Abstract:

According to the properties of wheat disease (wheat stripe rust and wheat leaf rust) images in complex background, we propose a segmentation method based on the Kmeans clustering segmentation method and the Otsu threshold algorithm. The proposed method is comprised of three main steps. Firstly, by utilizing the difference between the background and the a*b* of leaves, the k-means clustering segmentation method deletes the irrelevant background. Secondly, we use the Otsu dynamic threshold segmentation method to perform binarization and combine the mathematical morphological method with the area threshold method to separate the wheat leaf image with disease from the complex background. Finally, the Kmeans segmentation method is used again and eventually the disease spot of the image is divided. Experimental results show that the correct extraction rate can reach 95%, and it can segment the diseased regions from the whole color image  with good robustness and good accuracy, thus offering an effective method for wheat disease detection and diagnosis.

Key words: wheat stripe rust;wheat leaf rust;image segmentation;complex background;K-means