一种复杂背景下多尺度多视角的人体检测方法

针对复杂背景下的人体检测技术所面临的噪声干扰、背景复杂、相互遮挡等问题,设计一种多尺度多视角人体检测算法。针对传统的梯度方向直方图目标特征提取方法特征维数大、有遮挡时检测率低等缺陷,分别使用扩展多尺度方向特征和经WTA hash编码的多尺度梯度方向直方图特征提取,并使用弱分类器和贪婪算法进行特征选择以获得图像的粗特征和精特征。然后使用线性平移合成多视角样本,使用多层级联的Adaboost算法和支持向量机作为分类器进行人体目标检测,结合复杂背景处理、特征重装等方法提高检测精度。使用INRIA公共测试集的实验结果表明,该算法可精确检测出复杂背景下相互遮挡情况下多视角、多姿态的人体目标,与传统的人体检测算法相比,具有更高的检测效率和检测精度。     

基于Adaboost和SVM的人头实时检测

针对复杂场景图像中的人头检测问题,提出一种Adaboost与支持向量机(SVM)相结合的检测算法.该算法重点对Adaboost特征进行了改进,用Adaboost对人头进行快速检测,并引入级联的SVM分类器对Adaboost检测结果进行逐级筛选,从而实现对人头的精确检测.实验表明,该方法降低了Adaboost运算复杂度,提高了特征分类能力,引入级联SVM分类器在保证高检测率的同时,降低了误检率,对复杂场景具有较强的鲁棒性.     

复杂背景下人眼的快速定位

采用基于改进Adaboost算法的级联式人脸和人眼分类器检测人脸和眼睛的候选位置,再用支持向量机（SVM）分类器验证并确定人眼的最佳位置;实现了在复杂背景图像中快速、准确的眼睛定位.与传统的Adaboost算法相比,改进的Adaboost算法构建分类器所需的特征数目大大减少,提高了眼睛定位速度;同时利用人脸几何特征的SVM分类器验证,提高了定位精度.实验结果表明该算法具有很好的精确性和实时性.     

基于Gabor变换和Adaboost算法的人体目标检测分类器

针对在图像中检测人体目标,提出一种基于Gabor变换和Adaboost算法的检测方法.首先利用二维Gabor小波变换进行特征提取,然后利用Adaboost算法对Gabor特征进行选取并训练强分类器.为了提高检测精度,提出采用单一正样本集合与多个负样本集合分别进行训练,形成多个强分类器级联的层级检测分类器.实验结果表明了该方法的有效性,同时显示该方法须与其它辅助手段相结合,才能提高检测的实时性.     

复杂背景下人体检测算法

静态图像的人体识别技术在实际视频监控场景中, 面临背景复杂、图像分辨率低、光照影响和噪声干扰等问题, 这些多变性以及可能发生的遮挡给单一图像中的人体识别技术提出了挑战. 设计和实现一种复杂背景下的人体检测算法, 基于HOG人体检测算法, 使用积分直方图计算HOG特征, 并用级联SVM分类器对样本进行训练. 实验结果表明, 该算法在复杂视频监控场景中进行人体检测比其它人体检测算法具有更高的准确率和更快的检测速度.     

基于多尺度特征和嵌套级联Adaboost的车辆检测

为提高车辆检测的速度和准确性,提出了一种基于多尺度方向梯度直方图(HOG)和多尺度多块局部二进制模式(MB-LBP)两种特征与嵌套级联Gentle Adaboost的车辆检测算法.分别使用积分直方图和积分图像加速提取多尺度HOG和多尺度MB-LBP特征.基于两种特征为Gentle Adaboost构建两种弱分类器,并采用嵌套级联Gentle Adaboost分类器提高检测率和检测速度.仿真实验结果表明:相比于现有的几种车辆检测算法,提出的算法检测速度更快,且检测精度和召回率更高.     

