Mountney Peter, Stoyanov Danail, Davison Andrew, Yang Guang-Zhong
Royal Society/Wolfson Foundation Medical Image Computing Laboratory, Imperial College, London SW7 2BZ, UK.
Med Image Comput Comput Assist Interv. 2006;9(Pt 1):347-54. doi: 10.1007/11866565_43.
Minimally Invasive Surgery (MIS) has recognized benefits of reduced patient trauma and recovery time. In practice, MIS procedures present a number of challenges due to the loss of 3D vision and the narrow field-of-view provided by the camera. The restricted vision can make navigation and localization within the human body a challenging task. This paper presents a robust technique for building a repeatable long term 3D map of the scene whilst recovering the camera movement based on Simultaneous Localization and Mapping (SLAM). A sequential vision only approach is adopted which provides 6 DOF camera movement that exploits the available textured surfaces and reduces reliance on strong planar structures required for range finders. The method has been validated with a simulated data set using real MIS textures, as well as in vivo MIS video sequences. The results indicate the strength of the proposed algorithm under the complex reflectance properties of the scene, and the potential for real-time application for integrating with the existing MIS hardware.
微创手术（MIS）具有减少患者创伤和恢复时间的公认优势。在实际操作中，由于3D视觉的丧失以及摄像头提供的狭窄视野，MIS手术面临诸多挑战。受限的视野会使在人体内部进行导航和定位成为一项具有挑战性的任务。本文提出了一种强大的技术，用于构建场景的可重复长期3D地图，同时基于同步定位与地图构建（SLAM）恢复相机运动。采用了一种仅基于视觉的顺序方法，该方法提供6自由度相机运动，利用可用的纹理表面并减少对测距仪所需强平面结构的依赖。该方法已通过使用真实MIS纹理的模拟数据集以及体内MIS视频序列进行了验证。结果表明了所提出算法在场景复杂反射特性下的优势，以及与现有MIS硬件集成进行实时应用的潜力。
