In IBVS (image based visual servoing), the error signal in image space should be transformed into the control signal in the input space quickly. To avoid the iterative adjustment and complicated inverse solution of im...In IBVS (image based visual servoing), the error signal in image space should be transformed into the control signal in the input space quickly. To avoid the iterative adjustment and complicated inverse solution of image Jacobian, CMAC (cerebellar model articulation controller) neural network is inserted into visual servo control loop to implement the nonlinear mapping. Two control schemes are used. Simulation results on two schemes are provided, which show a better tracking precision and stability can be achieved using scheme 2.展开更多
A new visual servo control scheme for a robotic manipulator is presented in this paper, where a back propagation （BP） neural network is used to make a direct transition from image feature to joint angles without req...A new visual servo control scheme for a robotic manipulator is presented in this paper, where a back propagation （BP） neural network is used to make a direct transition from image feature to joint angles without requiring robot kinematics and camera calibration. To speed up the convergence and avoid local minimum of the neural network, this paper uses a genetic algorithm to find the optimal initial weights and thresholds and then uses the BP Mgorithm to train the neural network according to the data given. The proposed method can effectively combine the good global searching ability of genetic algorithms with the accurate local searching feature of BP neural network. The Simulink model for PUMA560 robot visual servo system based on the improved BP neural network is built with the Robotics Toolbox of Matlab. The simulation results indicate that the proposed method can accelerate convergence of the image errors and provide a simple and effective way of robot control.展开更多
An effective blind digital watermarking algorithm based on neural networks in the wavelet domain is presented. Firstly, the host image is decomposed through wavelet transform. The significant coefficients of wavelet a...An effective blind digital watermarking algorithm based on neural networks in the wavelet domain is presented. Firstly, the host image is decomposed through wavelet transform. The significant coefficients of wavelet are selected according to the human visual system （HVS） characteristics. Watermark bits are added to them. And then effectively cooperates neural networks to learn the characteristics of the embedded watermark related to them. Because of the learning and adaptive capabilities of neural networks, the trained neural networks almost exactly recover the watermark from the watermarked image. Experimental results and comparisons with other techniques prove the effectiveness of the new algorithm.展开更多
Unsupervised learning plays an important role in the neural networks. Focusing on the unsupervised mechanism of neural networks, a novel generalized goodness criterion for the unsupervised neural learning of visual pe...Unsupervised learning plays an important role in the neural networks. Focusing on the unsupervised mechanism of neural networks, a novel generalized goodness criterion for the unsupervised neural learning of visual perception based on the martingale measure is proposed in the paper. The differential geometrical structure is used as the framework of the whole inference and spatial statistical description with adaptive attribute is embedded in the corresponding nonlinear functional space. Consequently the integration of optimization process and computational simulation with the NeoDarwinian paradigm is obtained. And the generalization of the guidance for the evolutionary learning in the neural net framework, the convergence of the goodness and process of the evolution guaranteed by the mathematical features are discussed. This criterion has generic significance in the field of machine vision and visual pattern classification.展开更多
目的通过机器学习分析“舌边白涎”舌象特性,对舌象进行局部特征识别研究,探讨卷积神经网络算法在舌象识别应用中的性能。方法使用Python进行图像预处理,搭建用于舌象识别的视觉几何组16层(visual geometry group 16,VGG16)卷积神经网...目的通过机器学习分析“舌边白涎”舌象特性,对舌象进行局部特征识别研究,探讨卷积神经网络算法在舌象识别应用中的性能。方法使用Python进行图像预处理,搭建用于舌象识别的视觉几何组16层(visual geometry group 16,VGG16)卷积神经网络模型,分析其对“舌边白涎”舌象鉴别分析的效果,并结合热力图分析“舌边白涎”典型舌象表现。结果基于PyTorch框架,进行卷积神经网络的舌象鉴别研究,VGG16及残差网络50层(residual network 50,ResNet50)模型验证准确率均较高,达到80%以上,且ResNet50模型优于VGG16模型,可为舌象识别提供一定参考。基于加权梯度类激活映射(gradient-weighted class activation mapping,Grad-CAM)技术,通过舌苔舌色差异分布的网络可视化,有助于直观进行模型评估分析。结论基于卷积神经网络模型对舌象数据库进行分析,实现“舌边白涎”舌象识别,有助于临床诊疗的客观化辅助分析,为舌诊智能化发展提供一定借鉴。展开更多
This paper proposes a technique for synthesizing a pixel-based photo-realistic talking face animation using two-step synthesis with HMMs and DNNs. We introduce facial expression parameters as an intermediate represent...This paper proposes a technique for synthesizing a pixel-based photo-realistic talking face animation using two-step synthesis with HMMs and DNNs. We introduce facial expression parameters as an intermediate representation that has a good correspondence with both of the input contexts and the output pixel data of face images. The sequences of the facial expression parameters are modeled using context-dependent HMMs with static and dynamic features. The mapping from the expression parameters to the target pixel images are trained using DNNs. We examine the required amount of the training data for HMMs and DNNs and compare the performance of the proposed technique with the conventional PCA-based technique through objective and subjective evaluation experiments.展开更多
基金This project is supported by National Natural Science Foundation of China (No.59990470).
