杂乱场景下小物体抓取检测研究

目的 杂乱场景下的物体抓取姿态检测是智能机器人的一项基本技能。尽管六自由度抓取学习取得了进展,但先前的方法在采样和学习中忽略了物体尺寸差异,导致在小物体上抓取表现较差。方法 提出了一种物体掩码辅助采样方法,在所有物体上采样相同的点以平衡抓取分布,解决了采样点分布不均匀问题。此外,学习时采用多尺度学习策略,在物体部分点云上使用多尺度圆柱分组以提升局部几何表示能力,解决了由物体尺度差异导致的学习抓取操作参数困难问题。通过设计一个端到端的抓取网络,嵌入了提出的采样和学习方法,能够有效提升物体抓取检测性能。结果 在大型基准数据集GraspNet-1Billion上进行评估,本文方法取得对比方法中的最优性能,其中在小物体上的抓取指标平均提升了7%,大量的真实机器人实验也表明该方法具有抓取未知物体的良好泛化性能。结论 本文聚焦于小物体上的抓取,提出了一种掩码辅助采样方法嵌入到提出的端到端学习网络中,并引入了多尺度分组学习策略提高物体的局部几何表示,能够有效提升在小尺寸物体上的抓取质量,并在所有物体上的抓取评估结果都超过了对比方法。     

Efficient feature tracking of time-varying surfaces using multi-scale motion flow propagation

This paper presents a framework for efficient feature tracking of time-varying surfaces. The framework can not only capture the dynamic geometry features on time-varying surfaces, but can also compute the accurate boundaries of the geometry features. The basic idea of the proposed approach is using the multi-scale motion flow and surface matching information to propagate the feature frame on time-varying surfaces. We first define an effective multi-scale geometry motion flow for the time-varying surfaces, which efficiently propagates the geometry features along the time direction of the time-varying surfaces. By combining both the approximately invariant signature vectors and geometry motion flow vectors, we also incorporate the shape matching into the system to process feature tracking for time-varying surfaces in large deformation while with low frame sampling rate. Our approach does not depend on the topological connection of the underlying surfaces. Thus, it can process both mesh-based and point-based time-varying surfaces without vertex-to-vertex correspondence across the frames. Feature tracking results on different kinds of time-varying surfaces illustrate the efficiency and effectiveness of the proposed method.     

自由形态物体多视点深度数据配准方法研究

针对多视点数据配准问题提出了一种基于几何匹配的定位方法,该方法是在最近点迭代匹配算法（Ｉｔｅｒａｔｉｖｅ Ｃｌｏｓｅｓｔ Ｐｏｉｎｔ Ａｌｇｏｒｉｔｈｍ）的基础上,对参与最佳旋转矩阵计算的对应点对进行了限制,因而扩大了算法的适用范围,同时又对用结构光方法重建获得的三维物体表面数据进行了加权处理,从而减少了曲面重建所产生的误差对配准精度的影响,并通过实验证明了该方法的有效性。     

Model-Based Object Recognition Using Geometric Invariants of Points and Lines

In this paper, we derive new geometric invariants for structured 3D points and lines from single image under projective transform, and we propose a novel model-based 3D object recognition algorithm using them. Based on the matrix representation of the transformation between space features (points and lines) and the corresponding projected image features, new geometric invariants are derived via the determinant ratio technique. First, an invariant for six points on two adjacent planes is derived, which is shown to be equivalent to Zhu's result [1], but in simpler formulation. Then, two new geometric invariants for structured lines are investigated: one for five lines on two adjacent planes and the other for six lines on four planes. By using the derived invariants, a novel 3D object recognition algorithm is developed, in which a hashing technique with thresholds and multiple invariants for a model are employed to overcome the over-invariant and false alarm problems. Simulation results on real images show that the derived invariants remain stable even in a noisy environment, and the proposed 3D object recognition algorithm is quite robust and accurate.     

