基于PSO优化神经网络和空间网格的机器人位姿标定方法

该文提出一种将机器人的位置和姿态拆分开,分别进行标定的机器人位姿标定方法。采用空间精度控制网格标定机器人定位误差,粒子群优化算法(particle swarm optimization,PSO)优化神经网络标定机器人定姿误差。该方法以指数积公式(product of exponentials,POE)为基础建立机器人正向运动学模型,用映射法建立空间网格,用三坐标测量臂测量机器人位姿,用空间网格精度标定定位误差,用PSO优化的神经网络标定定姿误差。其优点在于既标定机器人工具中心点(TCP)的定位误差,又标定机器人工具坐标系的姿态误差,使得机器人定位、定姿误差都得到补偿。实验结果表明机器人的定位、定姿均方根误差减小接近一个数量级。     

Binocular vision system calibration based on a one-dimensional target

This paper proposes an adjustment method for binocular vision measurement to calibrate a camera's internal and external parameters based on a one dimensional (1D) target in the field of view. A 1D target with two feature points lying randomly in the field of view is used to get the images of the feature points. The distance between the two feature points is known. The internal and external parameters can be acquired by solving equations combining the photograph measurement collinear equations and the feature points' distance equations. To solve these equations, we use linearization of nonlinear equations and the adjustment method. During the process, we deal with the equations as measurement equations and the internal/external parameters and the 3D target points as the unknown parameters to calculate them. According to field experiment results, in about a 600 mm×600 mm field of view, the relative error of the distance of two points is less than two ten-thousandths, obtained by using the calculated results of the binocular vision system. The calibration process is simple, convenient, and suitable for calibrating a camera on the spot.     

基于机器视觉的神经网络在茶叶鉴别中的应用

名优茶(芽茶、一芽一叶、一芽两叶)在外观检测分类中,主要依靠人为的主观判断,这样既耗时又难以给出客观的评价标准。针对这个问题,提出一种基于机器视觉的分类方法,首先对原始图像进行剪裁、预处理,然后利用灰度共生矩阵和Tamura方法提取茶鲜叶图像的十二种纹理特征(粗糙度、对比度、方向度、平滑度、一致性等),并采用主成分分析方法建立更加有效精简的特征库,最后根据图像的特征利用BP神经网络对茶叶进行分类,仿真实验能够达到理想的效果。     

基于神经网络的地形匹配方法研究

提出了基于神经网络实现多特征融合的地形匹配算法，充分利用地形的各种不同的统计特征和几何特征，构造了一种地形匹配网络模型．通过对实时图和基准图的分析，给出了计算网络节点之间的权值函数，建立了网络系统能量方程，通过求系统的最小能量得到最佳匹配位置．由于网络能融合地形的不同统计特征和几何特征，所以算法大大提高了系统的抗干扰能力和定位精度，适合于实时图容易发生畸变的地形匹配领域．实验结果表明，定位精度和抗干扰能力均优于传统的地形匹配方法．     

线结构光视觉传感器机器人手眼关系标定

为实现对工业机器人手眼关系的标定,提出一种基于线结构光视觉传感器的手眼关系标定方法。该方法在标定时,将一个平面靶标作为参考物固定在工业机器人工作空间内,控制工业机器人末端运动以带动线结构光视觉传感器作多组变位姿运动,获取在不同位姿状态下的平面靶标图像并对其进行图像处理。通过对图像上固定特征点的测量,以及建立线结构光视觉传感器模型和手眼关系模型实现对线结构光内参数和手眼关系的标定。用棋盘格标定板进行测量实验验证,实验结果表明该方法准确度为0.036 mm,即优于40μm,可用于工业机器人的测量应用。     

基于神经网络的图像边缘检测方法

提出了一种基于神经网络的图像边缘检测新方法.该方法首先基于邻域灰度极值提取边界候选图像,然后以边界候选象素及其邻域象素的二值模式作为样本集,输入边缘检测神经网络进行训练.边缘检测神经网络采用BP网络,为加快网络的训练速度,采用了滚动训练和权值随机扰动的方法.实验表明,该方法提高了神经网络的学习效率,获得的边缘图像封闭性好,边缘描述真实.     

