Previously, we introduced a passengers' counting algorithm in public rail transport. The main disadvantage of that algorithm is it lacks automatic event detection. In this article, we implement two automatic wavelet-based passengers counting algorithms. The new algorithms employ the spatial-domain Laplacian-of-Gaussian-based wavelet, and the frequency-domain applied Non-Linear Difference of Gaussians-based wavelet bandpass video scene filters to extract illumination invariant scene features and to combine them efficiently into the background reference frame. Manual segmentation of the scene into rectangles and tiles for detecting an object as seated is no longer needed as we now apply a boundary box tracker on the segmented moving objects' blobs. A scene map is combined with the wavelet-based methods and the boundary box for multi-camera object registration. We have developed a novel holistic geometrical approach for exploiting the scene map and the recorded video sequences from both cameras installed in each train coach to separate the detected objects and locate their positions on the scene map. We test all the algorithms with several video sequences recorded from the both cameras installed in each train coach. We compare the previously developed non-automatic passengers' counting algorithm with the two new automatic wavelet-based passengers' counting algorithms, and an additional spatial-domain automatic non-wavelet based Simple Mixture of Gaussian Models algorithm.