NEURAL NETWORK TESTING FOR SPOT-APPLICATION OF PHYTOSANITARY SUBSTANCES IN VEGETABLE CROPS USING A SELF-PROPELLED ELECTRICAL SPRAYER

Abstract

For negative effects minimization generated by agriculture on the environment, there were established a series of measures regarding the reduction of the amount of fertilizers and phytosanitary substances used. Thus, one of the innovative technologies appeared on the market is represented by the usage of some automated equipment for selective spraying of targeted plants, this way significantly reducing the amount of active substances used. The paper presents the usage of a technique specific to artificial intelligence for identification of target crops and their proper treatment. Thus, was developed a convolutional neural network formed of six neuron layers, which was used for analysis of crop field images recorded with a LOGITECH HD Pro C92.0 video camera. The network was developed in C++ programming language, using function libraries from OpenCV, and has run on a Dell laptop, with Intel i8 processor. Following images analysis and targeted plants identification, from laptop there are sent ON/OFF commands through an Arduino microcontroller toward the electrical microvalves mounted on the nozzles of a self-propelled electric spraying machine having a working width of 8 m, with the purpose of spot-spraying the crop plants and reducing the amount of used substances. In this paper are presented the experiments done for testing the neural network efficiency.