基于视觉仿生机理的成像目标检测和识别方法及感知计算

目标检测和识别是图像分析和理解的核心问题,构建了一种仿人眼视觉特性的视觉检测和目标识别体系结构及感知计算模式。借鉴人眼视觉信息获取与处理的变空间分辨率机理和稀疏性,构建大场景(LF)子系统和小场景(SF)子系统分别获取多分辨率、多尺度和不同精细粒度的初级视觉特征信息。提出了一种在小波域下受视觉注意力机制引导的LF子系统感知场景整体统计特性的目标检测和定位方法,由SF子系统集中对目标形成凝视并提取细粒度特征信息,对特征进行整合,形成兴趣图,然后采用非均匀采样、多尺度分析和胜者为王(Winner-take-all机制)产生目标间的竞争实现分类识别。仿真实验结果表明,统计分析方法降低了信息冗余,快速准确地检测出感兴趣目标区域,而基于注意机制的目标识别在多类目标分类中达到94.40%的总准确率。     

一种基于梯度特征列表的机器视觉目标匹配

机器视觉目标匹配中,匹配相似性测量应用特征列表相关算法计算,可有效减少处理时间,并且匹配具有较高的峰值系数和峰值信噪比,可以清晰识别目标。所提出的基于梯度特征列表的机器视觉目标匹配方法,采用梯度特征列表描述图像,对特征像素点非均匀采样,匹配相似测量基于特征像素点梯度归一化互相关,可有效改善特征列表算法性能。     

神经形态视觉传感器的研究进展及应用综述

神经形态视觉传感器具有高时域分辨率、高动态范围、低数据冗余和低功耗等优势,近年来在自动驾驶、无人机视觉导航、工业检测及视觉监控等领域(尤其在涉及高速运动和极端光照等条件下)产生了巨大的应用前景.神经形态视觉是一个包含硬件、软件、生物神经模型等的视觉感知系统,终极目标是模拟生物视觉感知结构与机理,以硅视网膜达到、扩展或超越人类智能.作为神经形态工程的重要分支,神经形态视觉是计算神经科学与计算机视觉领域的交叉学科与研究热点.本文从生物视觉采样模型、神经形态视觉传感器的采样模型及类型、视觉信号处理与特征表达、视觉任务应用等视角进行了系统性地回顾与综述,展望了该领域未来研究的技术挑战与可能发展方向,同时探讨了其对未来机器视觉和人工智能领域的潜在影响.     

SD大鼠光刺激视觉诱发电位提取研究

在生物神经学领域,从脑电信号中提取出诱发电位对动物视觉图像恢复的研究具有重要作用。以Sprague-Dawley（SD）大鼠的脑电信号为研究对象,采用均值法和改进的Mallat多分辨率小波快速变换算法相结合的方法对SD大鼠脑电信号进行提取处理,得到了光刺激视觉诱发电位的主要特征。通过和SD大鼠视觉诱发脑电信号的拟合度仿真表明算法所得到的大鼠光刺激视觉诱发电位主要特征可以用于SD大鼠视觉图像恢复的处理。     

基于纹理图像分析的生物视觉模型不变性评价

基于纹理图像,从计算机视觉角度对生物视觉模型——视皮层目标识别的标准模型进行定量分析与评价。对原始图像分别进行尺度、旋转及仿射等变化,利用标准模型提取变化后图像的生物视觉特征,再根据提取的生物视觉特征对纹理图像进行分类,采用图像分类结果的曲线下面积来定量分析和评价生物视觉模型是否具有不变性。大量与局部二元模式特征的对比实验表明,该模型提取的生物视觉特征对于纹理图像具备优良的尺度、旋转与仿射不变性。     

轻量级的全息道路交通状态视觉检测的研究

针对道路拥堵检测难、交通基本参数获取计算复杂度高等问题,提出了一种轻量级的全息道路交通状态视觉检测方法。为了能在嵌入式系统上同时实现道路拥堵状态和各种交通基本参数的视觉自动化检测,首先通过定制道路区域并自动生成均匀分布的采样点,采用以点代面的设计思想,来减少图像处理的计算资源和存储资源;其次,采用背景差法和帧间差法相结合的处理方法分别得到非存在采样点、存在采样点、移动存在采样点和静止存在采样点;接着,采用非存在采样点实现精准快速的道路背景建模,根据存在采样点的空间分布情况获取一些重要的交通基本参数,并利用静止存在采样点的空间排列情况进行拥堵分析。实验结果表明,文中提出的检测算法具有计算效率高、耗费资源少、检测范围广、鲁棒性强等优点,能快速并准确地检测出各种交通基本参数和道路拥堵状态。     

