Vector Extraction Research Articles

Ensemble of multiple kNN classifiers for societal risk classification

Societal risk classification is a fundamental and complex issue for societal risk perception. To conduct societal risk classification, Tianya Forum posts are selected as the data source, and four kinds of representations: string representation, term-frequency representation, TF-IDF representation and the distributed representation of BBS posts are applied. Using edit distance or cosine similarity as distance metric, four k-Nearest Neighbor (kNN) classifiers based on different representations are developed and compared. Owing to the priority of word order and semantic extraction of the neural network model Paragraph Vector, kNN based on the distributed representation generated by Paragraph Vector (kNN-PV) shows effectiveness for societal risk classification. Furthermore, to improve the performance of societal risk classification, through different weights, kNN-PV is combined with other three kNN classifiers as an ensemble model. Through brute force grid search method, the optimal weights are assigned to different kNN classifiers. Compared with kNN-PV, the experimental results reveal that Macro-F of the ensemble method is significantly improved for societal risk classification.

The Bivariate Empirical Mode Decomposition and Its Contribution to Grinding Chatter Detection

Grinding chatter reduces the long-term reliability of grinding machines. Detecting the negative effects of chatter requires improved chatter detection techniques. The vibration signals collected from grinders are mainly nonstationary, nonlinear and multidimensional. Hence, bivariate empirical mode decomposition (BEMD) has been investigated as a multiple signal processing method. In this paper, a feature vector extraction method based on BEMD and Hilbert transform was applied to the problem of grinding chatter. The effectiveness of this method was tested and validated with a simulated chatter signal produced by a vibration signal generator. The extraction criterion of true intrinsic mode functions (IMFs) was also investigated, as well as a method for selecting the most ideal number of projection directions using the BEMD algorithm. Moreover, real-time variance and instantaneous energy were employed as chatter feature vectors for improving the prediction of chatter. Furthermore, the combination of BEMD and Hilbert transform was validated by experimental data collected from a computer numerical control (CNC) guideway grinder. The results reveal the good behavior of BEMD in terms of processing nonstationary and nonlinear signals, and indicating the synchronous characteristics of multiple signals. Extracted chatter feature vectors were demonstrated to be reliable predictors of early grinding chatter.

An identification framework for print-scan books in a large database

In this paper, we propose an identification framework to determine copyright infringement in the form of illegally distributed print-scan books in a large database. The framework contains following main stages: image pre-processing, feature vector extraction, clustering, and indexing, and hierarchical search. The image pre-processing stage provides methods for alleviating the distortions induced by a scanner or digital camera. From the pre-processed image, we propose to generate feature vectors that are robust against distortion. To enhance the clustering performance in a large database, we use a clustering method based on the parallel-distributed computing of Hadoop MapReduce. In addition, to store the clustered feature vectors efficiently and minimize the searching time, we investigate an inverted index for feature vectors. Finally, we implement a two-step hierarchical search to achieve fast and accurate on-line identification. In a simulation, the proposed identification framework shows accurate and robust in the presence of print-scan distortions. The processing time analysis in a parallel computing environment gives extensibility of the proposed framework to massive data. In the matching performance analysis, we empirically and theoretically find that in terms of query time, the optimal number of clusters scales with O(N) for N print-scan books.

Vision algorithms for fixed-wing unmanned aerial vehicle landing system

Autonomous landing has become a core technology of unmanned aerial vehicle (UAV) guidance, navigation and control system in recent years. As a novel autonomous landing approach, computer vision has been studied and applied in rotary-wing UAV landing successfully. This paper aims to fixed-wing UAV and focus on two problems: how to find runway only depending on airborne front-looking camera and how to align UAV with the designated landing runway. The paper can be divided into two parts to solve above two problems respectively. In the first part, the paper firstly presents an algorithm of region of interest (ROI) detection, which is based on spectral residual saliency map, and then an algorithm of feature vector extraction based on sparse coding and spatial pyramid matching (SPM) is proposed, finally, ROI including designated landing runway is recognized by a linear support vector machine. In the second part, the paper presents an approach of relative position and pose estimation between UAV and landing runway. Estimation algorithm firstly selects five feature points on the runway surface, and then establishes a new earth-fixed reference frame, finally uses orthogonal iteration to estimate landing parameters including three parameters of distance, height and offset, and three pose parameters of roll, yaw, pitch. The experimental results verify the effectiveness of the algorithms proposed in this paper.

