关键词： information theory   pattern recognition theory   information quantity   information value   probability   fuzzy sets   fuzzy logic   queueing systems  

摘要： A new approach based on the fuzzy set theory is proposed, which allows one to quantify the value of information. Different approaches to definition and calculation of basic concepts of information theory are considered, in particular, amount of information and its evaluation based on statistical considerations (classical approach), theory of algorithms (algorithmic approach), and pattern recognition theory (image approach). Approaches to processing of fuzzy information under incomplete definition of the vector of input attributes based on fuzzy set theory are proposed. Their analysis is carried out, the limits of their use and the fields of efficient application are established.

关键词： Fiber-reinforced composites   Acoustic emission   Frequency component   Grinding damage pattern recognition   Effective voltage value   Event number percentage of AE peak frequency  

摘要： Current researches of acoustic emission (AE) mainly put the focus on fault diagnosis of traditional isotropic materials machining process or analysis of fiber-reinforced composite (FRC) tensile or bending strength, while there are merely studies on AE research of FRC grinding process. Numerous kinds of damage occur during FRC grinding process owing to their complicated structure. The main purpose of this paper is to extract proper index to estimate AE signals of FRC grinding and to recognize damage patterns, thus realizing FRC processing on-line detection. AE signals are obtained by single-grain grinding experiments of quartz fiber-reinforced silicon dioxide matrix composite (SiO₂/SiO₂). The AE signal features are discussed, and the outstanding character of frequency components is proposed. The frequency of each damage pattern is analyzed and verified. AE effective voltage value (EVV) and event number percentage (ENP) of peak frequency (PF) of AE signals with processing parameters are researched. The results show that for the same kind of FRCs, frequency components of AE signals are only affected by damage patterns rather than processing parameters or grinding directions, thus being a proper estimate index. There are four main frequency bands during SiO₂/SiO₂ grinding. The frequency 6.4-9.8 KHz corresponds to fiber fracture, 14.8-17.9 KHz is fiber debonding, 23.6-26.4 KHz is debris rubbing with workpiece and tool, and 34-35.5 KHz is matrix crack. EVV has a similar changing trend to grinding force with machining parameters. AE ENP of PF that the maximum peak amplitude (PA) corresponds to could quantitatively confirm the main damage modes under each processing condition.

关键词： Acoustic sensors   energy harvesting   event-driven sensors   microbial fuel cells (MFCs)   underwater monitoring  

摘要： Minimizing the power consumption of always-on sensors is crucial for extending the lifetime of battery-operated devices that are required to monitor events continuously and for long periods. This paper proposes a novel programmable mu W event-driven acoustic detector featuring "always-on" audio pattern recognition. The event-driven detector detects up to eight programmable spectral-temporal features extracted with a low-power single-channel analog circuit and classifies the features by an onboard microcontroller. The event-driven detector is combined with novel microbial fuel cells (MFCs) to achieve self-sustainability in an underwater scenario. Experimental results demonstrate that the power consumption of the detector is only 26.89 mu W during always-on mode, achieving up to 59-dB sound pressure level of sensitivity. High detection accuracy of up to 95.89% in recognizing acoustic patterns has been experimentally verified. Accurate measurements with commercial MFCs demonstrate the capability to achieve self-sustainability in always-on monitoring.

关键词： Manufacturing process   Process pattern recognition   Deep learning   Stacked denoising autoencoder   Feature learning  

摘要： Recognition of various defect patterns exhibited in discrete manufacturing processes can significantly reduce the diagnostic processes, and increase manufacturing process stability and quality. Thus the effective recognizers are in great demand to improve the performance of process pattern recognition (PPR). Deep learning has been widely applied in image and visual analysis with great successes. However, the application of deep learning in feature learning for process control is still few. This paper presents an effective and reliable deep learning method known as stacked denoising autoencoder (SDAE) for PPR in manufacturing processes. This paper will concentrate on developing an SDAE model to learn effective features from the process signals and then implementing an effective PPR through a deep network architecture. Feature visualization is also performed to explicitly present the feature representation of the proposed SDAE model. The effectiveness of the proposed PPR method is verified through a big simulation dataset and Tennessee Eastman process. The result shows that the proposed method obtains good feature learning and PPR performance. (C) 2019 Elsevier Ltd. All rights reserved.

