Our previous work on road detection for autonomous road vehicles suggests the usage of high-level symbolic knowledge about the road structure. In this paper, we present our new approach to symbolic road recognition. We explain feature extraction, model representation, and the tree search-based matching processes and discuss performance evaluation results.
{"title":"Symbolic Road Perception-based Autonomous Driving in Urban Environments","authors":"Mike Foedisch, R. Madhavan, C. Schlenoff","doi":"10.1109/AIPR.2006.38","DOIUrl":"https://doi.org/10.1109/AIPR.2006.38","url":null,"abstract":"Our previous work on road detection for autonomous road vehicles suggests the usage of high-level symbolic knowledge about the road structure. In this paper, we present our new approach to symbolic road recognition. We explain feature extraction, model representation, and the tree search-based matching processes and discuss performance evaluation results.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132192139","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 automated identification and mapping of the constituent materials in a hyperspectral image is a problem of considerable interest. A significant issue is that the spectra at many pixels in such an image are actually mixtures of the spectra of the pure constituents. I review methods of "unmixing" spectra into their pure constituents, both when a "spectral library" of the pure constituents is available, and where no such library is available. Our own algorithms in both these areas are exemplified with a mineral and a biological example.
{"title":"Some Unmixing Problems and Algorithms in Spectroscopy and Hyperspectral Imaging","authors":"M. Berman","doi":"10.1109/AIPR.2006.37","DOIUrl":"https://doi.org/10.1109/AIPR.2006.37","url":null,"abstract":"The automated identification and mapping of the constituent materials in a hyperspectral image is a problem of considerable interest. A significant issue is that the spectra at many pixels in such an image are actually mixtures of the spectra of the pure constituents. I review methods of \"unmixing\" spectra into their pure constituents, both when a \"spectral library\" of the pure constituents is available, and where no such library is available. Our own algorithms in both these areas are exemplified with a mineral and a biological example.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123639665","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}
I have developed covariance matrix expressions for the registration Cramer-Rao lower bound for images processed with a linear filter. These results also generalize a previous registration CRLB by accounting for Poisson noise as well as read noise. Expressions shown here have been translated into a Fortran 90 code currently being tested.
{"title":"Cramer-Rao Lower Bound Calculations for Registration of Linearly-Filtered Images","authors":"D. Tyler","doi":"10.1109/AIPR.2006.19","DOIUrl":"https://doi.org/10.1109/AIPR.2006.19","url":null,"abstract":"I have developed covariance matrix expressions for the registration Cramer-Rao lower bound for images processed with a linear filter. These results also generalize a previous registration CRLB by accounting for Poisson noise as well as read noise. Expressions shown here have been translated into a Fortran 90 code currently being tested.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122255684","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}
Administration of clinical photographs taken by commonly used digital camera often requires troublesome manual operation. In this paper, we made a prototype scheme of automatic photographed area identification from clinical images to help or reduce administration task. A total of 8047 clinical photographs taken in department of dermatology, Keio University Hospital, were classified into 11 categories; head, hair, upper limb, lower limb, trunk, palm, sole, back of hand, back of foot, finger & detent and genital; to meet request by several dermatologists and we developed separate linear classifiers for each body region. The developed classifiers achieved an 82.8% in sensitivity (SE) and an 82.0% of specificity (SP) in average. In addition, integration of these classifiers with consideration of the feature space of each body region improved SP of 2.3% and precision (PR) of 3.0% at a maximum when the classification threshold was set to around 75% in SE. The proposed scheme requires only photographs to identify the photographed area and therefore it can be easily applied for DICOM (digital image and communication in medicine) system that is commonly used in clinical practice or other medical database systems.
