sunflower

Purpose. To work out methodological approaches to the use of quadcopters for weeds assesment. Methods. The shooting was carried out using DJI Phantom Vision 2+ and LadyBug Copper Dot. The LadyBug was shoted in the visible and near-infrared range using the 12-megapixel S100 NDVI UAV-Kit camera with elevations: 20 m, 40 m and 60 m. The DJI Phantom Vision 2+ was shot in the visible range of the GoPro 14 megapixel camera altitudes: 10 m, 15 m, 30 m and 60 m. Decryption of photographs was carried out using the controlled classification method in QGIS and TNTmips programs. Weed accounting was performed on control sites 1m2 by weight method, taking into account their qualitative composition. Results. It is shown that the best results of weed recognition during decoding of images was obtained by the use of controlled classification according to the maximum likelihood method under conditions of shooting from heights up to 40 m. In order to improve the recognition of weeds and separate their image from images of cultivated plants, it is expedient to use the object-oriented analysis. At the stage of sunflower budding, about 30% of the weeds are closed from the remote observation, which led to an automatic underestimation of number of weeds. Conclusions. In order to evaluate the crop contamination, it is possible to successfully use the data from UAVs in a visible range of electromagnetic waves under low altitudes (up to 40 meters) and the use of a controlled classification method for decoding images. For the recognition of weeds, the images in the infrared range do not have advantages over images in the visible range. It is necessary to additionally apply ground-based control of weeds to assess the proportion of "hidden" from remote observation of weeds.

Purpose. Evaluate the use kvadrokopteriv for evaluation of weed-infested crops of sunflower. Methods. Aerial survey using drones, object-oriented image analysis. Results. In the article are given the results of as-sessment of a contamination of crops of sunflower by results of decryption of the pictures made by means of the UAV in the visible range.It is shown that the best results of decoding of photo-images are received when using supervised classification by a method of the maximum plausibility. Conclusions. For improving of recognition of weeds and separation of their image from images of cultural plants it is expedient to use the object-oriented analysis.