摘要:

AbstractThis paper proposes a distributed algorithm for reconstructing a minimum‐weight spanning treeT′ of a networkN′ when link addition and deletion occur in a networkNwith a minimum‐weight spanning treeT.In this algorithm each processor uses information whose adjacent links belong toTin order to constructT′ efficiently. The communication complexity and ideal time complexity of the algorithm areO(nlog(f + t) + m)andO(nlog(f + t) + n), respectively, wherenandeare the number of processors and that of links inN′,tis the number of added links, andfrepresents that of deleted links belonging toT.Here,m=n+twhenf= 0 andm = eotherwise.This paper also presents a distributed algorithm for reconstructing a minimum‐weight spanning treeT+ when links and processors are added and deleted. The communication complexity and ideal time complexity of the algorithm areO(nlog(g + h) + randO(nlog(g + h)+n), respectively, wheregis the total number of added links (including links incident to added processors) andhrepresents the number of deleted links belonging toT(including links incident to deleted processors). Thus,r = n + gwhenh= 0 an

ISSN:0882-1666    

DOI:10.1002/scj.4690230301

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

年代:1992

数据来源: WILEY

摘要:

AbstractA control system for an unmanned vehicle is constructed by combining dynamic image processing and fuzzy inference. The features of this system are that an object on the road can be recognized with a high speed, and the unmanned vehicle runs according to the result of recognition.First, the object to follow is extracted by the image processor from a tremendous amount of image data containing noises. The result is then transformed, using the improved Hough transform, into the information concerning the position and the orientation, which are needed in the steering of the vehicle.The information is then compared to the expertise of the vehicle steering in the fuzzy controller, and the handle variables are determined. The image processor executes the pipeline processing using the analog color extractor and the logical filtering processor with the local parallel pipeline architecture. Sixty images per second can be processed in real‐time. The fuzzy inference has the fuzzy rulebase with the motion information of the vehicle as the input parameter so that the handle variables can be determined even for a moving vehicle.By an LSI implementation of the image processor, the proposed system is realized as a small‐scale system which can be mounted on the vehicle (A4 size × 10 cm). In the unmanned control experiment using a vehicle with the proposed system, a high‐speed processing of 100 ms on the average from the image input to the handle operation is verified. The follow‐up to the series of objects within the error of 5 cm is verified, indicating the practical usefulness of the propose

ISSN:0882-1666    

DOI:10.1002/scj.4690230302

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

年代:1992

数据来源: WILEY

摘要:

AbstractThis paper proposes an RMD‐tree (region multidimensional tree), which is a method of data management based on the tree structure, aiming at the efficient management and retrieval of spatial objects with nonzero sizes.In the proposed method, the spatial object in theN‐dimensional space corresponds to a point in the 2N‐dimensional space, being represented by its centroid and the extent along each axis. The coordinate transformation is applied to the point data and the data are managed by the tree structure, so that retrieval such as the point location problem or the range search in then‐dimensional space can be performed only by the search for the point data in the hyper‐rectangle range.By this approach, the space splitting and data management are made possible considering the size of the object as well as the centroid. By the coordinate transformation the object data are represented by the maximum and the minimum in the originalN‐dimensional space, multiplied by a certain factor.It is indicated by simulation experiments that the speed of search is improved by this method by a factor of 2 to 4, compared to the conventional method when the size of the object to be managed is enlarged. When points and objects with small sizes are to be handled, the performance is almost the same as that of the conventional method. In the experiment using the actual printed circuit board data, the speed of search also is improved by approximately a

ISSN:0882-1666    

DOI:10.1002/scj.4690230303

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

年代:1992

数据来源: WILEY

摘要:

AbstractIf information concerning speech and hand gestures is handled in an integrated way, a more user‐friendly interface will be realized. These are defined here as the “hand gesture” as the “shape and motion of hands to transmit the intention of the speaker.” They aim at the realization of the hand gesture interface.Using the hand shape input device Data GloveTM, we investigate the coding of the hand shape/motion information, a series of hand shapes in a continuous motion and the recognition method for the hand motion. For the case of hand shape/motion in the alphabet gesture in the sign language, a recognition experiment is executed. It is verified that some 20 hand gestures can be recognized, including the dynamic hand gestures. As the interface for the map guidance system, the instruction by speech and hand gesture is

ISSN:0882-1666    

DOI:10.1002/scj.4690230304

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

年代:1992

数据来源: WILEY

摘要:

