Among ITS applications, it is very important to acquire detailed statistics of traffic flows. For that purpose, vision sensors have an advantage because of their rich information compared to such spot sensors such as loop detectors or supersonic wave sensors. However, for many years, vehicle tracking in traffic images has suffered from the problems of occlusion effect and illumination effect. In order to resolve occlusion problems, we have been proposing the Spatio-Temporal Markov Random Field model(S-T MRF) for segmentation of Spatio-Temporal images. This S-T MRF model optimizes the segmentation boundaries of occluded vehicles and their motion vectors simultaneously by referring to textures and segment labeling correlations along the temporal axis as well as the spatial axis. Consequently, S-T MRF has been proven to be successful for vehicle tracking even against severe occlusions found in low-angle traffic images with complicated motions, such at highway junctions. In addition, in this paper, we define a method for obtaining illumination-invariant images by estimating MRF energy among neighbor pixel intensities. These illumination-invariant images are very stable even when sudden variations in illumination or shading effect are occurred in the original images. We then succeeded in seamlessly integrating the method for MRF energy images into our S-T MRF model. Thus, vehicle tracking was performed successfully by S-T MRF, even against sudden variations in illumination and against shading effects . Finally, in order to verify the effectiveness of our tracking algorithm based on the S-T MRF for practical uses, we developed an automated system for acquiring traffic statistics out of a flow of traffic images. This system has been operating continuously for ten months, and thus effectiveness of the tracking algorithm based on S-T MRF model was proven.

Pedestrian dead reckoning (PDR) is an effective positioning means that can be used in urban-canyon environments where the accuracy of GPS is significantly degraded. Magnetic disturbances caused by artificial objects affect the accuracy of positioning if the PDR system uses a magnetometer to estimate the heading direction. Conventional PDR systems consider magnetic disturbances as unpredictable error sources, but the error becomes predictable and removable if the amount of the deviation in the magnetic field can be calculated at any position. In this study, we propose a method to correct the heading direction by referring to a map of magnetic deviation. The experimental results show that our method reduced the error in the heading direction caused by magnetic disturbances. Our approach removed the error components that differ depending on the position, and consequently, the resultant trajectory represented better the shape of the true trajectory.

ITS R&D includes wide variety of research area such as mechanical engineering, road engineering, traffic engineering, information and communication engineering, and electrical engineering. In spite of initiatives across the variety of engineering is essential to solve the problems of practical social systems, it is difficult to collaborate among engineering. Based on the joint research of the Japan Society of Civil Engineers and the Institute of Electrical Engineers held at the Great East Japan Earthquake, this paper discusses about necessity of collaboration among academies on ITS R&D. International collaboration is also important for ITS R&D. Asian countries could share the same problems and solutions, since many of mega cities exist in Asia region and they suffers from heavy traffics. Therefore, we need to discuss the common solution to our problems.

We have developed a dedicated onboard “sensor” utilizing wireless communication devices for collision avoidance around road intersections. The “sensor” estimates the positions of transmitters on traffic participants by comparing the strengths of signals received by four ZigBee receivers installed at the four corners of a vehicle. On-board sensors involving cameras cannot detect objects in non line-of-sight (NLOS) area caused by buildings and other vehicles. Although infrastructure sensors for vehicle-to-infrastructure (V2I) cooperative systems can detect such hidden objects, they are substantially more expensive than on-board sensors. The on-board wireless “sensor” developed in this work would function as an alternative tool for collision avoidance around intersections. Herein, we extend our previous work by considering a road surface reflection model to improve the estimation accuracy. By using this model, we succeeded in reducing the error mismatches between the observed data and the calibration data of the estimation algorithm. The proposed system will be realized on the basis of these enhancements.

Autonomous driving is not only required to detect pedestrians around vehicles, but also expected to understand the behaviors of pedestrians. Pedestrian body orientation and head orientation are the relevant indicators of the pedestrian intention. This paper proposes an accurate estimation system to recognize the pedestrian body orientation and the pedestrian head orientation from on-board camera and inertial sensors. The proposed system discretizes the body orientation and the head orientation into 16 directions. In order to achieve the accurate orientation estimation, a novel training database is established, which includes strongly labeled data and weakly labeled data. Semi-Supervised Learning method is employed to annotate the weakly labeled data, and to generate the accurate classifier based on the proposed training database. In addition, the temporal constraint and the human physical model constraint are considered in orientation estimation, which are beneficial to the reasonable and stable result of orientation estimation for the pedestrian in image sequences. This estimated result is the orientation in camera space. The comprehension of the pedestrian behavior needs to be conducted in the real world space. Therefore, this paper proposes to model the motion of the host vehicle using inertial sensor, then transforms the estimated orientation from camera space to the real world space by considering the vehicle and pedestrian motion. The represented orientation indicates the behavior of the pedestrian more directly. Finally, a series of experiments demonstrate the effectiveness of the proposed pedestrian orientation system.

