David Page - Academia.edu (original) (raw)

Papers by David Page

2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops

X-ray luggage inspection systems play an important role in ensuring air travelers' security. Howe... more X-ray luggage inspection systems play an important role in ensuring air travelers' security. However, the false alarm rate of commercial systems can be as high as 20% due to less than perfect image processing algorithms. In an effort to reduce the false alarm rate, this paper proposes a combinational scheme to fuse, de-noise and enhance dual-energy X-ray images for better object classification and threat detection. The fusion step is based on the wavelet transform. Fused images generally reveal more detail information; however, background noise often gets amplified during the fusion process. This paper applies a backgroundsubtraction-based noise reduction technique which is very efficient in removing background noise from fused X-ray images. The de-noised image is then processed using a new enhancement technique to reconstruct the final image. The final image not only contains complementary information from both source images, but is also background-noise-free and contrastenhanced, therefore easier to segment automatically or be interpreted by screeners, thus reducing the false alarm rate in X-ray luggage inspection.

Computational Imaging and Vision

This research is motivated towards the deployment of intelligent robots for under vehicle inspect... more This research is motivated towards the deployment of intelligent robots for under vehicle inspection at checkpoints , gate-entry terminals and parking lots. Using multi-modality measurements of temperature, range, color, radioactivity and with future potential for chemical and biological sensors, our approach is based on a modular robotic "sensor brick" architecture that integrates multisensor data into scene intelligence in 3D virtual reality environments. The remote 3D scene visualization capability reduces the risk on close-range inspection personnel, transforming the inspection task into an unmanned robotic mission. Our goal in this chapter is to focus on the 3D range "sensor brick" as a vital component in this multi-sensor robotics framework and demonstrate the potential of automatic threat detection using the geometric information from the 3D sensors. With the 3D data alone, we propose two different approaches for the detection of anomalous objects as potential threats. The first approach is to perform scene verification using a 3D registration algorithm for quickly and efficiently finding potential changes to the undercarriage by comparing previously archived scans of the same vehicle. The second 3D shape analysis approach assumes availability of the CAD models of the undercarriage that can be matched with the scanned real data using a novel perceptual curvature variation measure (CVM). The definition of the CVM, that can be understood as the entropy of surface curvature, describes the under vehicle scene as a graph network of smooth surface patches that readily lends to matching with the graph description of the apriori CAD data. By presenting results of real-time acquisition, visualization, scene verification and description, we emphasize the scope of 3D imaging over several drawbacks with present day inspection systems using mirrors and 2D cameras.

Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001

In this paper, we describe a robust method for the estimation of curvature on a triangle mesh, wh... more In this paper, we describe a robust method for the estimation of curvature on a triangle mesh, where this mesh is a discrete approximation of a piecewise smooth surface. The proposed method avoids the computationally expensive process of surface fitting and instead employs normal voting to achieve robust results. This method detects crease discontinuities on the surface to improve estimates near those creases. Using a voting scheme, the algorithm estimates both principal curvatures and principal directions for smooth patches. The entire process requires one user parameter-the voting neighborhood size, which is a function of sampling density, feature size, and measurement noise. We present results for both synthetic and real data and compare these results to an existing algorithm developed by Taubin.

SPIE Proceedings, 2000

Towards photo-realistic 3D scene reconstruction from range and color images, we present a statist... more Towards photo-realistic 3D scene reconstruction from range and color images, we present a statistical technique for multimodal image registration. Statistical tools are employed to measure the dependence of two images, considered as random distributions of pixels, and to find the pose of one imaging system relative to the other. The similarity metrics used in our automatic registration algorithm are based on the chi-squared measure of dependence, which is presented as an alternative to the standard mutual information criterion. These two criteria belong to the class of information-theoretic similarity measures that quantify the dependence in terms of information provided by one image about the other. This approach requires the use of a robust optimization scheme for the maximization of the similarity measure. To achieve accurate results, we investigated the use of heuristics such as genetic algorithms. The retrieved pose parameters are used to generate a texture map from the color image, and the occluded areas in this image are determined and labeled. Finally the 3D scene is rendered as a triangular mesh with texture.

