Image signatures are generated from the comparison of segments contained within an image to a database of segments collected over a large variety of images. It is impossible to retain all of the segments from all of the images so the segments are clustered becomes an image primitive as each cluster contains a unique set of similar segments. The size of the image signature is NK where N is the number of segments and K is the number of clusters. These numbers are significantly smaller than the dimensions of the image and so a signature is a condensed representation of the contents of the image.
{"title":"Image primitive signatures","authors":"J. Kinser","doi":"10.1109/AIPR.2004.28","DOIUrl":"https://doi.org/10.1109/AIPR.2004.28","url":null,"abstract":"Image signatures are generated from the comparison of segments contained within an image to a database of segments collected over a large variety of images. It is impossible to retain all of the segments from all of the images so the segments are clustered becomes an image primitive as each cluster contains a unique set of similar segments. The size of the image signature is NK where N is the number of segments and K is the number of clusters. These numbers are significantly smaller than the dimensions of the image and so a signature is a condensed representation of the contents of the image.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"45 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":"115678416","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}
Successful unmanned ground vehicle (UGV) navigation in urban areas requires the competence of the vehicle to cope with Global Positioning System (GPS) outages and/or unreliable position estimates due to multipathing. At the National Institute of Standards and Technology (NIST) we are developing registration algorithms using LADAR (LAser Detection And Ranging) data to cope with such scenarios. In this paper, we present a building detection and recognition (BDR) algorithm using LADAR range images acquired from UGVs towards reliable and efficient registration. We verify the proposed algorithms using field data obtained from a Riegl LADAR range sensor mounted on a UGV operating in a variety of unknown urban environments. The presented results show the robustness and efficacy of the BDR algorithm.
{"title":"Robust detection and recognition of buildings in urban environments from LADAR data","authors":"R. Madhavan, T. Hong","doi":"10.1109/AIPR.2004.40","DOIUrl":"https://doi.org/10.1109/AIPR.2004.40","url":null,"abstract":"Successful unmanned ground vehicle (UGV) navigation in urban areas requires the competence of the vehicle to cope with Global Positioning System (GPS) outages and/or unreliable position estimates due to multipathing. At the National Institute of Standards and Technology (NIST) we are developing registration algorithms using LADAR (LAser Detection And Ranging) data to cope with such scenarios. In this paper, we present a building detection and recognition (BDR) algorithm using LADAR range images acquired from UGVs towards reliable and efficient registration. We verify the proposed algorithms using field data obtained from a Riegl LADAR range sensor mounted on a UGV operating in a variety of unknown urban environments. The presented results show the robustness and efficacy of the BDR algorithm.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"1 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":"124738723","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}
Inspired by recent advances in real-time vision for certain applications, we propose a framework for developing and implementing systems that are capable of detecting and recognizing a large number of objects in real time on a top desktop workstation with field programmable gate array (FPGA) devices. To avoid explicit segmentation, detection and recognition is performed by scanning through local windows of input scenes at multiple scales. This is achieved by using a new feature family (named as topological local spectral histogram (ToLoSH) features, consisting of histograms of local regions of filtered images) and a lookup table decision tree (i.e. a decision tree where each node is implemented as lookup tables) as the classifier to reduce the average time per local window while achieving high accuracy. We show through analysis and empirical studies that ToLoSH features are effective to discriminate a large number of object classes and can be computed using only three instructions. Given the choice of the ToLoSH feature family and lookup table decision tree classifiers, the problem of real-time scene interpretation becomes a joint optimization problem of learning an optimal classifier and associated optimal ToLoSH features. To show the feasibility of the proposed framework, we have constructed a decision lookup table tree for a dataset consisting of textures, faces, and objects. We argue that the proposed framework may reconcile some of the fundamental issues in visual recognition modeling.
{"title":"A computational framework for real-time detection and recognition of large number of classes","authors":"Li Tao, V. Asari","doi":"10.1109/AIPR.2004.1","DOIUrl":"https://doi.org/10.1109/AIPR.2004.1","url":null,"abstract":"Inspired by recent advances in real-time vision for certain applications, we propose a framework for developing and implementing systems that are capable of detecting and recognizing a large number of objects in real time on a top desktop workstation with field programmable gate array (FPGA) devices. To avoid explicit segmentation, detection and recognition is performed by scanning through local windows of input scenes at multiple scales. This is achieved by using a new feature family (named as topological local spectral histogram (ToLoSH) features, consisting of histograms of local regions of filtered images) and a lookup table decision tree (i.e. a decision tree where each node is implemented as lookup tables) as the classifier to reduce the average time per local window while achieving high accuracy. We show through analysis and empirical studies that ToLoSH features are effective to discriminate a large number of object classes and can be computed using only three instructions. Given the choice of the ToLoSH feature family and lookup table decision tree classifiers, the problem of real-time scene interpretation becomes a joint optimization problem of learning an optimal classifier and associated optimal ToLoSH features. To show the feasibility of the proposed framework, we have constructed a decision lookup table tree for a dataset consisting of textures, faces, and objects. We argue that the proposed framework may reconcile some of the fundamental issues in visual recognition modeling.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"34 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":"126310465","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 Intelligent Systems Division of the National Institute of Standards and Technology has been engaged for several years in developing real-time systems for autonomous driving. A road detection program is an essential part of the project. Previously we developed an adaptive road detection system based on color histograms using a neural network. This, however, still required human involvement during the initialization step. As a continuation of the project, we have expanded the system so that it can adapt to the new environment without any human intervention. This system updates the neural network continuously based on the road image structure. In order to reduce the possibility of misclassifying road and non-road, we have implemented an adaptive road feature acquisition method.
