电液伺服加载系统的神经网络逆控制

针对液压加载系统及试件的特殊要求,采用基于径向基函数网络(RBFN)的逆控制器,给出了系统NARMA模型及逆模型存在的条件。该方法使用实际系统的输入输出信号以及径向基函数网络实现系统建模,并利用系统神经网络模型离线训练系统的逆动态作为控制器,以克服由于实际系统所受的扰动而可能引起的控制器(逆模型)不收敛。实际控制表明该系统对期望加载轨线具有良好的跟踪能力,同时对系统干扰和不确定性具有较强的鲁棒性。

3-D visual tracking based on CMAC neural network and Kalman filter

In this paper, the Kalman filter is used to predict image feature positionaround which an image-processing window is then established to diminish feature-searching area andto heighten the image-processing speed. According to the fundamentals of image-based visual servoing(IBVS), the cerebellar model articulation controller (CMAC) neural network is inserted into thevisual servo control loop to implement the nonlinear mapping from the error signal in the imagespace to the control signal in the input space instead of the iterative adjustment and complicatedinverse solution of the image Jacobian. Simulation results show that the feature point can bepredicted efficiently using the Kalman filter and on-line supervised learning can be realized usingCMAC neural network; end-effector can track the target object very well.

A method for workpiece surface small-defect detection based on CutMix and YOLOv3

Surface small defects are often missed and incorrectly detected due to their small quantity and unapparent visual features.A method named CSYOLOv3,which is based on CutMix and YOLOv3,is proposed to solve such a problem.First,a four-image CutMix method is used to increase the small-defect quantity,and the process is dynamically adjusted based on the beta distribution.Then,the classic YOLOv3 is improved to detect small defects accurately.The shallow and large feature maps are split,and several of them are merged with the feature maps of the predicted branch to preserve the shallow features.The loss function of YOLOv3 is optimized and weighted to improve the attention to small defects.Finally,this method is used to detect 512×512 pixel images under RTX 2060Ti GPU,which can reach the speed of 14.09 frame/s,and the mAP is 71.80%,which is 5%-10%higher than that of other methods.For small defects below 64×64 pixels,the mAP of the method reaches 64.15%,which is 14%higher than that of YOLOv3-GIoU.The surface defects of the workpiece can be effectively detected by the proposed method,and the performance in detecting small defects is significantly improved.

Vision-based behavior prediction of ball carrier in basketball matches

A new vision-based approach was presented for predicting the behavior of the ball carrier—shooting, passing and dribbling in basketball matches. It was proposed to recognize the ball carrier’s head pose by classifying its yaw angle to determine his vision range and the court situation of the sportsman within his vision range can be further learned. In basketball match videos characterized by cluttered background, fast motion of the sportsmen and low resolution of their head images, and the covariance descriptor, were adopted to fuse multiple visual features of the head region, which can be seen as a point on the Riemannian manifold and then mapped to the tangent space. Then, the classification of head yaw angle was directly completed in this space through the trained multiclass LogitBoost. In order to describe the court situation of all sportsmen within the ball carrier’s vision range, artificial potential field (APF)-based information was introduced. Finally, the behavior of the ball carrier—shooting, passing and dribbling, was predicted using radial basis function (RBF) neural network as the classifier. Experimental results show that the average prediction accuracy of the proposed method can reach 80% on the video recorded in basketball matches, which validates its effectiveness.

Automated retinal blood vessels segmentation based on simplified PCNN and fast 2D-Otsu algorithm

According to the characteristics of dynamic firing in pulse coupled neural network (PCNN) and regional configuration in retinal blood vessel network, a new method combined with simplified PCNN and fast 2D-Otsu algorithm was proposed for automated retinal blood vessels segmentation. Firstly, 2D Gaussian matched filter was used to enhance the retinal images and simplified PCNN was employed to segment the blood vessels by firing neighborhood neurons. Then, fast 2D-Otsu algorithm was introduced to search the best segmentation results and iteration times with less computation time. Finally, the whole vessel network was obtained via analyzing the regional connectivity. Experiments implemented on the public Hoover database indicate that this new method gets a 0.803 5 true positive rate and a 0.028 0 false positive rate on an average. According to the test results, compared with Hoover algorithm and method of PCNN and 1D-Otsu, the proposed method shows much better performance.