28th AIPR Workshop: 3D Visualization for Data Exploration and Decision Making, 2000

In this paper, we present a method for automatically registering a 3D range image and a 2D color ... more In this paper, we present a method for automatically registering a 3D range image and a 2D color image using the χ 2similarity metric. The goal of this registration is to allow the reconstruction of a scene using multi-sensor information. Traditional registration algorithms use invariant image features to drive the registration process. This approach limits the applicability to multi-modal data since features of interest may not appear in each modality. However, the χ 2-similarity metric is an intensity-based approach that has interesting multi-modal characteristics. We explore this metric as a mechanism to govern the registration search. Using range data from a Perceptron laser camera and color data from a Kodak digital camera, we present results using this automatic registration with the χ 2-similarity metric.

Unmanned Ground Vehicle Technology VII, 2005

Our research efforts focus on the deployment of 3D sensing capabilities to a multi-modal under ve... more Our research efforts focus on the deployment of 3D sensing capabilities to a multi-modal under vehicle inspection robot. In this paper, we outline the various design challenges towards the automation of the 3D scene modeling task. We employ laser-based range imaging techniques to extract the geometry of a vehicle's undercarriage and present our results after range integration. We perform shape analysis on the digitized triangle mesh models by segmenting them into smooth surface patches based on the curvedness of the surface. Using a region-growing procedure, we then obtain the patch adjacency. On each of these patches, we apply our definition of the curvature variation measure (CVM) as a descriptor of surface shape complexity. We base the information-theoretic CVM on shape curvature, and extract shape information as the entropic measure of curvature to represent a component as a graph network of patches. The CVM at the nodes of the graph describe the surface patch. We then demonstrate our algorithm with results on automotive components. With apriori manufacturer information about the CAD models in the undercarriage we approach the technical challenge of threat detection with our surface shape description algorithm on the laser scanned geometry.

SPIE Proceedings, 2004

The current threats to U.S. security both military and civilian have led to an increased interest... more The current threats to U.S. security both military and civilian have led to an increased interest in the development of technologies to safeguard national facilities such as military bases, federal buildings, nuclear power plants, and national laboratories. As a result, the Imaging, Robotics, and Intelligent Systems (IRIS) Laboratory at The University of Tennessee (UT) has established a research consortium, known as SAFER (Security Automation and Future Electromotive Robotics), to develop, test, and deploy sensing and imaging systems for unmanned ground vehicles (UGV). The targeted missions for these UGV systems include-but are not limited to-under vehicle threat assessment, stand-off checkpoint inspections, scout surveillance, intruder detection, obstaclebreach situations, and render-safe scenarios. This paper presents a general overview of the SAFER project. Beyond this general overview, we further focus on a specific problem where we collect 3D range scans of under vehicle carriages. These scans require appropriate segmentation and representation algorithms to facilitate the vehicle inspection process. We discuss the theory for these algorithms and present results from applying them to actual vehicle scans.

Lecture Notes in Computer Science

In comparison with 2D face images, 3D face models have the advantage of being illumination and po... more In comparison with 2D face images, 3D face models have the advantage of being illumination and pose invariant, which provides improved capability of handling changing environments in practical surveillance. Feature detection, as the initial process of reconstructing 3D face models from 2D uncalibrated image sequences, plays an important role and directly affects the accuracy and robustness of the resulting reconstruction. In this paper, we propose an automated scene-specific selection algorithm that adaptively chooses an optimal feature detector according to the input image sequence for the purpose of 3D face reconstruction. We compare the performance of various feature detectors in terms of accuracy and robustness of the sparse and dense reconstructions. Our experimental results demonstrate the effectiveness of the proposed selection method from the observation that the chosen feature detector produces 3D reconstructed face models with superior accuracy and robustness to image noise.

2008 19th International Conference on Pattern Recognition, 2008

Most existing sensor planning algorithms find it difficult to tackle the discrepancy between a PT... more Most existing sensor planning algorithms find it difficult to tackle the discrepancy between a PTZ camera's limited instant field of view (FOV) and panoramic achievable FOV. In this paper, we introduce the probability of camera overload to resolve this discrepancy and present a sensor planning algorithm for PTZ cameras under the same framework as static cameras. The resulting camera placement achieves the optimal balance between coverage and handoff success rate. Furthermore, our algorithm is able to incorporate the target's dynamics into sensor planning. As a result, the system's handoff success rate can be maintained in environments with various target densities. Experimental results and comparisons with a reference algorithm proposed by Erdem and Sclaroff verify the effectiveness of our algorithm via a significantly improved handoff success rate.

