Development of System for Deterrence of Animals using Artificial Intelligence

Abstract

This thesis is part of the Mechatronics masters program at the University of Agder (UiA). It describes the development of an animal deterrence system, using computer vision with embedded systems for edge computing. Household cats can run freely around in neighborhoods, but all neighbors may not appreciate it. This project seeks to develop a harmless and accurate system for keeping cats away. The solution is to use a state-of-the-art object detector to look for cats in an area. Using an embedded system to run the object detector and tracking method. Design a two-axis turret with servo motors and a camera. The object detector used for this project was YOLOv4-tiny. It was trained to detect cats, dogs, humans, and hedgehogs. The YOLOv4-tiny object detector was implemented on an NVIDIA Jetson Nano development board and ran at an average of 23 FPS. Robot Operating System (ROS) was used to run the control algorithm servo motors and the object detector. A two-axis turret, which uses two servo motors, was designed and built to make the camera track the cat. A two-axis turret, which uses two servo motors, was designed and built to make the camera track the cat. A component housing was also designed and built to have the whole system in a small enclosure.