摘要
景物描述是计算机视觉的一个重要内容。提出了一个利用三维信息对未知景物的描述方法。该方法对三维场景进行物体级和结构级两级描述;根据三维物体分割的结果,首先计算单个物体的最大尺度、形状类型和边集等整体特征,继而推测它们之间的相互位置关系。从数学上推导了“简化边框矩”和“平面惯量张量”的计算公式,并给出了实验结果。
3D scene description is identified as an important component in the understanding of images. In this paper we present a technique for 3D segmentation and scene description on unknown scenes. Object edges are grouped on the basis of their geometrical and connectional relationships. Gross shape features as well as corresponding edge - set of individual object are stored in an object record that acts as an object level descriptor. This information is then used for computing structural relationships between objects under the governing of a set of rules and a rule implementer-a production system. The system fulfills the task for scene description by creating a series of scene records. Experiments have demonstrated its usefulness in certain applications, especially for the obstacle avoidance and path planning of mobile robots in an indoor scene.
出处
《南京邮电学院学报（自然科学版）》
2005年第4期16-23,共8页
Journal of Nanjing University of Posts and Telecommunications
关键词
景物描述
三维分割
图像理解
Scene description
3D segmentation
Image understanding