Multi-Resolution Feature Fusion model for coal rock burst hazard recognition based on Acoustic Emission data

Abstract The Acoustic Emission (AE) technique is an important nondestructive testing method used to detect coal rock burst hazards in coal mine strata to ensure the safety of the lives of miners. It is, however, still difficult to accurately detect coal rock bursts due to the complex mechanism of AE propagation underground. Therefore, this study proposes a new Multi-Resolution Feature Fusion SVM recognition (MRFF-SVM) approach to compute a comprehensive feature vector for coal rock burst hazard recognition and forecasting. The proposed approach contains the following three improved processes: the Coiflet Wavelet Transform (CWT) to decompose AE waveforms into multiple perspectives for feature vectors extraction, the Multi-Resolution Feature Fusion (MRFF) method to fuse these selected feature vectors into an enhanced MRFF feature vector and the following Support Vector Machines (SVM) to recognize the coal rock burst conditions. Several innovative experiments are carried out to evaluate the coal rock conditions according to the safe, crisis and burst categories. The results indicate that the proposed MRFF feature vectors can retain AE signatures well to recognize coal rock conditions and forecast sandstone failure. Thus, MRFF-SVM provides an effective analysis tool for coal rock burst hazard detection.

Novel feature extraction technique for content-based image recognition with query classification

A surge of concern has been witnessed to manage the growing size of image information available from various sources namely internet and digital image capturing devices. The rich content of information available with image data has proved to be useful for analytical decision-making process. Content-based image recognition has been considered as an effective measure to identify the object of interest. The success of aforementioned procedure has largely been influenced by the method of feature extraction from the image content. Image binarisation has proved to be an efficient tool for feature vector extraction using various threshold selection techniques. The authors have proposed a novel feature extraction technique based on local threshold selection and have evaluated the technique on 17,021 images for performance assessment. The precision results for classification and retrieval have shown an increment of 17% and 13.1% respectively when compared to state-of-the-art techniques. A statistical test has also been conducted to establish the significance of the proposed method over the existing techniques.

Low-dimensional Feature Vector Extraction from Motion Capture Data by Phase Plane Analysis

Low-dimensional Feature Vector Extraction from Motion Capture Data by Phase Plane Analysis

EMD技术在船舶液压设备故障诊断中的应用研究

本文在分析传统液压设备故障诊断技术及信号处理方法在船舶液压系统故障诊断方面存在不足的基础上，论述了振动分析在船舶液压系统故障诊断领域应用的可行性及前景，分析了经验模态分解(EMD)技术在处理非平稳信号方面的工程应用优势；分别以经验模态分解算法和径向基函数网络为基础，结合船舶液压系统应用实际，构建了船舶液压系统振动信号特征向量提取模型和故障诊断模型；最后以某型船艇装备液压舵机系统实测数据为基础，对构建的特征向量提取模型和故障诊断模型进行了验证。 Based on the analysis of the shortcomings of the traditional hydraulic equipment fault diagnosis technology and signal processing signal processing methods in the field of marine hydraulic system fault diagnosis, the feasibility and prospect of vibration analysis method in the field of marine hy-draulic system fault diagnosis are discussed, the advantages of EMD in dealing with non-stationary signals are analyzed. On the basis of EMD and RBF network, combined with the application of ma-rine hydraulic system, the vibration signal characteristic vector extraction model and fault diagno-sis model of marine hydraulic system are constructed. Finally, the model of characteristic vector extraction and fault diagnosis model are validated on the basis of measured data of hydraulic steering gear equipped with a certain type of watercraft.

Based On Intrinsic Mode Function Energy Tracking Method of Circuit Breaker Vibration Signal Feature Extraction Studies

In order to detect a mechanical type of structural failure of the circuit breaker, the characteristics of the circuit breaker mechanical vibration signal is analyzed in this paper. A combination of medium voltage circuit breaker based on empirical mode decomposition (EMD) amount of energy and support vector machine (SVM) theory vibration signal feature vector extraction and analysis of fault classification method is proposed. First, the vibration signal of the circuit breaker is decomposed by EMD, then intrinsic mode function (IMF) is obtain. The major fault feature information intrinsic mode functions the amount of energy of the component is obtained by discrete sampling points and the amount of energy. Using the amount of energy of IMF component as a feature vector, the failure of the test sample signal as input feature vector into trained “BT-SVM” support vector machine classification mechanism for fault classification. The differences and fault type of vibration signals can be identified by this method through the experimental analysis.

