Pulverização de plantas daninhas em tempo real utilizando visão computacional

Abstract

Within agriculture, one of the greatest difficulties is to reduce the growth of weeds, this involves the application of chemical herbicides to ensure good quality and good production of the crop of interest, however, excessive application of these herbicides cause damage to the environment and high costs to the producer, about 20% of the fallow area is infested with weeds, corresponding to 80% of herbicide waste, applying only to the target, can achieve a gain of 90% reduction in cost. Soybean corresponds to 60% of herbicides on total investment with agricultural pesticides. This work aims to reduce losses in the application, with applied in the total area. The automation of the sprayer with an algorithm using the specific OpenCV library written in python programming language connected to arduin nano and other components electronic devices, performing the spray where it detects the green color, will be evaluated the lowest detected leaf area and repeatability of the target application and accuracy of application. The development, design, construction, and testing in laboratory and field were accomplished and conducted at the Laboratory of Intelligent Agriculture of the Agricultural and Environmental Engineering, at the Institute of Agrarian and Technological Sciences of the Federal University of Rondonopolis (UFR). The equipment 100% in controlled ´ environment to detect static green color reaching area with a minimum of 36 mm2, with nozzle height of 0.50 m minimum with 312 mm2, with spout bar nozzle height of 1,12 m, however, when exposed in soil conditions Red Latosol there was a loss of precision reaching 70% with the nozzle height 0.50 m from the surface and the tractor speed 1.67 m/s and in the condition of the asphaltic floor had the precision of 70% with the height of 0.50 m of the surface and tractor speed 1.67 m / s.