This paper presents a novel framework to generate scenarios for the evaluation of autonomous vehicle safety. The test are generated by sampling parameters from probabilistic model based on naturalistic driving data, which consist kinetic information and traffic risk index. proposed also allows estimate comprehensiveness scenario with respect dataset. Further, it flexibly adopts arbitrary definitions relationships between To evaluate effectiveness method, we conducted experiments using...

In unsignalized intersections with poor visibility, proactive driving hazard anticipation is required in order to avoid collisions other traffic participants from a blind corner. However, for elderly drivers and novice drivers, it difficult recognize potential hazardous area select an appropriate speed pass the safely. To assist such driver model which can recommend by learning data of expert based on statistical approach useful assistance system. The proposed method automatically estimates...

This paper presents a comparison of driving behavior modeling methods based on hidden Markov models (HMMs) with driver's eye-gaze measurement and ego-vehicle localization. Original HMMs are sometimes insufficient to model real-world scenarios. To overcome these limitations, extended have been proposed, e.g., autoregressive input-output (AIOHMMs). first details AIOHMMs ways use them for modeling. We compare the performance maneuver discrimination six types HMMs. The data this work was...

Most of the recent automated driving systems assume accurate functioning localization. Unanticipated errors cause localization failures and result in driving. An exact failure detection is necessary to ensure safety driving; however, challenging because sensor measurement assumed be independent each other process. Owing assumption, entire relation ignored. Consequently, it difficult recognize misalignment between map when partial overlaps with map. This paper proposes a method for using...

This paper presents a longitudinal and lateral motion planning method for driver assistance systems in urban scenarios. We proposed Bayesian network based planner to generate the trajectory, including positions velocities path through multiple traffic participants. To design probabilistic models which represent lane keeping maneuver an obstacle avoidance maneuver, we collect analyze natural driving data. Then, it is difficult collision data real world. Therefore, inverse probability from...

[abstFig src='/00280006/11.jpg' width='300' text='Road detection method with HOG and SVM' ] This paper describes a road area using support vector machine (SVM) histogram of oriented gradient (HOG) features. The boundary lines have many features, such as changes in height, color, brightness, but these are sensitive to noise. In terms robustness, it is difficult match the on 2D maps. Localization methods texture matching accurate, they disadvantages related adapting environment. We therefore...

In this paper, we propose a method of generating individual facial models based on generic model with not only skin but also additional data such as skull model, expression muscle etc. When the is transformed into by GFFD (Generalized Free-Form Deformation), are automatically altered suitable state for faces at same time. By using method, can easily do simulation or hairstyle without making each model.

Traffic accidents and near miss incident data are important to extract remarkable driving scenarios for the development of autonomous vehicles improvement transportation systems. In order analyze accident situations, it is design an annotation format that can comprehensively describe behavior traffic participants environments. The authors have been doing project collect incidents provide database over 10 years. keywords in this consists compound words easily understood by humans, however,...

This paper presents motion prediction model of cyclist based on potential field for a hazard-anticipatory collision avoidance braking system to enhance the performance and secure smoothness driving. The target situation is chosen as scene that overtakes pedestrian or another slowcyclist traffic survey. 1st order predictionmethod reaches its limit under runs towards conducted current position velocity within finite time horizon. If overtake action can be predicted before changes moving...

This paper proposes an interactive 3D animation system specifically aiming efficient control of human motion. However there are various commercial products for creating movies and game contents, those still difficult to deal with non-professional users. To ease the creation process encourage utilize general users, e.g., in field such as education, medicine so on, we propose a using Kinect. The data skeleton models motion estimated by Kinect is processed generate Motion Graph finally...

This paper proposes a driving safety evaluation method based on the parameter analysis of driver model. Since abilities, such as reaction, recognition and judgment time decrease with aging, it is difficult for elderly drivers to avoid traffic accidents instantly. Therefore, proactive important prevent accidents. Then, criteria behavior are useful encourage keep driving. And these develop an advanced assistance system drivers. To evaluate behavior, we use parameters risk potential model which...

We propose a novel method of generating 3D virtual foot model with use measuring technique based on multiple camera image data. Our costs very low and for each person can be generated easily rapidly acceptable error range. have to estimate the cause errors that occur in whole process endeavor lower errors. The are mainly classified into two kinds. One occurs transforming our original free deformation called GFFD other obtaining coordinates feature points. So, we devoted ourselves do enough...

Recently importance of tablet devices with touch interface increases significantly, because they attract not only mobile phone users, but also people all generations including small kids to elderly people. One key technology on those is 3D graphics; some them exceed the ability desktop PCs. Although currently graphics are used for game applications, it has a great potential other purposes. For example, an interactive animation system been developed recently and attracts many With system,...

Nowadays tablet devices are considered most important category of information-processing devices, because they attract not only PC users, but also all generation people including small kids to elderly people. However, 3D graphics is mostly used by games and few well-known applications exist. In this paper, we propose a new content creation system for interactive animation, which dynamically changes the story dependent on users' gesture. We introduce state transition model efficiently achieve...

Advanced driving assistance systems (ADAS) for various scenarios are developed. To develop and to evaluate the ADAS, a large quantity high quality database of actual data is needed. In this paper, we propose trajectory estimation method based on 3D point cloud. The registration methods cloud (e.g. ICP) has advantages in accuracy local (short range) motion. However, since estimate relative motions frame by frame, errors accumulated it difficult apply long distance analysis. solve problem,...

This paper describes the development of a new localization method using simple 2D map. The general for map is to match between boundary line on and detected boundaries in real world, i.e. lane markers, walls building, or curbs. It necessary create differential process order detect lines road. If these methods try small changes, false-detection rate increases due enhanced noises; if they reduce effect noise, are misdetected. estimation result this deteriorates drastically misdetection occurs....

This paper describes an estimation method of ego-vehicle’s position using a 2D map. The general approach for localization is to match the white lines between map and real world. However, such approaches suffer from changing environmental conditions painting clearance. If considerable changes occur, false-detection rate increases due noises miss-detected lines. In order solve this issue, based on holistic road area detection proposed. First, extracted predefined boundary Next, world detected...

This paper proposes a novel method of generating 3D virtual nail model from multiple camera image data based on the GFFD deformation method. models for individuals can be generated easily and rapidly with considerably low cost. The outline whole procedures is as follows. First, we take three pictures an individual's nail. Second, extract feature points in each image, which arranged surface Third, obtain corresponding positions using stereo computing Finally, deform generic into one obtained...