A modified algorithm for image registration by ring Zernike moments is presented, which has rotation-invariance as well as the method based on the original Zernike moments. Ring Zernike moments are rotation-invariant and orthogonal similarly to the original ones. A Zernike moment consists of ring Zernike moments defined on separated ring regions of the common center. We can construct a fast algorithm for registration of rotated images by utilizing a few ring Zernike moments as a feature vector and evaluation of dissimilarity be-tween feature vectors. Fundamental characteristics and stability of ring Zernike moments are experimentally investigated. From experiments with synthesized and measured images, it has been shown that speed-up of more than 95% of computation time could be achieved by use of the proposed ring Zernike moments.

抄録全体を表示

We have previously introduced several approaches of 3D-vision system for industrial part handling. Every vision system of them uses our unique stereo ranging technique based on area based correlation with two cameras. After then, we have improved our stereo ranging algorithm to use two step area based correlation procedures calling “hierarchical stereo matching algorithm” so that we have established a novel high accurate stereo ranging technique. This paper describes our unique approach to detect and locate industrial parts for handling system and our improvements and refinements on our stereo ranging technique.

抄録全体を表示

In order to strengthen the surveillance of jellyfishes interrupting the intake of cooling water in power stations, we have developed the system that detects an inflow of jellyfish by image recognition using existing monitor cameras. Monitor cameras are set on land in the way to look down the surface of the sea. Their images are changing by various factors, such as wave patterns, shadows and reflections of structures affected by the sun and the weather.
We developed the image processing method to detect jellyfishes in distinction from other phenomena. Jellyfish candidate regions are extracted by edge enhancement and thresholding, and are divided into jellyfishes and the others by their shapes and histogram features. This process features the automatic masking operation that exclude regions of image which are inappropriate for jellyfish detection such as regions except sea-surface from the object region of processing.
We have tested experimental system at two thermal stations and confirmed that the system is of practical use.

抄録全体を表示

In this paper, we propose a new method for the visual inspection such as defects detections of textile sur-faces; selectively removing the textural information from textile surfaces. For this purpose, we use Wavelet Shrinkage, which was originally proposed by Donoho & Johnstone as a method to remove Gaussian white noise. We also propose a modification of Wavelet Shrinkage to selectively remove the textural information from textile surfaces. Once the textural information is removed from textile surfaces, the remaining process-ings of inspection can be conducted in the image domain. So we can utilize the image processing methods and techniques developed so far for the visual inspections of other industrial products.

抄録全体を表示

In this paper, we propose a novel library system using images of books and borrower's faces. The images of lent are recorded on the lending list. A returned book is identified from a list of lent books by processing an image of the book cover and comparing it to books that have been lent. The administration of lending can be checked on a browser. Four directional features in the low resolution are used to identify of the book. To identify books of series which are similar but different in small area, the partial feature is used to distinguish these books. The experiments show that the four directional features and the partial features are effective for the book discrimination. Face recognition is used for easy administration. The availability of face recognition for the system is shown.

抄録全体を表示

It is necessary for a man-machine interface and tele-control of a robot to acquire a wide field view and a high degree of three-dimensional information of dynamic environment. In this paper, we proposed a stereo omnidirectional system, named SOS. The system can take all directional views and depth maps from an observation point, in real-time, by using the component constructed of sixty cameras. The cameras are sepa-rated into twenty stereo vision units. Each unit is composed of three cameras that are arranged horizontally and vertically. These units are arranged on each face of a regular icosahedron and each stereo vision unit is arranged so that it does not obstruct the view of other units. As a result, we succeeded in reducing the size of the sensor, securing the base line length to generate the depth value with enough precision to allow the simultaneous acquisition of the color images and depth maps omnidirectionally in three dimensions.

抄録全体を表示

This paper describes a measurement method of the relative position between the spreader of a quayside container crane and a container for an automatic container handling system. It is essential to realize an efficient loading and unloading task of quayside container cranes. First we describe our basic strategy of measurement method in order to realize an efficient loading and unloading task. Next we explain a fast and reliable measurement method of the relative position concretely. The measurement method is realized with an image processing method based on a template matching. Then we consider the adjustment method of prepared template by taking some images. Finally we confirm the validity of our proposed method with some experiments.