Sensor Review, 2008

PurposeThis paper seeks to present a novel X‐ray system and associated image segmentation algorit... more PurposeThis paper seeks to present a novel X‐ray system and associated image segmentation algorithm for imaging the below‐ground root structures of plants.Design/methodology/approachA matched filter design for segmenting the important root structures from the background clutter in the X‐ray images was presented.FindingsThe feasibility of root imaging and the applicability of matched filters to this problem domain have been demonstrated.Originality/valueThis research offers a novel approach over existing methods for in situ monitoring of root structures through the application of matched filters for image segmentation.

SPIE Proceedings, 2006

3D models of real world environments are becoming increasingly important for a variety of applica... more 3D models of real world environments are becoming increasingly important for a variety of applications: Vehicle simulators can be enhanced through accurate models of real world terrain and objects; Robotic security systems can benefit from as-built layout of the facilities they patrol; Vehicle dynamics modeling and terrain impact simulation can be improved through validation models generated by digitizing real tire/soil interactions. Recently, mobile scanning systems have been developed that allow 3D scanning systems to undergo the full range of motion necessary to acquire such real-world data in a fast, efficient manner. As with any digitization system, these mobile scanning systems have systemic errors that adversely affect the 3D models they are attempting to digitize. In addition to the errors given by the individual sensors, these systems also have uncertainties associated with the fusion of the data from several instruments. Thus, one of the primary foci for 3D model building is to perform the data fusion and post-processing of the models in such a manner as to reconstruct the 3D geometry of the scanned surfaces as accurately as possible, while alleviating the uncertainties posed by the acquisition system. We have developed a modular scanning system that can be configured for a variety of application resolutions, as well as the algorithms necessary to fuse and process the acquired data. This paper presents the acquisition system and the tools utilized for constructing 3D models under uncertain real-world conditions, as well as some experimental results on both synthetic and real 3D data.

2008 IEEE Conference on Computer Vision and Pattern Recognition, 2008

The estimation of the fundamental matrix is the key step in feature-based camera ego-motion estim... more The estimation of the fundamental matrix is the key step in feature-based camera ego-motion estimation for applications in scene modeling and vehicle navigation. In this paper, we present a new method of analyzing and further reducing the risk in the fundamental matrix due to the choice of a particular feature detector, the choice of the matching algorithm, the motion model, iterative hypothesis generation and verification paradigms. Our scheme makes use of model-selection theory to guide the switch to optimal methods for fundamental matrix estimation within the hypothesis-and-test architecture. We demonstrate our proposed method for vision-based robot localization in large-scale environments where the environment is constantly changing and navigation within the environment is unpredictable.

2007 IEEE International Conference on Image Processing, 2007

Interest point detectors are the starting point in image analysis for depth estimation using epip... more Interest point detectors are the starting point in image analysis for depth estimation using epipolar geometry and camera ego-motion estimation. With several detectors defined in the literature, some of them outperforming others in a specific application context, we introduce Multi-Feature Sample Consensus (MuFeSaC) as an adaptive and automatic procedure to choose a reliable feature detector among competing ones. Our approach is derived based on model selection criteria that we demonstrate for mobile robot self-localization in outdoor environments consisting of both man-made structures and natural vegetation.

2008 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, 2008

Humans perceive some objects more complex than others and learning or describing a particular obj... more Humans perceive some objects more complex than others and learning or describing a particular object is directly related to the judged complexity. Towards the goal of understanding why the geometry of some 3D objects appear more complex than others, we conducted a psychophysical study and identified contributing attributes. Our experiments conclude that surface variation, symmetry, part count, simpler part decomposability, intricate details and topology are six significant dimensions that influence 3D visual shape complexity. With that knowledge, we present a method of quantifying complexity and show that the informational aspect of Shannon's theory agrees with the human notion of shape complexity.

