Scott K. Ralph, J. Irvine, M. R. Stevens, M. Snorrason, D. Gwilt
Present methods of quantifying the performance of ATR algorithms involves the use of large video datasets that must be truthed by hand, frame-by-frame, requiring vast amounts of time. To reduce this cost, we have developed an application that significantly reduces the cost by only requiring the operator to grade a relatively sparse number of data "keyframes". A correlation-based template matching algorithm computes the best position, orientation and scale when interpolating between keyframes. We demonstrate the performance of the automated truthing application, and compare the results to those of a series of human operator test subjects. The START-generated truth is shown to be very close to the mean truth data given by the human operators. Additionally the savings in labor producing the results is also demonstrated.
{"title":"Assessing the performance of an automated video ground truthing application","authors":"Scott K. Ralph, J. Irvine, M. R. Stevens, M. Snorrason, D. Gwilt","doi":"10.1109/AIPR.2004.15","DOIUrl":"https://doi.org/10.1109/AIPR.2004.15","url":null,"abstract":"Present methods of quantifying the performance of ATR algorithms involves the use of large video datasets that must be truthed by hand, frame-by-frame, requiring vast amounts of time. To reduce this cost, we have developed an application that significantly reduces the cost by only requiring the operator to grade a relatively sparse number of data \"keyframes\". A correlation-based template matching algorithm computes the best position, orientation and scale when interpolating between keyframes. We demonstrate the performance of the automated truthing application, and compare the results to those of a series of human operator test subjects. The START-generated truth is shown to be very close to the mean truth data given by the human operators. Additionally the savings in labor producing the results is also demonstrated.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"89 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127191990","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}
We create a model of joint color changes in face images due to lighting variations. This is done by observing how colors of an individual's face with fixed pose and expression are mapped to new colors under different lighting conditions. One of the challenges we are dealing with in this work is that the scenes are not constant for different lighting. Hence we cannot observe the joint color changes of the scenes. However all the scenes have a human subject with approximately frontal pose, so we use the color changes observed on a human subjects face to learn the color mapping. The joint color mappings are represented in a low dimensional subspace obtained using singular value decomposition (SVD). Using these maps the detected face from a new image can be transformed to appear as if taken under canonical lighting condition.
{"title":"Illumination invariant faces","authors":"Rajkiran Gottumukkal, V. Asari","doi":"10.1109/AIPR.2004.27","DOIUrl":"https://doi.org/10.1109/AIPR.2004.27","url":null,"abstract":"We create a model of joint color changes in face images due to lighting variations. This is done by observing how colors of an individual's face with fixed pose and expression are mapped to new colors under different lighting conditions. One of the challenges we are dealing with in this work is that the scenes are not constant for different lighting. Hence we cannot observe the joint color changes of the scenes. However all the scenes have a human subject with approximately frontal pose, so we use the color changes observed on a human subjects face to learn the color mapping. The joint color mappings are represented in a low dimensional subspace obtained using singular value decomposition (SVD). Using these maps the detected face from a new image can be transformed to appear as if taken under canonical lighting condition.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"143 7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129470020","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}
Persistent ISR (intelligence surveillance and reconnaissance) has proven its value as a tactic for national defense. This activity can collect, in particular, information necessary for executing an important concept of operations: wide-area autonomous change detection over long time intervals. Here we describe the remarkable potential of hyperspectral remote sensing systems for enabling such missions, using either visible or thermal infrared wavelengths. First we describe blind change detection, in which no target knowledge is assumed. Targets that have moved can nevertheless be distinguished from naturally occurring background radiometric changes through the use of multivariate statistics informed by simple physics. Detection relies on the ability of hyperspectral algorithms to predict certain conserved properties of background spectral patterns over long time intervals. We also describe a method of mitigating the most worrisome practical engineering difficulty in pixel-level change detection, image misregistration. This has led, in turn, to a method of estimating spectral signature evolution using multiple-scene statistics. Finally, we present a signature-based detection technique that fuses two discrimination mechanisms: use of some prior knowledge of target spectra, and the fact that a change has occurred.
