Методика кластеризации сельскохозяйственных полей по RGB-изображениям беспилотных летательных аппаратов

Abstract

The Earth remote sensing (ERS) technologies, that are being developed on the basis of satellite sensing and, more recently, unmanned aerial vehicles (UAVs), provide a significant potential for smart (precision) farming applications. Sensors (for example, digital RGB cameras) provide real-time data about the environment of the study area in the form of images. The sensors installed on unmanned aerial vehicles can be used in various applications related to assessing the quality of plowing, winter wheat seedlings, growing crops, etc. It becomes possible because they capture large areas with a set of images with a low time (several hours per 1000 hectares), but with a high spatial resolution (several centimeters). The remote sensing technology is expected to revolutionize the agriculture by enabling faster decision-making limited by few days only, as well as reducing costs and increasing yields. Despite the significant development, one of the areas of use of UAVs in smart agriculture is not yet as reliable and accurate as expected, mainly due to problems arising in the collection, processing and analysis of images. The main problem is that there is still no standardized workflow that includes the steps from collection to visualization of results when using UAVs in agricultural applications. One of the weak points of many image processing technologies is the insufficient quality of clustering, where each cluster found is associated with a certain type of surface. This article discusses the use of images obtained using UAVs for solving the challenges of smart agriculture. A clustering technique related to the parametric representation of histograms of brightness of color spaces and Greenness Indexes is discussed. The results of image processing are presented.