Naoki Suganuma

An Obstacle Extraction Method Using Virtual Disparity Image

The driving support system is one of most important research field in intelligent transport system (ITS). In this paper, we address an obstacle extraction method for driving support system. The stereovision system is one of most suit sensor to recognize details of environment. On the other hand, a disparity image obtained by stereovision system has quite a lot of information. Therefore an efficient algorithm to analyze obtained disparity image is strongly demanded. If the road surface is extracted, obstacles can be easily extracted by evaluating whether one object touch on a road or not. In this paper, we propose a novel method to estimate three-dimensional road surface position by using virtual disparity image. Moreover, an obstacle extraction method is expressed.

Precise position estimation of autonomous vehicle based on map-matching

Localization of ego-vehicle is one of the most important technologies for autonomous driving. GNSS/INS systems, which integrate GNSS position and inertial measurement, are very useful for autonomous driving, since the system always estimate smooth and real-time measurement. However, there was a problem that the system produces significant drift after long GNSS signal outage. Therefore, we propose a localization method using both GNSS/INS and lane marker detection. In this system, drift error of GNSS/INS measurement is compensated by lane marker measurement. From some experiments, it was confirmed that robust and smooth localization can be achieved by using this system.

Obstacle detection using Virtual Disparity Image for non-flat road

The driving support is one of most important research area in intelligent transport system (ITS). Moreover, obstacle extraction system is one of most important system. In our previous report, we proposed an obstacle extraction method using stereovision system. In this system, traditional ldquov-disparityrdquo approach was extended to more flexible system by using virtual disparity image. Hereby obstacles can be extracted even if the vehicle has large roll movement. However, our system has still problem when a road shape cannot be approximated as a flat plane. In this paper, we propose road shape recognition method using dynamic programming (DP) and our method is extended to a new method suitable for a non-flat road.

Obstacle Detection Based on Occupancy Grid Maps Using Stereovision System

We previously reported on an obstacle detection method using a stereovision system. The system generated disparity images that include three-dimensional spatial information. Using these images, obstacles could be detected, but some false positives were generated. In this paper, we attempt to eliminate this problem and propose a method that generates Occupancy Grid Maps based on measurements from a stereovision system which leads to robust obstacle detection. Furthermore, it is confirmed that high distance accuracy can be achieved by using our method.

Localization for autonomous driving on urban road

Localization is one of the core techniques of autonomous vehicle. In this paper, we proposed the localization algorithm for autonomous vehicle, and show it is more suitable than the GPS/IMU system.

Simultaneous state recognition for multiple traffic signals on urban road

Automated vehicle researches move on to the public road experiments. This study focuses on a traffic signal detection based on mono-camera, predefined map database and accurate vehicle pose which is estimated by a localization module. By using the map data and the vehicle pose, an image ROI (Region-Of-Interest) can be calculated. This paper handles a situation with multiple signals appeared in an ROI especially for far traffic signals. We propose a probabilistic method for an image location of the target traffic signal in order to realize simultaneous traffic signal recognition for urban driving. In addition, an adaptive contrast updating strategy has been proposed to enhance the contrast of the extracted traffic signal using its probability. Experiments show the performance of the proposed method for real driving data.

Development of autonomous vehicle - Overview of autonomous driving demonstration in ITS world Congress 2013

Recently, fully automated autonomous vehicles have been developed, and field examinations in public road have also been conducted, especially in United States. In this paper, preparation of our laboratory toward field examination of the autonomous vehicle is reported. Additionally, overview of demonstration in the ITS world Congress 2013 (ITSWC2013) is reported.

Fast dynamic object extraction using stereovision based on Occupancy Grid Maps and optical flow

The driving support system is most important research areas in intelligent transport system (ITS). Moreover, obstacle detection is one of the key technologies, and we have proposed such system based on stereovision system. Additionally, to assist driving safely, it is necessary to extract dynamic objects and alert driver faster. In our previous report, we proposed dynamic objects extraction method based on Occupancy Grid Maps. However we found that it takes a long time to detect it. So, in this paper, we propose a method to analyze motion of dynamic objects used 6D information comprised of 3D position and motion of objects, and extract the dynamic objects faster.

Obstacle map generation using virtual disparity image

The driving support system is one of the most important research fields in intelligent transport system (ITS). In this paper, we address an obstacle maps generation method for driving support system. The stereovision system is the one of most suit sensor to recognize details of environment. On the other hand, a disparity image obtained by stereovision system has quite a lot of information. Therefore an efficient algorithm to analyze obtained disparity image is strongly demanded. If the road surface is extracted, obstacles can be easily extracted by evaluating whether one object touch on a road or not. In this paper, we propose a novel method to estimate three-dimensional road surface position by using virtual disparity image. Moreover, an obstacle extraction method is expressed.

Driver's Head Pose Measurement and Corner Center Detection

In this paper, we describe development of drivers head pose measurement and cornea center detection system. Similar researches have been performed so far. Most researches are based on feature point detection by the template matching. In general, the template matching has the weakness to the change in the expansion and contraction, rotation and brightness. To overcome these faults, three-dimensional shape alignment is introduced. Three-dimensional shape alignment is performed between drivers face model and shape data obtained in each time. Then, three-dimensional shapes are analyzed by stereovision. Moreover, shape alignment is performed by ICP registration. ICP registration can achieve high accuracy alignment, but needs huge computational effort, in general. To solve this problem, this paper introduces a reduction in volume of data based on pattern information in the image, coarse-to-fine search, and SIMD computing. Furthermore, this paper describes the cornea center estimation technique. At first, this technique extracts the eye region image from result of the head pose measurement. Afterward, the cornea center is estimated by recognition of eye contour and the ellipse fitting of cornea. It was confirmed by experiments that the proposed technique can measure in high accuracy, and the proposed technique is available in real environment