DEVELOPMENT OF THINNING AND INTRAROW WEEDING MACHINE FOR DIRECT SOWING SUGAR BEET USING IMAGE PROCESSING

在甜菜栽培中减少除草剂的使用和工时是一个迫切需要解决的问题。为此，开展了利用图像处理技术的甜菜直播自动间伐除草机的开发研究。通过摄像机拍摄的植物图像的颜色和形态进行了甜菜和杂草的鉴别，并实验得出了疏除机理。所得结果如下：1)在植物与背景土壤的分离方法上，通过实验，在不同的外界光照条件下，可以在RGB空间中确定最佳的分离平面。利用图像的颜色信息对甜菜和杂草进行了判别，利用YIQ颜色系统的q值可以区分出55 ~ 78%的杂草。2)在形状对甜菜和杂草的判别中，利用7个形状特征的线性判别函数能将91.7%的植物区分为甜菜和杂草。然而，在实际应用中，诸如拍摄时的光环境控制和挡风要求，以及叶片部分特征的提取方法等问题还有待解决。3)实验研制了计算机控制的疏除草机构。它有两个摆臂，摆臂上安装有旋转的圆盘刀片，摆臂操作由计算机控制。在绿纸组成的作物行模型上进行了性能测试，其精度在【正负】10mm以内。然而，为了实际应用，对旋转盘进行改进是必要的。

The reduction of herbicide use and working hours in the sugar beet cultivation are an urgent problem. Therefore, the development research of the automatic thinning and weeding machine for direct sowing sugar beet using image processing was carried out. While the discrimination of sugar beet and weed was tried by color and form of the plant image photographed by the video camera, and thinning and weeding mechanism was produced experimentally. Results obtained were as follows;1) On the method for separating the plant from the background soil, it was possible to decide the optimum separation plane in the RGB space as a result of the experiment under various outside light conditions. As a result of trying the discrimination of sugar beet and weed using the color information of the image, 55〜78% weed was able to be distinguished by using Q-value of the YIQ color system.2) In the discrimination of sugar beet and weed by the shape, the linear discriminant function using seven shape characteristic was able to separate 91.7% of plant to sugar beet and weed. However, problems such as light environment control and wind shield requirement in photographing, and extraction method of partial feature of the leaf remained for the practical application.3) Computer controlled thinning and weeding mechanism was produced experimentally. It has two swing arms in which rotating disk blade is installed, and the computer controls the swinging operation. As a result of the performance test in crop row model composed of green paper, it was able to be operated at the accuracy within 【plus-minus】10mm. It is, however, necessary to improve the rotating disk for the practical application.