UNLOCKING THE POSSIBILITIES OF USING MULTI-SPECTRAL IMAGES FOR ACCURATE CROP ASSESSMENT

Abstract

The paper describes the theoretical and technical aspects of obtaining information on the condition of cereal crops based on medium resolution space multispectral imagery. Experimental studies have been carried out to create digital index maps for fields with grain crops. The high efficiency of the methodology for studying the state of fields using multispectral satellite images with the use of L1C and higher L2A processing level products is proved, the features and limitations of the methodology are shown. The theoretical and methodological foundations for processing multispectral satellite imagery results have been implemented in many application services and software. They allow, first of all, to generate such important products for explaining the state of vegetation in fields as vegetation index maps. Sentinel-2 L1C and Sentinel-2 L2A space imagery is a valuable and inexpensive source of information on crop condition. During the pilot study, Sentinel- 2 L1C satellite image products were processed to bring them up to the level of Sentinel-2 L2A products. This is a complex step related to the correct selection of the atmospheric correction model and the conversion of pixel values into reflectivity of objects on the Earth's surface. Another important step in this process is the ortho-correction of the images using global terrain models. Such correction can be important for the further generation of task maps for soil and crop management. The creation of a time series of vegetation index maps for a field of winter wheat is a very substantive way to build a field history that fully confirms the theory of remote field monitoring. The information obtained allows us to adapt the processing algorithm to refine the measurement and prediction of quantitative indicators (biomass calculation, yield prediction, etc.). The experiment also revealed some positive features and drawbacks of the method of creating index maps. For example, under certain weather conditions in summer, correcting the images for atmospheric influence does not significantly affect the calculated index values. On the contrary, with the specified acquisition period (5 days), it may turn out that there are no more than 3-5 cloudless days in a month during the spring growing season, which significantly affects the efficiency of mapping. Accordingly, more reliable and accurate, but significantly more expensive, is the multispectral field survey from unmanned aerial vehicles.

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.