Research And Design Of The Stage Recognition System Of Corn Borer

Maize,an important cereal crop in China,has a very extensive cultivation area in the country.However,with the impact of climate change,geographical and ecological damage in recent years,the phenomenon of corn borer infestation has been increasing in existing maize production and harvesting.As one of the major pests affecting maize yields,the corn borer can reduce yields by one-tenth at spring maize harvest and three-tenths at summer maize harvest,causing huge losses to China’s maize industry.Under conventional conditions,the identification of corn borers in field testing is more often done by plant protection experts and experienced farmers,but this method is not guaranteed in terms of accuracy and is prone to disadvantages such as low recognition efficiency due to missed and wrong detections.This thesis focuses on the effective recognition of corn borer pests in different periods of time,and develops a corn borer staging recognition system through deep learning and image recognition technology.The details of the research are as follows:Firstly,the dataset images required for the experiment were obtained by both field collection in maize fields and online crawling.The dataset images consisted of images of corn borer eggs,corn borer larvae,corn borer pupae,corn borer moths,Athetis lepigone larvae,Beet armyworm eggs and Beet armyworm larvae stages.The resulting images were pre-processed and collated into a dataset for subsequent experimental analysis.Secondly,three convolutional neural network models were selected for cross-sectional comparison,and experiments were conducted with the Alex Net model,VGG-16 model and Mobile Net-V2 model.The Mobile Net-V2 model,with an overall recognition accuracy of 95.99% for the test set of pest images,was selected as the recognition model for the image recognition system through comparative analysis of the test results.Finally,based on the above work,a system requirements analysis was conducted to build and design a corn borer staging image recognition system.The system provides a fully functional and easy-to-use user interface that allows for the accurate and effective recognition of corn borer pests,providing an aid to the recognition and control of the pests and facilitating the daily use of agricultural staff.