Recent years there has been a remarkable development in VLSI technology. A proposal has made for the realization of VLSI algorithm performance in Image Processing fields.
The Systolic Array is an example for VLSI architecture. Systolic Array has a control system of parallel processing abilities. Pre-processing and transformation characteristic are very useful for industrial applications. In this paper we present a discussion about the method of realization of parallel image processing using Systolic Array Processor (Geometric Arithmetic Parallel Processor chips; GAPP) and LSAPS (Large Scale Systolic Array Processor System). Particulary, here, we discuss in detail about pre-processing and transformation characteristics of images, in which high speed processing capability are essential to deal with large volume data.

View full abstract

This paper proposes a parallel processing scheme for continuous- and discrete-time control systems simulation on multiprocessor system. The scheme allows ordinary users to easily simulate control systems represented by S domain and Z domain transfer functions, state space equations and block diagrams on a shared memory multiprocessor system in real time. A control system simulator based on the proposed scheme has been implemented on a multiprocessor system named OSCAR. The simulator automatically generates a simulation program from a block diagram input on X-Window, compiles the program into parallel machine codes, executes them on a multiprocessor and outputs the results graphically.

View full abstract

This paper proposes a parallel processing scheme for solution of non-linear equations on a multiprocessor system. As an instance of non-linear equations, load flow calculation using Newton-Raphson method is chosen. Solution of non-linear equations using Newton-Raphson method includes iterated solution of sparse linear equations. These sparse linear equations are solved by a direct method or a relaxation method. The relaxation method has high parallelism but a problem in convergence. Therefore the proposed scheme parallelizes a direct solution method of non-linear equations with LU-decomposition. For efficient parallel processing on multiprocessor system, the proposed scheme generates optimized parallel machine code that minimizes data transfer overhead and synchronization overhead by adopting static multiprocessor scheduling algorithm considering data transfers among processors. The effectiveness and practicality of the proposed scheme are demonstrated on an actual multiprocessor system named OSCAR that has been designed to take full advantages of the static scheduling.

View full abstract

To reduce the load of switching devices and elongate the lifetime of the circuits, we have successfully developed a new type of the excitation circuit named “pre-charging circuit”. In this circuit, peaking capacitor is divided in two stages which are connected in series. The one capacitor is charged pulsively, while the another by DC. Increasing DC voltage, we are able to reduce the pulse voltage, resulting in the reduction of the load at the switch. The performance of this circuit is evaluated at the gas pressure of 760 Torr. The laser output of_??_4.5 J and total efficiency of_??_11% are obtained. With a semiconductor switch, we have obtained_??_1.3 J of the laser output and_??_6% of the total efficiency at the gas pressure of 400 Torn

View full abstract

Medical ultrasonic echocardiogram contains two undesirable components. One is the drift or offset in low frequency range which shapes the background of whole part of image. The other is small amplitude noise in high frequency range which becomes the factor of comparatively local degradates. This paper desribes a method of the improvement of image quality by removing the above-mentioned two factors. First, we detect the background in the low frequency range by using linear programming, and we suppress the drift and offset by substracting the background from original image. Second, we remove the low amplitude noise by using spline filter which belongs to non-linear filter. At the end of this paper, the usefulness of this method is demonstrated to show some examples.

View full abstract

A mapping system using a commercialized residential map is getting attention as an effective management tool in public service sectors. Utilization of a residential map by linking with an existing customer database allows efficient round inspection and quick response for customer's claims.
A conventional matching method for linking between an customer database and a residential map is simple comparison of customer's names and resident's names. This simple method is applicable only for a limited number of data, but it is not feasible in linkage of a large number of data because of extremely long computational time and low matching ratio.
In this paper, we propose an inferential matching method which enables automatic matching for a large volume database. This method rationalizes matching process by adaptive focusing of candidate-data. These data are selected in corporate with calculated matching certitude. We applied the method for 200, 000 customer data and the result shows the significant improvement of matching ratio from 30% to 60% in comparison with the conventional method.

