• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.2747-2752
• /
• 2003

This paper has develops a new framework to understand the human’s driving maneuver based on the expression as HDS focusing on the driver’s stopping maneuver. The driving data has been collected by using the three-dimensional driving simulator based on CAVE, which provides three-dimensional visual information. In our modeling, the relationship between the measured information such as distance to the stop line, its first and second derivatives and the braking amount has been expressed by the PWPS model, which is a class of HDS. The key idea to solve the identification problem was to formulate the problem as the MILP with replacing the switching conditions by binary variables. From the obtained results, it is found that the driver appropriately switches the ‘control law’ according to the following scenario: At the beginning of the stopping behavior (just after finding the stopping point), the driver decelerate the vehicle based on the acceleration information, and then switch to the control law based on the distance to the stop line.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권9호
• /
• pp.3583-3597
• /
• 2020

In this paper, a parallel deep learning model using a convolutional neural network and a dilated convolutional neural network is proposed to classify Alzheimer's disease with high accuracy in PET/CT images. The developed model consists of two pipelines, a conventional CNN pipeline, and a dilated convolution pipeline. An input image is sent through both pipelines, and at the end of both pipelines, extracted features are concatenated and used for classifying Alzheimer's disease. Complimentary abilities of both networks provide better overall accuracy than single conventional CNNs in the dataset. Moreover, instead of performing binary classification, the proposed model performs three-class classification being Alzheimer's disease, mild cognitive impairment, and normal control. Using the data received from Dong-a University, the model performs classification detecting Alzheimer's disease with an accuracy of up to 95.51%.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권9호
• /
• pp.3762-3781
• /
• 2020

Infrared pedestrian target recognition is a vital problem of significant interest in computer vision. In this work, a novel infrared pedestrian target recognition method that uses heterogeneous features with multi-kernel transfer learning is proposed. Firstly, to exploit the characteristics of infrared pedestrian targets fully, a novel multi-scale monogenic filtering-based completed local binary pattern descriptor, referred to as MSMF-CLBP, is designed to extract the texture information, and then an improved histogram of oriented gradient-fisher vector descriptor, referred to as HOG-FV, is proposed to extract the shape information. Second, to enrich the semantic content of feature expression, these two heterogeneous features are integrated to get more complete representation for infrared pedestrian targets. Third, to overcome the defects, such as poor generalization, scarcity of tagged infrared samples, distributional and semantic deviations between the training and testing samples, of the state-of-the-art classifiers, an effective multi-kernel transfer learning classifier called MK-TrAdaBoost is designed. Experimental results show that the proposed method outperforms many state-of-the-art recognition approaches for infrared pedestrian targets.

• International journal of advanced smart convergence
• /
• 제6권2호
• /
• pp.65-71
• /
• 2017

Fingerprints are increasingly used for user authentication in small devices such as mobile phones. Therefore, it is important for Fingerprint authentication systems in personal devices to protect the user's fingerprint information while performing efficiently with a lightweight matching algorithm. In this paper, we propose a new method to extract strong minutiae with unique numbers from fingerprint images. Strong minutiae are at all times obtained from fingerprint images, and can be useful for secure and rapid fingerprint authentication. The binary information of strong minutiae of a fingerprint can be transformed securely and can create cancelable fingerprint templates. Also the bit-strings of strong minutiae decrease computing time necessary for the matching procedure between two fingerprints due to the simplicity of bitwise operations. First, we enroll several fingerprints images of a finger. From these images we select a reference fingerprint and put a number on each minutia. Following this procedure, we search for mated-minutiae between the reference fingerprint and other fingerprints one by one. Finally we derive unique numbers of strong minutiae of the finger. In the experiment with the FVC2004 fingerprint database, we show that using the proposed method, strong minutiae can be extracted successfully.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제7권5호
• /
• pp.1166-1179
• /
• 2013

In wireless sensor network (WSN) environments, environmental noises are generated by, for example, small passing animals, crickets chirping or foliage blowing and will interfere target detection if the noises are higher than the sensor threshold value. For accurate tracking by acoustic WSNs, these environmental noises should be filtered out before initiating track. This paper presents the effect of environmental noises on target tracking and proposes a new algorithm for the noise mitigation in acoustic WSNs. We find that our noise mitigation algorithm works well even for targets with sensing range shorter than the sensor separation as well as with longer sensing ranges. It is also found that noise duration at each sensor affects the performance of the algorithm. A detection algorithm is also presented to account for the Doppler effect which is an important consideration for tracking higher-speed ground targets. For tracking, we use the weighted sensor position centroid to represent the target position measurement and use the Kalman filter (KF) for tracking.