抄録全体を表示

It is important to know the projective transformation between two images in order to construct the image mosaics with multiple images. The conventional methods to obtain the projective transform use the geometrical features or the intensities of the image. These methods execute the iterative computation, so that the total amount of computation to extract the correspondence and to induce the parameters of the transform is too much.
In this paper, we propose the projective registration algorithm using the quasi-feature points, which are based on the intensity. We define the quasi-feature point as a central pixel of a block that has enough texture information. To determine these feature points, we extract the overlap image between two images using the Fourier method and divide this overlap image of the reference into the four subareas. Then we choose the central pixel of the block which has the maximum local variance in each subarea. We use the block matching algorithm, which is considered of both the camera movement and the variation of the illumination conditions, in order to match these four feature points with the four points in the target image. The proposed algorithm is applied to various images to estimate the performance and the simulation results present that this method finds the correct correspondence.

抄録全体を表示

A main problem with dynamical associative memories (DAMs) is that when memory patterns are stored, pseudo-memories (false fixed points and limit cycles) are also generated and they hinder proper association of input patterns. To overcome this problem, Hassoun proposed a heuristic method of reducing pseudo-memories. In this method, DAMs are constructed such that a zero vector called “ground state” as well as stored patterns is stabilized and sparsely activated states (sparse patterns) converge to the ground state. Such dynamical properties of neural networks can be described with linear inequalities, and connection weights of networks are obtained by solving these inequalities using the Ho-Kashyap algorithm. In this paper, we propose an extended Hassoun model in which network dynamics are modified such that dense patterns, mix-ture patterns and inhibition patterns are also converged to the ground state. In simulations, we compare association performance of this extended Hassoun model with conventional associative memory models, and demonstrate the usefulness of our proposed model as a dynamical associative memory.

抄録全体を表示

A dual plane comb-line filter consisting of conductor-backed CPW is proposed in this paper. It is shown through numerical simulations by means of the FD-TD method that this filter has an attenuation pole just below the pass band in contrast to conventional uniplane-based-CPW comb-line filters having an attenuation pole on the other side of the pass band. We also investigate the filtering characteristics from both numerical simulations and experiments when the resonators are formed in the step-impedance structure. The validity of FD-TD simulations has been comfirmed through a careful comparison between the numerical results and measurements.

抄録全体を表示

In this paper, a new algorithm for extracting a human face and facial parts from images based on skin color and motion information, and the extracting method of motion information of the eye and mouth regions are presented. For the efficient extraction of the face area, two cases are assumed. If the moving area is large, the facial area is extracted based on moving area and color information. On the contrary, it is judged that no movement exists when the moving area is small, the facial area is extracted based on the facial area which has already been extracted in the previous frame. For precise recognition of facial expressions, we propose a new scheme for extracting an opening and closing ratio of the eyes and mouth, and the efficacy of the proposed method is examined experimentally using several images.

抄録全体を表示

We have newly developed two types of 15" high resolution non-glasses 3-D displays, the one is no cross-talk 3-D display using the double image splitter technology and the other is 2-D/3-D full compatible display using the image splitter and the controllable scattering shutter technology. In this time, in the 2-D/3-D full compatible display, controllable scattering shutter is united with image splitter in order to realize simple construction and short viewing distance of the non-glasses 3-D image. Those displays adopted newly developed head tracking system. The head tracking system detects the viewer's head position without special attachment and adjusts the images automatically. So it can expands the stereoscopic area. With this combination of larger screen, higher resolution and head tracking system, the display enables viewer to have more realistic and enjoyable images.

抄録全体を表示

This paper describes theory of estimation method of displacement and strain distribution on a 2-D object using relative optical flow. Two sequential images of the object are mapped on to a normalized rectangular region using time yawing function coordinate systems. Relative optical flow is estimated on the normalized rectangular domain. Displacement and strain distribution on the object are obtained parametrically by transforming from relative optical flow. The proposed method may estimate strain distribution on the object without contact, even if that is moving or changing the shape of itself. Finally, usefulness of the proposed method is demonstrated applying to an example.

