Full Utilization of a Single Image by Characterizing Multiple Regions of Interest for Line Tracing

Abstract

This paper presents a new method of image processing for the line tracing task, which is one of the simple and fundamental tasks that has been applied to an unmanned system, utilizing multiple regions of interest to draw information from the entire image which was discarded in traditional image processing method for more accurate and flexible line trace. This new method divides the acquired image by machine vision into 3 regions: feedback region, prediction region, and inspection region. And different process was applied to each region to acquire parameters depending on the characteristics of each region that can enhance line tracing performance. In this paper, parameters of the new method are applied to the proportional control method and implemented to the robot arm and the camera and evaluated with the basic proportional control by comparing adaptability to a sharp curve. Consequently, the new method provided more adaptability in line tracing compared to the traditional single region of interest method.