Abstract
The purpose of this study is to develop a crop-row detector which can be applied to row-following control systems. In this paper, the method for crop-row detection was developed. In this method, first, a crop-row image is divided into far and near fields of vision, and brightness (Green and Red) of pixels is integrated vertically. Secondly, integrated values are calculated (G-R or G/R). As a result, peaks of these graphs correspond to positions of crop, and a line drawn from far peak point to near peak point becomes the detected crop-row. In the experiment on crop-row detection, two systems were tested. One is the line-sensor system (hardware processing system) that was designed for this study, and the other is the area-sensor system (software processing system) that was assembled from general devices such as a CCD camera. As a result, the software processing system was superior to the hardware processing system. The software processing system would be accurate and fast enough for practical use, though some cases of wrong detection occurred.
