Crop identification by machine learning algorithm and Sentinel -2 data

There is a growing need for remote identification of the crop types, which is directly related to vegetation density and it is a valuable tool for agricultural inspectors and government agencies. The serious issue that has arisen in recent years for policy makers and statistical accountants is the degree of validity of information concerning the type and area of each crop. Information on the spatial distribution of arable land and crop types can also assist in the accurate statistical estimation of crop area and the improval of agricultural policy planning. In this study remote sensing data was utilized by Sentinel-2 mission, from which NDVI values were calculated based on Band 4 (0.665μm) and Band 8 (0.842μm) for the period 2017-2020. Three different crops were analyzed: cotton, rice and olive trees. The data used for the experiment were preprocessed and monthly average NDVI values were calculated for each month. Preprocess included the typical method of average. Next, a machine learning algorithm was developed and training was accomplished utilizing monthly average NDVI values. Python programming language and KNN machine learning module on a Pycharm shell were used for the development of the machine learning algorithm. The algorithm was saved on a joblib file when training was completed. Subsequently, a second algorithm was developed for inserting used defined NDVI values and processing using the trained algorithm (joblib). Through this process, the identification of crop types was accomplished based on NDVI values, through the algorithm that has been developed. From the literature so far, it is concluded that a machine learning system can respond to crop recognition learning and further identification.