This paper presents a system for fully automatic recognition and reconstruction of 3D objects in image databases. We pose the object recognition problem as one of finding consistent matches between all images, subject to the constraint that the images were taken from a perspective camera. We assume that the objects or scenes are rigid. For each image, we associate a camera matrix, which is parameterised by rotation, translation and focal length. We use invariant local features to find matches between all images, and the RANSAC algorithm to find those that are consistent with the fundamental matrix. Objects are recognised as subsets of matching images. We then solve for the structure and motion of each object, using a sparse bundle adjustment algorithm. Our results demonstrate that it is possible to recognise and reconstruct 3D objects from an unordered image database with no user input at all.
{"title":"Unsupervised 3D object recognition and reconstruction in unordered datasets","authors":"Matthew A. Brown, D. Lowe","doi":"10.1109/3DIM.2005.81","DOIUrl":"https://doi.org/10.1109/3DIM.2005.81","url":null,"abstract":"This paper presents a system for fully automatic recognition and reconstruction of 3D objects in image databases. We pose the object recognition problem as one of finding consistent matches between all images, subject to the constraint that the images were taken from a perspective camera. We assume that the objects or scenes are rigid. For each image, we associate a camera matrix, which is parameterised by rotation, translation and focal length. We use invariant local features to find matches between all images, and the RANSAC algorithm to find those that are consistent with the fundamental matrix. Objects are recognised as subsets of matching images. We then solve for the structure and motion of each object, using a sparse bundle adjustment algorithm. Our results demonstrate that it is possible to recognise and reconstruct 3D objects from an unordered image database with no user input at all.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116886760","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}
Peiran Liu, Xiaojun Shen, N. Georganas, Gerhard Roth
The computational cost of a collision detection (CD) algorithm on polygonal surfaces depends highly on the complexity of the models. A novel "locally refined" approach is introduced in this paper for fast CD in haptic rendering applications, e.g. haptic surgery and haptic sculpture simulations. Exact interference detections are performed on proposed locally refined meshes, which are in multiresolution representation. The meshes are generated using mesh simplification and space partition. A new BVH algorithm called "active bounding tree", or AB-tree, handling collision queries is introduced. At runtime the meshes are dynamically refined to higher resolution in areas that are most likely to collide with other objects. The algorithms are successfully demonstrated in an interactive haptic environment. Compared to existing CD algorithms on single resolution models, noticeable performance improvement has been observed in terms of the precision of collision queries, frame rate, and memory usage.
{"title":"Multi-resolution modeling and locally refined collision detection for haptic interaction","authors":"Peiran Liu, Xiaojun Shen, N. Georganas, Gerhard Roth","doi":"10.1109/3DIM.2005.58","DOIUrl":"https://doi.org/10.1109/3DIM.2005.58","url":null,"abstract":"The computational cost of a collision detection (CD) algorithm on polygonal surfaces depends highly on the complexity of the models. A novel \"locally refined\" approach is introduced in this paper for fast CD in haptic rendering applications, e.g. haptic surgery and haptic sculpture simulations. Exact interference detections are performed on proposed locally refined meshes, which are in multiresolution representation. The meshes are generated using mesh simplification and space partition. A new BVH algorithm called \"active bounding tree\", or AB-tree, handling collision queries is introduced. At runtime the meshes are dynamically refined to higher resolution in areas that are most likely to collide with other objects. The algorithms are successfully demonstrated in an interactive haptic environment. Compared to existing CD algorithms on single resolution models, noticeable performance improvement has been observed in terms of the precision of collision queries, frame rate, and memory usage.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131086200","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}
Multiperspective imaging has been used to recover the structure of a scene. Although several algorithms for structure recovery have been developed as typified by stereo panoramas, there exists no common framework which subsumes various camera motions to capture stereo images. This paper presents a framework for stereo by multiperspective imaging, which is general in that it can handle 6 degree-of-freedom (DOF) camera motion. We derive geometric constraints, equation for structure recovery and that for an epipolar curve by modeling the acquisition of stereo images using push-broom cameras (line sensors). We consider a class of camera motion called a vertical view plane class and demonstrate that several previous results are really special cases of our results. Numerical examples are given to show the correctness of the derived equations.
