Detection of Agricultural Pests Based on Improved YOLOv5 Algorithm

An improved YOLOv5 model is designed to identify and locate small target pests for agricultural pest analysis. For the existing data set, firstly, the data enhancement operation was carried out on the images, and then the target background analysis was combined to reduce the brightness of half of the data to increase the training difficulty. To solve the problem that small and medium targets in the target detection model are easy to lose target information with the deepening of the convolutional neural network, a feature extraction layer was added to the model to enrich the feature information and location information of small targets in the target detection model. The model adopts a cutting image re-detection method, which reduces the calculation of the model and improves the detection accuracy. Finally, several experiments were designed and compared to verify the validity of the model, with an average detection accuracy of 92%.