EXTRACTION OF CROP INFORMATION USING GEOMATICS AND MACHINE LEARNING

Abstract

The main objective of the present research work is to extract the agricultural information such as discrimination of specific crop type and estimation of crop yield as well as its by-products by the sophisticated use of remote sensing and machine learning techniques. The application of data science, machine learning and information technology in agricultural research is becoming essential to adapt statistical formula or complex model in digital form for simple and precision agriculture. Technological innovations are currently being exploited for automation and to expand Decision Support Systems (DSS) for agricultural production and fortification research. Recently, remote sensing technology and Geographic Information Systems (GIS) have developed a capable role in agricultural research. Predominated research is in crop area and yield prediction, in addition to crop suitability studies and site-specific resource allocation. Therefore, it is required to use the recent advancements of data science, information technology, remote sensing, and sophisticated statistical methods to extract crucial information from the agricultural databases. A framework of crop growth modelling for the sugarcane is presented in this research work. Information extracted from remotely sensed data as well as ancillary data is used as input datasets to capture the knowledge about the crops of the study area. The blending of the spectral, spatial, multi-temporal and ground truth data plays a crucial role in the information extraction. The models are focused on facilitating the user to extract information related to the crop growth stage, discrimination of specific crop variety, estimation of crop yield and by-products. The proposed work is based on the hypothesis that remotely sensed data and machine learning algorithms could be used as an e ective tool to extract agricultural information. Agricultural analysts may use this information for sustainable and precision agriculture. Additionally, the work is hypothesized around the temporal variations during the photosynthesis activity in the agricultural fields. Another assumption is that satellite data and the extracted information is essential to predict the unknown attributes. Many techniques and models have been developed to extract information from the remotely sensed data, but it remains an exigent problem due to the accuracy, reliability and timeliness parameters. It motivates us to explore the current and powerful methods of extracting the crop information from the spectral and field data. The importance of pre-processing and feature selection is also explored in this work. The challenges related to technical informatics and the agricultural domain have been investigated in the present work. In the initial stage of the present work, the description, importance and the process of the data collected during the field visits is presented.