抄録全体を表示

In this paper, we propose a tool creating support system using evolutionary techniques. Here, a tool is defined as an object which consists of multiple parts, and is encoded to tree structure. The proposed system is composed of two phases: the fundamental-parts-generation-phase where each fundamental part is generated, and the structuring-phase where the generated parts are combined. On each phase, evolutionary techniques which utilize user's evaluations as decision of fitness are employed. Therefore it is possible to obtain desirable shapes for users. We have implemented the system on the O2 workstation, and have evaluated the proposed system. It is confirmed that various novel shapes can be obtained.

抄録全体を表示

As demand grows for wireless communication systems, the problem of frequency assignment becomes in-creasingly important. This paper therefore investigates the frequency assignment problem (FAP) as it relates to the mobile communication systems of the Kyushu Electric Power Company. A number of algorithms for solving the FAP have been proposed. However the FAP investigated in this paper is more complex than the benchmark problems used in these studies. Thus we propose a hybrid algorithm of genetic and greedy algorithms to solve the problem. A global optimal assignment can be obtained in a very short time using the algorithm. Thus we are able to reduce the time required for finding the optimal assignment.

抄録全体を表示

This paper describes a method of time constant change by gain control and its application for equalizers and automatic level controllers (ALC) for AC biased audio recording circuit.
Bilinear characteristics are used for these equalizers. The proposed method utilizes the fact that the pole is controlled by feedback gain and the null by feedforward gain. In ALC, the gain control per operating period is changed in attack and recovery condition, so that a different time constant can be obtained.
In the case of a bipolar IC for VTR, 10 time constants were used as the external components. However, the experimental results verify a significant reduction of these time constants to 2. That is, it was confirmed that this method has an equivalent performance to the above example: bandwidth is 10.5 kHz for SP (Standard Play Mode) and 5 kHz for EP (Extended Play Mode) with the recovery time being 1.8 s. Moreover, the attack time could be shortened to 1/20, namely 0.3 ms.
The proposed method is suitable for a system LSIs with a BiCMOS process.

抄録全体を表示

It is an important problem in multi area power systems to attenuate the frequency deviation due to la change. Since power systems are spatially dispersive, decentralized control is more practical than lump control. In this paper, we propose two robust decentralized control schemes for this problem. The difficul with this problem is that the linearized model of a multi area power system with frequency deviations outputs is an MIMO system that has a transmission zero at s=0. The key idea is to change the output one subsystem so that the resulting new plant does not have any zero at s=0. The new outputs are chos, carefully so that the regulation of the new outputs can guarantee the regulation of all frequency deviatio and tie-line power. Further, parameter uncertainty is taken into account in the controller design to achie robustness. Simulation shows the proposed methods are effective.

抄録全体を表示

A pixel composed of several classes in remote sensing data is called mixed pixel (mixel), and the mixel exists a lot due to resolution of data. When a mixel is in the boundary area, it can be divided into pure pixel of landcover class set in proportion to the class mixture ratio. Therefore, this paper proposes a new image resolution algo-rithm for mixel in remote sensing data. The approach consists of three steps. To begin with, component class number and class mixture ratio are estimated with the fuzzy reasoning. Next, it is assumed that class mixture ratio of vicinity pixels shows the intensity of the relation between the extension pixels and vicinity eight pixels. Degree of association between each extension pixel and vicinity pixels is computed based on their position in-formation and the assumption. Finally, using the computation draws an extension image.
As a result of experience, it became clear that the proposed approach was able to accurately magnify the mixel in comparison with the previous method and conventional method.

抄録全体を表示

We propose a new method for pattern recognition based on fuzzy encoding. Fuzzy inference rules (FIRs) to recognize a given pattern are generated with the set of encoded patterns. A given pattern is recognized to carry out defuzzification of the patterns from the FIRS. Computational experiments were carried out, and the results of the experiments showed that the performance of our new method is much better than that of neural networks.

抄録全体を表示

Iterative decoding is a key feature of turbo code and each decoding results in additional computation and decoding delay. In this paper, we present a simple and dffective criterion for stopping the iteration process in turbo decoding with a negligible degradation of the error performance. Simulation results show the effectiveness of the proposed dynamic decoding method.

抄録全体を表示