Structure guided interior scene synthesis via graph matching

We present a method for reshuffle-based 3D interior scene synthesis guided by scene structures. Given several 3D scenes, we form each 3D scene as a structure graph associated with a relationship set. Considering both the object similarity and relation similarity, we then establish a furniture-object-based matching between scene pairs via graph matching. Such a matching allows us to merge the structure graphs into a unified structure, i.e., Augmented Graph (AG). Guided by the AG, we perform scene synthesis by reshuffling objects through three simple operations, i.e., replacing, growing and transfer. A synthesis compatibility measure considering the environment of the furniture objects is also introduced to filter out poor-quality results. We show that our method is able to generate high-quality scene variations and outperforms the state of the art.     

Stereo vision during adverse weather — Using priors to increase robustness in real-time stereo vision

Stereo vision can deliver a dense 3D reconstruction of the environment in real-time for driver assistance as well as autonomous driving. Semi-Global Matching (SGM) is a popular method of choice for solving this task which is already in use for production vehicles. Despite the enormous progress in the field and the high level of performance of modern stereo methods, one key challenge remains: robust stereo vision in automotive scenarios during rain, snow and darkness. Under these circumstances, current methods generate strong temporal noise, many disparity outliers and false positives on object level. These problems are addressed in this work by regularizing stereo vision via prior information. We formulate a temporal prior and a scene prior, which we apply to SGM in order to overcome the deficiencies. The temporal prior integrates knowledge from the previous disparity map to exploit the high temporal correlation, the scene prior exploits knowledge of a representative traffic scene. Using these priors, the object detection rate improves significantly on a driver assistance dataset of 3000 frames including bad weather while reducing the rate of erroneous object detections. We also outperform the ECCV Robust Vision Challenge 2012 winner, iSGM, on this dataset. In addition, results are presented for the KITTI dataset, even showing improvements under good weather conditions when exploiting the temporal prior.We also show that the temporal and scene priors are easy and efficient to implement on a hybrid CPU/reconfigurable hardware platform. The use of these priors can be extended to other application areas such as mobile robotics.     

BETWEEN LOCATION AWARENESS AND AWARE LOCATIONS: WHERE TO PUT THE INTELLIGENCE

This article presents a review on vision-aided systems and proposes an approach for visual rehabilitation using stereo vision technology. The proposed system utilizes stereo vision, image processing methodology, and a sonification procedure to support blind mobilization. The developed system includes wearable computer, stereo cameras as vision sensor, and stereo earphones, all molded in a helmet. The image of the scene in front of the visually handicapped is captured by the vision sensors. The captured images are processed to enhance the important features in the scene in front for mobilization assistance. The image processing is designed as a model of human vision by identifying the obstacles and their depth information. The processed image is mapped onto musical stereo sound for the blind's understanding of the scene in front. The developed method has been tested in the indoor and outdoor environments and the proposed image processing methodology is found to be effective for object identification.     

3D Object Recognition in Cluttered Environments by Segment-Based Stereo Vision

We propose a new method for 3D object recognition which uses segment-based stereo vision. An object is identified in a cluttered environment and its position and orientation (6 dof) are determined accurately enabling a robot to pick up the object and manipulate it. The object can be of any shape (planar figures, polyhedra, free-form objects) and partially occluded by other objects. Segment-based stereo vision is employed for 3D sensing. Both CAD-based and sensor-based object modeling subsystems are available. Matching is performed by calculating candidates for the object position and orientation using local features, verifying each candidate, and improving the accuracy of the position and orientation by an iteration method. Several experimental results are presented to demonstrate the usefulness of the proposed method.     

