The use of technical vision in the robotic cash counting area

Abstract

In this research we have installed a computer vision system on a single board computer named ODROID™ XU4. This project is a modification of a more complex and bigger system – the robotic cash counting area. Before our modification a trained neural network was installed on a single board computer, but this caused the computer’s CPU (central processing unit) to be overloaded and soft was shutting down. We implemented and compared two methods of object detection in an image and video, namely Shi-Tomasi and FAST (Features from Accelerated Segment) methods. The comparison criteria were accuracy of detection and CPU load.