Semantic Segmentation using K-means Clustering and Deep Learning in Satellite Image
Manami Barthakur, Kandarpa Kumar Sarma · 2019
In this paper, a deep learning based method, aided by certain clustering algorithm for use in semantic segmentation of satellite images in complex background is proposed. The work considers the formation and training of SegNet in which the output of K-means clustering algorithm is used as input and the label of the particular region of interest (ROI) in the image are used as target. The method does not require any feature extraction, region growing or splitting methods to configure and train the SegNet, which is a deep convolutional Encoder-Decoder architecture trained with (error) Back Propagation learning. The method is tested with different satellite images. The method is also compared with the results obtained when trained with SegNet without selecting the ROI using K-means algorithm and evaluated using metrics such as accuracy, mean IoU and weighted IoU. The normalized confusion matrix is also plotted. The experimental results show that the method is reliable and suitable for real world situations.