• Smart Structures and Systems
• /
• 제18권6호
• /
• pp.1233-1250
• /
• 2016

Cracks in plate-like structures are some of the main reasons for destruction of the entire structure. In this study, a novel two-stage methodology is proposed for damage detection of flexural plates using an optimized artificial neural network. In the first stage, location of damages in plates is investigated using curvature-moment and curvature-moment derivative concepts. After detecting the damaged areas, the equations for damage severity detection are solved via Bat Algorithm (BA). In the second stage, in order to efficiently reduce the computational cost of model updating during the optimization process of damage severity detection, multiple damage location assurance criterion index based on the frequency change vector of structures are evaluated using properly trained cascade feed-forward neural network (CFNN) as a surrogate model. In order to achieve the most generalized neural network as a surrogate model, its structure is optimized using binary version of BA. To validate this proposed solution method, two examples are presented. The results indicate that after determining the damage location based on curvature-moment derivative concept, the proposed solution method for damage severity detection leads to significant reduction of computational time compared with direct finite element method. Furthermore, integrating BA with the efficient approximation mechanism of finite element model, maintains the acceptable accuracy of damage severity detection.

• Journal of Information Processing Systems
• /
• 제12권2호
• /
• pp.263-274
• /
• 2016

The exceptional development of electronic device technology, the miniaturization of mobile devices, and the development of telecommunication technology has made it possible to monitor human biometric data anywhere and anytime by using different types of wearable or embedded sensors. In daily life, mobile devices can collect wireless body area network (WBAN) data, and the co-collected location data is also important for disease analysis. In order to efficiently analyze WBAN data, including location information and support medical analysis services, we propose a geohash-based spatial index method for a location-aware WBAN data monitoring system on the NoSQL database system, which uses an R-tree-based global tree to organize the real-time location data of a patient and a B-tree-based local tree to manage historical data. This type of spatial index method is a support cloud-based location-aware WBAN data monitoring system. In order to evaluate the proposed method, we built a system that can support a JavaScript Object Notation (JSON) and Binary JSON (BSON) document data on mobile gateway devices. The proposed spatial index method can efficiently process location-based queries for medical signal monitoring. In order to evaluate our index method, we simulated a small system on MongoDB with our proposed index method, which is a document-based NoSQL database system, and evaluated its performance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권2호
• /
• pp.742-762
• /
• 2015

In this study, we propose a distance metric learning approach called discriminant metric learning (DML) for face verification, which addresses a binary-class problem for classifying whether or not two input images are of the same subject. The critical issue for solving this problem is determining the method to be used for measuring the distance between two images. Among various methods, the large margin nearest neighbor (LMNN) method is a state-of-the-art algorithm. However, to compensate the LMNN's entangled data distribution due to high levels of appearance variations in unconstrained environments, DML's goal is to penalize violations of the negative pair distance relationship, i.e., the images with different labels, while being integrated with LMNN to model the distance relation between positive pairs, i.e., the images with the same label. The likelihoods of the input images, estimated using DML and LMNN metrics, are then weighted and combined for further analysis. Additionally, rather than using the k-nearest neighbor (k-NN) classification mechanism, we propose a verification mechanism that measures the correlation of the class label distribution of neighbors to reduce the false negative rate of positive pairs. From the experimental results, we see that DML can modify the relation of negative pairs in the original LMNN space and compensate for LMNN's performance on faces with large variances, such as pose and expression.

• Korean Chemical Engineering Research
• /
• 제56권4호
• /
• pp.536-541
• /
• 2018

Excess molar volumes ($V_m^E$) of binary mixtures of 1-propanol or 2-propanol (1) + cyclohexane or n-hexane (2) were measured with V-shaped dilatometer at 303.15 K. The $V_m^E$ data for these mixtures varied as: 2-propanol > 1-propanol and were higher for cyclohexane than n-hexane for both propanol systems. The experimental data were correlated with Redlich-Kister polynomial. The $V_m^E$ data were interpreted qualitatively as well as quantitatively in terms of Flory-Treszczanowicz-Benson model and Prigogine-Flory-Patterson theory. Both models correctly described the sign and shape of $V_m^E$ vs $x_1$ curves. The values calculated by both the models agree well with the experimental data.

• 한국건축시공학회지
• /
• 제11권3호
• /
• pp.238-246
• /
• 2011

Choosing the best-performance materials is a crucial task for the successful completion of a project in the construction field. In general, the process of material selection is performed through the use of information by a highly experienced expert and the purchasing agent, without the assistance of logical decision-making techniques. For this reason, the construction field has considered various artificial intelligence (AI) techniques to support decision systems as their own selection method. This study proposes the application of a systematic and efficient support vector machine (SVM) model to select optimal exterior materials. The dataset of the study is 120 completed construction projects in South Korea. A total of 8 input determinants were identified and verified from the literature review and interviews with experts. Using data classification and normalization, these 120 sets were divided into 3 groups, and then 5 binary classification models were constructed in a one-against-all (OAA) multi classification method. The SVM model, based on the kernel radical basis function, yielded a prediction accuracy rate of 87.5%. This study indicates that the SVM model appears to be feasible as a decision support system for selecting an optimal construction method.