Technology for recognizing the state of vegetation via ERS airborne (including obtained from UAVs) and satellite images

Input data: color images of agricultural vegetation fields of different spatial resolution.

The technology will provide the formation of maps of agricultural vegetation for monitoring the state of vegetation:

- identification of the disease manifested by a violation of photosynthesis;

- weed areas determination.

To improve the monitoring accuracy, it is proposed to use:

- combined informative features of multispectral images, combining vegetation indices, color, textural and fractal characteristics of images;

- intelligent analysis of the vegetation state using deep neural networks

Scientific and technical groundwork.

The combining method has been developed. This method combines informative features of multispectral images for assessing the state of agricultural vegetation, the model of neural networks for the identification and classification of objects in the form of a neocognitron, self-organizing Kohonen maps, convolutional networks and their combinations, the method for choosing optimal neural networks and methods for their training, the technique for using ensembles of neural networks for the identification of remote sensing objects which improve stability and reduce the computational complexity of the learning process with fuzzy information about objects.

The results were tested in the implementation of a number of international projects carried out under grants from the BRFFR and the SCST.

1. Ganchenko, V. Agricultural Vegetation Monitoring Based on Aerial Data Using Convolutional Neural Networks V. Ganchenko, A. Doudkin // Optical Memory and Neural Networks. – 2019. – Vol. 28, No. 2, pp. 129–134. © Allerton Press, Inc., 2019

2. Ганченко, В.В. Построение карт сельскохозяйственных полей по аэрофотоснимкам различного спектрального диапазона для систем точного земледелия / В.В. Ганченко, А.А. Дудкин, Е.Е. Марушко // Вестник БрГТУ. – 2017. - №5(107). - С. 45–48.

3. Дудкин А.А. Подсистема параллельной обработки изображений для мониторинга состояния лесных угодий / А.А. Дудкин // Известия ЮФУ. Технические науки – 2014. − №12 (161). – С. 129–137.