Nuanced Collision Detection for Parked Vehicles

Abstract

Today’s omnipresence of smartphones enables the development of systems detecting hazardous situations and collisions in road traffic as well as recognizing and predicting individual driving behavior. This is accelerated by comprehensive equipment of sensors as well as the availability of mobile data transmission networks. Commonly, they rely on motion sensors in combination with frequency analysis and distance algorithms, e.g., dynamic time warping. Moreover, they also use audio and video features for support. Despite a whole range of work in the area of analyzing vehicle’s signals and to the best of our knowledge, there is no reliable procedure for nuanced collision detection in combination with a smartphone as a sensor platform known to us. A segmented detection of collisions itself would be of great value, e.g., in order to relief customers of a car-sharing or car rental service. After a short introduction of related work, we frame the underlying problem of this paper and present our concept for a nuanced collision detection on specific segments of a parked RC-vehicle under the usage of a static-placed smartphone as the only sensor platform. Therefore, we want to create acceleration signal patterns during an offline phase by analyzing recorded sensor data for impacting pushes on specific vehicle segments in x-,y-, and z-direction. Our work, based on repeated collisions with a RC-vehicle, might function as a proof of concept which may be applied to real vehicles later on.

Publication
12. GI/ITG KuVS Fachgespräch Ortsbezogene Anwendungen und Dienste (LBAS 2015)