Harmonic Line Association 기반 특징벡터 추출에 의한 드론 음향 식별 및 분류

Harmonic Line Association 기반 특징벡터 추출에 의한 드론 음향 식별 및 분류

Precise on-machine extraction of the surface normal vector using an eddy current sensor array

To satisfy the requirements of on-machine measurement of the surface normal during complex surface manufacturing, a highly robust normal vector extraction method using an Eddy current (EC) displacement sensor array is developed, the output of which is almost unaffected by surface brightness, machining coolant and environmental noise. A precise normal vector extraction model based on a triangular-distributed EC sensor array is first established. Calibration of the effects of object surface inclination and coupling interference on measurement results, and the relative position of EC sensors, is involved. A novel apparatus employing three EC sensors and a force transducer was designed, which can be easily integrated into the computer numerical control (CNC) machine tool spindle and/or robot terminal execution. Finally, to test the validity and practicability of the proposed method, typical experiments were conducted with specified testing pieces using the developed approach and system, such as an inclined plane and cylindrical and spherical surfaces.

Deep learning for constructing microblog behavior representation to identify social media user’s personality

Due to the rapid development of information technology, the Internet has gradually become a part of everyday life. People would like to communicate with friends to share their opinions on social networks. The diverse behavior on socials networks is an ideal reflection of users’ personality traits. Existing behavior analysis methods for personality prediction mostly extract behavior attributes with heuristic analysis. Although they work fairly well, they are hard to extend and maintain. In this paper, we utilize a deep learning algorithm to build a feature learning model for personality prediction, which could perform an unsupervised extraction of the Linguistic Representation Feature Vector (LRFV) activity without supervision from text actively published on the Sina microblog. Compared with other feature extractsion methods, LRFV, as an abstract representation of microblog content, could describe a user’s semantic information more objectively and comprehensively. In the experiments, the personality prediction model is built using a linear regression algorithm, and different attributes obtained through different feature extraction methods are taken as input of the prediction model, respectively. The results show that LRFV performs better in microblog behavior descriptions, and improves the performance of the personality prediction model.

발화구간 검출을 위해 학습된 CNN 기반 입 모양 인식 방법

소음환경에서의 음성인식 문제점으로 인해 1990년대 중반부터 음성정보와 영양정보를 결합한 AVSR(Audio Visual Speech Recognition) 시스템이 제안되었고, Lip Reading은 AVSR 시스템에서 시각적 특징으로 사용되었다. 본 연구는 효율적인 AVSR 시스템을 구축하기 위해 입 모양만을 이용한 발화 단어 인식률을 극대화하는데 목적이 있다. 본 연구에서는 입 모양 인식을 위해 실험단어를 발화한 입력 영상으로부터 영상의 전처리 과정을 수행하고 입술 영역을 검출한다. 이후 DNN(Deep Neural Network)의 일종인 CNN(Convolution Neural Network)을 이용하여 발화구간을 검출하고, 동일한 네트워크를 사용하여 입 모양 특징 벡터를 추출하여 HMM(Hidden Markov Mode)으로 인식 실험을 진행하였다. 그 결과 발화구간 검출 결과는 91%의 인식률을 보임으로써 Threshold를 이용한 방법에 비해 높은 성능을 나타냈다. 또한 입모양 인식 실험에서 화자종속 실험은 88.5%, 화자 독립 실험은 80.2%로 이전 연구들에 비해 높은 결과를 보였다. Due to speech recognition problems in noisy environment, Audio Visual Speech Recognition (AVSR) system, which combines speech information and visual information, has been proposed since the mid-1990s,. and lip reading have played significant role in the AVSR System. This study aims to enhance recognition rate of utterance word using only lip shape detection for efficient AVSR system. After preprocessing for lip region detection, Convolution Neural Network (CNN) techniques are applied for utterance period detection and lip shape feature vector extraction, and Hidden Markov Models (HMMs) are then used for the recognition. As a result, the utterance period detection results show 91% of success rates, which are higher performance than general threshold methods. In the lip reading recognition, while user-dependent experiment records 88.5%, user-independent experiment shows 80.2% of recognition rates, which are improved results compared to the previous studies.

