Abstract: "Three-dimensional object recognition problem addresses a number of significant research issues in computer vision: representation of a 3D object, identification of the object, robust estimation of its pose, and registration of multiple views of the object for automatic model construction. This paper surveys the previous work in three important topics of computer vision: representation, matching, and pose estimation of a 3D object. It also presents an overview of the free-form surface matching problem, and describes Cosmos, our framework for representing and recognizing free-form objects. This vision system recognizes arbitrarily curved 3D rigid objects from a single view using dense surface data. We present both the theoretical aspects of this work and the experimental results of a prototype recognition system based on Cosmos."
One of the grand challenges of artificial intelligence is to enable computers to interpret 3D scenes and objects from imagery. This book organizes and introduces major concepts in 3D scene and object representation and inference from still images, with a focus on recent efforts to fuse models of geometry and perspective with statistical machine learning. The book is organized into three sections: (1) Interpretation of Physical Space; (2) Recognition of 3D Objects; and (3) Integrated 3D Scene Interpretation. The first discusses representations of spatial layout and techniques to interpret physical scenes from images. The second section introduces representations for 3D object categories that account for the intrinsically 3D nature of objects and provide robustness to change in viewpoints. The third section discusses strategies to unite inference of scene geometry and object pose and identity into a coherent scene interpretation. Each section broadly surveys important ideas from cognitive science and artificial intelligence research, organizes and discusses key concepts and techniques from recent work in computer vision, and describes a few sample approaches in detail. Newcomers to computer vision will benefit from introductions to basic concepts, such as single-view geometry and image classification, while experts and novices alike may find inspiration from the book's organization and discussion of the most recent ideas in 3D scene understanding and 3D object recognition. Specific topics include: mathematics of perspective geometry; visual elements of the physical scene, structural 3D scene representations; techniques and features for image and region categorization; historical perspective, computational models, and datasets and machine learning techniques for 3D object recognition; inferences of geometrical attributes of objects, such as size and pose; and probabilistic and feature-passing approaches for contextual reasoning about 3D objects and scenes. Table of Contents: Background on 3D Scene Models / Single-view Geometry / Modeling the Physical Scene / Categorizing Images and Regions / Examples of 3D Scene Interpretation / Background on 3D Recognition / Modeling 3D Objects / Recognizing and Understanding 3D Objects / Examples of 2D 1/2 Layout Models / Reasoning about Objects and Scenes / Cascades of Classifiers / Conclusion and Future Directions
3D object recognition is a very difficult and important problem in computer vision, arising in a wide range of applications. Typically in 3D object recognition, interest points are extracted from images and then matched. A shortcoming of this approach is that points only carry local visual information. Therefore, there could be many similar interest points, and these similar points could confuse the matching algorithm. In this thesis, we focus on region-based image matching for 3D object recognition. The main steps that we take are: 1) Extract regions from the image segmentations 2) Represent the objects with the aspect graph of regions 3) Match the given images based on region properties and special layout of regions. Our experiments on a challenging dataset show that we are able to match regions of the objects under 3D transforms. Our main contribution is that while traditional approaches concentrate on matching points, we focus on matching regions of objects. We are the first to show that matching segments of these challenging images is possible under 3D transforms.
This thesis is dedicated to the problem of object recognition in the three-dimensional space. Instead of using exclusively the information typically transported by a two-dimensional image, the concept of this work additionally incorporates the third dimension, namely the depth. The depth data itself is captured by sensors capable of measuring the distance from the device's position to those objects residing inside its field of view. The actual recognition process is implemented in analogy to the Path Similarity Skeleton Graph Matching (PSSGM). Basically, this method represents a 2D object by its skeleton and uses the idea of shortest paths to describe it. Finally, the similarity between two objects is calculated based on the Hungarian method. The contribution of the current work maps this approach into the three-dimensional space and applies it to 3D objects. While one of the experiments aims at the recognition of 3D chairs and tables, another one is devoted to the registration of fully segmented vascular structures. Excellent and promising recognition results are achieved in challenging evaluation setups showing that the 3D version of the PSSGM has the potential to solve complex recognition tasks.
An in-depth description of the state-of-the-art of 3D shape analysis techniques and their applications This book discusses the different topics that come under the title of "3D shape analysis". It covers the theoretical foundations and the major solutions that have been presented in the literature. It also establishes links between solutions proposed by different communities that studied 3D shape, such as mathematics and statistics, medical imaging, computer vision, and computer graphics. The first part of 3D Shape Analysis: Fundamentals, Theory, and Applications provides a review of the background concepts such as methods for the acquisition and representation of 3D geometries, and the fundamentals of geometry and topology. It specifically covers stereo matching, structured light, and intrinsic vs. extrinsic properties of shape. Parts 2 and 3 present a range of mathematical and algorithmic tools (which are used for e.g., global descriptors, keypoint detectors, local feature descriptors, and algorithms) that are commonly used for the detection, registration, recognition, classification, and retrieval of 3D objects. Both also place strong emphasis on recent techniques motivated by the spread of commodity devices for 3D acquisition. Part 4 demonstrates the use of these techniques in a selection of 3D shape analysis applications. It covers 3D face recognition, object recognition in 3D scenes, and 3D shape retrieval. It also discusses examples of semantic applications and cross domain 3D retrieval, i.e. how to retrieve 3D models using various types of modalities, e.g. sketches and/or images. The book concludes with a summary of the main ideas and discussions of the future trends. 3D Shape Analysis: Fundamentals, Theory, and Applications is an excellent reference for graduate students, researchers, and professionals in different fields of mathematics, computer science, and engineering. It is also ideal for courses in computer vision and computer graphics, as well as for those seeking 3D industrial/commercial solutions.
