Volume 49 Number 1 (2018.3)
Special Feature
Spatial Information Technology towards Intelligent Vehicle Systems
Special Feature
Overview
Research Reports
-
2. Automatic Lane-level Map Generation Using Low-cost Sensors in Urban Traffic Scenarios
(3625kB)pages 1-8
Chunzhao Guo, Kiyosumi Kidono and Junichi Meguro
This work proposes an automatic lane-level map generation approach using low-cost sensors. Specifically, it integrates the road surface images acquired from a rear-view camera into large synthetic orthographic images, and subsequently constructs the lane-level map from the integrated images together with a large amount of vehicle trajectories without manual processing.
-
3. Precise Dead-reckoning Based on Multi-sensor Fusion
(3066kB)pages 9-20
Kojiro Takeyama and Yoshiko Kojima
In this study, the accuracy of visual-odometry has been improved via time-series tightly coupled integration of GPS Doppler and INS that increases the availability of absolute heading estimation in an urban environment. The experiment showed the error of visual-odometry has been reduced to about 1/4 that of the conventional method.
-
pages 21-31
Takeo Kato, Chunzhao Guo, Kiyosumi Kidono, Yoshiko Kojima and Takashi Naito
An effective and low-cost sparse feature interaction descriptor (SpaFIND) is proposed for pedestrian protection systems in economy cars, which have limited computational power. SpaFIND selectively computes the correlations among adjacent components of the histogram of oriented gradients. The balance between detection performance and computational load is demonstrated through experiments.
-
5. Compact Imaging LIDAR with CMOS SPAD
(2098kB)pages 33-40
Hiroyuki Matsubara, Mitsuhiko Ohta, Mineki Soga, Isamu Takai and Masaru Ogawa
We studied a light detection and ranging (LIDAR) system with a high resolution and a long distance range to sense the surroundings of a vehicle during automated driving. We realized a high-resolution compact LIDAR using a single-photon avalanche diode (SPAD) and a coaxial scanning optical system.
-
6. Free-viewpoint Image Reconstruction for ADAS Virtual Assessment
(1757kB)pages 41-48
Takashi Machida, Satoru Nakanishi and Tomokazu Sato
We propose a virtual assessment methodology for Advanced Driving Assistance Systems (ADASs) that operates by reconstructing images at an arbitrary viewpoint from a three-dimensional (3D) point cloud and an omnidirectional image sequence. Our results show that free-viewpoint images are similar to real images with respect to image processing algorithm outputs.
-
7. Layered Vehicle Control Architecture Coordinated between Multiple Edge Servers
(1016kB)pages 49-57
Kengo Sasaki, Naoya Suzuki, Satoshi Makido and Akihiro Nakao
We propose a remote vehicle control architecture using Mobile Edge Computing (MEC). Our proposed architecture is composed of Lower Edge Servers and Upper Edge Servers, and deploys these servers at different locations in the mobile network. To use the features of both edge servers, the proposed architecture switches control between the edge servers according to the communication delay.