A moderate-size video project.  Anybody interested in automatic segmentation of video sequence, have a look at how it's done here.  It uses video segmenter + audio merger: segmentation using video information tends to cut too much (such as a news program where an anchor person is speaking while many different scenes are showing - they are about the same subject, so should not be segemented, but using only visual information it cuts all different shots).  So after this too much cut, use audio analysing and merge some cuts back to one unit.  Not many systems use audio information to complement indexing process.  Link to VISION