中文摘要：

为了获取更加清晰、更多细节的轮廓特征，充分利用Kinect传感器获取的RGB-D图像信息，将结构化的随机森林作为分类器，提出一种更加精确的轮廓提取器。首先，将RGB-D图像的多种信息利用数学公式表示出来；然后利用BSD500数据集以及NYU深度数据集训练结构化的随机森林算法，核心是将给定节点的结构化标签映射到一组离散标签；最后，利用该随机森林算法对RGB-D图像信息进行分类，得到图像轮廓。针对细节不同的四种场景图像进行对比实验，结果表明，经改进后的算法得到的轮廓效果更加清晰、准确。

英文摘要：

In order to obtain more clear and more detailed contour features, and fully use the RGB-D image information from Kineet, this paper proposed a more precise contour extractor using structured random forests as a classifier. First, it expressed the various information from RGB-D image with mathematical formulas. Then, it used the BSD500 segmentation dataset and NYU depth dataset to train the structured random forests. The core idea was to map all the structured labels at a given node into a discrete set of labels. Finally,it used the random forests to classify the RGB-D image information and got the image contours. In the experiments, it took four images with different details. The results of the experiments demonstrate significant improve- ments of this improved algorithm over the state-of-the-art.