Mutsuhiro Terauchi

Lane Detection on the iPhone

A robust and efficient lane detection system is an essential component of Lane Departure Warning Systems, which are commonly used in many vision-based Driver Assistance Systems (DAS) in intelligent transportation. Various computation platforms have been proposed in the past few years for the implementation of driver assistance systems (e.g., PC, laptop, integrated chips, PlayStation, and so on). In this paper, we propose a new platform for the implementation of lane detection, which is based on a mobile phone (the iPhone). Due to physical limitations of the iPhone w.r.t. memory and computing power, a simple and efficient lane detection algorithm using a Hough transform is developed and implemented on the iPhone, as existing algorithms developed based on the PC platform are not suitable for mobile phone devices (currently). Experiments of the lane detection algorithm are made both on PC and on iPhone.

Robust Vehicle Detection and Distance Estimation Under Challenging Lighting Conditions

Avoiding high computational costs and calibration issues involved in stereo-vision-based algorithms, this paper proposes real-time monocular-vision-based techniques for simultaneous vehicle detection and inter-vehicle distance estimation, in which the performance and robustness of the system remain competitive, even for highly challenging benchmark datasets. This paper develops a collision warning system by detecting vehicles ahead and, by identifying safety distances to assist a distracted driver, prior to occurrence of an imminent crash. We introduce adaptive global Haar-like features for vehicle detection, tail-light segmentation, virtual symmetry detection, intervehicle distance estimation, as well as an efficient single-sensor multifeature fusion technique to enhance the accuracy and robustness of our algorithm. The proposed algorithm is able to detect vehicles ahead at both day or night and also for short- and long-range distances. Experimental results under various weather and lighting conditions (including sunny, rainy, foggy, or snowy) show that the proposed algorithm outperforms state-of-the-art algorithms.

Low-Level Image Processing for Lane Detection and Tracking

Lane detection and tracking is a significant component of vision-based driver assistance systems (DAS). Low-level image processing is the first step in such a component. This paper suggests three useful techniques for low-level image processing in lane detection situations: bird’s-eye view mapping, a specialized edge detection method, and the distance transform. The first two techniques have been widely used in DAS, while the distance transform is a method newly exploited in DAS, that can provide useful information in lane detection situations. This paper recalls two methods to generate a bird’s-eye image from the original input image, it also compares edge detectors. A modified version of the Euclidean distance transform called real orientation distance transform (RODT) is proposed. Finally, the paper discusses experiments on lane detection and tracking using these technologies.

Online learning of virtual impedance parameters in non-contact impedance control using neural networks

Impedance control is one of the most effective methods for controlling the interaction between a manipulator and a task environment. In conventional impedance control methods, however, the manipulator cannot be controlled until the end-effector contacts task environments. A noncontact impedance control method has been proposed to resolve such a problem. This method on only can regulate the end-point impedance, but also the virtual impedance that works between the manipulator and the environment by using visual information. This paper proposes a learning method using neural networks to regulate the virtual impedance parameters according to a given task. The validity of the proposed method was verified through computer simulations and experiments with a multijoint robotic manipulator.

Vehicle Detection Based on Multi-feature Clues and Dempster-Shafer Fusion Theory

On-road vehicle detection and rear-end crash prevention are demanding subjects in both academia and automotive industry. The paper focuses on monocular vision-based vehicle detection under challenging lighting conditions, being still an open topic in the area of driver assistance systems. The paper proposes an effective vehicle detection method based on multiple features analysis and Dempster-Shafer-based fusion theory. We also utilize a new idea of Adaptive Global Haar-like (AGHaar) features as a promising method for feature classification and vehicle detection in both daylight and night conditions. Validation tests and experimental results show superior detection results for day, night, rainy, and challenging conditions compared to state-of-the-art solutions.

