结合DBSCAN聚类的室内场景分割

针对 RGB-D 图像具有丰富的三维几何特征,复杂度高这一具有挑战性的难题,提出一种针对室内场景RGB-D 图像的分割算法.首先,经过 RGB-D 图像过分割生成超像素,并基于超像素之间的距离度量测量超像素之间的相似性;然后,采用 DBSCAN 算法将具有相似的颜色信息和几何信息的超像素聚类到一个分类中.在该聚类过程中,通过限制扩散区域来降低计算复杂度.在室内场景 RGB-D 图像库上大量实验结果表明,文中算法分割精确度和速率均超过了其他算法,证明了其高效性和准确性.

Segmentation for Indoor Scenes Based on DBSCAN Clustering

Aiming at the challenging problems of RGB-D images with rich 3D geometric features and high com- plexity, this paper proposes a segmentation algorithm for RGB-D images of indoor scenes. Firstly, generating superpixels by over-segmentation of RGB-D images and measuring the similarity of two superpixels based on the distance. Then, the DBSCAN algorithm is used to cluster the superpixels with similar color and geo- metric information into the same classification. In the clustering process, we restrict the diffusion area to reduce computational complexity. A lot of experimental results on the database of RGB-D images show that the segmentation accuracy and rate of our algorithm exceed the other algorithms, which proves our algo- rithm’s efficiency and accuracy.