{"title":"Automatic Identification of Shot Body Region from Clinical Photographies","authors":"H. Iyatomi, H. Oka, Masaru Tanaka, K. Ogawa","doi":"10.1109/AIPR.2006.17","DOIUrl":"https://doi.org/10.1109/AIPR.2006.17","url":null,"abstract":"Administration of clinical photographs taken by commonly used digital camera often requires troublesome manual operation. In this paper, we made a prototype scheme of automatic photographed area identification from clinical images to help or reduce administration task. A total of 8047 clinical photographs taken in department of dermatology, Keio University Hospital, were classified into 11 categories; head, hair, upper limb, lower limb, trunk, palm, sole, back of hand, back of foot, finger & detent and genital; to meet request by several dermatologists and we developed separate linear classifiers for each body region. The developed classifiers achieved an 82.8% in sensitivity (SE) and an 82.0% of specificity (SP) in average. In addition, integration of these classifiers with consideration of the feature space of each body region improved SP of 2.3% and precision (PR) of 3.0% at a maximum when the classification threshold was set to around 75% in SE. The proposed scheme requires only photographs to identify the photographed area and therefore it can be easily applied for DICOM (digital image and communication in medicine) system that is commonly used in clinical practice or other medical database systems.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126706879","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. Vasile, Frederick R. Waugh, Daniel Greisokh, R. Heinrichs
We present an algorithm for the automatic fusion of city-sized, 2D color imagery to 3D laser radar imagery collected from distinct airborne platforms at different times. Our approach is to derive pseudo-intensity images from ladar imagery and to align these with color imagery using conventional 2D registration algorithms. To construct a pseudo-intensity image, the algorithm uses the color imagery's time of day and location to predict shadows in the 3D image, then determines ambient and sun lighting conditions by histogram matching the 3D-derived shadowed and non-shadowed regions to their 2D counterparts. A projection matrix is computed to bring the pseudo- image into 2D image coordinates, resulting in an initial alignment of the imagery to within 200 meters. Finally, the 2D intensity image and 3D generated pseudo-intensity image are registered using a modified normalized correlation algorithm to solve for rotation, translation, scale and lens distortion, resulting in a fused data set that is aligned to within 1 meter. Applications of the presented work include the areas of augmented reality and scene interpretation for persistent surveillance in heavily cluttered and occluded environments.
{"title":"Automatic Alignment of Color Imagery onto 3D Laser Radar Data","authors":"A. Vasile, Frederick R. Waugh, Daniel Greisokh, R. Heinrichs","doi":"10.1109/AIPR.2006.16","DOIUrl":"https://doi.org/10.1109/AIPR.2006.16","url":null,"abstract":"We present an algorithm for the automatic fusion of city-sized, 2D color imagery to 3D laser radar imagery collected from distinct airborne platforms at different times. Our approach is to derive pseudo-intensity images from ladar imagery and to align these with color imagery using conventional 2D registration algorithms. To construct a pseudo-intensity image, the algorithm uses the color imagery's time of day and location to predict shadows in the 3D image, then determines ambient and sun lighting conditions by histogram matching the 3D-derived shadowed and non-shadowed regions to their 2D counterparts. A projection matrix is computed to bring the pseudo- image into 2D image coordinates, resulting in an initial alignment of the imagery to within 200 meters. Finally, the 2D intensity image and 3D generated pseudo-intensity image are registered using a modified normalized correlation algorithm to solve for rotation, translation, scale and lens distortion, resulting in a fused data set that is aligned to within 1 meter. Applications of the presented work include the areas of augmented reality and scene interpretation for persistent surveillance in heavily cluttered and occluded environments.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116217287","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}
S. Lee, K. Price, R. Nevatia, T. Heinze, J. Irvine
The production of geospatial information from overhead imagery is generally a labor-intensive process. Analysts must accurately delineate and extract important features, such as buildings, roads, and landcover from the imagery. Automated feature extraction (AFE) tools offer the prospect of reducing analyst's workload. This paper presents a new tool, called iMVS, for extracting buildings and discusses user testing conducted by the National Geospatial-Intelligence Agency (NGA). Using a semi-automated approach, iMVS processes two or more images to form a set of hypothesized 3-D buildings. When the user clicks on one of the building vertices, the system determines which hypothesis is the best fit and extracts the building. A set of powerful editing tools support rapid clean-up of the extraction, including extraction of complex buildings. User testing of iMVS provides an assessment of the benefits and identifies areas for system improvement.
