• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.6
• /
• pp.227-233
• /
• 2019

Recently, along with the recent development of deep learning technique, neural networks are achieving success in computer vision filed. Convolutional neural network have shown outstanding performance in not only for a simple image classification task, but also for tasks with high difficulty such as object segmentation and detection. However many such deep learning models are based on supervised-learning, which requires more annotation labels than image-level label. Especially image semantic segmentation model requires pixel-level annotations for training, which is very. To solve these problems, this paper proposes a weakly-supervised semantic segmentation method which requires only image level label to train network. Existing weakly-supervised learning methods have limitations in detecting only specific area of object. In this paper, on the other hand, we use multi-classifier deep learning architecture so that our model recognizes more different parts of objects. The proposed method is evaluated using VOC 2012 validation dataset.

• The Journal of Korean Association of Computer Education
• /
• v.5 no.4
• /
• pp.43-52
• /
• 2002

A theory of binary wavelets which are performed over binary field has been recently proposed. Binary wavelet transform (BWT) of binary images can be used as an alternative to the real-valued wavelet transform of binary images in image processing applications such as compression, edge detection, and recognition. The BWT, however, requires large amount of computations for binary wavelet reconstruction since its operation is accomplished by matrix multiplication. In this paper, an efficient binary wavelet reconstruction method which utilizes filtering operation instead of matrix multiplication is presented. Experimental results show that the proposed algorithm can significantly reduce the computational complexity of the BWT. For the reconstruction of an $N{\times}N$ image, the proposed technique requires only $2MN^2$ multiplications and $2N(M-1)^2$ additions when the filter length M, while the BWT needs $2N^3$ multiplications and $2N(N-1)^2$ additions.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.35S no.12
• /
• pp.55-62
• /
• 1998

Matching two images is an essential step for many computer vision applications. A new approach to the scale and rotation invariant scene matching, using linear features, is presented. Scene or model images are described by a set of linear features approximating edge information, which can be obtained by the conventional edge detection, thinning, and piecewise linear approximation. A set of candidate parameters are hypothesized by mapping the angular difference and a new distance measure to the Hough space and by detecting maximally consistent points. These hypotheses are verified by a fast linear feature matching algorithm composed of a single-step relaxation and a Hough technique. The proposed method is shown to be much faster than the conventional one where the relaxation process is repeated until convergence, while providing matching performance robust to the random alteration of the linear features, without a priori information on the geometrical transformation parameters.

• Journal of Positioning, Navigation, and Timing
• /
• v.7 no.3
• /
• pp.127-138
• /
• 2018

Spoofing attacks including meaconing can provide a bogus position to a victim GPS receiver, and those attacks are notably difficult to detect at the point of view on the receiver. Several countermeasure techniques have been studied to detect, classify, and cancel the spoofing signals. Based on the countermeasure techniques, we have developed an anti-spoofing equipment that detects and mitigates or eliminates the spoofing signal based on raw measurements. Although many anti-spoofing techniques have been studied in the literatures, the evaluation test system is not deeply studied to evaluate the anti-spoofing equipment, which includes detection, mitigation, and elimination of spoofing signals. Each study only has a specific test method to verify its anti-spoofing technique. In this paper, we propose the performance evaluation test system that includes both spoofing signal injection system and its injection scenario with the constraints of stand-alone anti-spoofing techniques. The spoofing signal injection scenario is designed to drive a victim GPS receiver that moves to a designed position, where the mitigation and elimination based anti-spoofing algorithms can be successively evaluated. We evaluate the developed anti-spoofing equipment and a commercial GPS receiver using our proposed performance evaluation test system. Although the commercial one is affected by the test system and moves to the designed position, the anti-spoofing equipment mitigates and eliminates the injected spoofing signals as planned. We evaluate the performance of anti-spoofing equipment on the position error of the circular error probability, while injecting spoofing signals.

• The Bulletin of The Korean Astronomical Society
• /
• v.42 no.2
• /
• pp.41.2-41.2
• /
• 2017

Discovered as the second highest S/N detection of the Planck SZ survey, PLCK G287.0+32.9 is a massive galaxy cluster that belongs to a rare collection of merging clusters that exhibit two radio relics and a radio halo. A feature that makes this cluster even more unique is the separation of the radio relics with one $\sim 400$ kpc to the north-west of the X-ray peak and the other $\sim 2.8$ Mpc to the south-east. This asymmetric configuration requires a complex merging scenario. A key to gaining insight into the events that caused the formation of the merging features is to understand the dark matter mass distribution. Using a weak-lensing technique on deep Subaru and Hubble Space Telescope observations, we map the dark matter mass distribution of PLCK G287.0+32.9. Our investigation detects five significant mass structures. The mass is dominated by a primary structure that is centered near the X-ray peak of the intracluster medium. Four lesser mass structures are detected with two located within $\sim 1\arcmin$ of the primary mass structure, a third to the north-west, and a fourth near the south-east radio relic. Along with these detections, we estimate the mass of each structure and relate their distributions to the intracluster medium and galaxy distributions. In addition, we discuss the relation of the mass structures to the formation of the relics and plausible merging scenarios.

