Pests Detection and Classification on National Crop Based on Deep Learning

Abstract

Pests are a major threat to crop production worldwide. In many countries, pests cause significant economic losses, affecting food security and livelihoods. A significant amount of the harvest could be saved if these damages are prevented, increasing agricultural productivity. Pests not only have the potential to diminish the yield but also to harm equipment and machinery. The main objective of our project is to propose a deep learning based method for detecting and classifying pests on national crops. Our approach utilizes YOLOv4, YOLOv5, YOLOv7 and YOLOv8 models and involve three key steps: > Image preprocessing, > Feature extraction. > Classification. To train our models, we curated a dataset consisting of crop images infected with pests. The dataset was split into training, validation and testing parts, with 70% of the images allocated for training, 20% for validation and 10% for testing. Additionally, we applied data augmentation techniques to enhance the training set and the aim is to enhance the model's capacity to generalize to unfamiliar data instances.The proposed approach has the potential to provide a more efficient and accurate method for pest’s detection and classification on national crops. This could lead to improved crop yields and reduced use of pesticides. The trained model can be integrated into a mobile application, allowing farmers to easily detect and identify pests in their crops. Addressing this issue will have a positive impact on the agricultural sector's economy and reduce the need for excessive utilization of natural resources to achieve food production targets. The initial step in preventing agricultural damage caused by pests is the ability to identify and classify insects accurately, distinguishing between beneficial and harmful species. The continuous and costly monitoring efforts currently required for this purpose can be alleviated through our proposed method.