Year: 2023
Type: article
Abstract: The purpose of the research is to develop a hyperspectral method for assessing the condition of plants in conditions of protected soil, to identify internal changes in plants that affect the yield. The article presents the results of the investigation of vegetation condition assessment based on the use of hyperspectral images (HSI) acquired with a scanning slit-lamp hyperspectral camera with Offner optical scheme. The classifier used is a multi-stage GSI processing with initial compensation of lighting variations, selection of the vegetation area by the index method and further evaluation based on a trained spatial spectral convolutional neural network. The shooting process was organized in such a way that in one frame it was possible to obtain a set of spectrum characteristics corresponding to only one column of the hyperspectral image being formed. The slit diaphragm restricts the area of space from which the light is decomposed into a spectrum. It allows simultaneously to receive up to 2040 spectra in a single frame. In the pro-cess of space scanning and serial synchronous imaging, spectra of sequentially changing regions of space were recorded. Based on the results of the research, an algorithm for classifying high-resolution hyperspectral images was developed, taking into account both spatial and spectral features. A convolutional neural network with an architecture modified to better account for changes in scene lighting was used as a classifier. To prepare the training data, it is proposed to use vegetation indexes, which allow to per-form primary binary segmentation of the hyperspectral image. It was shown that the proposed approach is effective. The accuracy of the classification obtained by the level of chlorosis was more than 83%. The conducted experiments show the effectiveness of the proposed approach for assessing the condition of plants in protected soil conditions.
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDGs: Life on land, Zero hunger
Open Access status: hybrid
OpenAlex ID: W4409703637