• Journal of Broadcast Engineering
• /
• v.20 no.3
• /
• pp.398-407
• /
• 2015

Compressive sensing (CS) has drawn much interest as a novel sampling technique that enables sparse signal to be sampled under the Nyquitst/Shannon rate. By noting that the block-based CS can still keep spatial correlation in measurement domain, this paper proposes to adapt sampling rate of each block in frame according to its characteristic defined by edge information. Specifically, those blocks containing more edges are assigned more measurements utilizing block-wise correlation in measurement domain without knowledge about full sampling frame. For natural image, the proposed adaptive rate allocation shows considerable improvement compared with fixed subrate block-based CS in both terms of objective (up to 3.29 dB gain) and subjective qualities.

• Journal of Broadcast Engineering
• /
• v.16 no.1
• /
• pp.44-53
• /
• 2011

In this paper, we propose a 3D conversion algorithm of 2D video using depth layer partition method. In the proposed algorithm, we first set frame groups using cut detection algorithm. Each divided frame groups will reduce the possibility of error propagation in the process of motion estimation. Depth image generation is the core technique in 2D/3D conversion algorithm. Therefore, we use two depth map generation algorithms. In the first, segmentation and motion information are used, and in the other, edge directional histogram is used. After applying depth layer partition algorithm which separates objects(foreground) and the background from the original image, the extracted two depth maps are properly merged. Through experiments, we verify that the proposed algorithm generates reliable depth map and good conversion results.

• Journal of Korean Society for Geospatial Information Science
• /
• v.7 no.2 s.14
• /
• pp.143-153
• /
• 1999

GIS have become easier to use and very popular. In recent year digital photogrammetric systems ire becoming cost-effective tools to build and update GIS databases. In close-range photogrammetry for the acquisition of geospatial data, the bundle adjustment needs both initial approximate values and control points to solve the exterior orientation parameters. This paper gives a review of applied and potential algorithms for estimating Initial approximate values before the bundle adjustment, develope new algorithms for determine the exterior orientation parameters, and gives a cost-effective methods for close-range digital photogrammetric system Modifications of existing DLT algorithm were made in this study for providing an efficient, economic, and more accurate photogrammetric data reduction technique. These modifications include robust approaches for automatic detection and elimination of all lands of gross errors in the measurement data, and incorporation of GPS to reduce the number of control points necessary for a DLT solution. Also, this paper derives a new method for space resection from a monocular image. A major advantage of proposed method is that the solution can be uniquely and analytically determined without initial approximate values of exterior orientation parameters and without iterative computation.

• Atmosphere
• /
• v.22 no.2
• /
• pp.137-147
• /
• 2012

The technique of Brightness Temperature Difference (BTD) between 11 and $12{\mu}m$ separates yellow sand dust from clouds according to the difference in absorptive characteristics between the channels. However, this method causes consistent false alarms in many cases, especially over the desert. In order to reduce these false alarms, we should eliminate the background noise originated from surface. We adopted the Background BTD (BBTD), which stands for surface characteristics on clear sky condition without any dust or cloud. We took an average of brightness temperatures of 11 and $12{\mu}m$ channels during the previous 15 days from a target date and then calculated BTD of averaged ones to obtain decontaminated pixels from dust. After defining the BBTD, we subtracted this index from BTD for the Yellow Sand Index (YSI). In the previous study, this method was already verified using the geostationary satellite, MTSAT. In this study, we applied this to the polar orbiting satellite, MODIS, to detect yellow sand dust over Northeast Asia. Products of yellow sand dust from OMI and MTSAT were used to verify MODIS YSI. The coefficient of determination between MODIS YSI and MTSAT YSI was 0.61, and MODIS YSI and OMI AI was also 0.61. As a result of comparing two products, significantly enhanced signals of dust aerosols were detected by removing the false alarms over the desert. Furthermore, the discontinuity between land and ocean on BTD was removed. This was even effective on the case of fall. This study illustrates that the proposed algorithm can provide the reliable distribution of dust aerosols over the desert even at night.

• Journal of Power System Engineering
• /
• v.20 no.3
• /
• pp.22-28
• /
• 2016

Energy consumption rapidly increases with industrial development. therefore an interest in the efficiency of various power converters is increasing. Especially, the field of high-efficiency small motors widely distributed for industrial and family use is considered a very important field in terms of efficient energy usage, and accordingly, in the field of small electric equipment, the use of BLDC that allows high-efficiency drive in an inductor gradually increases. However, for the BLDC drive, information on the position of the rotor is essential. Both methods using a magnetic encoder and an optical encoder to detect the information on the position of the rotor obtain the information by three position sensitive devices, and if any one of these position sensitive devices fails to function, no positional information can be obtained, so there is a limitation in implementing a position sensor with high reliability. In the paper, proposes a new type of encoder that can obtain the positional information on the position of the rotor using a single position sensor in order to overcome the issues that it has to use numerous signal flows and that the reliability is reduced for the acquisition of positional information generated by using multiple position detectors. The encoder in the proposed method replaced the function of generating positional information from multiple sensors with the shape of the encoder plate and the capture function of MICOM. In order to verify the validity of the position detection technique by the proposed encoder, a prototype was produced, and an experiment using the capture function of DSP was conducted through this.