{"title":"Stereo by multiperspective imaging under 6 DOF camera motion","authors":"N. Ichimura","doi":"10.1109/3DIM.2005.76","DOIUrl":"https://doi.org/10.1109/3DIM.2005.76","url":null,"abstract":"Multiperspective imaging has been used to recover the structure of a scene. Although several algorithms for structure recovery have been developed as typified by stereo panoramas, there exists no common framework which subsumes various camera motions to capture stereo images. This paper presents a framework for stereo by multiperspective imaging, which is general in that it can handle 6 degree-of-freedom (DOF) camera motion. We derive geometric constraints, equation for structure recovery and that for an epipolar curve by modeling the acquisition of stereo images using push-broom cameras (line sensors). We consider a class of camera motion called a vertical view plane class and demonstrate that several previous results are really special cases of our results. Numerical examples are given to show the correctness of the derived equations.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128114946","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 novel method for digitizing the 3D shape of an object in real-time, which can be used for capturing live sequence of the 3D shape of moving or deformable objects such as faces. Two DMD (digital micro mirror) devices are used as high speed switches for modulating and demodulating light rays. One DMD is used to generate rays of light pulses, which are projected onto the object to be measured. Another DMD is used to demodulate the light reflected from the object illuminated by the light pulses into intensity image that describes the disparity. A prototype range finder implementing the proposed method has been built. The experimental results showed that the proposed method works and video sequences of disparity images can be captured in real time.
{"title":"A light modulation/demodulation method for real-time 3D imaging","authors":"Qian Chen, T. Wada","doi":"10.1109/3DIM.2005.9","DOIUrl":"https://doi.org/10.1109/3DIM.2005.9","url":null,"abstract":"This paper describes a novel method for digitizing the 3D shape of an object in real-time, which can be used for capturing live sequence of the 3D shape of moving or deformable objects such as faces. Two DMD (digital micro mirror) devices are used as high speed switches for modulating and demodulating light rays. One DMD is used to generate rays of light pulses, which are projected onto the object to be measured. Another DMD is used to demodulate the light reflected from the object illuminated by the light pulses into intensity image that describes the disparity. A prototype range finder implementing the proposed method has been built. The experimental results showed that the proposed method works and video sequences of disparity images can be captured in real time.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122372877","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}
Nowadays, with the exponential growing of 3D object representations in private databases or on the web, it is all the more required to match these objects from some views. To improve the results of their matching, we work on the characteristic views of an object. The aim of this study is to find how many characteristic views are required and what relative positions are optimal. This is the reason why the visual hulls are used. From some 2D masks, the nearest possible 3D mesh from the original object is computed. OpenGL views are used to build the visual hulls of 3D models from a given collection and then the distance between the visual hulls and the models are measured thanks to the Hausdorff distance. Then the best view parameters are deduced to reduce the distance. These shots show that three orthogonal views give results very close to the ones given by twelve views on a isocahedron. Some other results on the view resolution and the field of view are discussed.
{"title":"Determining characteristic views of a 3D object by visual hulls and Hausdorff distance","authors":"A. Theetten, Jean-Philippe Vandeborre, M. Daoudi","doi":"10.1109/3DIM.2005.31","DOIUrl":"https://doi.org/10.1109/3DIM.2005.31","url":null,"abstract":"Nowadays, with the exponential growing of 3D object representations in private databases or on the web, it is all the more required to match these objects from some views. To improve the results of their matching, we work on the characteristic views of an object. The aim of this study is to find how many characteristic views are required and what relative positions are optimal. This is the reason why the visual hulls are used. From some 2D masks, the nearest possible 3D mesh from the original object is computed. OpenGL views are used to build the visual hulls of 3D models from a given collection and then the distance between the visual hulls and the models are measured thanks to the Hausdorff distance. Then the best view parameters are deduced to reduce the distance. These shots show that three orthogonal views give results very close to the ones given by twelve views on a isocahedron. Some other results on the view resolution and the field of view are discussed.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115921004","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}
Today modeling from reality receives more and more attention. In this paper, we present a novel image-based 3D registration method. Compared with a previous one it does not require accurate starting position, albedo and geometric invariants, but has to address the more apparent mismatch problems. First, distinctive corner points, which act as salient features for subsequent image matching, are detected via the minimal eigenvalue of the auto-correlation matrix. Then, a verification scheme discards the potential mismatches by thresholds of the correlation coefficients and point-to-point distances. Experimental results demonstrate the superiority of our proposed verification scheme to the previous one using only correlation coefficients under the relaxed conditions.
