The LiDAR (Light Detection And Ranging) technology is becoming a more and more popular method to collect spatial information. The acquisition of 3D data by means of one or several laser scanners mounted on a mobile platform (car) could quickly provide large volumes of dense data with centimeter-level accuracy. This is, therefore, the ideal solution to obtain information about objects with elongated shapes (corridors), and their surroundings. Point clouds used by specific applications must fulfill certain quality criteria, such as quantitative and qualitative indicators (i.e. precision, accuracy, density, completeness).Usually, the client fixes some parameter values that must be achieved. In terms of the precision, this parameter is well described, whereas in the case of density point clouds the discussion is still open. Due to the specificities of the MLS (Mobile Laser Scanning), the solution from ALS (Airborne Laser Scanning) cannot be directly applied. Hence, the density of the final point clouds, calculated as the number of points divided by "flat" surface area, is inappropriate. We present in this article three different ways of determining and interpreting point cloud density on three different test fields. The first method divides the number of points by the "flat" area, the second by the "three-dimensional" area, and the last one refers to a voxel approach. The most reliable method seems to be the voxel method, which in addition to the local density values also presents their spatial distribution.