{"title":"Advanced algorithms for autonomous hyperspectral change detection","authors":"A. Schaum, A. Stocker","doi":"10.1109/AIPR.2004.10","DOIUrl":"https://doi.org/10.1109/AIPR.2004.10","url":null,"abstract":"Persistent ISR (intelligence surveillance and reconnaissance) has proven its value as a tactic for national defense. This activity can collect, in particular, information necessary for executing an important concept of operations: wide-area autonomous change detection over long time intervals. Here we describe the remarkable potential of hyperspectral remote sensing systems for enabling such missions, using either visible or thermal infrared wavelengths. First we describe blind change detection, in which no target knowledge is assumed. Targets that have moved can nevertheless be distinguished from naturally occurring background radiometric changes through the use of multivariate statistics informed by simple physics. Detection relies on the ability of hyperspectral algorithms to predict certain conserved properties of background spectral patterns over long time intervals. We also describe a method of mitigating the most worrisome practical engineering difficulty in pixel-level change detection, image misregistration. This has led, in turn, to a method of estimating spectral signature evolution using multiple-scene statistics. Finally, we present a signature-based detection technique that fuses two discrimination mechanisms: use of some prior knowledge of target spectra, and the fact that a change has occurred.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116987824","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}
In this paper, a morphological based method for recognition of handwritten middle Persian characters is presented. After pre-processing and noise cancellation, morphological erosion operator with many structure elements is applied. The structure elements are with variable length lines at directions 0, 45, 90, 135 degrees. A five element feature set has been defined so: (1) relative energy of eroded version with respect to the original image energy (REL/spl I.bar/ENG),(2) displacement of the center of mass (CM/spl I.bar//spl I.bar/DIS), (3) minimum eigenvalue (EIG/spl I.bar/MIN), (4) maximum eigenvalue (EIG/spl I.bar/MAX) and (5) its direction (EIG-DIR). These features are used to design a feedforward neural network with one hidden layer. The best classification error is about 2.39% (97.61% recognition rate), and is achieved with 150 neurons for the hidden layer.
{"title":"An efficient selected feature set for the middle age Persian character recognition","authors":"S. Alirezaee, H. Aghaeinia, M. Ahmadi, K. Faez","doi":"10.1109/AIPR.2004.12","DOIUrl":"https://doi.org/10.1109/AIPR.2004.12","url":null,"abstract":"In this paper, a morphological based method for recognition of handwritten middle Persian characters is presented. After pre-processing and noise cancellation, morphological erosion operator with many structure elements is applied. The structure elements are with variable length lines at directions 0, 45, 90, 135 degrees. A five element feature set has been defined so: (1) relative energy of eroded version with respect to the original image energy (REL/spl I.bar/ENG),(2) displacement of the center of mass (CM/spl I.bar//spl I.bar/DIS), (3) minimum eigenvalue (EIG/spl I.bar/MIN), (4) maximum eigenvalue (EIG/spl I.bar/MAX) and (5) its direction (EIG-DIR). These features are used to design a feedforward neural network with one hidden layer. The best classification error is about 2.39% (97.61% recognition rate), and is achieved with 150 neurons for the hidden layer.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116175316","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}
The objective of this research is to discover fuzzy semantic and syntactic relationships among English words at various levels of abstraction without using any other sources of semantic or syntactical reference information (e.g. dictionaries, lexicons, grammar rules, etc...) An agglomerative clustering algorithm is applied to the co-occurrence space formed by subsets of target words and training words the output of which is a set of semantic or syntactic classes. The fuzzy-relationships (membership coefficients) between test words and the semantic and syntactic classes are estimated by the non-negative least-squares solution to the system of linear equations defined. Experiments using raw text in 218 unrelated novels have yielded promising results. It is expected that larger and/or more narrowly focused training sets would yield even better and more diverse results.
