Abstract-This paper reports on the problem of map-based visual localization in urban environments for autonomous vehicles. Self-driving cars have become a reality on roadways and are going to be a consumer product in the near future. One of the most significant road-blocks to autonomous vehicles is the prohibitive cost of the sensor suites necessary for localization. The most common sensor on these platforms, a three-dimensional (3D) light detection and ranging (LIDAR) scanner, generates dense point clouds with measures of surface reflectivity-which other state-of-the-art localization methods have shown are capable of centimeter-level accuracy. Alternatively, we seek to obtain comparable localization accuracy with significantly cheaper, commodity cameras. We propose to localize a single monocular camera within a 3D prior groundmap, generated by a survey vehicle equipped with 3D LIDAR scanners. To do so, we exploit a graphics processing unit to generate several synthetic views of our belief environment. We then seek to maximize the normalized mutual information between our real camera measurements and these synthetic views. Results are shown for two different datasets, a 3.0 km and a 1.5 km trajectory, where we also compare against the state-of-the-art in LIDAR map-based localization.
This paper reports on a fast multiresolution scan matcher for local vehicle localization of self-driving cars. State-of-the-art approaches to vehicle localization rely on observing road surface reflectivity with a 3D light detection and ranging (LIDAR) scanner to achieve centimeter-level accuracy. However, these approaches can often fail when faced with adverse weather conditions that obscure the view of the road paint (e.g. puddles and snowdrifts), poor road surface texture, or when road appearance degrades over time. We present a generic probabilistic method for localizing an autonomous vehicle equipped with a three-dimensional (3D) LIDAR scanner. This proposed algorithm models the world as a mixture of several Gaussians, characterizing the [Formula: see text]-height and reflectivity distribution of the environment—which we rasterize to facilitate fast and exact multiresolution inference. Results are shown on a collection of datasets totaling over 500 km of road data covering highway, rural, residential, and urban roadways, in which we demonstrate our method to be robust through heavy snowfall and roadway repavements.
This paper reports on a fast multiresolution scan matcher for vehicle localization in urban environments for selfdriving cars. State-of-the-art approaches to vehicle localization rely on observing road surface reflectivity with a threedimensional (3D) light detection and ranging (LIDAR) scanner to achieve centimeter-level accuracy. However, these approaches can often fail when faced with adverse weather conditions that obscure the view of the road paint (e.g., puddles and snowdrifts) or poor road surface texture. We propose a new scan matching algorithm that leverages Gaussian mixture maps to exploit the structure in the environment; these maps are a collection of Gaussian mixtures over the z-height distribution. We achieve real-time performance by developing a novel branch-andbound, multiresolution approach that makes use of rasterized lookup tables of these Gaussian mixtures. Results are shown on two datasets that are 3.0 km: a standard trajectory and another under adverse weather conditions.
Localization is a key task for autonomous cars; systems such as the Google driverless car rely on precise and detailed maps for safe operation. Light detection and ranging (LIDAR) sensors are capable of providing rich informationincluding metric range and point appearance. Robust methods can use this data for vehicle localization by extracting the ground-plane for alignment to a prior map, as in [2].Vision sensors as part of the localization pipeline can be a great enabler for autonomous platforms. Contrary to LIDAR methods, identifying the ground-plane from a camera image is a more challenging task. In our previous work [3], we considered localizing with a monocular camera by aligning the image to a prior map. As we demonstrated, this can be difficult as the groundplane can be obscured by obstacles within view of the camera. In this work, we are interested in partitioning an image stream into obstacles and prior map, as shown in Fig. 1, so we can mask out obstacles during registration.Similar to previous work [1, 4], we use a 1D-Markov random field (MRF) to model a horizontal image partition between obstacles and ground-plane, as in Fig. 1. However, rather than formulating our MRF potentials using image appearance alone (using learned [1] or hand-tuned features [4]), we instead consider the temporal stream of images and inferred parallax. We probabilistically evaluate optical flow against expected optical flow derived from known scene structure and camera egomotion, as in Fig. 2.Our approach is evaluated on a challenging urban dataset with grayscale imagery, where lighting is non-uniform. We demonstrate our proposed algorithms by looking at errors with respect to hand-labeled groundtruth and present results showing improved image registration when obstacle masks are used.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.