• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.48 no.5
• /
• pp.45-51
• /
• 2011

In this paper, we propose an unsupervised learning method for modeling motion trajectory patterns effectively. In our approach, observations of an object on a trajectory are treated as words in a document for latent dirichlet allocation algorithm which is used for clustering words on the topic in natural language process. This allows clustering topics (e.g. go straight, turn left, turn right) effectively in complex scenes, such as crossroads. After this procedure, we learn patterns of word sequences in each cluster using Baum-Welch algorithm used to find the unknown parameters in a hidden markov model. Evaluation of abnormality can be done using forward algorithm by comparing learned sequence and input sequence. Results of experiments show that modeling of semantic region is robust against noise in various scene.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1302-1308
• /
• 2019

This paper presents an approach for classifying image made by acquired position data from a 2D Lidar sensor with a convolutional neural network (CNN). Lidar sensor has been widely used for unmanned devices owing to advantages in term of data accuracy, robustness against geometry distortion and light variations. A CNN algorithm consists of one or more convolutional and pooling layers and has shown a satisfactory performance for image classification. In this paper, different types of CNN architectures based on training methods, Gradient Descent(GD) and Levenberg-arquardt(LM), are implemented. The LM method has two types based on the frequency of approximating Hessian matrix, one of the factors to update training parameters. Simulation results of the LM algorithms show better classification performance of the image data than that of the GD algorithm. In addition, the LM algorithm with more frequent Hessian matrix approximation shows a smaller error than the other type of LM algorithm.

• Journal of the Microelectronics and Packaging Society
• /
• v.12 no.3 s.36
• /
• pp.267-274
• /
• 2005

This paper presents the fabrication of surface acoustic wave (SAW)-based pressure sensor for long-term stable mechanical compression force measurement. SAW pressure sensor has many attractive features for practical pressure measurement: no battery requirement, wireless pressure detection especially at hazardous environments, and easy other functionality integrations such as temperature, humidity, and RFID. A $41^{\circ}$ YX $LiNbO_3$ piezoelectric substrate was used because of its high SAW propagation velocity and large values of electromechanical coupling factors $K^2$. A silicon substrate with $\~200{\mu}m$ deep cavity was bonded to the diaphragm with epoxy, in which gold was covered all over the inner cavity in order to confine electromagnetic energy inside the sensor, and provide good isolation of the device from its environment. The reflection coefficient $S_{11}$ was measured using network analyzer. High S/N ratio, sharp reflected peaks, and clear separation between the peaks were observed. As a mechanical compression force was applied to the diaphragm from top with extremely sharp object, the diaphragm was bended, resulting in the phase shifts of the reflected peaks. The phase shifts were modulated depending on the amount of applied mechanical compression force. The measured $S_{11}$ results showed a good agreement with simulated results obtained from equivalent admittance circuit modeling.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.43 no.1 s.307
• /
• pp.67-74
• /
• 2006

Contort-aware computing technology is one of the key technology of ubiquitous computing in the mobile device environment. Context recognition computing enables computer applications that automatically respond to user's everyday activity to be realized. In this paper, We use accelerometer could sense activity states of the object and apply to mobile devices. This method for estimating human motion states utilizes various statistics of accelerometer data, such as mean, standard variation, and skewness, as features for classification, and is expected to be more effective than other existing methods that rely on only a few simple statistics. Classification algorithm uses simple decision tree instead of existing neural network by considering mobile devices with limited resources. A series of experiments for testing the effectiveness of the our context detection system for mobile applications and ubiquitous computing has been performed, and its result is presented.

• Journal of Broadcast Engineering
• /
• v.18 no.4
• /
• pp.643-646
• /
• 2013

We present an improved AdaBoost algorithm to avoid overfitting phenomenon. AdaBoost is widely known as one of the best solutions for object detection. However, AdaBoost tends to be overfitting when a training dataset has noisy samples. To avoid the overfitting phenomenon of AdaBoost, the proposed method divides positive samples into K clusters using k-means algorithm, and then uses only one cluster to minimize the training error at each iteration of weak learning. Through this, excessive partitions of samples are prevented. Also, noisy samples are excluded for the training of weak learners so that the overfitting phenomenon is effectively reduced. In our experiment, the proposed method shows better classification and generalization ability than conventional boosting algorithms with various real world datasets.