{"title":"Semi-automated 3-D Building Extraction from Stereo Imagery","authors":"S. Lee, K. Price, R. Nevatia, T. Heinze, J. Irvine","doi":"10.1109/AIPR.2006.36","DOIUrl":"https://doi.org/10.1109/AIPR.2006.36","url":null,"abstract":"The production of geospatial information from overhead imagery is generally a labor-intensive process. Analysts must accurately delineate and extract important features, such as buildings, roads, and landcover from the imagery. Automated feature extraction (AFE) tools offer the prospect of reducing analyst's workload. This paper presents a new tool, called iMVS, for extracting buildings and discusses user testing conducted by the National Geospatial-Intelligence Agency (NGA). Using a semi-automated approach, iMVS processes two or more images to form a set of hypothesized 3-D buildings. When the user clicks on one of the building vertices, the system determines which hypothesis is the best fit and extracts the building. A set of powerful editing tools support rapid clean-up of the extraction, including extraction of complex buildings. User testing of iMVS provides an assessment of the benefits and identifies areas for system improvement.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128399237","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}
Analytical shaded relief is commonly used for visualization of digital elevation models (DEMs). Sometimes, the quality of unaltered analytical shaded relief can be lacking for identification of streams and water divides. Hydroshading is a technique that provides enhanced capabilities of visualization of hydrologically-meaningful topographical features. In this research, hydroshading algorithms are applied to NASA's Shuttle Radar Topography Mission (SRTM) DEM datasets. The visualization technique is applied to coastal and inland watersheds in Mississippi (Saint Louis Bay and Luxapallila, respectively). The testing of hydroshading in these two areas shows that the technique is more effective in areas with moderate topographical relief than in low relief terrain. Combining hydroshading with standard three- dimensional visualization identification of water Hydroshaded DEMs were used to manually delineate Luxapallila and Saint Louis Bay's Wolf River catchments. Delineation results are comparable to output of standard automated delineation produced by GIS software (BASINS).
{"title":"Advanced Techniques for Watershed Visualization","authors":"V. J. Alarcon, C. O'Hara","doi":"10.1109/AIPR.2006.10","DOIUrl":"https://doi.org/10.1109/AIPR.2006.10","url":null,"abstract":"Analytical shaded relief is commonly used for visualization of digital elevation models (DEMs). Sometimes, the quality of unaltered analytical shaded relief can be lacking for identification of streams and water divides. Hydroshading is a technique that provides enhanced capabilities of visualization of hydrologically-meaningful topographical features. In this research, hydroshading algorithms are applied to NASA's Shuttle Radar Topography Mission (SRTM) DEM datasets. The visualization technique is applied to coastal and inland watersheds in Mississippi (Saint Louis Bay and Luxapallila, respectively). The testing of hydroshading in these two areas shows that the technique is more effective in areas with moderate topographical relief than in low relief terrain. Combining hydroshading with standard three- dimensional visualization identification of water Hydroshaded DEMs were used to manually delineate Luxapallila and Saint Louis Bay's Wolf River catchments. Delineation results are comparable to output of standard automated delineation produced by GIS software (BASINS).","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"101 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124713852","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 primary mission of air assets is to detect and destroy enemy ground targets. In order to accomplish this mission, it is essential to detect, track, and classify contacts to determine which are valid targets. Traditional combat identification has been performed using all-weather sensors and processing algorithms designed specifically for such sensor data. Electro- optical (EO) sensors produce a very different type of data that does not lend itself to traditional combat identification algorithms. This paper will detail how we analyzed the visual and physical characteristics of a large number of potential targets. The results of this analysis were used to drive the requirements of a demonstration system. We will detail the test data we collected from the military and CAD models for likely targets, as well as overall requirements for system performance.
{"title":"An Automatic Target Classifier using Model Based Image Processing","authors":"D. Haanpaa, G. Beach, C. Cohen","doi":"10.1109/AIPR.2006.12","DOIUrl":"https://doi.org/10.1109/AIPR.2006.12","url":null,"abstract":"A primary mission of air assets is to detect and destroy enemy ground targets. In order to accomplish this mission, it is essential to detect, track, and classify contacts to determine which are valid targets. Traditional combat identification has been performed using all-weather sensors and processing algorithms designed specifically for such sensor data. Electro- optical (EO) sensors produce a very different type of data that does not lend itself to traditional combat identification algorithms. This paper will detail how we analyzed the visual and physical characteristics of a large number of potential targets. The results of this analysis were used to drive the requirements of a demonstration system. We will detail the test data we collected from the military and CAD models for likely targets, as well as overall requirements for system performance.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128176364","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}
This paper, describes a series of steps on how to perform model validation and how to interpret the results. The first step is to establish how much error is allowable. If this is not done, then the results of the validation will be very hard to interpret. Once this allowable error is established, the paper describes a procedure for how to perform the validation and analyze the results in a productive way.
