The subject of this paper is detection of railway infrastructure objects based on mobile laser scanning. The first important step, made before proceeding with correct detection, is to remove the measurement noise. In the case of mobile scanning noise is a key issue, since it introduces a large distortion of the measurement data. In the first part of the article a brief description of the programs in terms of simple geometric filters which both remove noise and carry out simple operations of the classification (for example, the separation of objects spaced by a certain depth). Object recognition is possible only after the removal of the noise This is a relatively new problem, opening a wide field for analysis and research. So far several methods have been described for the classification of the mobile data. Some algorithms based on methods derived from the image processing theory. Scattered cloud of points is stored in a regular grid, the pixel values correspond to the depth or intensity of the laser data. The image-based methods: algorithms using morphological filters and retrieval algorithms. Other methods are based on the detection of objects from the original cloud of points. An example is the method based on RANSAC algorithm. An analysis of filtering algorithms revealed that mobile laser scanning can be a reliable information source to extract objects.