Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521021
Heung-Nam Kim, M. J. Irwin, R. Owens
Since image processing tasks in machine perception are computationally intensive, massively parallel processing is often employed to exploit the inherent parallelism. Some pioneering researchers have developed massively parallel processing systems including the CLIP family, the MPP, and the CM. All of these systems however, are hard to use in machine perception systems because of their expensiveness and physical size. To tackle these problems we developed an inexpensive and easily portable massively parallel processing system, the MGAP (Micro Grained Array Processor). The MGAP is a very fine grained, massively parallel, programmable array processor which is designed to be used as a low cost massively parallel coprocessor board in a desktop workstation. We discuss the versatile applications of the MGAP for image processing subtasks including an efficient histogramming algorithm, the systolic 2D Discrete Cosine Transform (DCT), and the Dynamic Space Warping Algorithm (DSWA).
{"title":"MGAP applications in machine perception","authors":"Heung-Nam Kim, M. J. Irwin, R. Owens","doi":"10.1109/CAMP.1995.521021","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521021","url":null,"abstract":"Since image processing tasks in machine perception are computationally intensive, massively parallel processing is often employed to exploit the inherent parallelism. Some pioneering researchers have developed massively parallel processing systems including the CLIP family, the MPP, and the CM. All of these systems however, are hard to use in machine perception systems because of their expensiveness and physical size. To tackle these problems we developed an inexpensive and easily portable massively parallel processing system, the MGAP (Micro Grained Array Processor). The MGAP is a very fine grained, massively parallel, programmable array processor which is designed to be used as a low cost massively parallel coprocessor board in a desktop workstation. We discuss the versatile applications of the MGAP for image processing subtasks including an efficient histogramming algorithm, the systolic 2D Discrete Cosine Transform (DCT), and the Dynamic Space Warping Algorithm (DSWA).","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"97 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122897941","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521026
M. Mekkaoui, N. Benameur, K. Ouriachi
Discusses the intention to provide robots with a computer system which gives them the capability to build an abstract and condensed description of their environment from low level data provided by an artificial vision system. This intention rapidly faced a complexity barrier: on one hand, the importance of the volume of information conveyed by an image they manipulate; on the other hand, the access to pertinent information to validate decision making. To reach this goal, research is oriented towards extracting a set of computing tools learning in two ways: descriptive process of the objects to manipulate, and constructive process of the pertinent information for each operation to apply to an object.
{"title":"A description approach of information in a robotics-vision system","authors":"M. Mekkaoui, N. Benameur, K. Ouriachi","doi":"10.1109/CAMP.1995.521026","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521026","url":null,"abstract":"Discusses the intention to provide robots with a computer system which gives them the capability to build an abstract and condensed description of their environment from low level data provided by an artificial vision system. This intention rapidly faced a complexity barrier: on one hand, the importance of the volume of information conveyed by an image they manipulate; on the other hand, the access to pertinent information to validate decision making. To reach this goal, research is oriented towards extracting a set of computing tools learning in two ways: descriptive process of the objects to manipulate, and constructive process of the pertinent information for each operation to apply to an object.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130596567","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521019
R.P. Rogers, I. Macduff, S. Tanimoto
We describe a systolic cellular logic architecture implemented in the systolic cellular logic (SCL) VLSI chip. The SCL chip's hardware support for virtual processing is intended to provide low cost, high performance image processing. We present empirical performance data for the Abingdon Cross benchmark which indicate that the SCL chip successfully delivers high performance at a relatively low cost. We further substantiate the SCL's success with empirical performance data for a complete experimental shape extraction protocol based on mathematical morphology.