Third International Symposium on 3D Data Processing, Visualization, and Transmission (3DPVT'06), 2006

We propose a mathematical approach for quantifying shape complexity of 3D surfaces based on perce... more We propose a mathematical approach for quantifying shape complexity of 3D surfaces based on perceptual principles of visual saliency. Our curvature variation measure (CVM), as a 3D feature, combines surface curvature and information theory by leveraging bandwidth-optimized kernel density estimators. Using a part decomposition algorithm for digitized 3D objects, represented as triangle meshes, we apply our shape measure to transform the low level mesh representation into a perceptually informative form. Further, we analyze the effects of noise, sensitivity to digitization, occlusions, and descriptiveness to demonstrate our shape measure on laser-scanned real world 3D objects.

SAE Technical Paper Series, 2005

Three-dimensional models of real world terrain have application in a variety of tasks, but digiti... more Three-dimensional models of real world terrain have application in a variety of tasks, but digitizing a large environment poses constraints on the design of a 3D scanning system. We have developed a Mobile Scanning System that works within these constraints to quickly digitize large-scale real world environments. We utilize a mobile platform to move our sensors past the scene to be digitized-fusing the data from cm-level accuracy laser range scanners, positioning and orientation instruments, and high-resolution video cameras-to provide the mobility and speed required to quickly and accurately model the target scene.

Sensor Review, 2003

In this paper, we explore the technical challenges to automatically generate computer‐aided desig... more In this paper, we explore the technical challenges to automatically generate computer‐aided design models of existing vehicle parts using laser range imaging techniques. We propose a complete system that integrates data acquisition and model reconstruction. We discuss methods to resolve the occlusion problem and the associated registration problem. We also present our reconstruction algorithm. This range image‐based, computer‐aided reverse engineering system has a potential for faster model reconstruction over traditional reverse engineering technologies. Finally, we present results derived from the system.

Proceedings. International Conference on Image Processing

In this paper, we propose a novel algorithm to smooth and simplify simultaneously range images an... more In this paper, we propose a novel algorithm to smooth and simplify simultaneously range images and also triangle meshes derived from those images. These data sets often suffer from noise and over-sampling. To overcome these issues, smoothing from image processing and simplification from computer graphics attempt to minimize noise and reduce complexity, respectively. Typically, these algorithms are separate and distinct steps, but we combine them into one algorithm. We employ surface normal voting to generate robust orientation estimates and then extend the quadric error metric framework to smooth noise while simplifying the surface. We demonstrate the capabilities of this algorithm with both synthetic and real data. The proposed algorithm provides significant noise smoothing improvement when compared to the standard Garland and Heckbert quadric simplification algorithm.

2008 IEEE Conference on Computer Vision and Pattern Recognition, 2008

Most existing camera placement algorithms focus on coverage and/or visibility analysis, which ens... more Most existing camera placement algorithms focus on coverage and/or visibility analysis, which ensures that the object of interest is visible in the camera's field of view (FOV). However, visibility, a fundamental requirement of object tracking, is insufficient for persistent and automated tracking. In such applications, a continuous and consistently labeled trajectory of the same object should be maintained across different cameras' views. Therefore, a sufficient overlap between the cameras' FOVs should be secured so that camera handoff can be executed successfully and automatically before the object of interest becomes untraceable or unidentifiable. The proposed sensor planning method improves existing algorithms by adding handoff rate analysis, which preserves necessary overlapped FOVs for an optimal handoff success rate. In addition, special considerations such as resolution and frontal view requirements are addressed using two approaches: direct constraint and adaptive weight. The resulting camera placement is compared with a reference algorithm by Erdem and Sclaroff. Significantly improved handoff success rate and frontal view percentage are illustrated via experiments using typical office floor plans.

Optics and Lasers in Engineering, 2007

In this paper, we present our experience in building a mobile imaging system that incorporates mu... more In this paper, we present our experience in building a mobile imaging system that incorporates multi-modality sensors for road surface mapping and inspection applications. Our proposed system leverages 3D laser-range sensors, video cameras, global positioning systems (GPS) and inertial measurement units (IMU) towards the generation of photo-realistic, geometrically accurate, geo-referenced 3D models of road surfaces. Based on our summary of the state-of-the-art systems for a road distress survey, we identify several challenges in the real-time deployment, integration and visualization of the multi-sensor data. Then, we present our data acquisition and processing algorithms as a novel two-stage automation procedure that can meet the accuracy requirements with real-time performance. We provide algorithms for 3D surface reconstruction to process the raw data and deliver detail preserving 3D models that possess accurate depth information for characterization and visualization of cracks as a significant improvement over contemporary commercial video-based vision systems.

