Search

High-resolution satellite imagery provides the information about the planet's surface whose automated labelling helps in various practical applications. Segmentation of building footprint from remote sensing images is a challenging task. This paper proposes a novel architecture bi-channel bi-spatial (B2-CS) LinkNet for implementing the semantic segmentation of building footprint. The proposed network uses two parallel semantic segmentation architectures (LinkNet) of RGB and HSV channels with dual spatial and channel attention schemes. By enhancing the features specifically in various channels and positions, result of the segmentation scheme improves. The proposed model outperforms other models with different characteristics like scales, spatial resolution, and object shapes while the analysis was done in two state-of-art datasets: INRIA aerial image labelling dataset and Massachusetts buildings dataset. Accuracy of the proposed work is 97.92% for both datasets which is better than many of the existing methods. [ABSTRACT FROM AUTHOR]

Texture features play a predominant role in land cover classification of remotely sensed images. In this study, for extracting texture features from data intensive remotely sensed image, Gabor wavelet has been used. Gabor wavelet transform filters frequency components of an image through decomposition and produces useful features. For classification of fuzzy land cover patterns in the remotely sensed image, Adaptive Neuro Fuzzy Inference System (ANFIS) has been used. The strength of ANFIS classifier is that it combines the merits of fuzzy logic and neural network. Hence in this article, land cover classification of remotely sensed image has been performed using Gabor wavelet and ANFIS classifier. The classification accuracy of the classified image obtained is found to be 92.8%. [ABSTRACT FROM AUTHOR]

Land cover classification is a vital application area in satellite image processing domain. Texture is a useful feature in land cover classification. The classification accuracy obtained always depends on the effectiveness of the texture model, distance measure and classification algorithm used. In this work, texture features are extracted using the proposed multivariate descriptor, MFTM/MVAR that uses Multivariate Fuzzy Texture Model (MFTM) supplemented with Multivariate Variance (MVAR). The K_Nearest Neighbour (KNN) algorithm is used for classification due to its simplicity coupled with efficiency. The distance measures such as log likelihood, Manhattan, Chi squared, Kullback Leibler and Bhattacharyya were used and the experiments were conducted on IRS P6 LISS-IV data. The classified images were evaluated based on error matrix, classification accuracy and Kappa statistics. From the experiments, it is found that log likelihood distance with MFTM/MVAR descriptor and KNN classifier gives 95.29% classification accuracy. [ABSTRACT FROM AUTHOR]