Extraction of Effective Motion Vectors Based on Morphological Feature and Thinning Algorithm

This research proposed an algorithm conducting context awareness and tracking of a moving object. The research process is as follows: First, this study carried out preprocessing, such as binary coding, noise elimination and opening-closing arithmetic operations, after acquiring background subtraction image using BSM for moving object extraction. Second, this research eliminated the shadow of a moving object using three characteristics, namely chromatic image, brightness change and shadow direction in order to eliminate shadows, which a negative factor for image analysis. Third, this research calculated the skeleton line of a moving object by applying the thinning algorithm. Then, this research perceived the rapidly changing context of the moving object using the slope of the skeleton line. Lastly, the center point of the moving object was found by applying the pruning algorithm to the skeleton line, while it created moving direction and moving trajectory by comparing each frame central point of the moving object. As a result of the experiment, the rapid movement change of a moving object was extracted, and the moving direction was found through a central point comparison, while the method proposed by creating a moving trajectory was confirmed to be valid.

SIFT특징치를 이용한 다국적 지폐의 계층적 인식 알고리즘에 관한 연구

본 연구에서는 물체 인식 분야에서 잘 알려진 회전, 스케일, 조명의 변화에 강인한 특징치인 SIFT를 이용하여 지폐의 특징 벡터를 구하고 이를 ANN알고리즘에 의해 정합하여 다국적 지폐를 인식하는 방법에 관한 것으로 계층적 지폐인식 방법을 제안한다. 지폐마다 지니고 있는 특징치를 추출하여 국적 및 권종을 인식하기 위하여 자외선, 적외선, 및 백색 투과광 조명을 개발하고 조명 변화에 따라 촬영된 영상으로부터 SIFT특징치를 구하고 다국적 지폐의 국적과 권종을 인식하는 방법을 구현하였다. 한화, 달러화, 유로화에 관하여 회전 및 크기 변화가 있는 환경에서 제안한 알고리즘을 적용하였고 잘 작동하고 있음을 확인 하였다. In this paper, we not only take advantage of the SIFT features in banknote recognition, which has robustness to illumination changes, geometric rotation as well as scale changes, but also propose the hierarchical banknote recognition algorithm, which comprised of feature vector extraction from the frame grabbed image of the banknotes, and matching to the prepared data base of multinational banknotes by ANN algorithm. The images of banknote under the developed UV, IR and white illumination are used so as to extract the SIFT features peculiar to each banknotes. These SIFT features are used in recognition of the nationality as well as face value. We confirmed successful function of the proposed algorithm by applying the proposed algorithm to the banknotes of Korean and USD as well as EURO.

Hybrid of Statistical and Spectral Texture Features for Vehicle Object Classification System

Objectives: To increase the performance of the classifier for the vehicle object among a mixed and highly texture background using hybrid feature extraction method without pre-processing. Methods/Analysis: Vehicle Object recognition system performance is based on the hybrid of feature vector extraction method and artificial neural network classifier without pre-processing. Every image is divided into single blocks size with 20x20 each. The feature vector is extracted from each single size block of the picture. Normalization is done for the extracted feature vector of the vehicle object in the image. These feature vectors are given as input to the neural network classifier for classification. The feed forward Back Propagation Neural Network (BPNN) algorithm is used to train and test the input feature vector by using Neural Network Classifier (NNC) for the vehicle classification. Findings: The idea of the proposed method is that combining the two different literatures namely statistical and spectral texture features without pre-processing for classification. This method is trained and tested with Illinois at Urbana-Champaign (UIUC) standard databases. UIUC database contains car and non-car images with mixed and highly textured background. The findings indicate that the selected input feature vector improves the classification accuracy rate compared to the previous literature. Also the hybrid features maximize the correct classification and minimize the wrong classification. The improved performance results 90.1% of quantitative evaluation is compared with different literature methods of similar work. Applications/Improvements: In different applications, the proposed method plays vital part in surveillance, security for vehicles, monitoring the traffic, etc.

