Providing a knowledge-based method for distinguishing crops and estimating a cultivation area (Case study: The Moghan Plain)

Monitoring and knowing accurate and update statistics of a cultivation area and estimating agricultural production is essential in proper economic planning, import, and export.

A primary purpose of the research is to develop a knowledge-based method based on the concepts of objectivism in terms of spatial and temporal conditions. This method seems to improve the accuracy of crop discrimination and reduce the number of ground samplings.

The present study is carried out in three main stages. The first stage includes creating an object-oriented structure (extracting the characteristics of each farm and storing them). The second stage is building a knowledge base according to spectral signatures of crops and spatial-temporal conditions. The third stage includes discriminating crops in satellite images using the proposed method.Study area: Due to the importance and diversity of Moghan plain and its products, the proposed method was implemented using information (spatial and spectral) of the 2019-2020 cultivation season there.

The overall accuracy of the proposed method for discriminating crops at three different dates (stages of crop growth), including April (beginning of a wheat’s green growth), May, and mid-June (before a wheat’s harvest), were 94.66, 91.5, and 95.12 percent, respectively. That shows an improvement of more than 10% compared to the Maximum-likelihood Classification.

The results showed that using Spatio-temporal conditions and spectral behavior of crops from their planting to harvest time can increase crop discrimination accuracy compared to the maximum likelihood classification method.