{"title":"Unsupervised fuzzy-membership estimation of terms in semantic and syntactic lexical classes","authors":"David Portnoy, P. Bock","doi":"10.1109/AIPR.2004.48","DOIUrl":"https://doi.org/10.1109/AIPR.2004.48","url":null,"abstract":"The objective of this research is to discover fuzzy semantic and syntactic relationships among English words at various levels of abstraction without using any other sources of semantic or syntactical reference information (e.g. dictionaries, lexicons, grammar rules, etc...) An agglomerative clustering algorithm is applied to the co-occurrence space formed by subsets of target words and training words the output of which is a set of semantic or syntactic classes. The fuzzy-relationships (membership coefficients) between test words and the semantic and syntactic classes are estimated by the non-negative least-squares solution to the system of linear equations defined. Experiments using raw text in 218 unrelated novels have yielded promising results. It is expected that larger and/or more narrowly focused training sets would yield even better and more diverse results.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123944118","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}
Neural models of the mammalian visual cortex have been used in digital image processing for many years. The neural models have demonstrated a robust talent for extracting image segments that are inherent in the image. This paper explores the use of a simple neural model for the extraction of texture information. The neural firing patterns in the model are dependent upon the input texture and by examining the neural patterns it is possible to classify the texture in the input image. Two applications presented here are the comparison of performance to other texture analysis techniques using two standard databases and the classification of texture regions in a medical image.
{"title":"Texture discrimination and classification using pulse images","authors":"Guisong Wang, J. Kinser","doi":"10.1109/AIPR.2004.44","DOIUrl":"https://doi.org/10.1109/AIPR.2004.44","url":null,"abstract":"Neural models of the mammalian visual cortex have been used in digital image processing for many years. The neural models have demonstrated a robust talent for extracting image segments that are inherent in the image. This paper explores the use of a simple neural model for the extraction of texture information. The neural firing patterns in the model are dependent upon the input texture and by examining the neural patterns it is possible to classify the texture in the input image. Two applications presented here are the comparison of performance to other texture analysis techniques using two standard databases and the classification of texture regions in a medical image.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132374436","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}
One of the more difficult problems in image processing is segmentation. The human brain has an ability that is unmatched by any current technology for breaking down the world into distributed features and reconstructing them into distinct objects. Neurons encode information both in the number of spikes fired in a given time period, which indicates the strength with which a given local feature is present, and in the temporal code or relative timing of the spike, indicating whether the individual features are part of the same or different objects. Neurons that respond to contiguous stimuli produce synchronous oscillations, while those that are not fire independently. Thus, neural synchrony could be used as a tag for each pixel in an image indicating to which object it belongs. We have developed a simulation based on the primary visual cortex. We found that neurons that respond to the same object oscillate synchronously while those that respond to different objects fire independently.
{"title":"Neurally-based algorithms for image processing","authors":"Mark Flynn, H. Abarbanel, Garrett T. Kenyon","doi":"10.1109/AIPR.2004.34","DOIUrl":"https://doi.org/10.1109/AIPR.2004.34","url":null,"abstract":"One of the more difficult problems in image processing is segmentation. The human brain has an ability that is unmatched by any current technology for breaking down the world into distributed features and reconstructing them into distinct objects. Neurons encode information both in the number of spikes fired in a given time period, which indicates the strength with which a given local feature is present, and in the temporal code or relative timing of the spike, indicating whether the individual features are part of the same or different objects. Neurons that respond to contiguous stimuli produce synchronous oscillations, while those that are not fire independently. Thus, neural synchrony could be used as a tag for each pixel in an image indicating to which object it belongs. We have developed a simulation based on the primary visual cortex. We found that neurons that respond to the same object oscillate synchronously while those that respond to different objects fire independently.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115609615","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}
Designing or learning of templates for cellular neural networks constitutes one of the crucial research problems of this paradigm. In this work, we present the use of a particle swarm optimizer, a global search algorithm, to design a template set for a CNN. A brief overview of the algorithms and methods is given. Design of popular templates is performed using the search algorithm described.
{"title":"Designing templates for cellular neural networks using particle swarm optimization","authors":"H. Firpi, E. Goodman","doi":"10.1109/AIPR.2004.21","DOIUrl":"https://doi.org/10.1109/AIPR.2004.21","url":null,"abstract":"Designing or learning of templates for cellular neural networks constitutes one of the crucial research problems of this paradigm. In this work, we present the use of a particle swarm optimizer, a global search algorithm, to design a template set for a CNN. A brief overview of the algorithms and methods is given. Design of popular templates is performed using the search algorithm described.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"129 2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124587977","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}
G. Carpenter, S. Martens, E. Mingolla, Ogi J. Ogas, C. Gaddam
Ongoing research at Boston University has produced computational models of biological vision and learning that embody a growing corpus of scientific data and predictions. Vision models perform long-range grouping and figure/ground segmentation, and memory models create attentionally controlled recognition codes that intrinsically combine bottom-up activation and top-down learned expectations. These two streams of research form the foundation of novel dynamically integrated systems for image understanding. Simulations using multispectral images illustrate road completion across occlusions in a cluttered scene and information fusion from input labels that are simultaneously inconsistent and correct. The CNS Vision and Technology Labs (cns.bu.edu/visionlab and cns.bu.edu/iechlab) are further integrating science and technology through analysis, testing, and development of cognitive and neural models for large-scale applications, complemented by software specification and code distribution.