2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops

X-ray luggage inspection systems play an important role in ensuring air travelers' security. Howe... more X-ray luggage inspection systems play an important role in ensuring air travelers' security. However, the false alarm rate of commercial systems can be as high as 20% due to less than perfect image processing algorithms. In an effort to reduce the false alarm rate, this paper proposes a combinational scheme to fuse, de-noise and enhance dual-energy X-ray images for better object classification and threat detection. The fusion step is based on the wavelet transform. Fused images generally reveal more detail information; however, background noise often gets amplified during the fusion process. This paper applies a backgroundsubtraction-based noise reduction technique which is very efficient in removing background noise from fused X-ray images. The de-noised image is then processed using a new enhancement technique to reconstruct the final image. The final image not only contains complementary information from both source images, but is also background-noise-free and contrastenhanced, therefore easier to segment automatically or be interpreted by screeners, thus reducing the false alarm rate in X-ray luggage inspection.

Computational Imaging and Vision

This research is motivated towards the deployment of intelligent robots for under vehicle inspect... more This research is motivated towards the deployment of intelligent robots for under vehicle inspection at checkpoints , gate-entry terminals and parking lots. Using multi-modality measurements of temperature, range, color, radioactivity and with future potential for chemical and biological sensors, our approach is based on a modular robotic "sensor brick" architecture that integrates multisensor data into scene intelligence in 3D virtual reality environments. The remote 3D scene visualization capability reduces the risk on close-range inspection personnel, transforming the inspection task into an unmanned robotic mission. Our goal in this chapter is to focus on the 3D range "sensor brick" as a vital component in this multi-sensor robotics framework and demonstrate the potential of automatic threat detection using the geometric information from the 3D sensors. With the 3D data alone, we propose two different approaches for the detection of anomalous objects as potential threats. The first approach is to perform scene verification using a 3D registration algorithm for quickly and efficiently finding potential changes to the undercarriage by comparing previously archived scans of the same vehicle. The second 3D shape analysis approach assumes availability of the CAD models of the undercarriage that can be matched with the scanned real data using a novel perceptual curvature variation measure (CVM). The definition of the CVM, that can be understood as the entropy of surface curvature, describes the under vehicle scene as a graph network of smooth surface patches that readily lends to matching with the graph description of the apriori CAD data. By presenting results of real-time acquisition, visualization, scene verification and description, we emphasize the scope of 3D imaging over several drawbacks with present day inspection systems using mirrors and 2D cameras.

Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001

In this paper, we describe a robust method for the estimation of curvature on a triangle mesh, wh... more In this paper, we describe a robust method for the estimation of curvature on a triangle mesh, where this mesh is a discrete approximation of a piecewise smooth surface. The proposed method avoids the computationally expensive process of surface fitting and instead employs normal voting to achieve robust results. This method detects crease discontinuities on the surface to improve estimates near those creases. Using a voting scheme, the algorithm estimates both principal curvatures and principal directions for smooth patches. The entire process requires one user parameter-the voting neighborhood size, which is a function of sampling density, feature size, and measurement noise. We present results for both synthetic and real data and compare these results to an existing algorithm developed by Taubin.

SPIE Proceedings, 2000

Towards photo-realistic 3D scene reconstruction from range and color images, we present a statist... more Towards photo-realistic 3D scene reconstruction from range and color images, we present a statistical technique for multimodal image registration. Statistical tools are employed to measure the dependence of two images, considered as random distributions of pixels, and to find the pose of one imaging system relative to the other. The similarity metrics used in our automatic registration algorithm are based on the chi-squared measure of dependence, which is presented as an alternative to the standard mutual information criterion. These two criteria belong to the class of information-theoretic similarity measures that quantify the dependence in terms of information provided by one image about the other. This approach requires the use of a robust optimization scheme for the maximization of the similarity measure. To achieve accurate results, we investigated the use of heuristics such as genetic algorithms. The retrieved pose parameters are used to generate a texture map from the color image, and the occluded areas in this image are determined and labeled. Finally the 3D scene is rendered as a triangular mesh with texture.

