Automatically identifying objects and people left in the interior of vehicles is highly desirable because human monitoring has high running costs and low efficiency associated with it. A new Personal Rapid Transit (PRT) system currently being designed by Advanced Transport Systems Ltd (ATS) features many autonomous vehicles and therefore the task is of particular importance. This paper describes two approaches that use changes in the visual image of the interior to predict the likelihood of left objects and remaining people. The first approach is based on identifying structural differences. The second approach uses a shading model method. A variation of the shading model with information from the colour channels is also described. The results show that the modified shading model approach gives the best performance.
|Translated title of the contribution||Image Change Detection for a Personal Rapid Transit Application|
|Title of host publication||16th European Signal Processing Conference (EUSIPCO), Lausanne|
|Number of pages||4|
|Publication status||Published - Aug 2008|