Light Analysis for Optimized Object Detection with Cameras for Integrated Safety Systems

In order to achieve maximum vehicle safety, advanced driver assistance systems as well as integrated safety systems need to be established in the vehicles. Especially in the field of camera-based driver assistance systems, numerous systems have been introduced in order to detect the vehicle’s surroundings. However, the environmental sensors must be extremely reliable in detecting objects and tracking so that safety systems benefits. This paper presents an approach on how to analyze critical light conditions for cameras and extract disturbing parameters such as describing the glazing effect to improve object detection and tracking with cameras. In a model for detection of interfering and dazzling sources different methods of color and contrast processing were examined to determine critical regions. Subsequently, objects identified by a YOLO algorithm in the test images were compared with the critical areas and light parameters such as wavelength, intensity and color values were extracted for overlapping areas. In this way, on the one hand, the physical properties for reproducibility of the disruptive effect for test systems can be illustrated, and on the other hand, statements about correlations between objects and disruptions can be made.