文摘In IBVS (image based visual servoing), the error signal in image space should be transformed into the control signal in the input space quickly. To avoid the iterative adjustment and complicated inverse solution of image Jacobian, CMAC (cerebellar model articulation controller) neural network is inserted into visual servo control loop to implement the nonlinear mapping. Two control schemes are used. Simulation results on two schemes are provided, which show a better tracking precision and stability can be achieved using scheme 2.
文摘A new visual servo control scheme for a robotic manipulator is presented in this paper, where a back propagation （BP） neural network is used to make a direct transition from image feature to joint angles without requiring robot kinematics and camera calibration. To speed up the convergence and avoid local minimum of the neural network, this paper uses a genetic algorithm to find the optimal initial weights and thresholds and then uses the BP Mgorithm to train the neural network according to the data given. The proposed method can effectively combine the good global searching ability of genetic algorithms with the accurate local searching feature of BP neural network. The Simulink model for PUMA560 robot visual servo system based on the improved BP neural network is built with the Robotics Toolbox of Matlab. The simulation results indicate that the proposed method can accelerate convergence of the image errors and provide a simple and effective way of robot control.
基金Supported by the National Natural Science Foun-dation of China ( 60473015)
文摘An effective blind digital watermarking algorithm based on neural networks in the wavelet domain is presented. Firstly, the host image is decomposed through wavelet transform. The significant coefficients of wavelet are selected according to the human visual system （HVS） characteristics. Watermark bits are added to them. And then effectively cooperates neural networks to learn the characteristics of the embedded watermark related to them. Because of the learning and adaptive capabilities of neural networks, the trained neural networks almost exactly recover the watermark from the watermarked image. Experimental results and comparisons with other techniques prove the effectiveness of the new algorithm.
文摘Unsupervised learning plays an important role in the neural networks. Focusing on the unsupervised mechanism of neural networks, a novel generalized goodness criterion for the unsupervised neural learning of visual perception based on the martingale measure is proposed in the paper. The differential geometrical structure is used as the framework of the whole inference and spatial statistical description with adaptive attribute is embedded in the corresponding nonlinear functional space. Consequently the integration of optimization process and computational simulation with the NeoDarwinian paradigm is obtained. And the generalization of the guidance for the evolutionary learning in the neural net framework, the convergence of the goodness and process of the evolution guaranteed by the mathematical features are discussed. This criterion has generic significance in the field of machine vision and visual pattern classification.
文摘目的通过机器学习分析“舌边白涎”舌象特性,对舌象进行局部特征识别研究,探讨卷积神经网络算法在舌象识别应用中的性能。方法使用Python进行图像预处理,搭建用于舌象识别的视觉几何组16层(visual geometry group 16,VGG16)卷积神经网络模型,分析其对“舌边白涎”舌象鉴别分析的效果,并结合热力图分析“舌边白涎”典型舌象表现。结果基于PyTorch框架,进行卷积神经网络的舌象鉴别研究,VGG16及残差网络50层(residual network 50,ResNet50)模型验证准确率均较高,达到80%以上,且ResNet50模型优于VGG16模型,可为舌象识别提供一定参考。基于加权梯度类激活映射(gradient-weighted class activation mapping,Grad-CAM)技术,通过舌苔舌色差异分布的网络可视化,有助于直观进行模型评估分析。结论基于卷积神经网络模型对舌象数据库进行分析,实现“舌边白涎”舌象识别,有助于临床诊疗的客观化辅助分析,为舌诊智能化发展提供一定借鉴。
文摘This paper proposes a technique for synthesizing a pixel-based photo-realistic talking face animation using two-step synthesis with HMMs and DNNs. We introduce facial expression parameters as an intermediate representation that has a good correspondence with both of the input contexts and the output pixel data of face images. The sequences of the facial expression parameters are modeled using context-dependent HMMs with static and dynamic features. The mapping from the expression parameters to the target pixel images are trained using DNNs. We examine the required amount of the training data for HMMs and DNNs and compare the performance of the proposed technique with the conventional PCA-based technique through objective and subjective evaluation experiments.
