Vehicle Offline Localization Based on Computer Vision: an Approach Based on Image Matching & Retrieval Algorithms and Implementation

Abstract

Driver assistant systems have approved their essential role in increasing safety in the driving environment. One of their main features is the ability to provide information about location of the vehicle so that the driver can understand how and where to go. Moreover, automatically report the emergency in case of accidents to accelerate the rescue process. Unfortunately, poor Internet connection in many places and the inaccurate GPS information on traveling roads, in particular, make the localization process even harder. In this paper, we propose an offline localization system that can estimate the location of the vehicle in real-time using edge devices. Our proposed system consists of three main components; the first component is the image matching part to estimate the rotation and transition between two images; the second component is the dynamic objects (vehicles, pedestrians, and etc.) segmentation to filter out the matches from them; and finally, an image retrieval system to auto-correct the location using predefined coordinates of famous landmarks.

Keywords

• computer vision vehicle localization video-based gps