Brain Tumor Classification using Adaptive Neuro-Fuzzy Inference System from MRI

Detecting correct type of brain tumor is a crucial task for diagnosis and curing the tumor. Identifying the correct type of brain tumor can provide a fast and effective way to plan the diagnosis of tumor. The proposed system provides a fast and efficient way to identify the correct type of tumor and classify it to the respective class label. Our proposed system is comprised of multiple stages. In the first stage MRI image is taken as input and is normalized. The second stage includes extraction of feature vectors from the image which results in reducing redundancy of data and will serve as the input to the classifier. The classifier takes each tuple of feature extracted vector to produce classified output. Performance analysis shows that our proposed methodology has performed very efficiently and accurately. In our work we demonstrate the application of Fuzzy Inference System (FIS) based classifier known as Adaptive Neuro Fuzzy Inference System (ANFIS) to successfully classify the input tuples in comparison to other two selected classifiers namely: Artificial Neural Network with Backpropagation Learning Model and K-Nearest Neighbors.

AUTOMATIC RIVER NETWORK EXTRACTION FROM LIDAR DATA

National Geographic Institute of Spain (IGN-ES) has launched a new production system for automatic river network extraction for the Geospatial Reference Information (GRI) within hydrography theme. The goal is to get an accurate and updated river network, automatically extracted as possible. For this, IGN-ES has full LiDAR coverage for the whole Spanish territory with a density of 0.5 points per square meter. To implement this work, it has been validated the technical feasibility, developed a methodology to automate each production phase: hydrological terrain models generation with 2 meter grid size and river network extraction combining hydrographic criteria (topographic network) and hydrological criteria (flow accumulation river network), and finally the production was launched. The key points of this work has been managing a big data environment, more than 160,000 Lidar data files, the infrastructure to store (up to 40 Tb between results and intermediate files), and process; using local virtualization and the Amazon Web Service (AWS), which allowed to obtain this automatic production within 6 months, it also has been important the software stability (TerraScan-TerraSolid, GlobalMapper-Blue Marble , FME-Safe, ArcGIS-Esri) and finally, the human resources managing. The results of this production has been an accurate automatic river network extraction for the whole country with a significant improvement for the altimetric component of the 3D linear vector. <br><br> This article presents the technical feasibility, the production methodology, the automatic river network extraction production and its advantages over traditional vector extraction systems.

AUTOMATIC RIVER NETWORK EXTRACTION FROM LIDAR DATA

Abstract. National Geographic Institute of Spain (IGN-ES) has launched a new production system for automatic river network extraction for the Geospatial Reference Information (GRI) within hydrography theme. The goal is to get an accurate and updated river network, automatically extracted as possible. For this, IGN-ES has full LiDAR coverage for the whole Spanish territory with a density of 0.5 points per square meter. To implement this work, it has been validated the technical feasibility, developed a methodology to automate each production phase: hydrological terrain models generation with 2 meter grid size and river network extraction combining hydrographic criteria (topographic network) and hydrological criteria (flow accumulation river network), and finally the production was launched. The key points of this work has been managing a big data environment, more than 160,000 Lidar data files, the infrastructure to store (up to 40 Tb between results and intermediate files), and process; using local virtualization and the Amazon Web Service (AWS), which allowed to obtain this automatic production within 6 months, it also has been important the software stability (TerraScan-TerraSolid, GlobalMapper-Blue Marble , FME-Safe, ArcGIS-Esri) and finally, the human resources managing. The results of this production has been an accurate automatic river network extraction for the whole country with a significant improvement for the altimetric component of the 3D linear vector. This article presents the technical feasibility, the production methodology, the automatic river network extraction production and its advantages over traditional vector extraction systems.

Induction machine bearing fault diagnosis based on the axial vibration analytic signal

This paper deals with a new induction motor defects diagnosis using the Axial Vibration Analytical Signal (AVAS). The signal is generated by a bearing-defected induction machine. The calculation method may be divided into two main parts; the former is the Hilbert transform that consists in the first part normalization of the axial vibration and its comparison with the AVAS module. The second part consists in the extraction of feature vectors using the Signal Class Dependent Time Frequency Representation (TFRSCD) based on the Fisher contrast design of the non parametrically kernel. The Particle Swarm Optimization (PSO) is used to optimize the feature vectors size. The vibration severity caused by the bearing fault is investigated for different loads. This last decreases with the increasing level of the load. The obtained results are experimentally validated on a 5500 W induction motor test bench.