View full abstract

A real time synthesis system that generates and displays the lane-markings of a road, as seen from a moving vehicle, is presented. The system operates on static road map information and dynamic vehicle motion data. Road map includes parameters such as horizontal and vertical curvature, bank angle, and distance. The parameters for vehicle motion include velocity, displacement, elevation and angle (pitch, roll, yaw). The system generates 3-dimensional coordinates of lane-markings using vehicle motion data and road map information. These coordinates are then transformed to 2-dimensional images.
For accurate display and high speed calculation, polynomial expressions for 3-dimensional configuration of lane-markings which use a parametrized variable are proposed. The system is implemented on a high speed image processor. Continuous dynamic images can be displayed at a video rate of 16.7 ms per image. This system is adequate for testing the performance of a road image recognition system for automatic driving vehicles.

View full abstract

A creation support method which supplys suitable information from computer is proposed.
The main idea of this method is creation support by finding hypotheses for an item. Other items, which are connected by the hypotheses, are retrieved from the database. Then these items are given as the related information to a user. A network type database named condition-action network installed beforehand and a retrieval algorithm by hypothesis are proposed. The retrieval algorithm consists of a connecting process which connects items by hypotheses, and a selecting process. The selecting process selects the influential items among a lot of connected items, using an evaluation function.
A structural modeling support system installed hypothesis based creation support have been developed. Through the experience using this system, it is clarified that it helps us to hit on the various and unusual items.

View full abstract

In this report we discuss which method is best among several fuzzy reasoning methods to control a model car, using driving simulation of right or left turning for the car on the L-shaped driving course. Fuzzy reasoning methods based on Mamdani's or Tsukamoto's method are considered. In order to find the best control rules, the fuzzy sets of the consequent parts were tested and selected by the simulations. As a result, it is best for the driving control to use the reasoning method including algebraic product for the antecedent and conseqent operations, logical sum for the unification operation of the conclusion fuzzy sets and the median method for the defuzzification.

View full abstract

Although there are numerous methods for measuring the glossiness instead of the psychological glossiness, the glossiness obtained by these methods do not agree with the psychological glossiness when the specimen's color changes. In this study, the effect of color on glossiness is investigated. Many sheets of paper of the different color are used as the specimens. The specimen, on which a transparent film is pasted, is wound on a cylinder. The psychological glossiness is estimated by NAKAYA's method, which is one kind of the paired comparison method. The relationship between the psychological glossiness and the lightness, the saturation and the hue are examined. The results are as follows. According as the lightness increases, the psychological glossiness decreases. The increase of the saturation gives the increase of the psychological glossiness. The correlation between the psychological glossiness and the hue is small. From these findings, an approximate equation expressing the effect of color is obtained. The glossiness we proposed before is revised by this equation. The revised glossiness agrees well with the psychological glossiness when the specimens with the different color are used.

View full abstract

The structure of gloss perception of human vision is not clear in the present time. A model of gloss perception using the operation resembling retina is proposed in this paper. The various reflected-light-intensity-distributions projected on the photo-sensor corresponding to the human visual cell are simulated with a computer. The distribution is transformed into an internal signal corresponding with the output of visual cell by modeling the visual adaptation. The output-distribution is obtained by the convolution-integral of the internal signal with DOG function that is similar to the sensitivity of the receptive field of ganglion cell. As the results, it is found that the maximum value H_max of the output-distribution has the characteristic similar to the psychological glossiness as follows.
(1) The maximum value H_max increases according as the psychological glossiness increases, when the specimens with the same curvature radius are used.
(2) When the curvature radius changes, H_max also changes and has the maximum value at a curvature radius. According as the psychological glossiness of a specimen decreases, the curvature radius at which H_max becomes a maximum decreases. This phenomenon is similar to that of psychological glossiness. Furthermore, it is likely that the half width of reflected-light-intensity-distribution roughly equals to the size of on-response field in receptive field.

View full abstract