Evaluation of the application of the multi-temporal method in Sentinel 2 satellite images for the separation of agricultural products

Abstract

One of the ways to obtain information about the condition of the land is to produce land use maps. In this research, using different time series from Sentinel 2 satellite images, with the aim of choosing the appropriate classification method for the separation of land products in Ravansar city, Kermanshah province. Took based on the growing season, it was first prepared by referring to the agricultural calendar of different products of the region. By determining the time of planting, the peak of greenness, harvesting and plowing of different crops, information was collected and stored in the database for the necessary analysis to determine the time of the images based on the major crops of the study area, including (wheat, corn, barley, road, other-vegetation, other-plough-barren, water, peas, tomato) should be taken from Sentinel 2 images on two dates. By making the necessary corrections on the images, in the next step, the mentioned dates were done with the PCA method and then the classification was done with the maximum likelihood and minimum distance method. The results showed that the maximum likelihood classification was more accurate than the minimum distance method in multiple times with an overall accuracy of 94% and a kappa coefficient of 92%.