Language：Japanese   Publishing type：Research paper (scientific journal)  

<p>This paper presents motion prediction model of cyclist based on potential field for a hazard-anticipatory collision avoidance braking system to enhance the collision avoidance performance and secure the smoothness of driving. The target situation is chosen as the scene that a cyclist overtakes a pedestrian or another slowcyclist based on the traffic survey. The 1st order motion predictionmethod reaches its limit under the situation that a cyclist runs towards a pedestrian or another cyclist as the prediction is conducted based on the current position and velocity of cyclist within a finite time horizon. If the overtake action of cyclist can be predicted before the cyclist changes moving direction, the vehicle maneuver to avoid collision with cyclist can be executed in advance without activating harsh braking. The trajectories of cyclists overtakes a pedestrian and a slower cyclist is measured to find the characteristics in overtake action.Motion prediction model of cyclist based on potential field is constructed by considering the trajectory analysis. The effectiveness of the proposed prediction model in the target scenario is verified by comparing the measured trajectory with the calculated data based on 1st order prediction and the proposed method.</p>

Language：Japanese   Publishing type：Research paper (scientific journal)  

This paper presents a path planning method for driver assistance systems in mixed urban scenarios. We proposed an inverse collision probability model based motion planner to generate the trajectory, including the positions to path through multiple traffic participants. To design the probabilistic models which represent an obstacle avoidance maneuver, we collect and analyze driving data of expert driver. Then, we conceder the direction and the value of time to collision of the target traffic participant. Then, we show the difference of driving behaviors of the expert drivers according to the difference of the direction of the target. The proposed method generates the optimal trajectory plan by using the global optimization algorithm named Belief Propagation. Finally, we show the evaluation experiment that compares the difference between the trajectories generated by the proposed method and natural driving data.

CiNii Books

Language：Japanese   Publishing type：Research paper (scientific journal)   Publisher：The Japan Society of Mechanical Engineers  

<p>This paper describes the development of a new localization method using a simple 2D map. The general method for localization using a 2D map is to match between the boundary line on the 2D map and the detected boundaries in the real world, i.e. lane markers, walls of building, or curbs. It is necessary to create a differential process in order to detect boundary lines of the road. If these methods try to detect small changes, the false-detection rate increases due to enhanced noises; if they try to reduce the effect of noise, boundary lines are misdetected. The estimation result of this method deteriorates drastically if the false-detection or misdetection of the boundary line occurs. We propose the new localization method based on the road area detection. First, the road map is extracted from the boundary lines on a 2D map. Next, the road plane image is made by the road area detected using LiDAR in the real world. Finally, the road map and the road plane image are matched by image registration. We confirmed that the proposed method have accurate estimation performance with several noise and low-cost calculation from the simulation. And we conducted the performance validation of proposed method in the real world. As a result, we confirmed the same tendency as simulation results.</p>

DOI： 10.1299/transjsme.17-00014

Language：English   Publishing type：Research paper (other academic)  

Language：English  

Web of Science

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：English  

Web of Science

Language：Japanese   Publishing type：Research paper (scientific journal)  

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

Language：English  

Web of Science

Language：English  

Web of Science

Language：Japanese   Publishing type：Research paper (scientific journal)  

In this paper, we present an interactive 3D animation creation system that enables the user to capture, edit the moving objects and share them easily on the web. Recently, educational concept called active leaning has been becoming important in the field of education. The difference between the active leaning and usual education is that the learner actively absorbs the educational information not only from the lecturer, but also from other media contents such as videos and images. Therefore, various kinds of media contents aiming for active learning are required to help the learners. As the one of such media contents, we developed a system that can create 3D video data and share them on a web service. Concretely, in our system, the 3D contents are acquired by using Kinects on a graphical user interface. By using the proposed system, the user can edit and integrate the other contents such as music and images easily. The contents can be uploaded on the web server that provides the experience to manipulate the 3D animation intuitively through a touch interface. In this paper, we explain the overview of the proposed system and the details of the our algorithms to reconstruct 3D data. The experiment shows our system can obtain the 3D data efficiently by using a novel approach. As the conclusion, we show some demos to validate that the system is available for educational use and a user can operate the proposed system efficiently.

