Detection Of Wheat Scab Based On Image Processing

[Research background]Wheat scab is one of the major diseases of grain production,which has great damage to wheat yield.Therefore,the researches and applications of wheat disease monitoring and early warning are very important for the development of high-yield,high-quality efficient and scientific production.In order to control wheat scab timely and effectively,it is necessary to monitor the incidence of scab timely and accurately.At present,due to the limitation of production conditions,geographical factors and technical level,the method of field sampling survey is mainly used to evaluate the harm degree of scab.This method is not only difficult to effectively monitor large areas of disease producing areas,but also has high labor cost and slow information acquisition,which seriously affects the efficiency of scab controlling.[Materials and Methods]In this experiment,wheat head was inoculated with Fusarium graminearum to cause wheat head blight.The RGB image of wheat head blight was acquired by UAV and digital camera,and the head blight in wheat head blight image was detected and recognized by extracting image features.[Results and Analysis](1)Based on the color characteristic index,the ear of wheat in the image of scab was extracted.The detection rate was 90.5%in the early stage and 88.4%in the middle stage.(2)Based on the deep learning network deeplabv3+,a model of wheat ear detection and recognition was established.The accuracy of the model is 0.9692,the Loss rate is 0.1030,and the MIOU is 0.793.The performance of the model is good.Basically,it can detect and segment the wheat ear with scab.(3)Based on the improved deep learning network u-net,the detection and recognition model of wheat ear with scab was established.The accuracy of the model is 0.9694,the Loss rate is 0.0759,and the MIOU is 0.799.The model has good performance and can realize the recognition and segmentation of wheat ears with scab.[Conclusion]It is found that the model based on deep learning network can detect and recognize wheat head with high recognition accuracy,which provides technical support for the detection and recognition of wheat head blight.