{"title":"Validation Techniques for Image-Based Simulations","authors":"D. Fraedrich","doi":"10.1109/AIPR.2006.39","DOIUrl":"https://doi.org/10.1109/AIPR.2006.39","url":null,"abstract":"This paper, describes a series of steps on how to perform model validation and how to interpret the results. The first step is to establish how much error is allowable. If this is not done, then the results of the validation will be very hard to interpret. Once this allowable error is established, the paper describes a procedure for how to perform the validation and analyze the results in a productive way.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121896924","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}
Model misspecification has been a major concern in practical model based image analysis. The underlying assumptions of generative processes usually can not exactly describe real-world data samples, which renders the maximum likelihood estimation (MLE) and the Bayesian decision methods unreliable. In this work we study a robust adjusted likelihood (RAL) function that can improve image classification performance under misspecified models. The RAL is calculated by raising the conventional likelihood function to a positive power and multiplying it with a scaling factor. Similar to model parameter estimation, these two new RAL parameters, i.e. the power and the scaling factor, are estimated from the training data using minimum error rate method. In two-category classification case, this RAL is equivalent to a linear discriminant function in log-likelihood space. To demonstrate the effectiveness of this RAL, we first simulate a model misspecification scenario, in which two Rayleigh sources are misspecified as Gaussian distributions. The Gaussian parameters and the RAL parameters are estimated accordingly from the training data, and the two RAL parameters are studied separately. The simulation results show that the Bayes decisions based on maximum-RAL yield higher classification accuracy than the decisions based on conventional maximum-likelihood. We further apply the RAL in automatic target recognition (ATR) of SAR images. Two target classes, i.e. t72 and bmp2, from MSTAR SAR target dataset are used in this study. The target signatures are modeled using Gaussian mixture models (GMMs) with five mixtures for each class. Image classification results again demonstrate a clear advantage of the proposed approach.
{"title":"Robust Adjusted Likelihood Function for Image Analysis","authors":"Rong Duan, Wei Jiang, H. Man","doi":"10.1109/AIPR.2006.34","DOIUrl":"https://doi.org/10.1109/AIPR.2006.34","url":null,"abstract":"Model misspecification has been a major concern in practical model based image analysis. The underlying assumptions of generative processes usually can not exactly describe real-world data samples, which renders the maximum likelihood estimation (MLE) and the Bayesian decision methods unreliable. In this work we study a robust adjusted likelihood (RAL) function that can improve image classification performance under misspecified models. The RAL is calculated by raising the conventional likelihood function to a positive power and multiplying it with a scaling factor. Similar to model parameter estimation, these two new RAL parameters, i.e. the power and the scaling factor, are estimated from the training data using minimum error rate method. In two-category classification case, this RAL is equivalent to a linear discriminant function in log-likelihood space. To demonstrate the effectiveness of this RAL, we first simulate a model misspecification scenario, in which two Rayleigh sources are misspecified as Gaussian distributions. The Gaussian parameters and the RAL parameters are estimated accordingly from the training data, and the two RAL parameters are studied separately. The simulation results show that the Bayes decisions based on maximum-RAL yield higher classification accuracy than the decisions based on conventional maximum-likelihood. We further apply the RAL in automatic target recognition (ATR) of SAR images. Two target classes, i.e. t72 and bmp2, from MSTAR SAR target dataset are used in this study. The target signatures are modeled using Gaussian mixture models (GMMs) with five mixtures for each class. Image classification results again demonstrate a clear advantage of the proposed approach.","PeriodicalId":375571,"journal":{"name":"35th IEEE Applied Imagery and Pattern Recognition Workshop (AIPR'06)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127192698","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：