索引对象标识与特征文件管理

本文提出一种基于对象标识原理、时间戳排序技术及汉字联想思想实现特征文件有效组织和汉字支撑环境自动形成的方法，并讨论了面向文本数据库管理系统（ＦＩＭＳ）的文本对象索引系统的设计与实现问题。     

Signature verification using global and grid features

In this work, algorithms for extracting global geometric and local grid features of signature images were developed. These features were combined to build a multi-scale verification function. This multi-scale verification function was evaluated using statistical procedures. Results indicated that the multi-scale verification function yielded a lower verification error rate and higher reliability than the single-scale verification function using either global geometric or local grid feature representation. The correct verification rate of the multi-scale system was more than 90% in rejecting skilled forgeries and was perfect in rejecting simple forgeries based on a limited database.     

Contour extraction of gait recognition based on improved GVF Snake model

Contourlet transform can be used to captures smooth contours and edges at any orientation. In order to solve the initial active contour problem of Snake model, Contourlet transform is introduced into the GVF (Gradient Vector Flow) Snake model, which will provides a way to set the initial contour, as a result, will improves the edge detection results of GVF Snake model effectively. The multi-scale decomposition is handled by a Laplacian pyramid. The directional decomposition is handled by a directional filter bank. Firstly, the contours of the object in images can be obtained based on Contourlet transform, and this contours will be identified as the initial contour of GVF Snake model. Secondly, then GVF Snake model is used to detect the contour edge of human gait motion. Experimental results show that the proposed method can extract the edge feature accurately and efficiently.     

A Concise and Provably Informative Multi-Scale Signature Based on Heat Diffusion

We propose a novel point signature based on the properties of the heat diffusion process on a shape. Our signature, called the Heat Kernel Signature (or HKS), is obtained by restricting the well-known heat kernel to the temporal domain. Remarkably we show that under certain mild assumptions, HKS captures all of the information contained in the heat kernel, and characterizes the shape up to isometry. This means that the restriction to the temporal domain, on the one hand, makes HKS much more concise and easily commensurable, while on the other hand, it preserves all of the information about the intrinsic geometry of the shape. In addition, HKS inherits many useful properties from the heat kernel, which means, in particular, that it is stable under perturbations of the shape. Our signature also provides a natural and efficiently computable multi-scale way to capture information about neighborhoods of a given point, which can be extremely useful in many applications. To demonstrate the practical relevance of our signature, we present several methods for non-rigid multi-scale matching based on the HKS and use it to detect repeated structure within the same shape and across a collection of shapes.     

Embedded Implicit Stand-Ins for Animated Meshes: A Case of Hybrid Modelling

In this paper, we address shape modelling problems, encountered in computer animation and computer games development that are difficult to solve just using polygonal meshes. Our approach is based on a hybrid-modelling concept that combines polygonal meshes with implicit surfaces. A hybrid model consists of an animated polygonal mesh and an approximation of this mesh by a convolution surface stand-in that is embedded within it or is attached to it. The motions of both objects are synchronised using a rigging skeleton. We model the interaction between an animated mesh object and a viscoelastic substance, which is normally represented in an implicit form. Our approach is aimed at achieving verisimilitude rather than physically based simulation. The adhesive behaviour of the viscous object is modelled using geometric blending operations on the corresponding implicit surfaces. Another application of this approach is the creation of metamorphosing implicit surface parts that are attached to an animated mesh. A prototype implementation of the proposed approach and several examples of modelling and animation with near real-time preview times are presented.     

Shapes recognition using the straight line Hough transform: theoryand generalization

A shape matching technique based on the straight line Hough transform (SLHT) is presented. In the &thetas;-ρ space, the transform can be expressed as the sum of the translation term and the intrinsic term. This formulation allows the translation, rotation, and intrinsic parameters of the curve to be easily decoupled. A shape signature, called the scalable translation invariant rotation-to-shifting (STIRS) signature, is obtained from the &thetas;-ρ space by computing the distances between pairs of points having the same &thetas; value. This signature is invariant to translation and can be easily normalized, and rotation in the image space corresponds to circular shifting of the signature. Matching two signatures only amounts to computing a 1D correlation. The height and location of a peak (if it exists) indicate the similarity and orientation of the test object with respect to the reference object. The location of the test object is obtained, once the orientation is known, by an inverse transform (voting) from the &thetas;-ρ space to the x-y plane     

