Research on Grape Leaf Disease Recognition Method Based on Improved YOLOv8n Model

Huiping Guo^*Linrui RongJiarui CaoYi WangFuzeng YangWeiguo Zang

• 机械与电子工程学院, Northwest A&F University, Xianyang, China

In order to solve the problems of large volume and no classification of disease types in existing grape leaf disease recognition models, this study proposes a grape leaf disease recognition model based on improved YOLOv8n. Firstly, G-bneck was introduced into the backbone network of YOLOv8n to replace ConvModule. At the same time, simSPPF was used to replace the model SPPF to improve the calculation speed while maintaining the feature extraction ability. Secondly, UIB was introduced into the backbone network and neck network to replace the Bottleneck module in C2f, and the C2F-UIB module was obtained after removing the skip connection, which reduced the amount of parameters and calculation of C2f module. Finally, LInner-CIoU was introduced to replace the loss function LCIoU in the head network. At the same time, partial convolution and convolution parameter sharing mechanism were introduced in the detection head to accelerate the inference speed and effectively deal with irregular missing or occluded images. The experimental results show that the average accuracy value of the grape disease recognition model based on YOLOv8n is higher than that of YOLOv3-tiny, YOLOv5n, YOLOv6n and other four models, and the comprehensive performance is the best. The accuracy of the improved YOLOv8n grape disease recognition model reaches 97.3%, the model volume is 3.53MB, and the frames per second is 228.55. The model was deployed to the spraying device, and the average accuracy of the model was 89.3%, and the average time consumption was 5.18s. This study can realize the accurate and rapid identification of grape leaf diseases under the background of natural environment, and provide reference for the development of precise control technology and fine and efficient management of grape diseases.