• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.4
• /
• pp.461-468
• /
• 2003

In a sequence of images obtained by surveillance cameras, facial regions appear very small and their colors change abruptly by lighting condition. This paper proposes a new face detection scheme, robust on complex background, small size, and lighting conditions. The proposed method is consisted of three processes. In the first step, the candidates for the face regions are selected using face color distribution and motion information. In the second stage, the non-face regions are removed using face color ratio, boundary ratio, and average of column-wise intensity variation in the candidates. The face regions containing eyes and mouth are segmented and classified, and then they are scored using their topological relations in the last step. To speed up and improve a performance the above process, a block based image segmentation technique is used. The experiments have shown that the proposed algorithm detects faced regions with more than 91% of accuracy and less than 4.3% of false alarm rate.

• The Journal of the Korea Contents Association
• /
• v.5 no.6
• /
• pp.138-145
• /
• 2005

This paper describes a technique for detecting the boundary of the optic disk in digital image of the retina using inward and outward curve evolution. This paper offers medical information about glaucoma progresses. For accurate boundary detection, image inpainting based on texture synthesis removes blood vessels crossing the optic disk. For removing noises and preserving boundary of optic disk in image inpainting process, the anisotropic diffusion filtering is necessary. After pre-processing, the optic disk boundary is determined using inward and outward curve evolution. The experimental results show that the algorithm is effective for detection of optic disk boundary.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.370-373
• /
• 2005

We have developed Hybrid algorithm for yellow sand detection. Hybrid algorithm is composed of three methods using infrared bands. The first method used the differential absorption in brightness temperature difference between $11\mu m\;and\;12\mu m$ (BID _1), through which help distinguish the yellow sand from various meteorological clouds. The second method uses the brightness temperature difference between $3.7\mu m\;and\;11\mu m$ (BID_2). The technique would be most sensitive to dust loading during the day when the BID _2 is enhanced by reflection of $3.7\mu m$ solar radiation. The third one is a newly developed algorithm from our research, the so-called surface temperature variation method (STY). We have applied the three methods to MODIS for derivation of the yellow sand dust and in conjunction with the Principle Component Analysis (PCA), a form of eigenvector statistical analysis. PCI shows better results for yellow sand detection in comparison with the results from individual method. The comparison between PCI and MODIS aerosols optical depth (AOD) shows remarkable good correlations during daytime and relatively good correlations over the land.

• Journal of Korea Multimedia Society
• /
• v.6 no.7
• /
• pp.1189-1199
• /
• 2003

This paper presents a new method for the motion detection of vehicles using hierarchical motion estimation and parallel processing. It captures the road image by using a CMOS sensor. It divides the captured image into small blocks and detects the motion of each block by using a block-matching method which is based on a hierarchical motion estimation and parallel processing for the real-time processing. The parallelism is achieved by using tile pipeline and the data flow technique. The proposed method has been implemented by using an embedded system. The proposed block matching algorithm has been implemented on PLDs(Programmable Logic Device) and clustering algorithm has been implemented by ARM processor. Experimental results show that the proposed system detects the motion of vehicles in real-time.

• Journal of Institute of Control, Robotics and Systems
• /
• v.2 no.4
• /
• pp.337-343
• /
• 1996

In this paper, we investigate the defect detection by making use of pre-made reference image data and classify the defects by using the artificial neural network. The approach is composed of three main parts. The first step consists of a proper generation of two reference image data by using a low level morphological technique. The second step proceeds by performing three times logical bit operations between two ready-made reference images and just captured image to be tested. This results in defects image only. In the third step, by extracting four features from each detected defect, followed by assigning them into the input nodes of an already trained artificial neural network we can obtain a defect class corresponding to the features. All of the image data are formed in a bit level for the reduction of data size as well as time saving. Experimental results show that proposed algorithms are found to be effective for flexible defect detection, robust classification, and high speed process by adopting a simple logic operation.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.58 no.2
• /
• pp.136-141
• /
• 2009

