Robotická platforma pro monitorování stresu rostlin: tvorba obrazového datasetu a aplikace metod umělé inteligence

Abstract

This master thesis focuses on the design and implementation of a robotic platform for monitoring and classifying plant vitality. The main contribution is the integration of the Hailo-8L AI accelerator with the Raspberry Pi 5 control system, enabling the efficient application of artificial intelligence methods for image processing tasks using convolutional neural networks directly on the device. For the purposes of this work, the robot's hardware was modified, specifically through the integration of an industrial camera and sensor equipment. To classify plant status (healthy/stressed), a ResNet-18 model was trained on a custom image dataset. The software development includes a control application written in Python featuring a graphical user interface. The system's functionality was verified through real-world testing. The primary outcome of this work is a functional application enabling comprehensive platform control and automated plant health diagnosis.