水果采摘机器人视觉系统的目标提取

在田间对作物的果实图像进行实时、准确地目标识别提取，是采摘机器人视觉系统的关键技术，而目标提取的实质是图像分割。大部分水(蔬)果处于采摘期时，表面颜色与背景颜色存在较大差异。而同一品种果实表面颜色相近，体现为在色彩空间果实表面颜色和背景颜色存在着不同的分布特性。根据这一特性，提出了一种基于色彩空间参照表的适用于水果采摘机器人视觉系统果实目标提取的图像分割算法。该算法先由果实样本图像建立色彩空间参照表，再根据色彩空间参照表采用一种类似于“卷积”的方法进行图像分割。与现有其他方法比较，本方法基于彩色的信息处理，可将背景除去得更干净；对背景不做分割处理、无复杂运算，有利于机器人实时图像处理。采用该算法分别对草莓、橙子、西红柿的图像在L^*n^*6^*，Hsv，YCbCr色彩模型下进行了实验，结果显示该算法在这些色彩模型下均可取得理想的图像分割效果。

Object extraction for the vision system of fruit picking robot

The key technology of the vision system of fruit picking robot is to extract fruit object from images quickly and exactly. The substance of object extraction for the vision system of robot is image segmentation. Among lots of fruits and vegetables, there are a considerable color difference between their ripe fruits and their backgrounds, yet the fruits of the same type are of almost the same color. Then the distribution of color between the ripe fruits and the backgrounds is different in Color Space. Based on these features, an algorithm to extract fruits object from images, which is based on the approach of color space reference list, was put forward. A color space reference list was built from fruits swatch images in the algorithm. The fruit object was extracted by using the list. Compared with other image segmentation arithmetic, this arithmetic base on color images, then the background will be more clean when the image is processed. Because of that there is no complex operation, this arithmetic is propitious to real time image process. Experiments on strawberry, orange and tomato were respectively conducted by adopting several color models usually used in image processing. The results of these experiments showed that ,by using this algorithm, expected results can be achieved under L *a *b,HSV,YC b C r color models.