28th AIPR Workshop: 3D Visualization for Data Exploration and Decision Making, 2000

In this paper, we present a method for automatically registering a 3D range image and a 2D color ... more In this paper, we present a method for automatically registering a 3D range image and a 2D color image using the χ 2similarity metric. The goal of this registration is to allow the reconstruction of a scene using multi-sensor information. Traditional registration algorithms use invariant image features to drive the registration process. This approach limits the applicability to multi-modal data since features of interest may not appear in each modality. However, the χ 2-similarity metric is an intensity-based approach that has interesting multi-modal characteristics. We explore this metric as a mechanism to govern the registration search. Using range data from a Perceptron laser camera and color data from a Kodak digital camera, we present results using this automatic registration with the χ 2-similarity metric.

Unmanned Ground Vehicle Technology VII, 2005

Our research efforts focus on the deployment of 3D sensing capabilities to a multi-modal under ve... more Our research efforts focus on the deployment of 3D sensing capabilities to a multi-modal under vehicle inspection robot. In this paper, we outline the various design challenges towards the automation of the 3D scene modeling task. We employ laser-based range imaging techniques to extract the geometry of a vehicle's undercarriage and present our results after range integration. We perform shape analysis on the digitized triangle mesh models by segmenting them into smooth surface patches based on the curvedness of the surface. Using a region-growing procedure, we then obtain the patch adjacency. On each of these patches, we apply our definition of the curvature variation measure (CVM) as a descriptor of surface shape complexity. We base the information-theoretic CVM on shape curvature, and extract shape information as the entropic measure of curvature to represent a component as a graph network of patches. The CVM at the nodes of the graph describe the surface patch. We then demonstrate our algorithm with results on automotive components. With apriori manufacturer information about the CAD models in the undercarriage we approach the technical challenge of threat detection with our surface shape description algorithm on the laser scanned geometry.

SPIE Proceedings, 2004

The current threats to U.S. security both military and civilian have led to an increased interest... more The current threats to U.S. security both military and civilian have led to an increased interest in the development of technologies to safeguard national facilities such as military bases, federal buildings, nuclear power plants, and national laboratories. As a result, the Imaging, Robotics, and Intelligent Systems (IRIS) Laboratory at The University of Tennessee (UT) has established a research consortium, known as SAFER (Security Automation and Future Electromotive Robotics), to develop, test, and deploy sensing and imaging systems for unmanned ground vehicles (UGV). The targeted missions for these UGV systems include-but are not limited to-under vehicle threat assessment, stand-off checkpoint inspections, scout surveillance, intruder detection, obstaclebreach situations, and render-safe scenarios. This paper presents a general overview of the SAFER project. Beyond this general overview, we further focus on a specific problem where we collect 3D range scans of under vehicle carriages. These scans require appropriate segmentation and representation algorithms to facilitate the vehicle inspection process. We discuss the theory for these algorithms and present results from applying them to actual vehicle scans.

Lecture Notes in Computer Science

In comparison with 2D face images, 3D face models have the advantage of being illumination and po... more In comparison with 2D face images, 3D face models have the advantage of being illumination and pose invariant, which provides improved capability of handling changing environments in practical surveillance. Feature detection, as the initial process of reconstructing 3D face models from 2D uncalibrated image sequences, plays an important role and directly affects the accuracy and robustness of the resulting reconstruction. In this paper, we propose an automated scene-specific selection algorithm that adaptively chooses an optimal feature detector according to the input image sequence for the purpose of 3D face reconstruction. We compare the performance of various feature detectors in terms of accuracy and robustness of the sparse and dense reconstructions. Our experimental results demonstrate the effectiveness of the proposed selection method from the observation that the chosen feature detector produces 3D reconstructed face models with superior accuracy and robustness to image noise.

2008 19th International Conference on Pattern Recognition, 2008

Most existing sensor planning algorithms find it difficult to tackle the discrepancy between a PT... more Most existing sensor planning algorithms find it difficult to tackle the discrepancy between a PTZ camera's limited instant field of view (FOV) and panoramic achievable FOV. In this paper, we introduce the probability of camera overload to resolve this discrepancy and present a sensor planning algorithm for PTZ cameras under the same framework as static cameras. The resulting camera placement achieves the optimal balance between coverage and handoff success rate. Furthermore, our algorithm is able to incorporate the target's dynamics into sensor planning. As a result, the system's handoff success rate can be maintained in environments with various target densities. Experimental results and comparisons with a reference algorithm proposed by Erdem and Sclaroff verify the effectiveness of our algorithm via a significantly improved handoff success rate.