改进的随机Hough变换在检测多圆中的应用

提出了一种在视觉检测中用于多圆提取的改进随机Hough变换算法，该算法采用一种新的取点方法来进行计算，避免了随机Hough变换由于随机采样造成的大量无效累积。并在计算圆的几何参数时充分利用了圆的几何特征，因此检测速度快，占用内存空间小，精度高，具有良好的抗噪和抗干扰能力。最后文中还分别给出了人工合成图和实物图的实验结果。     

基于生物机制脉冲神经网络的特征提取

脉冲神经元可以被用于处理生物刺激并且可以解释大脑复杂的智能行为。脉冲神经网络以非常逼近生物的神经元模型作为处理单元,可以直接用来仿真脑科学中发现的神经网络计算模型,输出的脉冲信号还可与生物神经系统对接。而小波变换是一个非常有利的时频分析工具,它可以有效的压缩图像并且提取图像的特征。本文中将提出一种与人类视觉系统的开/关神经元阵列相结合的脉冲神经网络,来实现针对视觉图像的快速小波变换。仿真结果显示,这个脉冲神经网络可以很好地保留视觉图像的关键特征。     

基于二进制特征描述器的图像匹配算法

图像特征点匹配在视觉系统中有广泛的应用。针对加速分割测试特征FAST和二进制稳健基元独立特征BRIEF算法中存在的问题进行改进。首先,在FAST算法中使用简化模板提取图像特征点,通过构建图像金字塔实现尺度不变性。接着,根据人类视觉系统原理改进BRIEF算法的点对采样模式,并通过特征点方向的计算实现图像的旋转不变性。最后,使用易于计算的海明距离度量各特征点的相似度实现特征匹配。实验表明,提出的图像匹配算法性能优于其他算法,而且运行速度更快。     

自助重要性采样用于实时多目标视觉跟踪

多目标视觉跟踪的主要困难来自于多个目标交互(部分或完全遮挡)导致的歧义性. 马尔可夫随机场(Markov random field, MRF)可以消除这种歧义性且无需显式的数据关联. 但是, 通用概率推理算法的计算代价很高. 针对上述问题, 本文做出了3点贡献: 1)设计了新的具有"分散-集中-分散"结构的递归贝叶斯跟踪框架—自助重要性采样粒子滤波器, 它 使用融入当前时刻观测的重要性密度函数解决维数灾难问题, 将计算复杂度从指数增长变为线性增长; 2)提出了新的蒙特卡洛策略— 自助重要性采样, 利用MRF的因子分解性质进行重要性采样, 并使用自助法产生低成本高质量的样本、降低似然度计算次数和维持多模式分布; 3)采用了新的边缘化技术—使用辅助变量采样进行边缘化, 使用自助直方图对边缘后验分布进行密度估计. 实验结果表明, 本文提出的算法能够对大量目标进行实时跟踪, 能够处理目标间复杂的交互, 能够在目标消失后维持多模式分布.     

Augmented TP model transformation‐based parallel distributed compensation control design

The usual sampled hyper grid points of uniform sampling method are distributed equally in the TP model transformation, thus, the sampling results often omit the local extrema when the sampling step is not fine‐tuned. Then the resultant tensor which is used for controller design can not fully cover the state space, although the gain which is the feasible solution of the linear matrix inequalities. In this paper, we proposed a non‐uniform sampling method for tensor product model transformation, local extrema are considered in the sampling step, while the sampling step can vary dynamically for different function entries. In this paper, TP model transformation‐based parallel distributed compensation (PDC) controller is extended in three folds: (i) The existing TP model transformation‐based PDC controller with uniform sampling method is extended to TP model transformation‐based PDC tracking controller with the uniform sampling method and an extended signal. (ii) A new TP model transformation‐based PDC tracking controller is proposed based on a new sampling method, that is, the Hammersley sampling method. (iii) TP model transformation‐based PDC tracking controller is also proposed based on the non‐uniform sampling method. The proposed adaptive TP model transformation‐based PDC tracking controller is able to enhance the performance of the TP model transformation‐based PDC controller, and the adaptive TP model transformation‐based PDC controller obtains the best results due to the nearly exact sampling of the system.     