{"title":"A robust image-based method for 3D registration","authors":"Ting Li","doi":"10.1109/3DIM.2005.12","DOIUrl":"https://doi.org/10.1109/3DIM.2005.12","url":null,"abstract":"Today modeling from reality receives more and more attention. In this paper, we present a novel image-based 3D registration method. Compared with a previous one it does not require accurate starting position, albedo and geometric invariants, but has to address the more apparent mismatch problems. First, distinctive corner points, which act as salient features for subsequent image matching, are detected via the minimal eigenvalue of the auto-correlation matrix. Then, a verification scheme discards the potential mismatches by thresholds of the correlation coefficients and point-to-point distances. Experimental results demonstrate the superiority of our proposed verification scheme to the previous one using only correlation coefficients under the relaxed conditions.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117313517","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}
Raghavendra Donamukkala, Daniel F. Huber, A. Kapuria, M. Hebert
Most research on 3-D object classification and recognition focuses on recognition of objects in 3-D scenes from a small database of known 3-D models. Such an approach does not scale well to large databases of objects and does not generalize well to unknown (but similar) object classification. This paper presents two ideas to address these problems (i) class selection, i.e., grouping similar objects into classes (ii) class prototyping, i.e., exploiting common structure within classes to represent the classes. At run time matching a query against the prototypes is sufficient for classification. This approach will not only reduce the retrieval time but also will help increase the generalizing power of the classification algorithm. Objects are segmented into classes automatically using an agglomerative clustering algorithm. Prototypes from these classes are extracted using one of three class prototyping algorithms. Experimental results demonstrate the effectiveness of the two steps in speeding up the classification process without sacrificing accuracy.
{"title":"Automatic Class Selection and Prototyping for 3-D Object Classification","authors":"Raghavendra Donamukkala, Daniel F. Huber, A. Kapuria, M. Hebert","doi":"10.1109/3DIM.2005.22","DOIUrl":"https://doi.org/10.1109/3DIM.2005.22","url":null,"abstract":"Most research on 3-D object classification and recognition focuses on recognition of objects in 3-D scenes from a small database of known 3-D models. Such an approach does not scale well to large databases of objects and does not generalize well to unknown (but similar) object classification. This paper presents two ideas to address these problems (i) class selection, i.e., grouping similar objects into classes (ii) class prototyping, i.e., exploiting common structure within classes to represent the classes. At run time matching a query against the prototypes is sufficient for classification. This approach will not only reduce the retrieval time but also will help increase the generalizing power of the classification algorithm. Objects are segmented into classes automatically using an agglomerative clustering algorithm. Prototypes from these classes are extracted using one of three class prototyping algorithms. Experimental results demonstrate the effectiveness of the two steps in speeding up the classification process without sacrificing accuracy.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126367782","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 presents techniques for the merging of 3D data coming from different sensors, such as ground and aerial laser range scans. The 3D models created are reconstructed to give a photo-realistic scene enabling interactive virtual walkthroughs, measurements and scene change analysis. The reconstructed model is based on a weighted integration of all available data based on sensor-specific parameters such as noise level, accuracy, inclination and reflectivity of the target, spatial distribution of points. The geometry is robustly reconstructed with a volumetric approach. Once registered and weighed, all data is re-sampled in a multi-resolution distance field using out-of-core techniques. The final mesh is extracted by contouring the iso-surface with a feature preserving dual contouring algorithm. The paper shows results of the above technique applied to Verona (Italy) city centre.
