Noriaki Maru

Visual servoing of the manipulator using the stereo vision

Presents a new method of visual servoing using stereo vision to control the position and orientation of the manipulator with respect to an object. Conventional visual servoing techniques using monocular vision need a priori knowledge such as a model of the object or a parameter of distance to calculate a fixed image Jacobian matrix. This requirement is unnatural from a practical point of view. Moreover, a fixed image Jacobian matrix tends to slow the speed of convergence down. Using stereo vision makes it possible to calculate the exact image Jacobian matrix at any position without using a model of the target object, and results in better convergence characteristics. Both simulation and experimental results are presented to demonstrate the effectiveness of this method.<<ETX>>

Binocular tracking using log polar mapping

This paper describes a new binocular tracking method using log polar mapping (LPM) which approximately represents the mapping of the retina into the visual cortex in primate vision. Using LPM makes it possible not only to obtain both a high central resolution and a wide field of view, but also to significantly reduce the processing of image data. In this paper, LPM is performed in software by the lookup table method. Our tracking method utilizes a zero disparity filter (ZDF) for extracting the target object and virtual horopter method for estimating binocular disparities, respectively. The performance of both target extraction and disparity estimation is improved in comparison with the conventional methods, by using LPM. Some experimental results are also shown to demonstrate the effectiveness of the proposed method.

Manipulator control by using servoing with the stereo vision

A new method of visual servoing with the stereo vision to control the position and orientation of the manipulator with respect to an object is presented. Conventional control methods by visual servoing use a monocular camera and have several problems. For example, either shape information or desired distance of the target object from the camera must be given. Furthermore, a stability problem arises unless the initial positional error of features in the image is sufficiently small. These problems are caused by the image Jacobian matrix, that is, its provisional value at a desired position is used instead of the correct one. By using stereo vision, the image Jacobian matrix can be calculated correctly at any position, so neither shape information nor desired distance of the target object is required. Also, stability is assured even if the initial error is not small. Both simulation and experimental results are presented to demonstrate the effectiveness of this method.

Positioning control of the arm of the humanoid robot by linear visual servoing

This paper presents a positioning control of the arm of the humanoid robot by linear visual servoing. Linear visual servoing is based on the linear approximation between binocular visual space and joint space of the arm of the humanoid robot. It is very robust to calibration error, especially to camera angle errors and joint angle errors, because it uses neither camera angles nor joint angles to calculate feedback command. In this paper, we propose a method to expand work space of linear visual servoing by using neck joint. We obtain the linear approximation matrix in wide space and express it as a function of the neck angle by using the neural network. Some experimental results are presented to demonstrate the effectiveness of the proposed method.

3-D positioning control by linear visual servoing

This article describes the performance of 3-D positioning control by linear visual servoing using binocular visual space in a human-like hand-eye system which has a similar kinetic structure to a human being. We approximate the nonlinear time-variant mapping from a binocular visual space to the joint space of the manipulator as a linear time-invariant mapping. We also investigate the effect of binocular visual space in linear mapping by comparing it with linear mapping using Cartesian space. Some experimental results are presented using the human-like hand-eye system to demonstrate the performance of 3-D positioning control.

3-D tracking of a moving object by an active stereo vision system

We present a 3D tracking method of an object which is moving in a complicated scene by an active stereo vision system. The system uses binocular vision robot, which can simulate the human eye movements. Gaze holding on an target object with the controlled cameras keeps the targets stereo disparity small, and simplifies the visual processing to locate the target for pursuit control. The novel feature of our tracking method is the disparity-based segmentation method of the target object. The method utilizes zero disparity filter and correlation to separate the target object with small disparity from distracting background. Furthermore, using correlation method to estimate stereo disparity makes it possible to fixate on a surface of the target object. We show the experimental results with the complicated scene to demonstrate the effectiveness of the proposed method.<<ETX>>

Foveated Vision for Scene Exploration

In this paper, foveated vision for scene exploration is implemented. The peripheral and central vision are the basic capabilities of foveated vision. The informations obtained from the peripheral vision are used to determine the next gaze point. Due to the low resolution of the periphery, however, the determination is not always appropriate. To solve this problem, we propose to evaluate the target object by the central vision after gazing. We implement foveated vision based on the Log Polar Mapping (LPM) and construct an evaluation scheme of the target object in the central vision using LPM rotational-invariance. The peripheral vision is realized by Zero Disparity Filter for LPM stereo images. Some experimental results are also shown to demonstrate the effectiveness of the proposed method.

Binocular gaze holding of a moving object with the active stereo vision system

We present a binocular gaze holding method of an object which is moving in the complicated scene with the Active Stereo Vision System. The system uses a binocular vision robot, which can simulate the human eye movements. Holding gaze on a target object with the controlled cameras keeps the targets stereo disparity small, and simplifies the visual processing to locate the target for pursuit control. The novel point of our tracking method is the disparity-based segmentation method of the target object. The method utilizes a zero disparity filter (ZDF) and correlation to separate the target object with small disparity from distracting background. Furthermore, using the correlation method to estimate stereo disparity makes it possible to fixate on a surface of the target object. We show the experimental results with the complicated scene to demonstrate the effectiveness of the proposed method.<<ETX>>

Redundant Arm Control by Linear Visual Servoing Using Pseudo Inverse Matrix

We proposed a simple visual servoing scheme called linear visual servoing (indicated as LVS). It is based on the linearity of the transformation from binocular visual space to joint space of the arm of the humanoid robot which has a similar kinetic structure as a human being. LVS is very robust to calibration errors, especially to camera angle errors, because it uses constant Jacobian matrix with neither camera angles nor joint angles to calculate feedback command. Furthermore, the amount of calculation is very small compared to conventional visual servoing schemes. Hence, it is especially suitable for humanoid robots which use active stereo vision. But conventional LVS can not deal with redundant arm, because it is based on linear approximation of inverse kinematics. In this paper, we propose a redundant arm positioning control method by linear visual servoing based on linear approximation of forward kinematics. Simulation and experimental results are presented to demonstrate the effectiveness of the proposed method

Active binocular stereo

Passive binocular stereo tends to produce stereo correspondence errors and require much computation. A method which overcomes these drawbacks by moving the stereo camera actively is presented. The method utilizes a motion parallax acquired by a monocular motion stereo to restrict the search range of binocular disparity. Using only the uniqueness of disparity makes it possible to find reliable binocular disparity and occlusion very efficiently. Experimental results with complicated scenes are presented to demonstrate the effectiveness of this method.<<ETX>>