Isoparametric line sampling for the inspection planning of sculptured surfaces

This paper presents a new approach for the sampling of sculptured surfaces using continuous scanning coordinate measuring machine probe heads. This approach is based on scanning isoparametric lines on the sculptured surface. This paper addresses the issue of isoparametric line sampling. The problem is to determine the locations of the sample lines extracted from the surface CAD model. These lines are fitted to construct a substitute geometry of the surface. The accuracy of the sampling plan is characterised by the maximum deviation between the substitute geometry and the surface CAD model. Two new algorithms for sampling isoparametric lines are proposed. These are, automatic sampling, and surface curvature-based sampling. Both algorithms are constrained by the maximum number of scan lines, and the step over distance between subsequent sample lines. Automatic sampling uses the deviations between the substitute geometry, and the surface CAD model to determine the sample locations. Curvature change-based sampling uses the change in surface curvature to determine the sampling line locations. Both algorithms are compared to uniform iso-planar sampling. The algorithms, their implementation, and a case study are presented in this paper.     

基于参量模型估计的周期信号等效采样方法研究

针对现有的常规的等效时间采样方法必须具备准确的模拟触发电路,必须具备精确的定时电路或时长检测电路,并且波形重建时间长甚至不能完全重建的问题,提出一种基于参量模型估计的周期信号等效时间采样方法.该方法采用三个两两互质频率进行三轮采样,从三轮采样数据估计出被测周期信号的基频和其它参数,从而重构出被测周期信号波形.实验证明,该方法在信噪比较高时重构出被测周期信号波形的正确概率很高,重构误差很小.该方法已被成功运用到数字存储示波器的高速数据采集系统中.     

Marginal likelihoods for non-Gaussian models using auxiliary mixture sampling

Several new estimators of the marginal likelihood for complex non-Gaussian models are developed. These estimators make use of the output of auxiliary mixture sampling for count data and for binary and multinomial data. One of these estimators is based on combining Chib’s estimator with data augmentation as in auxiliary mixture sampling, while the other estimators are importance sampling and bridge sampling based on constructing an unsupervised importance density from the output of auxiliary mixture sampling. These estimators are applied to a logit regression model, to a Poisson regression model, to a binomial model with random intercept, as well as to state space modeling of count data.     

基于在线最小二乘支持向量机的广义预测控制

This paper proposes a practical generalized predictive control(GPC)algorithm based on online least squares support vector machines(LS-SVM)which can deal with nonlinear systems effectively.At each sampling period the algorithm recursively modifies the model by adding a new data pair and deleting the least important one out of the consideration on realtime property.The data pair deleted is determined by the absolute value of lagrange multiplier from last sampling period.The paper gives the recursive algorithm of model parameters when adding a new data pair and deleting an existent one,respectively,and thus the inversion of a large matrix is avoided and the memory can be controlled by the algorithm entirely.The nonlinear LS-SVM model is applied in GPC algorithm at each sampling period.The experiments of generalized predictive control on pH neutralizing process show the effectiveness and practicality of the proposed algorithm.     

3D model retrieval based on color + geometry signatures

Color plays a significant role in the recognition of 3D objects and scenes from the perspective of cognitive psychology. In this paper, we propose a new 3D model retrieval method, focusing on not only the geometric features but also the color features of 3D mesh models. Firstly, we propose a new sampling method that samples the models in the regions of either geometry-high-variation or color-high-variation. After collecting geometry + color sensitive sampling points, we cluster them into several classes by using a modified ISODATA algorithm. Then we calculate the feature histogram of each model in the database using these clustered sampling points. For model retrieval, we compare the histogram of an input model to the stored histograms in the database to find out the most similar models. To evaluate the retrieval method based on the new color + geometry signatures, we use the precision/recall performance metric to compare our method with several classical methods. Experiment results show that color information does help improve the accuracy of 3D model retrieval, which is consistent with the postulate in psychophysics that color should strongly influence the recognition of objects.     