{"title":"Multisensor fusion for volumetric reconstruction of large outdoor areas","authors":"M. Fiocco, G. Boström, J. Gonçalves, V. Sequeira","doi":"10.1109/3DIM.2005.60","DOIUrl":"https://doi.org/10.1109/3DIM.2005.60","url":null,"abstract":"This paper presents techniques for the merging of 3D data coming from different sensors, such as ground and aerial laser range scans. The 3D models created are reconstructed to give a photo-realistic scene enabling interactive virtual walkthroughs, measurements and scene change analysis. The reconstructed model is based on a weighted integration of all available data based on sensor-specific parameters such as noise level, accuracy, inclination and reflectivity of the target, spatial distribution of points. The geometry is robustly reconstructed with a volumetric approach. Once registered and weighed, all data is re-sampled in a multi-resolution distance field using out-of-core techniques. The final mesh is extracted by contouring the iso-surface with a feature preserving dual contouring algorithm. The paper shows results of the above technique applied to Verona (Italy) city centre.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124331785","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 presents a fast pose estimation algorithm of a 3D free form object in 2D images using 2D distance maps. One of the popular techniques of the pose estimation of 3D object in 2D image is the point-based method such as the ICP algorithm. However, the calculation cost for determining point correspondences is expensive. To overcome this problem, the proposed method utilizes a distance map on the 2D image plane, which is constructed quite rapidly by the fast marching method. For pose estimation of the object, contour lines of the 2D image and the projection of the 3D object are aligned using the distance map iteratively by the robust m-estimator. Some experimental results with simulated models and actual images of the endoscopic operation are successfully carried out.
{"title":"Fast alignment of 3D geometrical models and 2D color images using 2D distance maps","authors":"Y. Iwashita, R. Kurazume, T. Hasegawa, K. Hara","doi":"10.1109/3DIM.2005.39","DOIUrl":"https://doi.org/10.1109/3DIM.2005.39","url":null,"abstract":"This paper presents a fast pose estimation algorithm of a 3D free form object in 2D images using 2D distance maps. One of the popular techniques of the pose estimation of 3D object in 2D image is the point-based method such as the ICP algorithm. However, the calculation cost for determining point correspondences is expensive. To overcome this problem, the proposed method utilizes a distance map on the 2D image plane, which is constructed quite rapidly by the fast marching method. For pose estimation of the object, contour lines of the 2D image and the projection of the 3D object are aligned using the distance map iteratively by the robust m-estimator. Some experimental results with simulated models and actual images of the endoscopic operation are successfully carried out.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"109 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124764072","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper we present an adaptive dandelion model for reconstructing spherical terrain-like visual hull (VH) surfaces. The dandelion model represents a solid by a pencil of organized line segments emitted from a common point. The directions and the topology of the line segments are derived from the triangle facets of a geodesic sphere, which are recursively subdivided until the desired precision is achieved. The initial lines are cut by silhouettes in 2D and then lifted back to 3D to determine the ending points of the line segments defining sampling points on the spherical terrain-like VH surface. A mesh model can be easily constructed from the dandelion model. Our algorithm has the advantages of controllable precision, adaptive resolution, simplicity and speediness. We validate our algorithm by theories and experiments.
{"title":"An adaptive dandelion model for reconstructing spherical terrain-like visual hull surfaces","authors":"Xin Liu, H. Yao, Wen Gao","doi":"10.1109/3DIM.2005.17","DOIUrl":"https://doi.org/10.1109/3DIM.2005.17","url":null,"abstract":"In this paper we present an adaptive dandelion model for reconstructing spherical terrain-like visual hull (VH) surfaces. The dandelion model represents a solid by a pencil of organized line segments emitted from a common point. The directions and the topology of the line segments are derived from the triangle facets of a geodesic sphere, which are recursively subdivided until the desired precision is achieved. The initial lines are cut by silhouettes in 2D and then lifted back to 3D to determine the ending points of the line segments defining sampling points on the spherical terrain-like VH surface. A mesh model can be easily constructed from the dandelion model. Our algorithm has the advantages of controllable precision, adaptive resolution, simplicity and speediness. We validate our algorithm by theories and experiments.","PeriodicalId":170883,"journal":{"name":"Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2005-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129493887","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：