Three-dimensional mapping by direct estimation of small motion using range and color images

Abstract

In this paper, a high-speed three-dimensional map generation method using direct estimation of the motion parameters of the sensor is proposed. The method aligns range images between frames after estimating the motion parameters of the sensor using two kinds of information. One is the relationship of the range images and the motion parameters of a small motion. The other is the relationship of the intensity gradient information of the color images and the motion parameters. The method can create three-dimensional maps quickly because the method does not need to obtain the correspondences of features. Both the processing time required for the motion parameter estimation and the accuracy of the map are verified by experiments using a developed fast RGB-D sensor. It is also shown that the method is applicable to the Kinect sensor.

Keywords

• self localization
• three-dimensional mapping
• motion estimation
• range image
• color image