Language：English  

Web of Science

Language：Japanese   Publishing type：Research paper (scientific journal)  

In this paper, we propose a marker-less motion capture method for a human face. We detect the movement of a face as the deformation from the initial pose to other poses. In such cases, the non-rigid registration algorithms are commonly used to detect the movement of an object. However, since the features of the movement of a face vary in each facial part, it complicates the registration problem. In our method, we recognize the facial parts by using Random Forest algorithm to be simplified to solve the registration problem. Then, we classify the facial parts into five types : nose, mouth, eye, cheek and obstacle. And we define the feature vector of each point of the face by using Fast Point Feature Histograms and a normalized position. We train the Random Forest based on the feature vector to recognize the facial parts from an arbitrary point set data. After the recognition of the facial parts, we detect the movement of the each part by using a non-rigid registration algorithm. Finally, we combine the translations of the each point as the deformation of the face by using the deformation technique based on a Radial Basis Function. In our results, we show that the proposed method enable us to detect the motion of the face more accurately.

Language：English  

Web of Science

Language：English  

Web of Science

Language：English   Publishing type：Research paper (scientific journal)  

Web of Science

Authorship：Lead author   Language：English   Publishing type：Research paper (scientific journal)  

Publishing type：Research paper (scientific journal)  

This paper proposes a novel method of generating a 3D virtual foot model from multiple camera image data based on the GFFD deformation method. 3D virtual foot models for individual persons can be generated easily and rapidly with considerably low cost. By estimating the cause of errors that can occur in the whole process and endeavoring to lower the errors, we succeeded to achieve satisfactory accuracy. The errors are mainly classified into two kinds. One occurs in transforming process with use of the GFFD deformation method and the other in obtaining 3D coordinates of feature points. So, we devoted ourselves to eliminate the cause of both errors and repeated experiments again and again. As a result, we got to get a 3D virtual foot model within acceptable error range. The average error is 1.32&sim;1.72mm.

Publishing type：Research paper (scientific journal)  

In this paper, we propose a parallel processing technique for creating animation of swaying trees in the wind based on physical simulation. Our technique independently assigns each tree to computation nodes. The each node processes not only parallel rendering for tree shapes, but also parallel simulation of the wind and bending of branches. This independency reduces data traffics between nodes and realizes efficient parallel processing. When the each node independently simulates the wind, it is necessary to consider wind interactions between the upper and the lower side tree. Therefore, we realize cooperative wind simulation with independent cluster nodes, by inventing a novel method for estimates of wind velocity of lower-side trees from simulation results of upper-side trees. By synthesizing rendered images from each cluster, our system generates animations of swaying trees. Through many experiments using these techniques, it is shown that the efficiency of parallel processing for animation of swaying trees is verified.

Language：Japanese   Publishing type：Research paper (scientific journal)  

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

In this paper, we describe a non-photorealistic rendering (NPR) framework for generating background art in animation. The background art, in many cases that is a Landscape Painting, is generally used in animation productions. One background art and multiple layers of moving characters compose a scene. The background art is one of most important factor for making the impression of an animation. We propose, in particular, a NPR technique that can reproduce painting techniques characteristic of background art such as leaf painting based on just a few colors, shading to produce a spherical effect, and aerial perspective. This technique can generate dynamic background art having a swaying motion in the wind by applying the animation generation tool for swaying trees developed by the authors. The proposed method, in contrast, is based on a painting technique for background art, which enables it to pick up features of the target work of art and to reproduce its style.

Language：English  

Web of Science

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

This paper presents a series of techniques for generating animations of trees swaying in the wind, in consideration of the influences that the tree shapes and leaf sizes give to the air current. To do the simulation of the wind around a tree having a complicated shape, it is necessary to consider the influence that some objects obstructing the wind such as leaves or branches give. Generally, the following problem occurs when we use the incompressible Navier-Stokes equations in a physical simulation model of the wind. Computational complexity increases because of considering the details of tree shapes, so it is difficult to generate the animations in real-time. Therefore, this paper proposes a novel method that reduces the computational complexity and realizes an animation in real-time, by means of a boundary condition map expressing space distribution of resistances from tree models automatically. In this case, we make a model as simple resistances decreasing the wind velocity from the parts that have similar shapes like leaves and branches. And also, it has another advantage that the influences between a tree and others can be rapidly calculated by using a hierarchical calculation method. Finally, through many experiments using these methods, it is shown that real-time animations of swaying trees in the wind can be realized.