• Journal of Korean Society for Geospatial Information Science
• /
• v.13 no.3 s.33
• /
• pp.59-67
• /
• 2005

Building information is primary source in many applications such as mapping, telecommunication, car navigation and virtual city modeling. While aerial CCD images which are captured by passive sensor(digital camera) provide horizontal positioning in high accuracy, it is far difficult to process them in automatic fashion due to their inherent properties such as perspective projection and occlusion. On the other hand, LIDAR system offers 3D information about each surface rapidly and accurately in the form of irregularly distributed point clouds. Contrary to the optical images, it is much difficult to obtain semantic information such as building boundary and object segmentation. Photogrammetry and LIDAR have their own major advantages and drawbacks for reconstructing earth surfaces. The purpose of this investigation is to automatically obtain spatial information of 3D buildings by fusing LIDAR data with aerial CCD image. The experimental results show that most of the complex buildings are efficiently extracted by the proposed method and signalize that fusing LIDAR data and aerial CCD image improves feasibility of the automatic detection and extraction of buildings in automatic fashion.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.39 no.2
• /
• pp.235-240
• /
• 2015

Active Safety system is requested on the market of the medium and heavy duty commercial vehicle over 4.5ton beside the market of passenger car with advancement of the digital equipment proportionally. Unlike the passenger car, the mounting position of camera in case of the medium and heavy duty commercial vehicle is relatively high, it is disadvantaged conditions for lane recognition in contradiction to passenger car. In this work, we show the method of lane recognition through the Sobel edge, based on the spatial domain processing, Hough transform and color conversion correction. Also we suggest the low error method of front vehicles recognition in order to reduce the detection error through Haar-like, Adaboost, SVM and Template matching, etc., which are the object recognition methods by frontal camera vision. It is verified that the reliability over 98% on lane recognition is obtained through the vehicle test.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.47 no.1
• /
• pp.160-170
• /
• 2010

This paper proposes a Haar Wavelet Feature Detector (HWFD) based on the Haar wavelet transform and average box filter. By decomposing the original image using the Haar wavelet transform, the proposed detector obtains the variance information of the image, making it possible to extract more distinctive features from the original image. For detection of interest points that represent the regions whose variance is the highest among their neighbor regions, we apply the average box filter to evaluate the local variance information and use the integral image technique for fast computation. Due to utilization of the Haar wavelet transform and the average box filter, the proposed detector is robust to illumination change, scale change, and rotation of the image. Experimental results show that even though the proposed method detects fewer interest points, it achieves higher repeatability, higher efficiency and higher matching accuracy compared with the DoG detector and Harris corner detector.

• Journal of the Korean Geophysical Society
• /
• v.5 no.3
• /
• pp.233-245
• /
• 2002

The spectral-analysis-of-surface-waves (SASW) method is a nondestructive testing method based upon generation and detection of elastic stress waves. SASW is widely used as one of the techniques to determine stiffness profile in engineering geophysics. The essential steps involved are construction of an experimental dispersion curve from data collected in situ, and inversion of the dispersion curve to determine the stiffness profile. The main object of this study is to derive an analytical Jacobian for the inversion. If we set the subsurface to N homogeneous layer, it could save 2N times Jacobian calculation compared to numerical jacobian calculation during inversion. To reconstruct a stiffness profile, constrained damped least square method was applied for the inversion. The algorithm was tested for the numerical data and for the real asphalt and tunnel data, which were able to verify the stiffness profile. The stiffness profile reconstructed by the algorithm showed the possibility to appraise the soundness of tunnel with applications SASW.

• Journal of Convergence for Information Technology
• /
• v.7 no.3
• /
• pp.73-81
• /
• 2017

The generalized Hough transform (GHough) is a useful technique for detecting and locating 2-D model. However, GHough requires a 4-D parameter array and a large amount of time to detect objects of unknown scale and orientation because it enumerates all possible parameter values into a 4-D parameter space. Several n-to-1 mapping algorithms were proposed to reduce the parameter space from 4-D to 2-D. However, these algorithms are very likely to fail due to the random votes cast into the 2-D parameter space. This paper proposes to use internal gradient information in addition to the model boundary points to reduce the number of random votes cast into 2-D parameter space. Experimental result shows that our proposed method can reduce both the number of random votes cast into the parameter space and the execution time effectively.