An AC generator is an important components in producing a electric power and so it requires highly reliable protection relays to minimize the possibility of demage occurring under fault conditions. Conventionally, a DFT based RDR has been used for protecting the generator stator winding. However, when DFTs based on Fourier analysis are used, it has been pointed out that defects can occur during the process of transforming a time domain signal into a frequency domain one which can lead to loss of time domain information. This paper proposes the internal fault detection and fault type discrimination for the stator winding by applying the detailed coefficients by Daubechies Wavelet Transform to overcome the defects in the DFT process. For the case studies reported in the paper, a model system was established for the simulations utilizing the ATP, and this verified the effectiveness of the proposed technique through various off-line tests carried out on the collected data. The propose method is shown to be able to rapidly identify internal fault and did not operate a miss-operation for all the external fault tested.

• Journal of Korea Multimedia Society
• /
• v.18 no.11
• /
• pp.1289-1301
• /
• 2015

Pedestrian detection (PD) is an essential task in various applications and sliding window-based methods utilizing HOG (Histogram of Oriented Gradients) or HOG-like descriptors have been shown to be very effective for accurate PD. However, due to exhaustive search across images, PD methods based on sliding window usually require heavy computational time. In this paper, we propose a real-time PD method for embedded visual surveillance with fixed backgrounds. The proposed PD method employs HOG descriptors as many PD methods does, but utilizes selective search so that it can save processing time significantly. The proposed selective search is guided by restricting searching to candidate regions extracted from Adaptive Gaussian Mixture Model (AGMM)-based background subtraction technique. Moreover, approximate computation of HOG descriptor and implementation in fixed-point arithmetic mode contributes to reduction of processing time further. Possible accuracy degradation due to approximate computation is compensated by applying an appropriate one among three offline trained SVM classifiers according to sizes of candidate regions. The experimental results show that the proposed PD method significantly improves processing speed without noticeable accuracy degradation compared to the original HOG-based PD and HOG with cascade SVM so that it is a suitable real-time PD implementation for embedded surveillance systems.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.3C
• /
• pp.153-157
• /
• 2005

Frequency selective fading occurs due to the Doppler Effect in mobile communication systems. The performances of the systems are rapidly reduced due to effect of multiuser interference under frequency selective channels at DS-CDMA systems. To overcome these problems, we adopted the multi-carrier modulation techniques, and it is able to solve the frequency selective channel effects by means of these modulation techniques, and interference problems due to multiuser access are solved by means of multiuser detection techniques. In this paper, we proposed the blind MOE/PIC multiuser detection method which is composed of both the blind multiuser detection technique and parallel interference canceller. Thus, simulation results show that the proposed method performs better than conventional methods.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.9C
• /
• pp.897-904
• /
• 2006

Multiple input multiple output architectures, known to provide high spectral efficiencies, can provide the best performance in terms of the block error rate when a maximum likelihood (ML) detector is employed. The complexity of the ML detector, however, increases exponentially with the numbers of transmit antennas and signals in the constellation. The zero forcing (ZF) detector has been suggested as a reduced-complexity detection method at the cost of performance degradation. In order to improve the performance of the ZF detector while reducing the complexity of the ML detector, we propose a novel multistage decision method. Numerical results show that, despite the proposed detector has a lower complexity than the ML detector, the performance difference between the ML and proposed detectors is negligibly small at high SNR.

• Journal of Korean Society of Steel Construction
• /
• v.11 no.1 s.38
• /
• pp.79-88
• /
• 1999

Damages of a steel box girder bridge are detected using neural networks. Damage detection using neural networks has increasing momentum in structural engineering. It is a new effort to overcome the limitations of the conventional analytical approaches and applied to the damage detection of a steel box-girder bridge. Data sets for training neural networks are obtained from the acceleration response of the bridge under moving load. Finite element model is first defined and damages of 5, 10, 15 and 20% are assumed in the model. Not only the trained damages but untrained damages are detected in the assessment stage. The untrained damages can be detected with acceptable errors. Because the number of damaged locations are limited to a few parts, more researches are needed to put this technique into practice.