Sensor Review, 2008

PurposeThis paper seeks to present a novel X‐ray system and associated image segmentation algorit... more PurposeThis paper seeks to present a novel X‐ray system and associated image segmentation algorithm for imaging the below‐ground root structures of plants.Design/methodology/approachA matched filter design for segmenting the important root structures from the background clutter in the X‐ray images was presented.FindingsThe feasibility of root imaging and the applicability of matched filters to this problem domain have been demonstrated.Originality/valueThis research offers a novel approach over existing methods for in situ monitoring of root structures through the application of matched filters for image segmentation.

SPIE Proceedings, 2006

3D models of real world environments are becoming increasingly important for a variety of applica... more 3D models of real world environments are becoming increasingly important for a variety of applications: Vehicle simulators can be enhanced through accurate models of real world terrain and objects; Robotic security systems can benefit from as-built layout of the facilities they patrol; Vehicle dynamics modeling and terrain impact simulation can be improved through validation models generated by digitizing real tire/soil interactions. Recently, mobile scanning systems have been developed that allow 3D scanning systems to undergo the full range of motion necessary to acquire such real-world data in a fast, efficient manner. As with any digitization system, these mobile scanning systems have systemic errors that adversely affect the 3D models they are attempting to digitize. In addition to the errors given by the individual sensors, these systems also have uncertainties associated with the fusion of the data from several instruments. Thus, one of the primary foci for 3D model building is to perform the data fusion and post-processing of the models in such a manner as to reconstruct the 3D geometry of the scanned surfaces as accurately as possible, while alleviating the uncertainties posed by the acquisition system. We have developed a modular scanning system that can be configured for a variety of application resolutions, as well as the algorithms necessary to fuse and process the acquired data. This paper presents the acquisition system and the tools utilized for constructing 3D models under uncertain real-world conditions, as well as some experimental results on both synthetic and real 3D data.

2008 IEEE Conference on Computer Vision and Pattern Recognition, 2008

The estimation of the fundamental matrix is the key step in feature-based camera ego-motion estim... more The estimation of the fundamental matrix is the key step in feature-based camera ego-motion estimation for applications in scene modeling and vehicle navigation. In this paper, we present a new method of analyzing and further reducing the risk in the fundamental matrix due to the choice of a particular feature detector, the choice of the matching algorithm, the motion model, iterative hypothesis generation and verification paradigms. Our scheme makes use of model-selection theory to guide the switch to optimal methods for fundamental matrix estimation within the hypothesis-and-test architecture. We demonstrate our proposed method for vision-based robot localization in large-scale environments where the environment is constantly changing and navigation within the environment is unpredictable.

2007 IEEE International Conference on Image Processing, 2007

Interest point detectors are the starting point in image analysis for depth estimation using epip... more Interest point detectors are the starting point in image analysis for depth estimation using epipolar geometry and camera ego-motion estimation. With several detectors defined in the literature, some of them outperforming others in a specific application context, we introduce Multi-Feature Sample Consensus (MuFeSaC) as an adaptive and automatic procedure to choose a reliable feature detector among competing ones. Our approach is derived based on model selection criteria that we demonstrate for mobile robot self-localization in outdoor environments consisting of both man-made structures and natural vegetation.

2008 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, 2008

Humans perceive some objects more complex than others and learning or describing a particular obj... more Humans perceive some objects more complex than others and learning or describing a particular object is directly related to the judged complexity. Towards the goal of understanding why the geometry of some 3D objects appear more complex than others, we conducted a psychophysical study and identified contributing attributes. Our experiments conclude that surface variation, symmetry, part count, simpler part decomposability, intricate details and topology are six significant dimensions that influence 3D visual shape complexity. With that knowledge, we present a method of quantifying complexity and show that the informational aspect of Shannon's theory agrees with the human notion of shape complexity.

Third International Symposium on 3D Data Processing, Visualization, and Transmission (3DPVT'06), 2006

We propose a mathematical approach for quantifying shape complexity of 3D surfaces based on perce... more We propose a mathematical approach for quantifying shape complexity of 3D surfaces based on perceptual principles of visual saliency. Our curvature variation measure (CVM), as a 3D feature, combines surface curvature and information theory by leveraging bandwidth-optimized kernel density estimators. Using a part decomposition algorithm for digitized 3D objects, represented as triangle meshes, we apply our shape measure to transform the low level mesh representation into a perceptually informative form. Further, we analyze the effects of noise, sensitivity to digitization, occlusions, and descriptiveness to demonstrate our shape measure on laser-scanned real world 3D objects.

