CORRESPONDENCE-FREE RANGE CAMERA ORIENTATION
Three-dimensional Simultanous Localization and Mapping (SLAM) using range cameras is a very attactive technique for the aquisition of 3d models of (especially) indoor scenes. The localisation aspect refers to the use of an input device, such as a video camera or a high speed range camera that moves through the scene while recording, requiring continuous real-time updates of the orientation parameters. Whereas the modelling potential for SLAM with range camera input is much better than with video, the localization as presented in current literature is still similar to video frame co-registration. as far as it relies on detection of corresponding features in different (successive) frames of the sequence. This is a weakness. as detection of correspondences is time consuming and error-prone, calling for additional processing effort further in the pipeline. The current paper presents an algorithm for estimation of relative orientation parameters by least squares inversion of a geometric transformation model of range images, which takes relief displacement into account. Preliminary results shown are promising in terms of accuracy and have satisfactory performance, based on which a number of improvements are suggested. These are expected to lead to real-time orientation of range image sequences.
SLAM 3d data acquisition range cameras real time least squares relative orientation.
Ben Gorte
Optical and Laser Remote Sensing, Delft University of Technology, Kluyverweg l, 2629 HS Delft, the Netherlands
国际会议
武汉
英文
245-250
2011-06-26（万方平台首次上网日期，不代表论文的发表时间）