基于边缘方向直方图的Adaboost人脸检测

在传统的Adaboost分类器算法中，采用色彩直方图或空间直方图作为提取特征，但这并不能充分描述图像目标的：特征，导致了目标检测和跟踪出现偏差。提出了一种基于边缘方向直方图的Adaboost人脸检测算法，使用空间分布和纹理信息作为提取特征。实验结果表明，该方法与传统Adaboost方法相比，准确率明显提高．而速度相当。     

基于边缘方向直方图的Adaboost人脸检测

在传统的Adaboost分类器算法中,采用色彩直方图或空间直方图作为提取特征,但这并不能充分描述图像目标的特征,导致了目标检测和跟踪出现偏差。提出了一种基于边缘方向直方图的Adaboost人脸检测算法,使用空间分布和纹理信息作为提取特征。实验结果表明,该方法与传统Adaboost方法相比,准确率明显提高,而速度相当。     

利用SVM改进Adaboost算法的人脸检测精度

提出利用SVM分类方法改进Adaboost算法的人脸检测精度。该方法先通过Adaboost算法找出图像中的候选人脸区域,根据训练样本集中的人脸和非人脸样本训练出分类器支持向量机(SVM),然后通过SVM分类器从候选人脸区域中最终确定人脸区域。实验结果证明,SVM分类算法可以提高检测精度,使检测算法具有更好的检测效果。     

基于连续Adaboost算法的多视角人脸检测

提出了一种基于连续Adaboost算法的多视角人脸检测方法.人脸按其三维姿态被划分成若干个视点子类,针对每个子类使用Haar型特征设计了具有连续致信度输出的查找表型弱分类器形式,构造出弱分类器空间,采用连续Adaboost算法学习出基于视图的瀑布型人脸检测器.为了提高检测速度,使用了多分辨率搜索和姿态预估计策略.对于正面人脸检测,在CMU MIT的正面人脸测试集合上检测的正确率为94.5%,误报57个;对于多视角人脸检测,在CMU侧面人脸测试集合上检测的正确率为89.8%,误报221个.在一台PentiumⅣ2.4GHz的PC上,处理一幅大小为320×240的图片平均需80ms.实验结果表明该方法十分有效,具有明显的应用价值.     

Inessential Features, Ineliminable Features, and Modal Logics for Model Theoretic Syntax

While monadic second-order logic (MSO) has played a prominent role in model theoretic syntax, modal logics have been used in this context since its inception. When comparing propositional dynamic logic (PDL) to MSO over trees, Kracht (1997) noted that there are tree languages that can be defined in MSO that can only be defined in PDL by adding new features whose distribution is predictable. He named such features “inessential features”. We show that Kracht’s observation can be extended to other modal logics of trees in two ways. First, we demonstrate that for each stronger logic, there exists a tree language that can only be defined in a weaker logic with inessential features. Second, we show that any tree language that can be defined in a stronger logic, but not in some weaker logic, can be defined with inessential features. Additionally, we consider Kracht’s definition of inessential features more closely. It turns out that there are features whose distribution can be predicted, but who fail to be inessential in Kracht’s sense. We will look at ways to modify his definition.     

Learnability and Definability in Trees and Similar Structures

We prove upper bounds for combinatorial parameters of finite relational structures, related to the complexity of learning a definable set. We show that monadic second-order (MSO) formulas with parameters have bounded Vapnik–Chervonenkis dimension over structures of bounded clique-width, and first-order formulas with parameters have bounded Vapnik–Chervonenkis dimension over structures of bounded local clique-width (this includes planar graphs). We also show that MSO formulas of a fixed size have bounded strong consistency dimension over MSO formulas of a fixed larger size, for labeled trees. These bounds imply positive learnability results for the PAC and equivalence query learnability of a definable set over these structures. The proofs are based on bounds for related definability problems for tree automata.     

Multi-weighted Automata and MSO Logic

Weighted automata are non-deterministic automata where the transitions are equipped with weights. They can model quantitative aspects of systems like costs or energy consumption. The value of a run can be computed, for example, as the maximum, average, or discounted sum of transition weights. In multi-weighted automata, transitions carry several weights and can model, for example, the ratio between rewards and costs, or the efficiency of use of a primary resource under some upper bound constraint on a secondary resource. Here, we introduce a general model for multi-weighted automata as well as a multi-weighted MSO logic. In our main results, we show that this multi-weighted MSO logic and multi-weighted auto-mata are expressively equivalent both for finite and infinite words. The translation process is effective, leading to decidability results for our multi-weighted MSO logic.     

