control the shooting distance of the handheld image acquisition instrument when collecting images,the proportion of planthoppers targets of the same size under different shooting distances in the images varies greatly,resulting in a low precision and recall rate of the detection model.Since it is difficult to control the shooting distance of the handheld image acquisition instrument when collecting images,the proportion of planthoppers targets of the same size under different shooting distances in the images varies greatly,resulting in a low precision and recall rate of the detection model.The results show that under the same test set,when the image is preprocessed by using the image hybrid block mechanism,the Cascade R-CNN-Res Ne Xt-50 model can detect four different insect states(long-winged adult,short-winged adult,advanced nymph,The average accuracy of the young nymphs)planthoppers increased from 63.9% of the original model to 90.3%.(3)Developed the We Chat applet of the rice planthopper intelligent investigation and early warning system.Designed the We Chat applet of the rice planthoppers intelligent investigation and early warning system,and built an image detection server based on the rice planthoppers detection model;The We Chat applet developed in the VScode environment realizes the registration and login function,the intelligent investigation function of rice planthoppers,the prediction and early warning function of rice planthoppers,and the query function of rice planthoppers survey results.In this paper,according to the characteristics of rice planthoppers images collected by hand-held image collectors,a hybrid image segmentation strategy is designed to preprocess the images,which effectively improves the average accuracy of the Cascade R-CNN-Res Ne Xt-50 model for rice planthoppers detection.On this basis,the We Chat applet of the rice planthoppers intelligent investigation and early warning system is designed to allow grassroots investigators to use handheld image acquisition equipment to efficiently conduct investigations and give early warnings in the field.