Buckwheat Disease Recognition Based on Convolution Neural Network

Buckwheat is an important cereal crop with high nutritional and health value. disease greatly affects the quality yield of buckwheat. The real-time monitoring essential part ensuring development buckwheat industry. In this research work, we proposed automated way to identify diseases. It was achieved by integrating a convolutional neural network (CNN) image processing technology. Firstly, approach would detect area accurately. Then, improve accuracy classification, two-level inception structure added traditional for accurate feature extraction. also helps handle low-quality problems, which includes complex imaging environment leaf crossing in sampling image, etc. At same time, instead convolution, convolution based on cosine similarity adopted reduce influence uneven illumination during imaging. experiment proved that revised enabled better extraction within samples illumination. Finally, results showed accuracy, recall, F1-measure detection reached 97.54, 96.38, 97.82%, respectively. For identifying categories, mean values precision, were 84.86, 85.78, 85.4%. Our method has provided technical support realizing automatic recognition