When a real-time image processing application is implemented with a general-purpose computer, CPU (Central Processing Unit) is usually heavily loaded and in many cases that CPU alone cannot meet the real-time requirement at all. Most modern computers are equipped with powerful Graphics Processing Units (GPUs) to accelerate graphics operations. There is a trend that the power of GPU outgrows that of CPU. If we take advantage of the powerful GPU for more general operations other than pure graphics operations, the processing time can be reduced. In this study, we will present techniques that apply GPU to general operations such as image processing procedures. Our experiment results show that significant speed-up can be achieved by using GPU.

The intelligent robots that will be needed in the near future are human-friendly robots that are able to coexist with humans and support humans effectively. To realize this, robots need to recognize their position and posture in known environment as well as unknown environment. Moreover, it is necessary for their localization to occur naturally. It is desirable for a robot to estimate of his position by solving uncertainty for mobile robot navigation, as one of the best important problems. In this paper, we describe a method for the localization of a mobile robot using image information of a moving object. This method combines the observed position from dead-reckoning sensors and the estimated position from the images captured by a fixed camera to localize a mobile robot. Using a priori known path of a moving object in the world coordinates and a perspective camera model, we derive the geometric constraint equations which represent the relation between image frame coordinates for a moving object and the estimated robot's position. Also, the control method is proposed to estimate position and direction between the walking human and the mobile robot, and the Kalman filter scheme is used for the estimation of the mobile robot localization. And its performance is verified by the computer simulation and the experiment.

A depth measurement system that consists of a single camera, a laser light source and a rotating mirror is investigated. The camera and the light source are fixed, facing the rotating mirror. The laser light is reflected by the mirror and projected to the scene objects whose locations are to be determined. The camera detects the laser light location on object surfaces through the same mirror. The scan over the area to be measured is done by mirror rotation. Advantages are 1) the image of the light stripe remains sharp while that of the background becomes blurred because of the mirror rotation and 2) the only rotating part of this system is the mirror but the mirror angle is not involved in depth computation. This minimizes the imprecision caused by a possible inaccurate angle measurement. The detail arrangement and experimental results are reported.

In this study, stereo vision system is applied to a mobile manipulator for effective tasks. The robot can recognize a target and compute the potion of the target using a stereo vision system. While a monocular vision system needs properties such as geometric shape of a target, a stereo vision system enables the robot to find the position of a target without additional information. Many algorithms have been studied and developed for an object recognition. However, most of these approaches have a disadvantage of the complexity of computations and they are inadequate for real-time visual servoing. Color information is useful for simple recognition in real-time visual servoing. This paper addresses object recognition using colors, stereo matching method to reduce its calculation time, recovery of 3D space and the visual servoing.

In this paper, a novel algorithm capable of detecting generic obstacles on a flat surface is proposed. The algorithm fundamentally exploits a distortion phenomena taken place in remapping process of original stereo images to a virtual top-view. Based on the distortion phenomena, we construct stereo polar histograms of edge maps, detect peaks on them, and search for matched peaks on both histograms using a Dynamic Programming (DP). Eventually, the searched corresponding peaks lead to estimate obstacles' positions. The advantages of the proposed algorithm are that it is not largely affected by an intensity difference between a pair of stereo images and does not depend on the typical stereo matching methodologies. Furthermore, the algorithm identifies the obstacles' positions quite robustly.

This paper proposes a new edge detection method using a $3{\times}3$ ideal binary pattern and lookup table (LUT) for the mobile robot localization without any parameter adjustments. We take the mean of the pixels within the $3{\times}3$ block as a threshold by which the pixels are divided into two groups. The edge magnitude and orientation are calculated by taking the difference of average intensities of the two groups and by searching directional code in the LUT, respectively. And also the input image is not only partitioned into multiple groups according to their intensity similarities by the histogram, but also the threshold of each group is determined by fuzzy reasoning automatically. Finally, the edges are determined through non-maximum suppression using edge confidence measure and edge linking. Applying this edge detection method to the mobile robot localization using projective invariance of the cross ratio. we demonstrate the robustness of the proposed method to the illumination changes in a corridor environment.

In this paper, the maze navigation system using image recognition for autonomous mobile robot is proposed. The proposed maze navigation system searches the target by image recognition method based on ADALINE neural network. The infrared sensor system must travel all blocks to find target because it can recognize only one block information each time. But the proposed maze navigation system can reduce the number of traveling blocks because of the ability of sensing several blocks at once. Especially, due to the simplicity of the algorithm, the proposed method could be easily implemented to the system which has low capacity processor.

