Land Use and Land Cover Classification Using Recurrent Neural Networks with Shared Layered Architecture

• Published in: 2021 International Conference on Computer Communication and Informatics (ICCCI) pp. 1 - 6
• Main Authors: Vignesh, T., Thyagharajan, K.K., Jeyavathana, R. Beaulah, Kanimozhi, K.V.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 27.01.2021
• Subjects: Computer architecture; Deep learning; Feature extraction; Informatics; Land use and land cover classification; machine learning; Monitoring; Multispectral image classification; Recurrent neural network; Recurrent neural networks; Satellites; Shared layered architecture
• DOI: 10.1109/ICCCI50826.2021.9402638

Land use and land cover specifies, classifying the land cover areas into various land use and land cover classes. The purpose of land use and land cover classification is that monitoring and identifying the various land cover classes exactly. Because of that, we can prevent land cover objects from destruction. The machine learning and deep learning techniques plays very important role in land use and land cover classification. In this research work, Shared Layer Recurrent neural networks (SLRNN) are used to classify the LISS-IV images into various land cover classes. Specifically, as a proposed work shared layer architecture are used. The performance of shared layer Recurrent neural network is compared with coupled layer Recurrent neural network (CLRNN), uniform layer Recurrent neural network (ULRNN) and traditional recurrent neural network. Multitask learning can be supported by shared layer Recurrent neural network. Due to this multitask learning, SLRNN successfully classify the LISS-IV multispectral satellite images into various land cover classes with high accuracy. SLRNN provides 5% accuracy rates than CLRNN, % accuracy rates than ULRNN, and 10% accuracy rates than traditional RNN. The results show that, this proposed method, provides better results than other method.