{"title":"Systolic cellular logic: architecture and performance evaluation","authors":"R.P. Rogers, I. Macduff, S. Tanimoto","doi":"10.1109/CAMP.1995.521019","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521019","url":null,"abstract":"We describe a systolic cellular logic architecture implemented in the systolic cellular logic (SCL) VLSI chip. The SCL chip's hardware support for virtual processing is intended to provide low cost, high performance image processing. We present empirical performance data for the Abingdon Cross benchmark which indicate that the SCL chip successfully delivers high performance at a relatively low cost. We further substantiate the SCL's success with empirical performance data for a complete experimental shape extraction protocol based on mathematical morphology.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114616509","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521057
N. Ratha, Ameet K. Jain, M. Chung
Genetic algorithms (GA) are useful in solving complex optimization problems. By posing pattern clustering as an optimization problem, GAs can be used to obtain optimal minimum squared error partitions. In order to improve the total execution time, a distributed algorithm has been developed using the divide and conquer approach. Using a standard communication library called PVM, the distributed algorithm has been implemented on a workstation cluster: the GA approach gives better quality clusters for many data sets compared to a standard K-means clustering algorithm. We have achieved a near linear speedup for the distributed implementation.
{"title":"Clustering using a coarse-grained parallel genetic algorithm: a preliminary study","authors":"N. Ratha, Ameet K. Jain, M. Chung","doi":"10.1109/CAMP.1995.521057","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521057","url":null,"abstract":"Genetic algorithms (GA) are useful in solving complex optimization problems. By posing pattern clustering as an optimization problem, GAs can be used to obtain optimal minimum squared error partitions. In order to improve the total execution time, a distributed algorithm has been developed using the divide and conquer approach. Using a standard communication library called PVM, the distributed algorithm has been implemented on a workstation cluster: the GA approach gives better quality clusters for many data sets compared to a standard K-means clustering algorithm. We have achieved a near linear speedup for the distributed implementation.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127717052","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521024
J. Olk, P. P. Jonker
Typical real time computer vision tasks require huge amount of processing power, larger than can be achieved by current state of the art workstations. Parallel processing appears to be the only solution (economically) to obtain sufficient processing power for handling real time computer vision applications. The nature of processing in a typical computer vision algorithm usually ranges from many large small arithmetic operations (fine grain parallelism) to symbolic operations (coarse grain parallelism). Yet, normal general purpose parallel computers usually only suit one type of processing, not the whole range. The ESPRIT basic research project SM-IMP looks at a scalable combined SIMD-MIMD architecture for image processing, suiting both fine grain and coarse grain parallelism and capable of offering sufficient processing performance for real time computer vision applications. A programming model and simulation model for this SIMD-MIMD architecture are proposed.
{"title":"A programming and simulation model of a SIMD-MIMD architecture for image processing","authors":"J. Olk, P. P. Jonker","doi":"10.1109/CAMP.1995.521024","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521024","url":null,"abstract":"Typical real time computer vision tasks require huge amount of processing power, larger than can be achieved by current state of the art workstations. Parallel processing appears to be the only solution (economically) to obtain sufficient processing power for handling real time computer vision applications. The nature of processing in a typical computer vision algorithm usually ranges from many large small arithmetic operations (fine grain parallelism) to symbolic operations (coarse grain parallelism). Yet, normal general purpose parallel computers usually only suit one type of processing, not the whole range. The ESPRIT basic research project SM-IMP looks at a scalable combined SIMD-MIMD architecture for image processing, suiting both fine grain and coarse grain parallelism and capable of offering sufficient processing performance for real time computer vision applications. A programming model and simulation model for this SIMD-MIMD architecture are proposed.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133329520","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521040
M. Brooks, W. Chojnacki, A. van den Hengel
We consider the problem of recovering surface shape from image shading for the situation in which a distant overhead "sun" illuminates a Lambertian surface. An iterative scheme is presented which requires no prerequisite shape information. This scheme forms the basis for a parallel algorithm implemented on a CM-5. Performance of the CM-5 implementation is compared with that of a sequential implementation running on a Sun Sparc 2. Also considered are the complexity and scalability of the parallel algorithm as a function of image size and number of processors, respectively.
