A method for the reproduction of vehicle test drives for the simulation based evaluation of image processing algorithms

Abstract

Since the 1980's computer aided engineering (CAE) and simulation (CAS) methods are widely used in the car development process. First the focus was set on crash simulation and car body design. Since a few years due to the rapidly increasing complexity of new electronic systems these techniques are applied in the development and test processes of advanced driving assistance systems (ADAS). There are a lot of motivations, such as cost reduction and safety issues. As a result of these efforts it is possible to reduce the development cycle times and to increase the quality of the products drastically. In this publication the next step in computer aided engineering is taken. The reproduction of a real test drive is described and simulation quality is examined by a vehicle detection algorithm. First an introduction about the motivation for using synthetic generated images for the test of computer vision applications is given. The next section of this paper deals with vehicle detection in general and gives a more detailed introduction to appearance based methods, especially the used Haar-wavelet based approach. The method for the reproduction of a real test drive is described in detail and the application is shown. The results are evaluated by comparing the detection results of classifier between simulation and reality frame by frame. Further the relevant frames of the evaluation are analysed. Finally, a conclusion and outlook about the possible applications and improvements is made