V. CONCLUSIONS
Remote sensing image classification is an important part of the field of image processing, and the volcanic ash cloud classification is also an important application in the remote sensing image classification. FCM algorithm is an unsupervised clustering method that can achieve the automatic classification of different types of earth surface information for remote sensing images without any prior knowledge. At the same time, it is also sensitive to the cluster generic and cluster center, thus it not only fails to consider the role of neighborhood pixels in remote sensing, but also belongs to binary-class algorithm where the pixels are forced divided into one or the other. Aiming at solving the uncertainty of traditional fuzzy clustering method and characteristics of remote sensing images, a new fuzzy clustering remote sensing classification method with neighborhood distance constraint for volcanic ash cloud was proposed in this study. Compared to the traditional classification methods (i.e. SWTD and RGB PC) and other literatures (i.e. [19], [20]), the proposed method has a good image quality and high classification accuracy of volcanic ash cloud.
An important merit of this work is that it has provided a new fuzzy clustering method for volcanic ash cloud from remote sensing images via introducing the neighborhood pixels. It has compensated for the disadvantage of single methods of FCM. In this experiment, the Sangeang Api volcanic ash cloud case on 30 May 2014 was explored and discussed by proposed method. In the method, the suitable MODIS bands 30, 31 and 36 were first selected by PCA method; next, different clustering center generics of spectral feature points were determined by the proposed method in feature space and the volcanic ash cloud classification was further got in the two-dimensional spectral feature space; finally the Sangeang Api volcanic ash cloud case on 30 May 2014 was explored and discussed. Our experiments demonstrate that the proposed method can effectively classify the volcanic ash cloud from remote sensing images, and the overall classification accuracy and Kappa coefficient can reach 88.4% and 0.8064, respectively; it overcomes the deficiency of traditional volcanic ash cloud remote sensing classification to some extent and has good application prospect in volcano monitoring and aviation safety.
As an important detection method, the fuzzy clustering classification of volcanic ash cloud from remote sensing images remains to be bottleneck-restriction actual application of remote sensing technique. This work has presented a new fuzzy clustering method with neighborhood distance constraint, and enhanced the classification accuracy. However, this work mainly focuses on the remote sensing classification method of volcanic ash cloud and application test. There are obvious problems in the study, i.e., small sample, further refined of volcanic ash cloud and the result may not be accurate, and it still should be investigated and improved in the future study.