基于神经网络趋势分析

文章在分析研究了国内外现状的基础上 ,利用神经网络的非线性处理特性 ,提出了通过神经网络预测常见机械零件剩余寿命的方法 ,用实例验证了其有效性     

基于Kohonen神经网络算法的网络入侵聚类算法的测试研究

提出一种基于Kohonen网络的网络入侵聚类研究的方法,在阐述基本理论、原理和算法步骤基础上,利用Matlab软件平台对提出的网络入侵算法进行测试研究,并同其他方法进行仿真对比,发现Kohonen神经网络算法的网络入侵聚类在训练准确率、测试准确率和运行时间3个方面都优于PNN算法,其准确率可以达到98.1％.     

基于BP神经网络的导线脱冰跳跃高度预测模型

冰区导线脱冰振动会引起绝缘间隙减小,严重时甚至导致闪络和跳闸等电气事故.首先利用数值方法模拟得到各种参数条件下导线的脱冰动力响应,获得导线的最大脱冰跳跃高度.进而基于数值模拟结果和BP神经网络构建导线脱冰跳跃高度预测模型,将线路的导线分裂数、导线型号、档距、高差等结构参数以及初始应力、覆冰厚度和脱冰率等载荷参数作为输入...     

Modeling and multi-objective optimization of an M-cycle cross-flow indirect evaporative cooler using the GMDH type neural network

A model was presented to determine product air properties of dew-point indirect evaporative coolers with cross flow heat exchanger, M-cycle CrFIEC. In this regard, the most powerful statistical method known as the group method of data handling-type neural network (GMDH) was employed. Then the developed GMDH model was implemented for multi-objective optimization of a prototype CrFIEC and the average annual values of coefficient of performance (COP) and cooling capacity (CC) were maximized, simultaneously, while working to air ratio (WAR) and inlet air velocity were decision variables of optimization. Accordingly, features of the proposed system were optimized at twelve diverse climates of the world based on Koppen–Geiger's classification. Results implied that the optimized inlet air velocity for all climates varied between 1.796 and 1.957 m.s⁻¹, while the optimum WAR was 0.318 for “A” class cities. Moreover, the mean values of the COP and CC were improved 8.1% and 6.9%, respectively.     

Adaptable product configuration system based on neural network

In the rapidly diversifying and globalising market, product configuration is implemented in a dynamic environment with continuous change of configuration knowledge. The adaptability of the product configuration system, which is defined as the capability to adjust product configurator, human resources and organisational resources to fit a new environment, is becoming more and more crucial. To keep the adaptability, this research suggests an adaptable product configuration (APC) system which transforms the development of the configuration system in a dynamic environment from a straightforward process to a closed circle. In the existing research on product configuration, most issues are addressed separately by different approaches and most approaches lack a systematic view which considers the interaction among product configurator, resources and environment. The circle of APC is therefore divided into several isolated stages and involves intensive human work, consumes a lot of organisational resources and results in a long response time. To successfully implement APC, this research adopts an artificial neural network and a specific rule extraction mechanism to develop a product configuration system. The neural network is able to automatically acquire configuration knowledge from historical transaction data and then directly apply it without further knowledge programming. Rule extraction mechanism has the capability to interpret the behaviour of the trained neural network and make it comprehensible and adjustable. Finally, knowledge acquisition, representation and application in product configuration are incorporated into the same connectionist methodology. And consequently, the APC circle is accelerated and the adaptability of the product configuration system is improved. A case study of computer configuration is presented.     

基于卷积神经网络和Transformer网络的鸟声识别

针对传统鸟声识别算法中特征提取方式单一、分类识别准确率低等问题,提出一种结合卷积神经网络和Transformer网络的鸟声识别方法。该方法综合考虑网络局部特征学习和全局上下文依赖性构造,从原始鸟声音频信号中提取短时傅里叶变换(Short Time Fourier Transform,STFT)语谱图特征,将其输入到卷积神经网络(ConvolutionalNeural Network,CNN)中提取局部频谱特征信息,同时提取鸟声信号的对数梅尔特征及一阶差分、二阶差分特征用于合成梅尔频率倒谱系数(Mel Frequency Cepstrum Coefficient,MFCC)混合特征向量,将其输入到Transformer网络中获取全局序列特征信息,最后融合所提取的特征可得到更丰富的鸟声特征参数,通过Softmax分类器得到鸟声识别结果。在Birdsdata和xeno-canto鸟声数据集上进行实验,平均识别准确率分别达到了97.81%和89.47%。实验结果表明该方法相较于其他现有的鸟声识别模型具有更高的识别准确率。     

基于脉冲耦合神经网络的红外图像增强

在分析了脉冲耦合神经网络的工作机理和行为特性后,指出可以利用神经元的点火-熄灭特性对图像进行增强.为了区分神经元的点火方式,提出一种根据链接矩阵判定神经元点火方式的方法,并利用自然点火和捕获点火建立了能使图像得到增强的非线性映射.文中对算法参数的设置及其对增强图像的影响做了详细地讨论,实验结果表明该算法不仅能使图像的对比度和亮度得到适当的增强,而且能够有效地抑制图像中的椒盐噪声,尤其适用于对比度和亮度都较低的红外图像.     

