摘要:

AbstractWith respect to CMOS logic circuits, it is reported that a fault can occur cannot be covered by conventional classical fault model, and the current testing is considered to be interesting as a testing method to detect such a fault.This paper proposes a fault‐detection method for the CMOS logic circuit which can detect the stuck‐at fault and the stuck‐open fault based on the dynamic power supply current observed when a pseudorandom pattern is applied as the input. A built‐in testing circuit for realizing the principle is shown. The feature of the current testing is that the fault can be detected by observing the power supply current as long as the fault in the circuit is activated, and it is not required to propagate the fault information to the output. Based on that property, a random pattern with a small test generation cost is used as the test pattern.As the test pattern, the pseudorandom pattern is generated autonomously by utilizing the output signal of the circuit under test. The fault detection is made easier by this pattern generation since the effect of the fault on the power supply current is enhanced. In determining whether or not the power supply current is that of the normal circuit, a method based on the time‐course of the power supply current is used, with the power supply currents for various test patterns as the feature parameter. The proposed procedure is evaluated by a simulation, and the usefulness of the proposed method is dem

ISSN:0882-1666    

DOI:10.1002/scj.4690251101

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractThe concept of a Visual Software Agent (VSA), i.e., an anthropomorphic agent possessing a realistic human form and communication functions that create an interface with the user, is proposed. Research is proceeding on its implementation on a parallel computer. In this paper, we describe a method that is effective as the visual input function of the anthropomorphic agent. This method recognizes and extracts only one person (user) directly facing the camera in a moving image in which several people occasionally enter and leave. The recognition method can be executed in real time in a typical indoor environment. The background and the lighting do not have to be restricted and recognition can be achieved in real time by a parallel computer. In addition, this technique simulates the initial stage of human visual functions and is based on the parallel extraction and integration of multiple image features. The movements of visual attention of cognitive psychology can also be simulated. An anthropomorphic agent derived from this approach successively looks at multiple human faces, fixes its gaze only on a specific person, and synthesizes agent behavior that approximate that of humans. These moving image recognition and synthesis processes were implemented on a newly designed parallel computer.

ISSN:0882-1666    

DOI:10.1002/scj.4690251102

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractPerception of visual motion is thought to consist of two stages: generation of candidates and their interaction to determine true motion. In this paper, a three‐layered neural network is applied to detect local visual motion. The network, after learning, could categorize nine types of motion and obtained a motion detection algorithm that included the two states of perception. The internal representations for the first stage agree with the functions of one‐center and off‐center cells, and those for the second stage agree with the functions of lateral inhibition. We also tried to detect arbitrary motions by combining multi‐resolution representation of images with the neural

ISSN:0882-1666    

DOI:10.1002/scj.4690251103

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractA method is proposed for estimating shape, surface reflectance, illuminant direction and illuminant intensity from images with shading. First, ann x mimage data matrixIis constructed fromnpixel image inten‐sity data throughmframes by moving a light source arbitrarily. Using the Lambertian assumption, the image data matrixIcan be written as the product of two matricesNandS; Nrepresents the surface normal and surface reflectance, andSrepresents the illuminant direction and illuminant intensity. Based on this for‐mulation, a singular value decomposition technique is employed and one of the following two constraints is used to factorize the image data matrix.One is the constraint of surface reflectance which can be used when there are at least 6 pixels in the images, where the surface reflectance is constant. The other is the constraint of illuminant intensity which can be used when there are at least 6 frames among the images, where the illuminant intensity is constant.As the number of images with shading increases, the estimation errors become smaller even if the images are taken in a noisy environment. The proposed method can also treat self shadows and cast shadows without assumptions. The effectiveness of the proposed method is demonstrated through computer simulation, performance analysis and laboratory experim

ISSN:0882-1666    

DOI:10.1002/scj.4690251104

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractTo obtain the cross section of an object by using an X‐ray CT scanner, many projections of the object with various angles are required. This also takes time for scanning. If the object moves during this period, the projections would be disturbed causing motion artifact in the CT image. Understanding of their characteristics in relation to the movements of the objects will be useful for analyzing CT images and will supply new knowledge to medical diagnosis.This paper investigates a reconstructed image of a point object which revolves around the center of a scanning area. The results show that the reconstructed image of the point object forms an epicycloid or hypocycloid (a curve on a plane) and this reduces with time. The same type of artifact is simulated by a computer, and this justifies the analysi