关键词： Computer vision   Pattern recognition   Principal component analysis   Fourier transform   Precision agriculture  

摘要： Automatic locating of weeds from fields is an active research topic in precision agriculture. A reliable and practical plant identification technique would enable the reduction of herbicide amounts and lowering of production costs, along with reducing the damage to the ecosystem. When the seeds have been sown row-wise, most weeds may be located between the sowing rows. The present work describes a clustering-based method for recognition of plantlet rows from a set of aerial photographs, taken by a drone flying at approximately ten meters. The algorithm includes three phases: segmentation of green objects in the view, feature extraction, and clustering of plants into individual rows. Segmentation separates the plants from the background. The main feature to be extracted is the center of gravity of each plant segment. A tentative clustering is obtained piecewise by applying the 2D Fourier transform to image blocks to get information about the direction and the distance between the rows. The precise sowing line position is finally derived by principal component analysis. The method was able to find the rows from a set of photographs of size 1452 x 969 pixels approximately in 0.11 s, with the accuracy of 94 per cent.

关键词： Hierarchical clustering   Levenshtein distance   pattern recognition   seismogram   wavelet  

摘要： In a seismogram, there exist many kinds of wavelets. The reflected wavelet from the gas sand zone has a different shape with other layers. Usually, the information of each wavelet is weak and unknown, and the unsupervised classification method is applied to the clustering of the wavelets. Using the shape structure of the wavelet, syntactic pattern recognition is applied to the clustering. The extracted wavelets can be represented as strings of symbols. Levenshtein distance is used to calculate the distance between the two strings. Bottom-up and top-down hierarchical clustering methods are used in the construction of the dendrogram. The top-down hierarchical clustering by the recursive method is proposed. A new pseudo F-statistics is proposed to decide the optimal number of clusters. From the experimental results in simulated and real seismograms, the wavelets on the gas sand zone can be detected successfully. It can improve the seismic interpretation.

关键词： Mt.Etna volcano   Flank eruption   Monitoring data complex system   Neural networks   Supervised pattern recognition  

摘要： We investigated the relationship between the temporal monitoring series routinely recorded at Mt. Etna and the flank eruptions that occurred between January 2001 and April 2005 by the K-contractive map (K-CM) method pattern classifier with supervised learning. The reference dataset includes 28 variables and 1580 records collected over 52months for a total of 301 eruptive days. A two-step analysis was performed. In the first step analysis, we used the 28 parameters of each day to recognize anomalies heralding a flank eruption. K-CM estimated a sensitivity higher than 95% and a specificity close to 100%. In the second step analysis, we considered each record comprising the 28 variables for 6days as an input (for a total of 180 inputs) and the outcomes of the seventh day as an output to predict eruption or rest. In this case, K-CM showed sensitivity and specificity close to 98% and 100%, respectively. Results highlight the reliability of the K-CM method to build up a prediction algorithm able to alert the volcano experts a day before the occurrence of a potential flank eruption. The robustness of the two analyses was investigated by the behavior of the receiver operating characteristic curve. The relative area under the curve showed values close to 1, thus providing a valid measure of the performance of the classifier. Finally, a complete overview of the performance levels of the method used was explored analyzing the retrieved Molchan error diagram, in both cases, trajectories very close to the theoretical minimum.

关键词： Cyber-attack detection   SCADA systems   Pattern recognition   Cyber-physical security   Data-driven  

摘要： Critical infrastructures crucial to our modern life, such as electricity grids and water pumps, are controlled by Supervisory Control and Data Acquisition (SCADA) systems. Over the last two decades, connecting critical infrastructures to the Internet has become essential due to performance and commercial needs. The combination of Internet connections to systems with little if any security features and the fact that security by obscurity is not working anymore, has moved the topic of SCADA security into the forefront in the last few years. To address these challenges, in this paper we propose cyber-attack detection techniques based on temporal pattern recognition. Temporal pattern recognition methods do not only look for anomalies in the data transferred by the SCADA components over the network but also look for anomalies that can occur by misusing legitimate commands such that unauthorized and incorrect time intervals between them may cripple the system. Specifically, we propose two algorithms based on Hidden Markov Models (HMM) and Artificial Neural Networks (ANN). We evaluate the algorithms on real and simulated SCADA data with five different feature extraction methods;in each method, the algorithms consider different aspects of the raw data. The results show that temporal pattern recognition methods, especially those based on time feature extraction, can detect cyber-attacks, including those that involve legitimate functions, which are known in the literature as hard to detect. (C) 2019 Elsevier Ltd. All rights reserved.

关键词： 人工智能   模式识别   料线时序频谱的识别   全卷积神经网络   候选框的尺寸设置  

摘要： 针对雷达信号频谱中真实料面回波分量对应时序频谱的识别问题,使用改进后的Faster R-CNN来识别料线.将Faster R-CNN检测模型与3种不同特征提取的全卷积神经网络相结合,选择出最优特征提取网络,并结合正负样本均衡理论完成区域建议候选框的尺寸设置.结果显示,相比料线分割算法,料线检测的实时性和准确率更高.

关键词： Data handling  

摘要： "A time-series based pattern recognition method to analyze rollover kinematics in running." Footwear Science, 11(sup1), pp. S162–S163