基于Faster R-CNN的马铃薯晚疫病图像识别系统

Image recognition system of potato late blight based on Faster R-CNN

马铃薯晚疫病在我国南方降雨较多的省份十分常见,该病害严重时,马铃薯植株会提早枯死,产量受损严重。通过计算机对马铃薯晚疫病识别很有必要。对Faster R-CNN网络进行微调,对图片进行预处理,将马铃薯叶片病害图片进行多种旋转,改变亮度值和对比度,增加噪点,并且使用迁移学习,最后实现了马铃薯晚疫病图像的精准识别。

In recent years,potato has spread to China for hundreds of years,and now it has become the fourth largest cultivated crop in China.China's potato industry has an irreplaceable position in the world.But as the potato for our country to bring rich economic benefits at the same time,the epidemic of this disease is very strong,the harm caused is very serious,in the process of growth,a season may cause many infections.Potato late blight in the south of our country rainfall more provinces are very common,the disease serious,potato plants will die,early production is severely damaged,in turn,the recognition of the potato late blight by computer we system is very necessary for Faster-R-CNN network the fine-tuning,the image preprocessing,we rotate the image for a variety of potato leaf disease,change the brightness and contrast,increase the noise,and using the migration study,finally realizes the potato late blight accurate identification of the image.