Study On Segmentation Of Rice Disease And Insect Based On Image

In the process of the rice growth, a variety of pests and diseases influence the rice growth and production. Traditionally, the methods including capturing the pest by light, netting and picking the leaf consists the survey in the farm, reducing efficiency. There are many problems in the field for rice survey by relying on artificial recognition and counting, such as investigation task, lower efficiency and non real-time problem. With the mature of computer vision and image processing technology, we have successful recognition and diagnosis by the new technology. While, before the automatic identification and diagnosis, we should segment the background which means separating the target from the background, including pests, the pests’harm area and rice diseases. As the difference of the pests, the harm area and rice diseases, our research display a variety of segmentation algorithm on three aspects. As follows:(1) Study on segmentation of the rice pest on image. Firstly, we cut the leaf so that the pest’s area accounts for2/5-3/5of the whole image. Secondly, analyzing from graph theory, we interactive markup on the image and using unsupervised segmentation algorithm, in order to achieving the best segmentation. Lastly, we use the ROC curve to evaluate segmentation result.(2) Study on segmentation of the pests’harm on image. Firstly, we cut the harm leaf so that the harm area accounts for2/5-3/5of the whole image. Main pest of rice harm include Cnaphalocrocis Medinalis and Chilo Suppressalis Dudgeon, According to its characteristic, we use a variety of significant algorithm, in order to achieving the best segmentation. Lastly, we use the ROC curve to evaluate segmentation result.(3) Study on segmentation of the rice disease on image. Firstly, we cut the disease image so that the disease area accounts for2/5-3/5of the whole image. Secondly, we prepare to process the image, analysis the color space and track the boundary of the disease. Lastly, we use the ROC curve to evaluate segmentation result.