{"title":"Adaptive road detection through continuous environment learning","authors":"Mike Foedisch, A. Takeuchi","doi":"10.1109/AIPR.2004.9","DOIUrl":"https://doi.org/10.1109/AIPR.2004.9","url":null,"abstract":"The Intelligent Systems Division of the National Institute of Standards and Technology has been engaged for several years in developing real-time systems for autonomous driving. A road detection program is an essential part of the project. Previously we developed an adaptive road detection system based on color histograms using a neural network. This, however, still required human involvement during the initialization step. As a continuation of the project, we have expanded the system so that it can adapt to the new environment without any human intervention. This system updates the neural network continuously based on the road image structure. In order to reduce the possibility of misclassifying road and non-road, we have implemented an adaptive road feature acquisition method.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"2011 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":"128001607","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}
Among many practical factors that need to be considered for a reliable character recognition system in 3D space, the location of the visual angle of a camera might play a crucial role. Different viewpoints in 3D space produces distorted license plate images in a camera. For this reason, a method is developed to segment and to recognize characters of license plate objects undergoing variant perspective view. A method for segmenting license plate characters on a moving vehicle in actual outdoor environment is based upon object contours. And the proposed method for recognizing is constructed from a feature-based approach, parameterized by an affine invariant parameters and the affine invariant features. Experimental results show that the performance of the proposed method is simple and robust, particularly when objects are heavily distorted with strong perspective view.
{"title":"A simple OCR method from strong perspective view","authors":"Mi-Ae Ko, Young-Mo Kim","doi":"10.1109/AIPR.2004.8","DOIUrl":"https://doi.org/10.1109/AIPR.2004.8","url":null,"abstract":"Among many practical factors that need to be considered for a reliable character recognition system in 3D space, the location of the visual angle of a camera might play a crucial role. Different viewpoints in 3D space produces distorted license plate images in a camera. For this reason, a method is developed to segment and to recognize characters of license plate objects undergoing variant perspective view. A method for segmenting license plate characters on a moving vehicle in actual outdoor environment is based upon object contours. And the proposed method for recognizing is constructed from a feature-based approach, parameterized by an affine invariant parameters and the affine invariant features. Experimental results show that the performance of the proposed method is simple and robust, particularly when objects are heavily distorted with strong perspective view.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"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":"131300546","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}
Geoffrey Wall, Faizal Iqbal, J. Isaacs, Xiuwen Liu, S. Foo
In this paper we present a novel hardware/software approach to implement a highly accurate texture classification algorithm. We propose the use of field programmable gate arrays (FPGAs) to efficiently compute multiple convolutions in parallel that is required by the spectral histogram representation we employ. The combination of custom hardware and software allows us to have a classifier that is able to achieve results of over 99% accuracy at a rate of roughly 6000 image classifications per second on a challenging real texture dataset.
{"title":"Real time texture classification using field programmable gate arrays","authors":"Geoffrey Wall, Faizal Iqbal, J. Isaacs, Xiuwen Liu, S. Foo","doi":"10.1109/AIPR.2004.38","DOIUrl":"https://doi.org/10.1109/AIPR.2004.38","url":null,"abstract":"In this paper we present a novel hardware/software approach to implement a highly accurate texture classification algorithm. We propose the use of field programmable gate arrays (FPGAs) to efficiently compute multiple convolutions in parallel that is required by the spectral histogram representation we employ. The combination of custom hardware and software allows us to have a classifier that is able to achieve results of over 99% accuracy at a rate of roughly 6000 image classifications per second on a challenging real texture dataset.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"59 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":"123393109","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}
R. Kendrick, Eric H. Smith, D. Christie, D. Bennett, D. Theil, E. Barrett
The Lockheed Martin Advanced Technology Center (LMVATC) is actively investigating alternate applications of coherently phased sparse-aperture optical imaging arrays. Controlling the relative phasing of the apertures enables these arrays to function as imaging interferometers, providing high spectral resolution as well as high spatial resolution imagery. In this paper we: a) summarize the basic theory of multiple-aperture imaging interferometers; b) illustrate the theory with Fourier transform imaging spectrometer (FTIS) simulations, using the Rochester Institute of Technology hyper-spectral scene simulator (DIRSIG) as our source of simulated input data; c) validate the theory with experimental results derived with an LM/ATC optical FTIS tested.