Vision resolvability for visually servoed manipulation

This article introduces a sensor placement measure called vision resolvability. The measure provides a technique for estimating the relative ability of various visual sensors, including monocular systems, stereo pairs, multi-baseline stereo systems, and 3D rangefinders, to accurately control visually manipulated objects. The resolvability ellipsoid illustrates the directional nature of resolvability, and can be used to direct camera motion and adjust camera intrinsic parameters in real-time so that the servoing accuracy of the visual servoing system improves with camera-lens motion. The Jacobian mapping from task space to sensor space is derived for a monocular system, a stereo pair with parallel optical axes, and a stereo pair with perpendicular optical axes. Resolvability ellipsoids based on these mappings for various sensor configurations are presented. Visual servoing experiments demonstrate that vision resolvability can be used to direct camera-lens motion to increase the ability of a visually servoed manipulator to precisely servo objects. © 1996 John Wiley & Sons, Inc.     

Generalized pattern extraction from concept lattices

In this paper, we show how the existence of taxonomies on objects and/or attributes can be used in Formal Concept Analysis to help discover generalized concepts. To that end, we analyze three generalization cases ( ?, ?, and α) and present different scenarios of a simultaneous generalization on both objects and attributes. We also discuss the cardinality of the generalized pattern set against the number of simple patterns produced from the initial data set.     

Introduction to AGG and EMF Tiger by modeling a Conference Scheduling System

In this paper, we focus on the Conference Scheduling System, a case study at the Tool Contest of Graph-Based Tools (GraBaTs) 2008. We took part in the contest with our graph transformation tool AGG and the Eclipse-based EMF model transformation tool EMF Tiger. We present the features of both tools and evaluate their abilities to model the Conference Scheduling System and to deal with additional contest assignments like model instance generation, property verification, and interoperability.     

HumanTop: a multi-object tracking tabletop

In this paper, a computer vision based interactive multi-touch tabletop system called HumanTop is introduced. HumanTop implements a stereo camera vision subsystem which allows not only an accurate fingertip tracking algorithm but also a precise touch-over-the-working surface detection method. Based on a pair of visible spectra cameras, a novel synchronization circuit makes the camera caption and the image projection independent from each other, providing the minimum basis for the development of computer vision analysis based on visible spectrum cameras without any interference coming from the projector. The assembly of both cameras and the synchronization circuit is not only capable of performing an ad-hoc version of a depth camera, but it also introduces the recognition and tracking of textured planar objects, even when contents are projected over them. On the other hand HumanTop supports the tracking of sheets of paper and ID-code markers. This set of features makes the HumanTop a comprehensive, intuitive and versatile augmented tabletop that provides multitouch interaction with projective augmented reality on any flat surface. As an example to exploit all the capabilities of HumanTop, an educational application has been developed using an augmented book as a launcher to different didactic contents. A pilot study in which 28 fifth graders participated is presented. Results about efficiency, usability/satisfaction and motivation are provided. These results suggest that HumanTop is an interesting platform for the development of educational contents.     

增强现实中基于轮廓深度恢复的虚实遮挡方法研究

针对增强现实系统中的虚实场景遮挡问题,提出一种基于立体视觉与前景轮廓深度重建的虚实遮挡恢复新方法.首先设计了一种基于色彩与亮度相结合的前景物体轮廓提取方法,以避免阴影对轮廓提取所造成的负面影响,系统随后利用立体视觉原理与插值技术恢复出前景轮廓以及轮廓内部的深度信息,为恢复虚拟与真实场景之间的正确遮挡关系提供依据.实验结...     

A graph-theoretic approach to optimize keyword queries in relational databases

Keyword search can provide users an easy method to query large and complex databases without any knowledge of structured query languages or underlying database schema. Most of the existing studies have focused on generating candidate structured queries relevant to keywords. Due to the large size of generated queries, the execution costs may be prohibitive. However, existing studies lack the idea of a generalized method to optimize the plan of the large set of generated queries. In this paper, we introduce a graph-theoretic optimization approach. We propose a general graph model, Weighted Operator Graph, to address the costs of keyword query evaluation plans. The proposed model is flexible to integrate all of the cost-based plans in a uniform way. We define a Keyword Query Optimization Problem based on a theoretical cost model as a graph-theoretic problem and show it to be a NP-hard problem. We propose a greedy heuristic Maximum Propagation that reduces the size of the intermediate result as early as possible. The proposed algorithm allows us to achieve efficiency in terms of query evaluation costs. The experimental studies on both synthetic and real data set results show that our work outperforms the existing work.     