Real-time hand and eye coordination for flexible impedance control of robot manipulator

In recent years a lot of versatile robots have been developed to work in environments with human. However they are not sufficiently flexible nor safe in terms of interaction with human. In our approach we focused on hand and eye coordination in order to establish a flexible robot control, in which a robot recognizes its environment from the input camera images and equips a soft contacting strategy by impedance control. To recognize the environment, we adopt a method to reconstruct motion from a sequence of monocular images by using a pair of parallel straight line segments, which enables us to obtain linear equations to solve the problem. On the other hand the impedance control strategy conveys a flexible interaction between robots and humans. The strategy can be considered as a passive force control, when something contacts the end-effector of the robot. In order to avoid a collision, we introduce a virtual impedance control which can generate force prior to the contact. Neural networks (hereafter: NN) learning is used to decide parameters for impedance control, in which NNs can obtain parameters during the motion (aka: online learning). The validity of the proposed method was verified through experiments with a multijoint robot manipulator.

Motion estimation of a block-shaped rigid object for robot vision

Proposes a new conventional method to reconstruct 3D motion of rigid polyhedral objects from a sequence of monocular images. In general the problem is ill-posed, therefore additional information is required to recover depth. The authors utilize line correspondence between sequential images and existence of parallel line segments in the scene. The relation between the coordinates of points can be described only by a rotation matrix, if it is formulated relatively. Then if there exists a pair of parallel line segments, the matrix can be solved by using linear equations. After that the translation vector is computed. It is necessary that there exists at least one pair of parallel line segments in the scene in order to obtain motion parameters. They also propose the method to extract pairs of parallel line segments in the image. Finally some experimental results for simulated data are demonstrated.<<ETX>>

A linear algorithm for motion of rigid objects using features of parallel lines and optical flow

Various approaches have been proposed toward the problem of restoring three-dimensional (3-D) structures and motion of rigid bodies from image information. Ullman and Huang presented algorithms using point and line correspondences, in which they assume that the correspondence problems can be solved. Prazdny et al., on the other hand, presented an algorithm using optical flow, in which equations become nonlinear and thus the second derivative of velocity is required. This paper proposes an algorithm which combines optical flow and edge information. First, considering segments consisting of edges in an image, we derive an equation for optical flow. Then, making use of parallelism of line segments, we show that 3-D motion can be restored by using linear equations. To apply the algorithm there must exist two pairs of parallel line segments on an object. This paper presents an algorithm for extracting these pairs of parallel line segments. Finally, we verify the effectiveness of the algorithm by simulation.

Extraction of surface orientation from texture using the gray level difference statistics

The problem of extracting orientation of an object surface from a monocular image is one of the important tasks in computer vision. Most of the existing methods for extracting surface orientation are ones using the structural features of texture such as texel and edge. However, to represent texture features statistically is shown to be effective also in texture discrimination and segmentation. Thus, in this paper we propose a method for extracting surface orientation using the statistical feature of a texture image. First, we assume uniformity of a probability density function of difference statistic on object surface; then using the fact that the difference statistics depend on the geometric factor of length and orientation, we formulate the relationship between distortion of a density function in an image caused by perspective projection and the object surface. Then we derive an algorithm for finding the object surface orientation by search based on this formulation. In addition we apply this method to simulation images and real images to show its effectiveness. This enables us to extract object surface directly from a gray level image without extracting the texel or edge (whose extraction is required in the existing methods).

Simultaneous Learning of Robot Impedance Parameters Using Neural Networks

Impedance control is one of the most effective control methods for interaction between a robotic manipulator and its environment. Robot impedance control regulates the response of the manipulator to contact and virtual impedance control regulates the manipulator’s response before contact. Although these impedance parameters may be regulated using neural networks, conventional methods do not consider regulating robot impedance and virtual impedance simultaneously. This paper proposes a simultaneous learning method to regulate the impedance parameters using neural networks. The validity of the proposed method is demonstrated in computer simulations of tasks by a multi-joint robotic manipulator.