Exploring the synergistic use of multi-scale image object metrics for land-use/land-cover mapping using an object-based approach

This study presents a new method for the synergistic use of multi-scale image object metrics for land-use/land-cover mapping using an object-based classification approach. This new method can integrate an object with its super-objects’ metrics. The entire classification involves two object hierarchies: (1) a five-level object hierarchy to extract object metrics at five scales, and (2) a three-level object hierarchy for the classification process. A five-level object hierarchy was developed through multi-scale segmentation to calculate and extract both spectral and textural metrics. Layers representing the hierarchy at each of the five scales were then intersected by using the overlay tool, an intersected layer was created with metrics from all five scales, and the same geometric elements were retained as those of the objects of the lowest level. A decision tree analysis was then used to build rules for the classification of the intersected layer, which subsequently served as the thematic layer in a three-level object hierarchy to identify shadow regions and produce the final map. The use of multi-scale object metrics yielded improved classification results compared with single-scale metrics, which indicates that multi-scale object metrics provide valuable spatial information. This method can fully utilize metrics at multiple scales and shows promise for use in object-based classification approaches.     

Interactive modeling of complex geometric details based on empirical mode decomposition for multi-scale 3D shapes

In this paper we develop an interactive modeling system for complex geometric details transformation based on empirical mode decomposition (EMD) on multi-scale 3D shapes. Given two models, we aim to transfer geometric details from one model to another one in an interactive manner. Taking full advantages of the multi-scale representation computed via EMD, different-scale geometric details can be transferred individually or in a concerted way, which makes our algorithm much more flexible than cut-and-paste and cloning based methods in transferring geometry details. In this process, the target surface with new transferred details could be generated by a mesh reconstruction method widely used in Laplacian surface editing. With the original positions of target surface serving as the soft constraints, the overall shape of the target model will be fully preserved. Our method can also support real-time continuous details transfer. Compared with state-of-the-art algorithms, our method provides an easier-to-use modeling tool and produces varied modeling results. We demonstrate the effectiveness of our modeling tool with various applications, such as detail transfer and enrichment, model reuse and recreation, and detail recovery for shape completion.     

Fractals and Solid Modeling

Trying to combine fractal geometry and solid modeling seems to be a contradiction in itself, In this paper a new type of 3D objects is presented that accomplishes this combination in a specific way. Objects with a fractal macro structure and a 3D solid micro structure can be specified and rendered efficiently by using context free, attribute, geometric grammars. This new object type can be incorporated into the CSG-modeling technique (Constructive Solid Geometry) in two ways: a) using CSG for the specification of the micro structure of the new object type, b) using these fractal like objects as a new type of primitive in the CSG model. Ray tracing is used for generating high quality images of these geometrically complex objects.     

从运动中恢复目标结构的改进因子分解法

因子分解法是从图像序列中恢复刚体目标几何结构的重要方法。针对传统因子分解法基本过程中存在的不足,及其容易失效的缺点,提出一种改进的因子分解法。该方法避开传统方法中求解修正矩阵的复杂过程,利用旋转矩阵的特性,直接修正由传统方法奇异值分解(SVD)得到的每帧图像的旋转矩阵,然后根据观测矩阵和得到的旋转矩阵,直接利用线性最小二乘法求解目标的结构矩阵。仿真和实测数据的实验结果表明,本文方法能够有效地从序列图像中恢复目标的几何结构,相比传统因子分解法,在稳定性上有较大的提升。     

h-graphs: A new representation for tree decompositions of graphs

In geometric constraint solving, 2D well constrained geometric problems can be abstracted as Laman graphs. If the graph is tree decomposable, the constraint-based geometric problem can be solved by a Decomposition–Recombination planner based solver. In general decomposition and recombination steps can be completed only when steps on which they are dependent have already been completed. This fact naturally defines a hierarchy in the decomposition–recombination steps that traditional tree decomposition representations do not capture explicitly.In this work we introduce h-graphs, a new representation for decompositions of tree decomposable Laman graphs, which captures dependence relations between different tree decomposition steps. We show how h-graphs help in efficiently computing parameter ranges for which solution instances to well constrained, tree decomposable geometric constraint problems with one degree of freedom can actually be constructed.     