面向手机短信的隐私保护方案

针对手机短信存在的用户隐私泄露问题,设计一种面向手机短信的隐私保护方案.本方案结合非对称加密技术,通过使用信息接收者的公钥对原始短信进行加密获得短信密文,并结合预设的生命周期信息封装成短信自毁对象(Message Self-destructing Object,MSO)并通过运营商发送给接收者,接收者接收到MSO后对其生命周期进行验证,只有当前时间处于其生命周期内时,才能进一步使用其私钥对短信密文进行解密获取原始短信内容,一旦超过MSO的生命周期,则MSO将被自动删除以保护用户隐私安全.实验分析表明,本方案能够有效保护用户手机短信的隐私安全,实现生命周期控制并自动删除过期短信,并且对硬件系统要求低,开销合理,适合在人们日常生活中进行推广使用.     

基于CMDS-MSO的无线传感器网络定位算法

针对基于经典多维标度(CMDS)的定位算法中用最短路径距离替代节点间Euclidean距离造成的距离误差,结合质点弹簧模型(MSO)抑制误差的优点,提出了基于CMDS-MSO的新定位算法,该算法既有效抑制了CMDS算法的定位误差,又克服了质点弹簧优化模型算法易陷入局部最优的不足,有效提高了节点的定位精度。仿真结果表明,基于CMDS-MSO的新型定位算法的定位精度要高于基于CMDS的定位算法和基于MSO的定位算法,具有较强的鲁棒性。     

Existential MSO over two successors is strictly weaker than over linear orders

As is well known, a language of finite words, considered as labeled linear orders, is definable in monadic second-order logic (MSO) iff it is definable in the existential fragment of MSO, that is the quantifier alternation hierarchy collapses. Even more, it does not make a difference if we consider existential MSO over a linear order or a successor relation only. In this note we show that somewhat surprisingly the latter does not hold if we just add a second linear order and consider finite relational structures with two linear orders, so-called texts.     

DCFLA: A distributed collaborative-filtering neighbor-locating algorithm

Although collaborative filtering (CF) has proved to be one of the most successful techniques in recommendation systems, it suffers from a lack of scalability as the time complexity rapidly increases when the number of the records in the user database increases. As a result, distributed collaborative filtering (DCF) is attracting increasing attention as an alternative implementation scheme for CF-based recommendation systems. In this paper, we first propose a distributed user-profile management scheme using distributed hash table (DHT)-based routing algorithms, which is one of the most popular and effective approaches in peer-to-peer (P2P) overlay networks. In this DCF scheme, an efficient DCF neighbor-locating algorithm (DCFLA) is proposed, together with two improvements, most same opinion (MSO) and average rating normalization (ARN), to reduce the network traffic and time cost. Finally, we analyze the performance of one baseline and three novel CF algorithms are being proposed: (1) a traditional memory-based CF (baseline); (2) a basic DHT-based CF; (3) a DHT-based CF with MSO; and (4) a DHT-based CF with MSO and ARN. The experimental results show that the scalability of our proposed DCFLA is much better than the traditional centralized CF algorithm and the prediction accuracies of these two systems are comparable.     

Expressiveness and complexity of graph logic

We investigate the complexity and expressive power of a spatial logic for reasoning about graphs. This logic was previously introduced by Cardelli, Gardner and Ghelli, and provides the simplest setting in which to explore such results for spatial logics. We study several forms of the logic: the logic with and without recursion, and with either an exponential or a linear version of the basic composition operator. We study the combined complexity and the expressive power of the four combinations. We prove that, without recursion, the linear and exponential versions of the logic correspond to significant fragments of first-order (FO) and monadic second-order (MSO) Logics; the two versions are actually equivalent to FO and MSO on graphs representing strings. However, when the two versions are enriched with μ-style recursion, their expressive power is sharply increased.Both are able to express PSPACE-complete problems, although their combined complexity and data complexity still belong to PSPACE.     

一种增量式模块化回声状态网络

针对传统回声状态网络(ESN) 难以解决多振荡子叠加(MSO) 问题, 提出一种增量式模块化回声状态网络(IM-ESN). 该网络储备池由多个相互独立的子储备池组成. 利用矩阵的奇异值分解(SVD) 构造每个子储备池的权值矩阵, 并依据分块对角矩阵原理, 将子储备池逐一添加至网络中. 在网络增长过程中, IM-ESN 无需放缩权值矩阵便能保证网络的状态回声特性. MSO 问题的仿真结果表明, IM-ESN 能够自主确定与问题复杂度相匹配的网络规模, 具有较好的预测性能和鲁棒性.

     

The equivalence problem for deterministic MSO tree transducers is decidable

It is decidable for deterministic MSO definable graph-to-string or graph-to-tree transducers whether they are equivalent on a context-free set of graphs.