This paper presents a general algorithm for pedestrian detection by on-board monocular camera which can be applied to cameras of various view ranges. Under the assumption that motion of background can be nearly approximated as a linear function, the Spatio-Temporal MRF (S-T MRF) model segments foreground objects. The foreground objects contain both pedestrian and non-pedestrian urban objects, verification was conducted by a cascaded classifier. However, the segmentation results based on motion were not exactly fit into pedestrian on the image so that shrunk or inflated pedestrian were generated. This causes errors on extracting pedestrian trajectory. For precise positioning, we implemented two types of feedback algorithm for ROI correction using the Kalman filter and the voting result of Motion-classifier and HOG-classifier. We confirmed that those ROI Corrections successfully extract precise area of pedestrian and decrease the false negative rate. Elaborately extracted pedestrian trajectory could be used as a useful source for predicting collision to pedestrian.

The limitation of the GPS in urban canyon has led to the rapid development of Wi-Fi positioning system (WPS). The fingerprint-based WPS could be divided into calibration and positioning stages. In calibration stage, several grid points (GPs) are selected, and their position tags and featured access points (APs) are collected to build fingerprint database. In positioning stage, real time measurement of APs are compared with the feature of each GP in the database. The k weighted nearest neighbors (KWNN) algorithm is used as pattern matching algorithm to estimate the final positioning result. However, the performance of outdoor fingerprint-based WPS is not good enough for pedestrian navigation. The main challenge is to build a robust fingerprint database. The received number of APs in outdoor environments has large variation. In addition, positioning result estimated by GPS receiver is used as position tag of each GP to automatically build the fingerprint database. This paper studies the lifecycle of fingerprint database in outdoor environment. We also shows that using long time collected data to build database could improve the positioning accuracy. Moreover, a new 3D-GNSS (3D building models aided GNSS) positioning method is used to provide accurate position tags. In this paper, the fingerprint-based WPS has been developed in an outdoor environment near the center of Tokyo city. The proposed WPS can achieve around 17 meters positioning accuracy in urban canyon.

Accurate pedestrian navigation remains a challenge in urban environments. GNSS receiver behaves poorly because the reflection and blockage of the GNSS signals by buildings or other obstacles. Integration of GNSS positioning and Pedestrian Dead Reckoning (PDR) could provide a more smooth navigation trajectory. However, the integration system cannot present the satisfied performance if GNSS positioning has large error. This situation often happens in the urban scenario. This paper focuses on improving the accuracy of the pedestrian navigation in urban environment using a proposed altitude map aided GNSS positioning method. Firstly, we use consistency check algorithm, which is similar to receiver autonomous integrity monitoring (RAIM) fault detection, to distinguish healthy and multipath contaminated measurements. Afterwards, the erroneous signals are corrected with the help of an altitude map. We called the proposed method altitude map aided GNSS. After correcting the erroneous satellite signals, the positioning mean error could be reduced from 17 meters to 12 meters. Usually, good performance for integration system needs accurately calculated GNSS accuracy value. However, the conventional GNSS accuracy calculation is not reliable in urban canyon. In this paper, the altitude map is also utilized to calculate the GNSS localization accuracy in order to indicate the reliability of the estimated position solution. The altitude map aided GNSS and accuracy are used in the integration with PDR system in order to provide more accurate and continuous positioning results. With the help of the proposed GNSS accuracy, the integration system could achieve 6.5 meters horizontal positioning accuracy in urban environment.

Highly accurate pedestrian position information is required in many applications, especially in automatic driving system. Global Positioning System (GPS) developed by American has proven itself reliability in most of the environments. Unfortunately, urban areas contain the signal reflection, known as multipath and non-line-of-sight (NLOS) effects. In addition, the lake of line-of-sight (LOS) satellites caused by the blockage of skyscrapers also severely degrades the accuracy and availability of the GPS positioning. To solve these problems, a solution that interoperated several Global Navigation Satellite Systems (GNSSs) is proposed. However, the actual difficulty of satellite positioning in urban area is the distorted satellite distribution. This paper proposes a GPS with 3D map ray tracing positioning method to conquer the difficulty. The proposed method takes the advantage of the non-LOS (NLOS) and uses it as an additional measurement. Significantly, these measurements are sourced from the satellites that should be blocked. Thus, the dilution of precision (DOP) can be greatly improved. To verify the performance of the proposed method, real data is collected at Tokyo urban area. This paper compares the performance of GPS/GLONASS and the proposed GPS with 3D map ray tracing methods. The results reveals the proposed method is capable of identifying which side of street the pedestrian stands and the GPS+GLONASS method is not.