{"title":"Multiple-aperture imaging spectrometer: computer simulation and experimental validation","authors":"R. Kendrick, Eric H. Smith, D. Christie, D. Bennett, D. Theil, E. Barrett","doi":"10.1109/AIPR.2004.32","DOIUrl":"https://doi.org/10.1109/AIPR.2004.32","url":null,"abstract":"The Lockheed Martin Advanced Technology Center (LMVATC) is actively investigating alternate applications of coherently phased sparse-aperture optical imaging arrays. Controlling the relative phasing of the apertures enables these arrays to function as imaging interferometers, providing high spectral resolution as well as high spatial resolution imagery. In this paper we: a) summarize the basic theory of multiple-aperture imaging interferometers; b) illustrate the theory with Fourier transform imaging spectrometer (FTIS) simulations, using the Rochester Institute of Technology hyper-spectral scene simulator (DIRSIG) as our source of simulated input data; c) validate the theory with experimental results derived with an LM/ATC optical FTIS tested.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"73 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":"122791672","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}
A modified approach on modular PCA for face recognition is presented in this paper. The proposed changes aim to improve the recognition rates for modular PCA for face images with large variation in light and facial expression. The eyes form one of the most invariant regions on the face. A sub-image from this region is considered. Weight vectors from this region are appended to the existing weight vector for modular PCA. The accuracies for the modified method, the original method and PCA method are evaluated under conditions of varying pose, illumination and expressions using standard face databases.
{"title":"A multi-view approach on modular PCA for illumination and pose invariant face recognition","authors":"P. Sankaran, V. Asari","doi":"10.1109/AIPR.2004.4","DOIUrl":"https://doi.org/10.1109/AIPR.2004.4","url":null,"abstract":"A modified approach on modular PCA for face recognition is presented in this paper. The proposed changes aim to improve the recognition rates for modular PCA for face images with large variation in light and facial expression. The eyes form one of the most invariant regions on the face. A sub-image from this region is considered. Weight vectors from this region are appended to the existing weight vector for modular PCA. The accuracies for the modified method, the original method and PCA method are evaluated under conditions of varying pose, illumination and expressions using standard face databases.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"76 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":"132850318","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 main objective of this paper is to present the methodology used for measuring and monitoring the quality of fingerprint database and fingerprint match performance of a large user fingerprint identification system. The Department of Homeland Security's (DHS) biometric identification system is used as an example for this study. In addition the paper presents lessons learned during system performance testing and independent validation and verification analysis of large scale systems such as DHS's biometric system and recommend improvements for the current test methodology.
{"title":"Monitoring and reporting of fingerprint image quality and match accuracy for a large user application","authors":"Teddy Ko, R. Krishnan","doi":"10.1109/AIPR.2004.30","DOIUrl":"https://doi.org/10.1109/AIPR.2004.30","url":null,"abstract":"The main objective of this paper is to present the methodology used for measuring and monitoring the quality of fingerprint database and fingerprint match performance of a large user fingerprint identification system. The Department of Homeland Security's (DHS) biometric identification system is used as an example for this study. In addition the paper presents lessons learned during system performance testing and independent validation and verification analysis of large scale systems such as DHS's biometric system and recommend improvements for the current test methodology.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"14 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":"133292699","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}
Conventional image formation approaches rely on frequency domain Fourier methods to create images of objects. Most rely on integrating spatial resolution in the Fourier domain and do not accurately factor in the spatial aperture function to create an image because of the uniform spatial sampling necessary for the Fourier transform. We propose a multiresolution approach based on a Greens function inverse scattering method that allows us to solve for the object function directly in the time domain thereby allowing a more accurate rendering of the object in question.
{"title":"A multiresolution time domain approach to RF image formation","authors":"R. Bonneau","doi":"10.1109/AIPR.2004.5","DOIUrl":"https://doi.org/10.1109/AIPR.2004.5","url":null,"abstract":"Conventional image formation approaches rely on frequency domain Fourier methods to create images of objects. Most rely on integrating spatial resolution in the Fourier domain and do not accurately factor in the spatial aperture function to create an image because of the uniform spatial sampling necessary for the Fourier transform. We propose a multiresolution approach based on a Greens function inverse scattering method that allows us to solve for the object function directly in the time domain thereby allowing a more accurate rendering of the object in question.","PeriodicalId":120814,"journal":{"name":"33rd Applied Imagery Pattern Recognition Workshop (AIPR'04)","volume":"61 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":"116021654","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：