This paper presents a lane-departure identification (LDI) algorithm for a traveling vehicle on a structured road. The algorithm makes up for the weak points of the former method based on EDF[1] by introducing a Lane Boundary Pixel Extractor (LBPE), the well known Hough transform, and liner regression. As a filter to extract pixels expected to be on lane boundaries, the LBPE plays an important role in enhancing the robustness of LDI. Utilizing the pixels from the LBPE the Hough transform provides the lane-related parameters composed of orientation and distance, which are used in the LDI. The proposed LDI is based on the fact the lane-related parameters of left and right lane boundaries are symmetrical as for as the optical axis of a camera mounted on a vehicle is coincident with the center of lane; as the axis deviates from the center of lane, the symmetrical property is correspondingly lessened. In addition, the LDI exploits a linear regression of the lane-related parameters of a series of successive images. It plays the key role of determining the trend of a vehicle's traveling direction and minimizing the noise effect. Except for the two lane-related parameters, the proposed algorithm does not use other information such as lane width, a curvature, time to lane crossing, and of feet between the center of a lane and the optical axis of a camera. The system performed successfully under various degrees of illumination and on various road types.

In this paper, we suggest an integrated tactile display system that provides kinesthetic force, pressure distribution, vibration and slip/stretch. The system consists of two parts: a 2 DOF force feedback device for kinesthetic display and a tactile feedback device for displaying the normal stimulation to the skin and the skin slip/stretch. Psychophysical experiments measure the effects of fingerpad selection, the direction of finger movements and the texture width on tactile sensitivity. We also investigate the characteristics of lateral finger movement while subjects perceive different textures. From the experimental results, the principal parameters for designing a tactile display are suggested. A tactile display device, using eight piezoelectric bimorphs and a linear actuator, Is implemented and attached to a 2 DOF translational force feedback device to simultaneously simulate the texture and stiffness of the object. As a result, we find out that the capability of the suggested device is sufficient to display physical quantities to display the texture.

This paper describes three kinds of experiments and analysis of their results related to human tactile sensitivity using an integrated tactile display system. The device can provide vibration, normal pressure and lateral slip/stretch which are important physical quantities to sense texture. We have tried to find out the efficient method of stimulating, limitation of surface discrimination by kinesthetic farce feedback and the effectiveness of the combination of kinesthetic force and tactile feedback. Seven kinds of different stimulating methods were carried out and they are single or combination of the kinesthetic force, normal static pressure, vibration, active/passive shear and moving wave. Both prototype specimen and stimulus using tactile display were provided to all examinees and they were allowed to answer the most similar sample. The experimental results show that static pressure is proper stimulus for the display of micro shape of the surface and vibrating stimulus is more effective for the display of fine surface. And the sensitivities of active touch and passive touch are compared. Since kinesthetic force feedback is appropriate to display shape and stiffness of an object, but roughness display has a limitation of resolution, the concurrent providing methods of kinesthetic and tactile feedback are applied to simulate physical properties during touching an object.

An empirical way of a covariance matrix which expresses the odometry uncertainty of mobile robots is proposed. This method utilizes PC-method which removes systematic errors of odometry. Once the systematic errors are removed, the odometry error can be modeled using the Gaussian probability distribution, and the parameters of the distribution can be represented by the covariance matrix. Experimental results show that the method yields $5{\%}$ and $2.3{\%}$ offset for the synchro and differential drive robots.

This paper describes a method aimed at improving dead-reckoning accuracy with gyroscopes in mobile robots. The method is a precision calibration procedure for gyroscopes, which effectively reduces the ill effects of nonlinearity of the scale-factor and temperature dependency. This paper also describes the methods of gyro data collection fur all ambient temperature$(-40^{\circ}C{\~}+80^{\circ}C)$ using cubic spline interpolation and defining the error function. The sensor used was a vibrating gyroscope called the EWTS82NA21, which is low lost and commonly used in car navigation system, made by Panasonic. This angular rate sensor utilizes Coriolis force generated by a vibrating tuning fork. The paper also provides experimental results to check the performance and the effectiveness of the proposed method.

Recently, development of digital technology is occupying a large part of multimedia information like character, voice, image, video, etc. Research about video indexing and retrieval progresses especially in research relative to video. This paper proposes the novel notation in order to retrieve MPEG video in the international standards of moving picture encoding For realizing the retrieval-system, we detect DCT DC coefficient, and then we obtain shot to apply MVC(Mean Value Comparative) notation to image constructed DC coefficient. We choose the key frame for start-frame of a shot, and we have the codebook index generating it using feature of DC image and applying PCA(principal Component Analysis) to the key frame. Also, we realize the retrieval-system through similarity after indexing. We could reduce error detection due to distinguish shot from conventional shot detection algorithm. In the mean time, speed of indexing is faster by PCA due to perform it in the compressed domain, and it has an advantage which is to generate codebook due to use statistical features. Finally, we could realize efficient retrieval-system using MVC and PCA to shot detection and indexing which is important step of retrieval-system, and we using retrieval-system over the internet.