Neighborhood fusion-based hierarchical parallel feature pyramid network for object detection

In order to improve the detection accuracy of small objects,a neighborhood fusion-based hierarchical parallel feature pyramid network(NFPN)is proposed.Unlike the layer-by-layer structure adopted in the feature pyramid network(FPN)and deconvolutional single shot detector(DSSD),where the bottom layer of the feature pyramid network relies on the top layer,NFPN builds the feature pyramid network with no connections between the upper and lower layers.That is,it only fuses shallow features on similar scales.NFPN is highly portable and can be embedded in many models to further boost performance.Extensive experiments on PASCAL VOC 2007,2012,and COCO datasets demonstrate that the NFPN-based SSD without intricate tricks can exceed the DSSD model in terms of detection accuracy and inference speed,especially for small objects,e.g.,4%to 5%higher mAP(mean average precision)than SSD,and 2%to 3%higher mAP than DSSD.On VOC 2007 test set,the NFPN-based SSD with 300×300 input reaches 79.4%mAP at 34.6 frame/s,and the mAP can raise to 82.9%after using the multi-scale testing strategy.

Color Perception of the Textile and Clothing

The color perception is related to color objects, vision system and central nervous system. The methods of evaluation, classification and prediction of the color perception are investigated and analyzed in this paper. The artificial neural networks are used for color perception, clustering and predicting based on the given data obtained from both objective measurement and subjective evaluation.

Robustness Design for CNN Templates with Performance of Extracting Closed Domain

The cellular neural/nonlinear network （CNN） is a powerful tool for image and video signal processing, robotic and biological visions. This paper introduces a kind of CNNs with performance of extracting closed domains in binary images, and gives a general method for designing templates of such a kind of CNNs. One theorem provides parameter inequalities for determining parameter intervals for implementing prescribed image processing functions, respectively. Examples for extracting closed domains in binary scale images are given.

Automated visual inspection of surface defects based on compound moment invariants and support vector machine

The traditional inspection methods are mostly based on manual inspection which is very likely to make erroneous judgments due to personal subjectivity or eye fatigue, and can＇t satisfy the accuracy. To overcome these difficulties, we develop a machine vision inspection system. We first compare several kinds of methods for feature extraction and classification, and then present a real-time automated visual inspection system for copper strips surface （CSS） defects based on compound moment invariants and support vector machine （SVM）. The proposed method first processes images collected by hardware system, and then extracts feature characteristics based on grayscale characteristics and morphologic characteristics （Hu and Zernike compound moment invariants）. Finally, we use SVM to classify the CSS defects. Furthermore, performance comparisons among SVM, back propagation （BP） and radial basis function （RBF） neural networks have been involved. Experimental results show that the proposed approach achieves an accuracy of 95.8% in detecting CSS defects.

Visualization of flatness pattern recognition based on T-S cloud inference network

Flatness pattern recognition is the key of the flatness control. The accuracy of the present flatness pattern recognition is limited and the shape defects cannot be reflected intuitively. In order to improve it, a novel method via T-S cloud inference network optimized by genetic algorithm(GA) is proposed. T-S cloud inference network is constructed with T-S fuzzy neural network and the cloud model. So, the rapid of fuzzy logic and the uncertainty of cloud model for processing data are both taken into account. What's more, GA possesses good parallel design structure and global optimization characteristics. Compared with the simulation recognition results of traditional BP Algorithm, GA is more accurate and effective. Moreover, virtual reality technology is introduced into the field of shape control by Lab VIEW, MATLAB mixed programming. And virtual flatness pattern recognition interface is designed.Therefore, the data of engineering analysis and the actual model are combined with each other, and the shape defects could be seen more lively and intuitively.

Wear Debris Identification Using Feature Extraction and Neural Network

A method and results of identification of wear debris using their morphological features are presented. The color images of wear debris were used as initial data. Each particle was characterized by a set of numerical parameters combined by its shape, color and surface texture features through a computer vision system. Those features were used as input vector of artificial neural network for wear debris identification. A radius basis function (RBF) network based model suitable for wear debris recognition was established, and its algorithm was presented in detail. Compared with traditional recognition methods, the RBF network model is faster in convergence, and higher in accuracy.