Wavelet neural network (WNN) approach for calibration model building based on gasoline near infrared (NIR) spectra

In this paper we have compared the abilities of two types of artificial neural networks (ANN): multilayer perceptron (MLP) and wavelet neural network (WNN) — for prediction of three gasoline properties (density, benzene content and ethanol content). Three sets of near infrared (NIR) spectra (285, 285 and 375 gasoline spectra) were used for calibration models building. Cross-validation errors and structures of optimized MLP and WNN were compared for each sample set. Four different transfer functions (Morlet wavelet and Gaussian derivative – for WNN; logistic and hyperbolic tangent – for MLP) were also compared. Wavelet neural network was found to be more effective and robust than multilayer perceptron.     

Parallel-distributed blind deconvolution based on a self-organizing neural network

A parallel-distributed blind deconvolution method based on a self-organizing neural network is introduced. A large degraded image is segmented into smaller subpatterns. Each subpattern can be used to get a blur function. Moreover, we propose a two-step unsupervised learning method in the self-organizing neural network. The two-step learning method includes parallel learning and series learning operations. The series learning operation is similar to a typical learning operation in the self-organizing neural network. The parallel learning operation is used as a positive perturbation to let the learning operation leave a local minimum. Several improved blur functions can be estimated from the different subpatterns, and the optimized blur function is evolved by use of a genetic algorithm. As the blur function is estimated, the source image of the large degraded image can be easily restored by use of a Wiener-type filter or other deconvolution methods. Computer simulations show that the proposed parallel-distributed blind deconvolution method gives good reconstruction and that the two-step learning method in the self-organizing neural network can promote learning. Since the main computational cost is dependent on the size of the subpattern, the proposed method is effective for the restoration of the large image.     

Bayesian regularisation neural network based on artificial intelligence optimisation

Stock prediction is generally considered to be challenging and known for its high noise and strong nonlinearities in financial time series analysis. However, current forecasting models ignore the importance of model parameter optimisation and the use of recent data. In this article, a novel forecasting approach with a Bayesian-regularised artificial neural networks (BR-ANN) was proposed. The weight of the proposed model (BR-ANN) is determined by the particle swarm optimisation (PSO) algorithm. Daily market prices and financial technical indicators are utilised as inputs to predict the one day future closing price of the Shanghai (in China) composite index. The Bayesian-regularised network uses a probabilistic nature for the network weights and can reduce the potential for over-fitting and over-training. Our empirical study and the results of our K-line theory analysis indicate that PSO is determined to be an effective algorithm to optimise the parameters of the Bayesian neural network compared with other well-known prediction algorithms. In particular, the PSO model is more reliable than the simple Bayesian regularisation neural network near the local maximum value.     

Optical bipolar kth-order neural network based on inner-product representation

A new polarization-encoding scheme for a bipolar kth-order neural network based on inner-product representation is proposed. Bipolar data multiplication is achieved as the rotation of linearly polarized light. A compact architecture of the bipolar kth-order neural network is suggested. In the architecture, no subtractions are needed, and the threshold levels for the neurons are fixed. Also, we show that a liquid-crystal device such as a liquid-crystal television is acceptable as a polarization modulator in the proposed architecture by computer simulation.     

基于卷积神经网络的超声金属材料辨识

传统的金属材料辨识方法会给被检测样品带来一定程度的损伤。文章通过采集金属材料的超声回波时域信号,采用短时傅里叶变换对其进行时频分析,得到包含金属材料微观组织信息的超声时频谱。将目标样品的超声时频谱预处理后作为训练样本,输入到构建好的卷积神经网络中进行训练。然后采集目标样品以及干扰样品的超声频谱图,分别将其输入网络进行辨识。结果表明,神经网络在训练时收敛较快,损失函数在迭代200次后收敛,在经过100次迭代后训练集准确率趋于100%。训练完成的网络模型记录着对应训练样本的特征信息,利用该训练好的网络对待测样本进行辨识,最终可实现超声金属材料辨识。