Machine Learning Methods for Image Segmentation and Feature Extraction of Cotton Diseases and Pests

Cotton, as an important food crop, is related to people's lives. Cotton is susceptible to various pests and diseases during its growth and development, which greatly affects its production. This article collects cotton pest images containing different types of diseases and pests, and performs image processing under different environmental conditions. In order to eliminate interference in the image and improve its accuracy, preprocessing work such as noise reduction, scale adjustment, and contrast can be performed on the obtained image. In addition, this article uses convolutional neural network algorithm to effectively separate cotton pest and disease areas from the background. By utilizing ensemble learning methods, the prediction results of multiple models can be fused to improve the overall recognition accuracy. The final optimized integrated model only ran for 21 minutes on the test set, which is the shortest among other models, indicating that the model maintains high accuracy while also optimizing running efficiency. The method described in this article helps to quickly and accurately identify the types of cotton diseases, and carry out timely and accurate prevention and control to prevent the spread of diseases.