Fast and precise localization at stop intersections

Abstract

This article presents a practical solution for fast and precise localization of a vehicle's position and orientation with respect to stop sign controlled intersections based on video sequences and mapped data. It consists of two steps. First, an intersection map is generated offline based on street-level imagery and GPS data, collected by a vehicle driving through an intersection from different directions. The map contains both landmarks for localization and information about stop line positions. This information is used in the second step to precisely and efficiently derive a vehicle's pose in real-time when approaching a mapped intersection. At this point, we only need coarse GPS information to be able to load the proper map data.The experimental results show this approach is able to successfully localize a vehicle at stop intersections with decimeter accuracy under various weather conditions and after several months between the map data collection and the test drive.The proposed method enables a variety of functions for future driver assistance systems, such as stop line violation warning, advanced Adaptive Cruise Control/ Traffic Jam Assist functions in cities, or autonomous driving.

Lane-based vehicle localization in urban environments

“…In addition, Barthet al presented a method for localization of a vehicle's position and orientation with respect to stop lines at intersections based on video sequences and mapped data [3]. Brenner provided a landmark map consisting of extracted poles obtained using a mobile mapping van equipped with LIDAR [6].…”

Lane marking based vehicle localization using particle filter and multi-kernel estimation

“…Vehicle localization part uses map information to adjust vehicle location. Common methods in vehicle localization include Kalman Filter [6] [7], Extended Kalman Filter (EKF) [8], optimization algorithm [9], and methods relying on interval analysis [10]. In [11], multi-object localization is performed to improve vehicle localization.…”

“…Vehicle location is then adjusted according to this vector. In [7], all markings from both offline maps and visions at a stop intersection are compared using particle filter to estimate vehicle position. In [6], belief theory combines several criteria to implement map-matching.…”

A vision system for detecting and tracking of stop-lines

“…This is because it can be used to localize ego-vehicle position [1], [2], to understand traffic rules of intersections [3], and to build a map of urban driving environment [1], [4], [5]. For instance, Barth et al manually marked stop-line locations to build a map of stop-lines for localization [1]. Wu and Ranganathan transformed stereo camera images into inverse perspective images to detect road-markings (e.g., directional arrows, railroad crossings, etc.)…”

“…Most works on detecting on-road-surface traffic devices including ours use inverse perspective images because the spatial layout between lane-markings is recovered, e.g., two longitudinal lane-markings are appeared to be parallel. Our approach is different from existing ones in that 1) we examine the geometric layouts between the detected, longitudinal and lateral lane-markings, 2) we develop a Bayes filter to track the detected stop-lines, and 3) our testing scenes are more complex and challenging than those of two previous works [1], [4]. In particular, our testing images are more challenging for stop-line detection in that the lateral and longitudinal lane-markings for stop-lines and for laneboundaries are not always visible because of occlusions by neighboring vehicles and other urban structures, and obsolete paintings.…”