Development of Automatic Weeding Thinner for Direct Sowing Sugar Beet (Part 2)

Distinction between Sugar Beet and Weeds Using Video Image

Abstract

The sugar beets and weeds were detected and distinguished using the video images taken by a CCD video camera mounted on a tractor. The aim was automatically thinning and weeding out directly sowed sugar beets. An accurate distinction based only on the analysis of the shape characteristics was difficult. The reason for this was the difference in the growing size of the sugar beets which was caused by the beets being cultivated by direct sowing. Therefore, the color information, Q of the YIQ color system, was added to the analysis of the shape characteristics. As the result, correct distinction rates for the sugar beets and weeds were 89.7% and 91.0%, respectively. The processing time using the low spec CPU (Celeron, 400MHz) was 3Hz. This performance was sufficient for the real-time works in the field.