AbstractWhen a mutually connected neural network (Hopfield net) is applied to the minimization problem, the process may fall in a local minimum and not converge to the global minimum. To remedy this problem, several means have been proposed such as Boltzmann machine; however, this and other proposals take too much time.This paper proposes a neural net that permits multitransition (transition between states with two or more Hamming distances). It is shown that convergence to the global minimum can be realized escaping from the local minimum. It is shown first that the decrement of the energy when more than one element simultaneously changes can be calculated by simple addition‐subtractions of the weight factors for the inputs to the element and the coupling factors among elements. A probabilistic minimum search algorithm is presented based on the multitransition. The optimization problem for 100 elements was solved by a personal computer, and the process converged to the minimum within an hour in 398 trials of 400. The average convergence time is 10 min, and the convergence is achieved in 20 s in the fastest case. The computation time increases with the increase of elements with a much slower rate than in the case of all‐sea

ISSN:0882-1666    

DOI:10.1002/scj.4690230305

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

年代:1992

数据来源: WILEY

摘要:

AbstractIn MRI based on the time‐varying gradient magnetic field, the image must be reconstructed from the sampled values which are distributed ink‐space with nonuniform intervals. If there exists any static field inhomogeneity, gradient field nonlinearity, or patient motion, the image must be reconstructed considering these effects, thereby preventing the deterioration of the reconstructed image.This paper proposes a new MRI image reconstruction method to cope with the forementioned problems in a unified way. The proposed method of reconstruction is based on the principle whereby the imaging equation representing the relation between the spin density distribution and FID signal is solved directly by an iterative least‐squares method. The proposed reconstruction method strictly considers that mathematically the measured data are the finite number of sampled values of the FID signal. The minimum norm solution is obtained as the reconstructed image. Also, a method is proposed which improves the convergence of the iterative least‐squares method by utilizing the information concerning the density of the sampling point distribution ink‐space. An image reconstruction experiment is shown assuming the spiralk

ISSN:0882-1666    

DOI:10.1002/scj.4690230306

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

年代:1992

数据来源: WILEY

摘要:

AbstractThe authors previously proposed MAP (multiangled parallel) operation, which extracts the feature of an image by applying the erosion‐dilation to the directional field, and applies the method to the topographical map. This paper discusses the development of MAP operation from the viewpoint of the hardware and the improvement of the extraction algorithm.The hardware components for MAP operation are presented first. It is shown that the feature of the image can be extracted through cyclic processing by the four kinds of 8‐parallel logic circuits using the local parallel operations. The implementation on the image processor is discussed. As to the feature extraction algorithm, a method is presented which extracts the curve continuously by propagating the direction feature plane component among the planes. As an example of the feature extraction by the directional erosion‐dilation, the extraction of the ropeway symbol is demonst

ISSN:0882-1666    

DOI:10.1002/scj.4690230307

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

年代:1992

数据来源: WILEY

摘要:

AbstractFingerprints are an effective means of identifying individuals, and several fingerprint comparison systems have already been developed. However, since it becomes difficult to recognize fingerprints with ease as the number and diversity of fingerprint patterns increases, it becomes desirable to create an efficient database for fingerprint matching with the goal of developing an automatic fingerprint classification system. Here an automatic classification system is imposed for fingerprint images in which the discrimination system is a multilayer neural network composed of five subnetworks to carry out macrofeature extraction on the fingerprint image. A total of 50 training examples, 10 per category, and 500 testing samples were presented to the network. According to the obtained results, the network achieved an 86 percent classification rate for the first candidate and a 99 percent classification rate for the first two candidates. The state of the network was analyzed after training, and the reasons for the system effectiveness and its classification error were studied.

ISSN:0882-1666    

DOI:10.1002/scj.4690230308

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

年代:1992

数据来源: WILEY

摘要:

AbstractA sufficient condition for the state of a recurrent neural network to converge stably to an equilibrium state is the symmetry of the weights of connections between constituent units. However, generally, it imposes a strong restriction on the capability of the network. Although several stability conditions have been proposed for asymmetric recurrent networks, they are too strict and not useful for actual neural networks. Six new stability conditions are derived herein by using two types of Lyapunov functions. Some of them provide milder constraints on the connection weights than the conventional results, and others are particularly useful when the mutual connections between units have opposite signs of weights.

ISSN:0882-1666    

DOI:10.1002/scj.4690230309

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

年代:1992

数据来源: WILEY