{"title":"Solving the shape-from-shading problem on the CM-5","authors":"M. Brooks, W. Chojnacki, A. van den Hengel","doi":"10.1109/CAMP.1995.521040","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521040","url":null,"abstract":"We consider the problem of recovering surface shape from image shading for the situation in which a distant overhead \"sun\" illuminates a Lambertian surface. An iterative scheme is presented which requires no prerequisite shape information. This scheme forms the basis for a parallel algorithm implemented on a CM-5. Performance of the CM-5 implementation is compared with that of a sequential implementation running on a Sun Sparc 2. Also considered are the complexity and scalability of the parallel algorithm as a function of image size and number of processors, respectively.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115433601","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521042
G. Caputo, L. Lombardi
In the literature different approaches of artificial visual systems inspired by biological solutions can be found. They can be classified into two main families: multiresolution and foveated retina approaches; for each family some solutions are described.
{"title":"Attention mechanisms in computer vision systems","authors":"G. Caputo, L. Lombardi","doi":"10.1109/CAMP.1995.521042","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521042","url":null,"abstract":"In the literature different approaches of artificial visual systems inspired by biological solutions can be found. They can be classified into two main families: multiresolution and foveated retina approaches; for each family some solutions are described.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115676248","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521065
A. Anzalone, A. Machì
An automated inspection method is sketched and its parallel implementation on a MIMD multiprocessor is discussed. The method is based on a segmentation of gray level views of the test object and on the extraction and measurement of meaningful blobs from the segmented images. Segmentation is performed processing the histogram of pixel intensities; blob extraction is based on the application of the distance transformation to interest regions. The method is able to detect shape defects affecting a type of moulded plastic drippers; typical defects revealed are: incompleteness, excess of moulded material along the joint lines, incompleteness of a labyrinth-like mask moulded on the dripper surface. The algorithm consists of several steps that can be easily implemented on a MIMD architecture using both a shared or a distributed memory approach. The implementation on the hierarchical shared memory ViP multiprocessor is discussed; drippers can be efficiently analyzed on this kind of system, configured with 3 clusters of 4 processors each, at the production rate of two per second, making possible single piece quality certification.
{"title":"Real-time visual inspection of moulded plastic drippers","authors":"A. Anzalone, A. Machì","doi":"10.1109/CAMP.1995.521065","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521065","url":null,"abstract":"An automated inspection method is sketched and its parallel implementation on a MIMD multiprocessor is discussed. The method is based on a segmentation of gray level views of the test object and on the extraction and measurement of meaningful blobs from the segmented images. Segmentation is performed processing the histogram of pixel intensities; blob extraction is based on the application of the distance transformation to interest regions. The method is able to detect shape defects affecting a type of moulded plastic drippers; typical defects revealed are: incompleteness, excess of moulded material along the joint lines, incompleteness of a labyrinth-like mask moulded on the dripper surface. The algorithm consists of several steps that can be easily implemented on a MIMD architecture using both a shared or a distributed memory approach. The implementation on the hierarchical shared memory ViP multiprocessor is discussed; drippers can be efficiently analyzed on this kind of system, configured with 3 clusters of 4 processors each, at the production rate of two per second, making possible single piece quality certification.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125869007","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521060
V. di Gesú, F. Isgrò, B. Lenzitti, D. Tegolo
Algorithms based on information fusion are often embodied in visual perception systems. Distributed architectures have been recently proposed to perform integrated computation. The complexity of distributed systems regards both their design, and the software environment to develop applications. Visual and iconic programming style intends to provide expressive tools to implement, to debug, and to execute programs in distributed environments. Multilayer graph languages seem suitable for handling such complexity. The paper describes the design of a visual dynamic environment (VDE), which is based on a graph grammar. A new class of dynamic visual interfaces is also introduced, and its properties are described. The proposed VDE has been implemented on the first emulated version of the machine M-VIF (Machine Vision based on Information Fusion).