Object segmentation in hyperspectral images using active contours and graph cuts

The interest in object segmentation on hyperspectral images is increasing and many approaches have been proposed to deal with this area. In this project, we developed an algorithm that combines both the active contours and the graph cut approaches for object segmentation in hyperspectral images. The active contours approach has the advantage of producing subregions with continuous boundaries. The graph cut approach has emerged as a technique for minimizing energy functions while avoiding the problems of local minima. Additionally, it guarantees continuity and produces smooth contours, free of self-crossing and uneven spacing problems. The algorithm uses the complete spectral signature of a pixel and also considers spatial neighbourhood for graph construction, thereby combining both spectral and spatial information present in the image. The algorithm is tested using real hyperspectral images taken from a variety of sensors, such as Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and Hyperspectral Data Imagery Collection Experiment (HYDICE), and also taken by the SOC hyperspectral camera. This approach can segment different objects in an image. This algorithm can be applied in many fields and it should represent an important advance in the field of object segmentation.     

计算机动画关键帧插补技术综述

对计算机辅助动画制作过程中关键帧的插补技术进行了研究。线性插值法通过给定的起始帧和结束帧对应点之间的直线距离算出中间插值帧。移动点约束条件插补法利用随时间空间变化的曲线与彩体上的点相关联，来控制关键帧上对应控制点的轨迹以厦动态特征。骨架插补法是将图形抽象成骨架，然后进行插补。     

Articulated Object Reconstruction and Markerless Motion Capture from Depth Video

We present an algorithm for acquiring the 3D surface geometry and motion of a dynamic piecewise‐rigid object using a single depth video camera. The algorithm identifies and tracks the rigid components in each frame, while accumulating the geometric information acquired over time, possibly from different viewpoints. The algorithm also reconstructs the dynamic skeleton of the object, thus can be used for markerless motion capture. The acquired model can then be animated to novel poses. We show the results of the algorithm applied to synthetic and real depth video.     

Generic Transformation from Weakly to Strongly Unforgeable Signatures

Current techniques for transforming unforgeable signature schemes （the forged message has never been signed） to strongly unforgeable ones （the forged message could have been signed） require supplementary components to be added onto the original key pairs of the schemes. In addition, some of them can only be applied to a certain type of signature schemes. In this paper, we propose a new generic transformation technique which converts any unforgeable signature scheme into a strongly unforgeable one without modifying any component in the original key pair. This makes our technique especially compatible for practical use. Our technique is based on strong one-time signature schemes. We show that they can be constructed efficiently from any one-time signature scheme that is based on one-way functions. The performance of our technique also compares favorably with that of current ones. Besides, it is shown in this paper that our transformation can further be applied to schemes satisfying only a weak variant of unforgeability without any further modification. Furthermore, our technique can also be used for constructing strongly unforgeable signature schemes in other cryptographic settings which include certificateless signature, identity-based signature, and several others. To the best of our knowledge, similar extent of versatility is not known to be supported by any of those comparable techniques. Finally and of independent interest, we show that our generic transformation technique can be modified to an on-line/off-line signature scheme, which possesses a very efficient signing process.     

逆向工程中约束驱动数据点云曲面特征优化

为了获得产品原始设计意图,提高重构模型的整体质量,提出一种实用的逆向工程中约束驱动数据点云曲面特征优化方法,其中包括约束分解和有效的数值求解.在约束分解部分,通过设计结构矩阵分割算法消除几何约束系统中曲面特征间的耦合约束,提出了基于多尺度特征的凝聚算法来实现几何约束系统的简化和分解;在数值求解部分,基于罚函数法建立了约束优化的数学模型,采用BFGS法进行了数值求解.对优化后的逼近误差与约束满足误差进行分析的结果表明,采用文中方法可以低数量级的逼近误差的放大,实现约束满足误差的减小,获得一种全局优化的结果.