Finding Consistency in Rousseau

Several of Rousseau＇s critics begin with the presupposition that his writings are inconsistent or incoherent ann proceed to locate the ＂essence＂ of his philosophy in some of his writings while excluding others. Ernst Cassirer is among the few philosophers who have attempted to defend Rousseau＇s claim to consistency. Despite its broad influence, Cassirer＇s interpretation has remained largely unchallenged. The aim of this paper is twofold. Firstly, it aims to show that Cassirer＇s interpretation undermines （1） the important role Rousseau assigns to pity in both the state of nature and civil society and （2） the significant role the general will plays in his political theory. Secondly, it proposes an alternative interpretation that succeeds in uniting Rousseau＇s opus.

Research on High Resolution Satellite Image Classification Algorithm based on Convolution Neural Network

Artifi cial neural network is a kind of artificial intelligence method to simulate the function of human brain, and deep learning technology can establish a depth network model with hierarchical structure on the basis of artificial neural network. Deep learning brings new development direction to artificial neural network. Convolution neural network is a new artificial neural network method, which combines artificial neural network and deep learning technology, and this new neural network is widely used in many fields of computer vision. Modern image recognition algorithm requires classifi cation system to adapt to different types of tasks, and deep network and convolution neural network is a hot research topic in neural networks. According to the characteristics of satellite digital image, we use the convolution neural network to classify the image, which combines texture features with spectral features. The experimental results show that the convolution neural network algorithm can effectively classify the image.

Visual interpretability for deep learning：a survey

This paper reviews recent studies in understanding neural-network representations and learning neural networks with interpretable/disentangled middle-layer representations.Although deep neural networks have exhibited superior performance in various tasks,interpretability is always Achilles' heel of deep neural networks.At present,deep neural networks obtain high discrimination power at the cost of a low interpretability of their black-box representations.We believe that high model interpretability may help people break several bottlenecks of deep learning,e.g.,learning from a few annotations,learning via human–computer communications at the semantic level,and semantically debugging network representations.We focus on convolutional neural networks(CNNs),and revisit the visualization of CNN representations,methods of diagnosing representations of pre-trained CNNs,approaches for disentangling pre-trained CNN representations,learning of CNNs with disentangled representations,and middle-to-end learning based on model interpretability.Finally,we discuss prospective trends in explainable artificial intelligence.

Data mining-based detection of acupuncture treatment on juvenile myopia

OBJECTIVE:We applied data mining techniques to the study of acupuncture as a treatment for juvenile myopia,with the aim of identifying hidden patterns in the data.METHODS:Fifty patients with juvenile myopia were selected and treated with acupuncture,and data mining was used to analyze the effects of treatment and the influence of behavioral variables.Clustering analysis was used to divide myopia patients into two classifications before acupuncture treatment.Artificial neural network BP algorithm was adopted to analyze the roles of different factors in changes in diopters.An association algorithm was used to analyze factors associated with the subjective experience of acupuncture and average diopter.RESULTS:The two classification results were fully consistent with the understandings of the ophthalmic circles.The duration of using the Internet and watching TV every day was the main factor that affected vision.Acupuncture feelings and therapeutic effect have a strong correlativity.A good or above experience's score of acupuncture could slow the progression of juvenile myopia.CONCLUSION:Collecting data from patients with juvenile myopia by using data mining can extract hidden potential rules and knowledge from the research evidence.The decision support can be provided to improve the doctor's clinical acupuncture treatment effects.

Prediction of shelled shrimp weight by machine vision

The weight of shelled shrimp is an important parameter for grading process.The weight prediction of shelled shrimp by contour area is not accurate enough because of the ignorance of the shrimp thickness.In this paper,a multivariate prediction model containing area,perimeter,length,and width was established.A new calibration algorithm for extracting length of shelled shrimp was proposed,which contains binary image thinning,branch recognition and elimination,and length reconstruction,while its width was calculated during the process of length extracting.The model was further validated with another set of images from 30 shelled shrimps.For a comparison purpose,artificial neural network(ANN) was used for the shrimp weight predication.The ANN model resulted in a better prediction accuracy(with the average relative error at 2.67%),but took a tenfold increase in calculation time compared with the weight-area-perimeter(WAP) model(with the average relative error at 3.02%).We thus conclude that the WAP model is a better method for the prediction of the weight of shelled red shrimp.