[Truncated abstract] The aim of visual recognition is to identify objects in a scene and estimate their pose. Object recognition from 2D images is sensitive to illumination, pose, clutter and occlusions. Object recognition from range data on the other hand does not suffer from these limitations. An important paradigm of recognition is model-based whereby 3D models of objects are constructed offline and saved in a database, using a suitable representation. During online recognition, a similar representation of a scene is matched with the database for recognizing objects present in the scene . . . The tensor representation is extended to automatic and pose invariant 3D face recognition. As the face is a non-rigid object, expressions can significantly change its 3D shape. Therefore, the last part of this thesis investigates representations and matching techniques for automatic 3D face recognition which are robust to facial expressions. A number of novelties are proposed in this area along with their extensive experimental validation using the largest available 3D face database. These novelties include a region-based matching algorithm for 3D face recognition, a 2D and 3D multimodal hybrid face recognition algorithm, fully automatic 3D nose ridge detection, fully automatic normalization of 3D and 2D faces, a low cost rejection classifier based on a novel Spherical Face Representation, and finally, automatic segmentation of the expression insensitive regions of a face.
The design and construction of three-dimensional [3-D] object recognition systems has long occupied the attention of many computer vision researchers. The variety of systems that have been developed for this task is evidence both of its strong appeal to researchers and its applicability to modern manufacturing, industrial, military, and consumer environments. 3-D object recognition is of interest to scientists and engineers in several different disciplines due to both a desire to endow computers with robust visual capabilities, and the wide applications which would benefit from mature and robust vision systems. However, 3-D object recognition is a very complex problem, and few systems have been developed for actual production use; most existing systems have been developed for experimental use by researchers only. This edited collection of papers summarizes the state of the art in 3-D object recognition using examples of existing 3-D systems developed by leading researchers in the field. While most chapters describe a complete object recognition system, chapters on biological vision, sensing, and early processing are also included. The volume will serve as a valuable reference source for readers who are involved in implementing model-based object recognition systems, stimulating the cross-fertilisation of ideas in the various domains. The variety of topics on Image Communication is so broad that no one can be a specialist in all the topics, and the whole area is beyond the scope of a single volume, while the requirement of up to date information is ever increasing. This new closed-end book series is intended both as a comprehensive reference for those already active in the area of Image Communication, as well as providing newcomers with a foothold for commencing research. Each volume will comprise a state of the art work on the editor's/author's area of expertise, containing information until now scattered in many journals and proceedings.
Automatie object recognition is a multidisciplinary research area using con cepts and tools from mathematics, computing, optics, psychology, pattern recognition, artificial intelligence and various other disciplines. The purpose of this research is to provide a set of coherent paradigms and algorithms for the purpose of designing systems that will ultimately emulate the functions performed by the Human Visual System (HVS). Hence, such systems should have the ability to recognise objects in two or three dimensions independently of their positions, orientations or scales in the image. The HVS is employed for tens of thousands of recognition events each day, ranging from navigation (through the recognition of landmarks or signs), right through to communication (through the recognition of characters or people themselves). Hence, the motivations behind the construction of recognition systems, which have the ability to function in the real world, is unquestionable and would serve industrial (e.g. quality control), military (e.g. automatie target recognition) and community needs (e.g. aiding the visually impaired). Scope, Content and Organisation of this Book This book provides a comprehensive, yet readable foundation to the field of object recognition from which research may be initiated or guided. It repre sents the culmination of research topics that I have either covered personally or in conjunction with my PhD students. These areas include image acqui sition, 3-D object reconstruction, object modelling, and the matching of ob jects, all of which are essential in the construction of an object recognition system.
One of the grand challenges of artificial intelligence is to enable computers to interpret 3D scenes and objects from imagery. This book organizes and introduces major concepts in 3D scene and object representation and inference from still images, with a focus on recent efforts to fuse models of geometry and perspective with statistical machine learning. The book is organized into three sections: (1) Interpretation of Physical Space; (2) Recognition of 3D Objects; and (3) Integrated 3D Scene Interpretation. The first discusses representations of spatial layout and techniques to interpret physical scenes from images. The second section introduces representations for 3D object categories that account for the intrinsically 3D nature of objects and provide robustness to change in viewpoints. The third section discusses strategies to unite inference of scene geometry and object pose and identity into a coherent scene interpretation. Each section broadly surveys important ideas from cognitive science and artificial intelligence research, organizes and discusses key concepts and techniques from recent work in computer vision, and describes a few sample approaches in detail. Newcomers to computer vision will benefit from introductions to basic concepts, such as single-view geometry and image classification, while experts and novices alike may find inspiration from the book's organization and discussion of the most recent ideas in 3D scene understanding and 3D object recognition. Specific topics include: mathematics of perspective geometry; visual elements of the physical scene, structural 3D scene representations; techniques and features for image and region categorization; historical perspective, computational models, and datasets and machine learning techniques for 3D object recognition; inferences of geometrical attributes of objects, such as size and pose; and probabilistic and feature-passing approaches for contextual reasoning about 3D objects and scenes. Table of Contents: Background on 3D Scene Models / Single-view Geometry / Modeling the Physical Scene / Categorizing Images and Regions / Examples of 3D Scene Interpretation / Background on 3D Recognition / Modeling 3D Objects / Recognizing and Understanding 3D Objects / Examples of 2D 1/2 Layout Models / Reasoning about Objects and Scenes / Cascades of Classifiers / Conclusion and Future Directions
