Prediction of the Vigor and Health of Peach Tree Orchard

Abstract

New technologies are a great support for decision-making for agricultural producers. An example is the analysis of orchards by way of digital image processing. The processing of multispectral images captured by drones allows the evaluation of the health or vigor of the fruit trees. This work presents a proposal to evaluate the vigor and health of trees in a peach orchard using multispectral images, an algorithm for segmentation of trees canopy, and application of vegetable indexes. For canopy segmentation, the Faster R-CNN convolutional neural network model was used. To predict the health of the peach trees, the vegetable indexes NDVI, GNDVI, NDRE, and REGNDVI were calculated. The values of the NDVI, GNDVI, NDRE, REGNDVI indexes obtained for the healthiest tree were 0.94, 0.86, 0.58, and 0.57, respectively. With the application of this method, it was possible to conclude that the use of multispectral images together with image processing algorithms, artificial intelligence, and plant indexes, allows providing relevant information about the vigor or health of the cultures serving to support the decision making in agricultural activities, helping the optimization of resources, reduction of time and cost, maximizing production, facilitating the work of agricultural explorers.