{"title":"Visual dynamic environment for distributed systems","authors":"V. di Gesú, F. Isgrò, B. Lenzitti, D. Tegolo","doi":"10.1109/CAMP.1995.521060","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521060","url":null,"abstract":"Algorithms based on information fusion are often embodied in visual perception systems. Distributed architectures have been recently proposed to perform integrated computation. The complexity of distributed systems regards both their design, and the software environment to develop applications. Visual and iconic programming style intends to provide expressive tools to implement, to debug, and to execute programs in distributed environments. Multilayer graph languages seem suitable for handling such complexity. The paper describes the design of a visual dynamic environment (VDE), which is based on a graph grammar. A new class of dynamic visual interfaces is also introduced, and its properties are described. The proposed VDE has been implemented on the first emulated version of the machine M-VIF (Machine Vision based on Information Fusion).","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126656090","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1995-09-18DOI: 10.1109/CAMP.1995.521055
B. Modayur, L. Shapiro
The PERFORM matching method, introduced by B. Modayur and I.G. Shapiro (1994), solves the recognition problem under a bounded error noise model by establishing correspondences between model and image features. PERFORM evaluates correspondences by intersecting error regions in the image space. The article describes the parallel formulation of the PERFORM matching method and the implications of a shared memory, MIMD implementation. When a single solution is sought, the time complexity of the sequential matching algorithm using point features is of the order O(I/sup 2/NI) for 2D-2D matching and O(I/sup 3/NI) for 2D-3D matching, where N is the number of model features and I is the number of image features. The corresponding parallel algorithm using O(I/sup 2/) processors for 2D-2D matching and processors for 2D-3D matching has O(NI) complexity. When line features are used, the sequential complexity is of the order O(I NI) for 2D-2D matching and O(I/sup 2/ NI) for 2D-3D matching. The corresponding parallel algorithm utilizing O(I) processors for 2D-2D matching and O(I/sup 2/) processors for 2D-3D matching has O(NI) complexity. When implemented in parallel, the method requires minimal memory and obviates load balancing overheads and communication between processors. The article describes parallel implementations of 2D-2D matching on a shared memory, MIMD machine (KSR-I). Results show that significant, close to linear speedups are achievable using multiple processors.
{"title":"Parallel algorithm for object recognition and its implementation on a MIMD machine","authors":"B. Modayur, L. Shapiro","doi":"10.1109/CAMP.1995.521055","DOIUrl":"https://doi.org/10.1109/CAMP.1995.521055","url":null,"abstract":"The PERFORM matching method, introduced by B. Modayur and I.G. Shapiro (1994), solves the recognition problem under a bounded error noise model by establishing correspondences between model and image features. PERFORM evaluates correspondences by intersecting error regions in the image space. The article describes the parallel formulation of the PERFORM matching method and the implications of a shared memory, MIMD implementation. When a single solution is sought, the time complexity of the sequential matching algorithm using point features is of the order O(I/sup 2/NI) for 2D-2D matching and O(I/sup 3/NI) for 2D-3D matching, where N is the number of model features and I is the number of image features. The corresponding parallel algorithm using O(I/sup 2/) processors for 2D-2D matching and processors for 2D-3D matching has O(NI) complexity. When line features are used, the sequential complexity is of the order O(I NI) for 2D-2D matching and O(I/sup 2/ NI) for 2D-3D matching. The corresponding parallel algorithm utilizing O(I) processors for 2D-2D matching and O(I/sup 2/) processors for 2D-3D matching has O(NI) complexity. When implemented in parallel, the method requires minimal memory and obviates load balancing overheads and communication between processors. The article describes parallel implementations of 2D-2D matching on a shared memory, MIMD machine (KSR-I). Results show that significant, close to linear speedups are achievable using multiple processors.","PeriodicalId":277209,"journal":{"name":"Proceedings of Conference on Computer Architectures for Machine Perception","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131179740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：