SAE Technical Paper Series, 2005

Three-dimensional models of real world terrain have application in a variety of tasks, but digiti... more Three-dimensional models of real world terrain have application in a variety of tasks, but digitizing a large environment poses constraints on the design of a 3D scanning system. We have developed a Mobile Scanning System that works within these constraints to quickly digitize large-scale real world environments. We utilize a mobile platform to move our sensors past the scene to be digitized-fusing the data from cm-level accuracy laser range scanners, positioning and orientation instruments, and high-resolution video cameras-to provide the mobility and speed required to quickly and accurately model the target scene.

Sensor Review, 2003

In this paper, we explore the technical challenges to automatically generate computer‐aided desig... more In this paper, we explore the technical challenges to automatically generate computer‐aided design models of existing vehicle parts using laser range imaging techniques. We propose a complete system that integrates data acquisition and model reconstruction. We discuss methods to resolve the occlusion problem and the associated registration problem. We also present our reconstruction algorithm. This range image‐based, computer‐aided reverse engineering system has a potential for faster model reconstruction over traditional reverse engineering technologies. Finally, we present results derived from the system.

Proceedings. International Conference on Image Processing

In this paper, we propose a novel algorithm to smooth and simplify simultaneously range images an... more In this paper, we propose a novel algorithm to smooth and simplify simultaneously range images and also triangle meshes derived from those images. These data sets often suffer from noise and over-sampling. To overcome these issues, smoothing from image processing and simplification from computer graphics attempt to minimize noise and reduce complexity, respectively. Typically, these algorithms are separate and distinct steps, but we combine them into one algorithm. We employ surface normal voting to generate robust orientation estimates and then extend the quadric error metric framework to smooth noise while simplifying the surface. We demonstrate the capabilities of this algorithm with both synthetic and real data. The proposed algorithm provides significant noise smoothing improvement when compared to the standard Garland and Heckbert quadric simplification algorithm.

2008 IEEE Conference on Computer Vision and Pattern Recognition, 2008

Most existing camera placement algorithms focus on coverage and/or visibility analysis, which ens... more Most existing camera placement algorithms focus on coverage and/or visibility analysis, which ensures that the object of interest is visible in the camera's field of view (FOV). However, visibility, a fundamental requirement of object tracking, is insufficient for persistent and automated tracking. In such applications, a continuous and consistently labeled trajectory of the same object should be maintained across different cameras' views. Therefore, a sufficient overlap between the cameras' FOVs should be secured so that camera handoff can be executed successfully and automatically before the object of interest becomes untraceable or unidentifiable. The proposed sensor planning method improves existing algorithms by adding handoff rate analysis, which preserves necessary overlapped FOVs for an optimal handoff success rate. In addition, special considerations such as resolution and frontal view requirements are addressed using two approaches: direct constraint and adaptive weight. The resulting camera placement is compared with a reference algorithm by Erdem and Sclaroff. Significantly improved handoff success rate and frontal view percentage are illustrated via experiments using typical office floor plans.

Optics and Lasers in Engineering, 2007

In this paper, we present our experience in building a mobile imaging system that incorporates mu... more In this paper, we present our experience in building a mobile imaging system that incorporates multi-modality sensors for road surface mapping and inspection applications. Our proposed system leverages 3D laser-range sensors, video cameras, global positioning systems (GPS) and inertial measurement units (IMU) towards the generation of photo-realistic, geometrically accurate, geo-referenced 3D models of road surfaces. Based on our summary of the state-of-the-art systems for a road distress survey, we identify several challenges in the real-time deployment, integration and visualization of the multi-sensor data. Then, we present our data acquisition and processing algorithms as a novel two-stage automation procedure that can meet the accuracy requirements with real-time performance. We provide algorithms for 3D surface reconstruction to process the raw data and deliver detail preserving 3D models that possess accurate depth information for characterization and visualization of cracks as a significant improvement over contemporary commercial video-based vision systems.