{"title":"Biologically inspired approaches to automated feature extraction and target recognition","authors":"G. Carpenter, S. Martens, E. Mingolla, Ogi J. Ogas, C. Gaddam","doi":"10.1109/AIPR.2004.17","DOIUrl":"https://doi.org/10.1109/AIPR.2004.17","url":null,"abstract":"Ongoing research at Boston University has produced computational models of biological vision and learning that embody a growing corpus of scientific data and predictions. Vision models perform long-range grouping and figure/ground segmentation, and memory models create attentionally controlled recognition codes that intrinsically combine bottom-up activation and top-down learned expectations. These two streams of research form the foundation of novel dynamically integrated systems for image understanding. Simulations using multispectral images illustrate road completion across occlusions in a cluttered scene and information fusion from input labels that are simultaneously inconsistent and correct. The CNS Vision and Technology Labs (cns.bu.edu/visionlab and cns.bu.edu/iechlab) are further integrating science and technology through analysis, testing, and development of cognitive and neural models for large-scale applications, complemented by software specification and code distribution.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"88 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128933850","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}
J. Price, Carlos Maraviglia, William Seisler, E. Williams, M. Pauli
Using image processing techniques, the Gunfire Detection and Location (GDL) system detects gunfire and aims a suite of imagers at the muzzle flash point of origin. This detection and location function is critical for force and perimeter defense in densely populated areas as well as difficult operating environments such as a remote desert. This paper defines requirements of the GDL project. The GDL system is the result of research into using real-time image processing of mid-wave infrared imagery to detect gunfire and pin point its origin. Varieties of modern imagers are made available over the spectrum to aid an operator in assessing a detected signal. By using optical and acoustical methods, a design effort was launched to yield five vehicle based platforms. The hardware and algorithm used to implement the five basic functions is discussed in this paper. Issues such as component reliability, thermal issues, camera sensitivity operated during the daytime and nighttime, and optical design and bore sighting had to be united into a system designed to operate in the desert and powered from a high mobility multi-purpose wheeled vehicle (HMMWV). The design, construction and testing was conducted in nine months. The project has yielded a system architecture that will be further tested and refined in the next phase of this project. Experiences with the development phase of GDL and future directions are described in this paper.
{"title":"System capabilities, requirements and design of the GDL gunfire detection and location system","authors":"J. Price, Carlos Maraviglia, William Seisler, E. Williams, M. Pauli","doi":"10.1109/AIPR.2004.42","DOIUrl":"https://doi.org/10.1109/AIPR.2004.42","url":null,"abstract":"Using image processing techniques, the Gunfire Detection and Location (GDL) system detects gunfire and aims a suite of imagers at the muzzle flash point of origin. This detection and location function is critical for force and perimeter defense in densely populated areas as well as difficult operating environments such as a remote desert. This paper defines requirements of the GDL project. The GDL system is the result of research into using real-time image processing of mid-wave infrared imagery to detect gunfire and pin point its origin. Varieties of modern imagers are made available over the spectrum to aid an operator in assessing a detected signal. By using optical and acoustical methods, a design effort was launched to yield five vehicle based platforms. The hardware and algorithm used to implement the five basic functions is discussed in this paper. Issues such as component reliability, thermal issues, camera sensitivity operated during the daytime and nighttime, and optical design and bore sighting had to be united into a system designed to operate in the desert and powered from a high mobility multi-purpose wheeled vehicle (HMMWV). The design, construction and testing was conducted in nine months. The project has yielded a system architecture that will be further tested and refined in the next phase of this project. Experiences with the development phase of GDL and future directions are described in this paper.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2004-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129071902","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：