Monitoring the Land Use, Land Cover Changes of Roorkee Region (Uttarakhand, India) Using Machine Learning Techniques

1. Introduction

Urbanization is considered as a form of development. It is usually associated and resulted from modernization and industrialization, which is generally done for the benefit of the upliftment of the backward area or for the improvement of a particular region population. Although urbanization is regarded as advancement, it has many negative impacts also. Urbanization has led to environmental problems like, lack of fertile land for cultivation, scarcity of water etc. All these effects are not at all good for the upcoming generation as they will not get any type of healthy environment for their living. The increase in population is one of the main reasons of urbanization which results into construction of roads, decrease in the cultivation area and deforestation. In order to identify all these deterioration, the use of satellite images have played an important role. Satellite images have proved beneficial in monitoring the earth surface and in identifying the associated problems.

Satellite image classification has a crucial role in extraction of consecutive useful knowledge required for analysis of Earth’s surface. Machine learning has played a major part in the field of satellite image classification and analysis. Since past decades, they have proved beneficial in identifying Earth’s surface by their implementation on distinct bands and indices of satellite images. There are various machine learning based classifying techniques present like, Neural Network, Support Vector Machine, Maximum Likelihood, K-mean, Binary Encoding classification, Minimum Distance classification, Parallelepiped classification etc. Each of the machine learning technique has different working principle and significance. These techniques have been usually explored individually for satellite image analysis and have given satisfactory results. The work reporting the comparison and analysis between these techniques for performing land cover assessment and change analysis is still very less communicated.

Hence, the current research work focuses on applying four different machine learning methods: SVM, NN, MLC, and PP for satellite image interpretation. For performing the research, the Roorkee area of Uttarakhand (which comes under Devbhoomi-land of God) has been selected. Since last few decades this area has also observed a great enhance in urbanization. The research work also compares and analyse the result obtained by the application of the distinct learning techniques. The result obtained has been further utilized for identifying the class wise changes by applying the machine learning algorithms on a time series data. Hence, effort has been made to investigate the land use/cover changes of Roorkee region, area which is in district Haridwar, to identified the rate or manner in which the land has been utilized and the changes that have been happened in the precedent two decades using machine learning techniques on a time series data.

The remaining article is structured as follows: The section 2 provides an overview of relevant work done by the various writers. Section 3 discusses the work's associated theory or methodology, as well as its benefits and drawbacks. Section 4 lists the data and materials that were utilised in this work. Section 5 presents the data preparation, identification of LULC change detection, result and accuracy evaluation, and change analysis with explanations. The work comes to a close with Section 6.