Vol. 7, Issue 6 (2019)
Extended differential pattern-based large scale live active learning model for classification of remote sensing data
Author(s): MK Pradhan, BL Sinha and S Ramole
Abstract: Recently, the rapid growth of the space-oriented imaging techniques with the tremendous remote sensors facilitates the earth observation in the spatial and spectral domain. The Hyperspectral Images (HSI) have the ability to deliver the detailed information of earth in such domains. An accurate identification of objects from the acquisition system depends on the clear segmentation and classification. Traditionally, clustering and rules-based methods are adopted for classification according to the thresholding effect of image pixel intensity. The clustering process depends on the mean feature of the image pixels with the gray limit. With the spectral limitations, the multi-label segmentation problem affects the clustering and rules adversely. The variations in spectrum cause the changes in the number of rules each and every time that leads to computational complexity and misclassification. This paper proposes the novel classification method based on the textural information obtained from the Extended Differential Pattern (EDP). Initially, the Distributed Intensity Filtering (DIF) removes the noise present in the image and the application of Histogram Equalization (HE) enhances the image quality. The merging and classification of different labels for each image sample are performed through the Extended Differential Pattern (EDP) provides the textural information clearly. With these pattern set, the traditional active learning methods such as Relevance Vector Machine (RVM) and the multi-class SVM classify the HSI patterns that play the major role in remote sensing applications. The comparative analysis between the proposed EDP-AL with the existing algorithms regarding the various parameters overall accuracy, average accuracy and kappa statistics conveys the effectiveness of EDP-AL in remote sensing applications.
Pages: 1610-1620 | 252 Views 55 Downloads
download (10123KB)
How to cite this article:
MK Pradhan, BL Sinha, S Ramole. Extended differential pattern-based large scale live active learning model for classification of remote sensing data. Int J Chem Stud 2019;7(6):1610-1620.