ISSN:0882-1666    

DOI:10.1002/scj.4690251105

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractTo enhance the recognition rate of handwritten characters, one has to consider the effects of various changed forms in character patterns. In Reference [1], we have proposed a recognition algorithm using pattern transformation, which can deal flexibly with the changed forms in character patterns. In the fine classification, the algorithm makes three types of pattern transformations for the input pattern, and selects the transformed pattern that matches the standard pattern best. This paper improves that algorithm and constructs a recognition system. As the results of testing with ETL9B, a database of handwritten characters that contains 600,000 Japanese characters and a recognition rate of 96.32 percent was achieved, a new record for ETL9B.

ISSN:0882-1666    

DOI:10.1002/scj.4690251106

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractThis paper discusses the test for the automatic extraction of lung cancer candidate areas from the high‐resolution continuous chest CT image taken with the slice interval of 1 mm, by a three‐dimensional image processing technique. The detection of local suspicious regions (SR) is considered. A procedure is developed which discriminates the lung area from other images (mostly vessels) based on the differences of three‐dimensional shapes. The actual procedure in the processing is composed largely of two steps: (1) segmentation of lung area; and (2) SR extraction.In this study, the processing is realized by a relatively simple procedure, combining such basic processes as thresholding, fusion of three‐dimensional figures (or three‐dimensional distance transformation) and the three‐dimensional skeletonization. The usefulness of the method is demonstrated using an actual case of lung cancer. The study is significant as the first test that demonstrated the usefulness of the three‐dimensional image processing, using the three‐dimensional CT image of a practically suff

ISSN:0882-1666    

DOI:10.1002/scj.4690251107

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractTo produce a realistic image at a fast speed, a parallel graphic generator MAGG using transputers and a graphic generation software using a parallel ray tracing algorithm have been developed. A screen space subdivision is sued for the parallel processing. To avoid the limitation of a local memory due to the object data, a necessary part of the data alone is transmitted to the local memory.To avoid the reduction of the effect of the parallel processing due to the increase of the interprocessor communication, the overall object data are stored in the shared memory so that they are transmitted to the local memory at a high speed using a DMA. The load balancing is carried out by dividing the screen into many small sections, which are called subscreens and are sufficiently larger than that of the processors, and by allocating the load dynamically. This method requires an estimation of the appropriate number of subscreens before the execution. To avoid this difficulty, already allocated subscreens are redivided and reallocated.The control of the reallocation is carried out adaptively by a message exchange. The proposed method makes the distribution of the load uniform even though there are only a few small screens and the method improves both the image generation time and the parallel processing efficiency.

ISSN:0882-1666    

DOI:10.1002/scj.4690251108

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractA drawback of the error‐backpropagation method for a multilayer feedforward neural network is overlearning. The authors have already discussed this problem, and obtained necessary and sufficient conditions for over‐learning to occur. Using those conditions and the concept of a reproducing kernel, this paper proposes methods for choosing a training set which prevents over‐learning.First, the case is discussed in which pure training data are available. Methods are presented for choosing a training set so that amemorization learningdoes not causeWiener over‐learning.Next, the case in which noisy training data are available is discussed and methods for choosing a training set so that arote‐memorization learningdoes not cause Wiener overlearning

ISSN:0882-1666    

DOI:10.1002/scj.4690251109

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY

摘要:

AbstractThis paper proposes a method of modeling a natural terrain based on fractal geometry which can be applied to problems such as route planning for outdoor exploration by autonomous mobile robots. The method uses elevation data of a terrain which are derived from range data obtained by a scanning‐laser range‐finder at irregular positions. The method handles a problem of reconstruction of 3‐dimensional data with an arbitrary resolution and uncertainty.Assuming that the terrain is fractal shape, the method solves the following modeling problems: (a) estimation of terrain roughness based on fractal Brownian function model; (b) reconstruction of the terrain surface adaptive to the roughness (using interpolation of elevation data with MaximumA Posteriori(MAP) estimation); and (c) estimation of uncertainty distribution of the elevation map of terrain using the Monte Carlo method. The solution of each problem is given. The usefulness of the method is confirmed by simulations and experiments using observational

ISSN:0882-1666    

DOI:10.1002/scj.4690251110

出版商:Wiley Subscription Services, Inc., A Wiley Company    

年代:1994

数据来源: WILEY