GENERATING TRAIN SIDE VIEWS FROM VIDEO SEQUENCES FOR MICROPHONE ARRAY PASS-BY MEASUREMENTS
Johannes Stier, Michael Beitelschmidt · 2012
A significant field of application for microphone arrays is the pass-by measurement of moving objects, mainly trains. In a former article, a special time-domain beamforming algorithm has been introduced which is able to evaluate whole pass-by measurements with respect to computational performance and memory. The evaluation results in a sound pressure level mapping of the train passing. But this mapping only shows the sound sources and their positions represented by coordinate values resulting in interpretation difficulties. A possible solution is the overlay of the sound pressure level mapping with a sketch or a picture of the whole train side view. Since this kind of picture is only available in very few cases or only for single parts of the train, an approach was developed which allows the picture generation of a train side view independently from its composition. In addition to the array measurements, the train’s passing is recorded with a video camera simultaneously. By analyzing the video frames with the SIFT algorithm and combining them based on the gained information, a side view picture of the whole train can be created. The approach presented, including its advantages and requirements, is described in detail. Additionally, further developments are exposed and the capabilities of this approach are demonstrated with an example taken from a variety of measurements completed so far.