In this paper the Simultaneous Localization And Mapping (SLAM) problem in unknown indoor environments is addressed. A probabilistic approach integrating FastSLAM algorithm and a line feature map is developed and validated. Experi- mental validation is performed by a smart wheelchair equipped with proprioceptive and exteroceptive sensors in an office like environment where loop closing is achieved without any dedicated algorithm. Geometric hypothesis of orthogonal line features are considered to enhance the performance of the algorithm in the considered en- vironment. The proposed approach results in a computationally efficient solution to the SLAM problem and the high quality sensor measurements allow to main- tain a good localization of the mobile base and a compact representation of the environment.