Object Detection in Precision Viticulture Based on Uav Images and Artificial Intelligence

The system of precise viticulture is the basis for the improvement of conventional intensive viticulture production, while achieving yields of high quality and minimizing costs. Detection of vines in the vineyard is one of the key issues that can be solved with new technologies. Although the detection and the location of individual vine stems is of fundamental importance, there are no methods that can make this fully automatic. Various events such as diseases or mechanical damage led to the disappearance of plants over the years and the reduction of the initial number of vine stems per hectare. In order to accurately estimate the yield of the vineyard, it is necessary to identify and count the actual number of vines in order to estimate the yield based on them. A conventional method of such visual counting would be extremely time-consuming. This research aims to propose an automated model that can detect and locate living vine stems, using UAV imagery and artificial intelligence methods. The proposed model showed a high accuracy in the detection of vine stems. The advantage of the proposed model for the detection and identification of vines is that the results are obtained quite quickly and not much data is required (only RGB + NIR image of the vineyard from two periods without and with vegetation).