This paper presents an improved feature-based 3D SLAM approach for a mobile robot equipped with a rotating laser scanner. The features are represented within the SPmodel, with associated planar segment information based on decimated triangle sets. An Extended Kalman Filter is used to build a three-dimensional map of the environment and track the robot's pose along with involved uncertainties. As shown, the resulting maps are highly detailed, useful for higher-level robotic tasks and small in size.