基于场景层次的遮挡建模与立体匹配

通过在扫描线和图像分割2个层面上建立基于场景层次的遮挡模型,实现了一种全新的从前景到后景分层处理的立体匹配算法.以像素灰度值为依据进行匹配时,离摄像机最近的物体因为不受遮挡的影响,总是最容易匹配成功;稍远一些的景物可以利用已经计算出来的前景建立局部的遮挡模型指导匹配.这样逐层递推,可求得场景中所有物体的视差.图像分割技术将处理单位由像素提升到了图像块,既提高了算法效率,也降低了视差映射图的不连续性.使用标准数据集的测试结果表明,文中算法精度高、强壮性好,在无纹理区域、深度变化边界区域和遮挡区域都达到了比较高的识别率.     

立体图像对的生成

获取同一场景的立体图像对是实现双目立体成像的一个关键问题。提出了一种在三维场景已经建好的情况下生成立体图像对的方法。该方法根据双目立体视觉的原理，利用3DS MAX中的摄像机对象对场景中的物体进行坐标变换和透视投影变换，分别生成左眼视图和右眼视图。实验结果表明，两个目标摄像机与三维模型的位置关系以及基线长度是影响立体效果的重要因素，改变目标摄像机与三维模型的位置，可以分别生成正视差、负视差的立体图像对，当AB与CO的比例参数为0.05时，生成的立体图像对的立体效果较佳。     

Hermes: a simple and efficient algorithm for building the AOC-poset of a binary relation

Given a relation ?? ? ?? × ?? on a set ?? of objects and a set ?? of attributes, the AOC-poset (Attribute/Object Concept poset), is the partial order defined on the “introducers” of objects and attributes in the corresponding concept lattice. In this paper, we present Hermes, a simple and efficient algorithm for building an AOC-poset which runs in O(m i n{n m, n ^α }), where n is the number of objects plus the number of attributes, m is the size of the relation, and n ^α is the time required to perform matrix multiplication (currently α = 2.376). Finally, we compare the runtime of Hermes with the runtime of other algorithms computing the AOC-poset: Ares, Ceres and Pluton. We characterize the cases where each algorithm is the more relevant.     

Scene graph generation by multi-level semantic tasks

Understanding scene image includes detecting and recognizing objects, estimating the interaction relationships of the detected objects, and describing image regions with sentences. However, since the complexity and variety of scene image, existing methods take object detection or vision relationship estimate as the research targets in scene understanding, and the obtained results are not satisfactory. In this work, we propose a Multi-level Semantic Tasks Generation Network (MSTG) to leverage mutual connections across object detection, visual relationship detection and image captioning, to solve jointly and improve the accuracy of the three vision tasks and achieve the more comprehensive and accurate understanding of scene image. The model uses a message pass graph to mutual connections and iterative updates across the different semantic features to improve the accuracy of scene graph generation, and introduces a fused attention mechanism to improve the accuracy of image captioning while using the mutual connections and refines of different semantic features to improve the accuracy of object detection and scene graph generation. Experiments on Visual Genome and COCO datasets indicate that the proposed method can jointly learn the three vision tasks to improve the accuracy of those visual tasks generation.

     

On-line extraction of stable visual landmarks for a mobile robot with stereo vision

This paper proposes a method to extract on-line stable visual landmarks from sensory data obtained by stereo vision so as to adapt to changes of environment. Given a two-dimensional obstacle map, the robot first extracts vertical line segments which are distinct and inside planar surfaces not near boundary edges as they are expected to be observed reliably from various viewpoints. However, the extracted feature information such as position and length include uncertainty due to errors of vision and motion. The robot then reduces the uncertainty by matching the planar surface containing the features to the map. These processes are performed on-line in order to adapt to actual changes of lighting and the scene depending on the robot's view. Experimental results in real scenes show the validity of the proposed method.