Neurally-inspired stochastic algorithm for determining multi-stage multi-attribute sampling inspection plans

In manufacturing industries, sampling inspection is a common practice for quality assurance and cost reduction. The basic decisions in sampling inspection are how many manufactured items to be sampled from each lot and how many identified defective items in the sample to accept or reject each lot. Because of the combinatorial nature of alternative solutions on the sample sizes and acceptance criteria, the problem of determining an optimal sampling plan is NP-complete. In this paper, a neurally-inspired approach to generating acceptance sampling inspection plans is proposed. A Bayesian cost model of multi-stage-multi-attribute sampling inspections for quality assurance in serial production systems is formulated. This model can accommodate various dispositions of rejected lott such as scraping and screening. The model also can reflect the relationships between stages and among attributes. To determine the sampling plans based on the formulated model, a neurally-inspired stochastic algorithm is developed. This algorithm simulates the state transition of a primal-dual stochastic neural network to generate the sampling plans. The simulated primal network is responsible for generation of new states whereas the dual network is for recording the generated solutions. Starting with an arbitrary feasible solution, this algorithm is able to converge to a near optimal or an optimal sampling plan with a sequence of monotonically improved solutions. The operating characteristics and performance of the algorithm are demonstratedvia numerical examples.     

Agent and data mining based decision support system and its adaptation to a new customer-centric electronic commerce

Recently, as the Internet has become more widely used, Electronic Commerce (EC) has emerged and has developed a high-level business environment. The customer-centric EC model is important for the success of EC and this study presents a new customer-centric EC model in make-to-order (MTO) semiconductor manufacturing environment. In this study we proposed the EC model providing the process transparency of process sampling method that can provide online semiconductor customers with the performance information of available process sampling methods which can be used at all manufacturing process steps for their own products in MTO manufacturing environment, and then the capability to select a desirable one among them based on their purchase situations on EC web site. In the proposed EC model the customer can select a process sampling method that is most suitable to him/her according to the customer's purchase situation. In this model the use of intelligent decision support system called customized sampling decision support system (CSDSS) that can autonomously generate available customized sampling methods and provide the performance information of those methods to EC system is requisite. We implemented an Internet-based prototype of CSDSS which had an architecture based on intelligent agent technology and also the successful integration of data mining process for the generation of optimal sampling method into DSS framework by means of applying that technology.     

A less conservative stability criterion for sampled‐data system via a fractional‐delayed state and its state‐space model

This paper introduces a state‐space model based on a fractional‐delayed state for a sampled‐data system described as a linear system under aperiodic sampling intervals. The new state‐space model exploits the information on the sampled‐data system and the fractional‐delayed state, which is defined on the sampling interval. The fractional‐delayed state and its state‐space model are utilized to construct the Lyapunov functional, which consists of the traditional Lyapunov function and a looped functional whose boundary conditions are zeros at sampling instants. Based on the Lyapunov functional, a stability criterion and a robust stability criterion are derived in terms of linear matrix inequalities. Simulation results show the effectiveness of the proposed criterion.     

Sequential exploration-exploitation with dynamic trade-off for efficient reliability analysis of complex engineered systems

A new sequential sampling method, named sequential exploration-exploitation with dynamic trade-off (SEEDT), is proposed for reliability analysis of complex engineered systems involving high dimensionality and a wide range of reliability levels. The proposed SEEDT method is built based on the ideas of two previously developed sequential Kriging reliability methods, namely efficient global reliability analysis (EGRA) and maximum confidence enhancement (MCE) methods. It employs Kriging-based sequential sampling to build a surrogate model (i.e., Kriging model) that approximates the performance function of an engineered system, and performs Monte Carlo simulation on the surrogate model for reliability analysis. A new acquisition function, referred to as expected utility (EU), is developed to sequentially locate a computationally efficient set of sample points for constructing the Kriging model. The SEEDT method possesses three technical contributions: (i) defining a new utility function with several desirable properties that facilitates the joint consideration of exploration and exploitation over the course of sequential sampling; (ii) introducing a new exploration-exploitation trade-off coefficient that dynamically weighs exploration and exploitation to achieve a fine balance between these two activities; and (iii) developing a new convergence criterion based on the uncertainty in the prediction of the limit-state function (LSF). The effectiveness of the proposed method in reliability analysis is evaluated with several mathematical and practical examples. Results from these examples suggest that, given a certain number of sample points, the SEEDT method is capable of achieving better accuracy in predicting the LSF than the existing sequential sampling methods.