• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.1-25
• /
• 2020

In this paper, we suggest an application system architecture which provides accurate, fast and efficient automatic gasometer reading function. The system captures gasometer image using mobile device camera, transmits the image to a cloud server on top of private LTE network, and analyzes the image to extract character information of device ID and gas usage amount by selective optical character recognition based on deep learning technology. In general, there are many types of character in an image and optical character recognition technology extracts all character information in an image. But some applications need to ignore non-of-interest types of character and only have to focus on some specific types of characters. For an example of the application, automatic gasometer reading system only need to extract device ID and gas usage amount character information from gasometer images to send bill to users. Non-of-interest character strings, such as device type, manufacturer, manufacturing date, specification and etc., are not valuable information to the application. Thus, the application have to analyze point of interest region and specific types of characters to extract valuable information only. We adopted CNN (Convolutional Neural Network) based object detection and CRNN (Convolutional Recurrent Neural Network) technology for selective optical character recognition which only analyze point of interest region for selective character information extraction. We build up 3 neural networks for the application system. The first is a convolutional neural network which detects point of interest region of gas usage amount and device ID information character strings, the second is another convolutional neural network which transforms spatial information of point of interest region to spatial sequential feature vectors, and the third is bi-directional long short term memory network which converts spatial sequential information to character strings using time-series analysis mapping from feature vectors to character strings. In this research, point of interest character strings are device ID and gas usage amount. Device ID consists of 12 arabic character strings and gas usage amount consists of 4 ~ 5 arabic character strings. All system components are implemented in Amazon Web Service Cloud with Intel Zeon E5-2686 v4 CPU and NVidia TESLA V100 GPU. The system architecture adopts master-lave processing structure for efficient and fast parallel processing coping with about 700,000 requests per day. Mobile device captures gasometer image and transmits to master process in AWS cloud. Master process runs on Intel Zeon CPU and pushes reading request from mobile device to an input queue with FIFO (First In First Out) structure. Slave process consists of 3 types of deep neural networks which conduct character recognition process and runs on NVidia GPU module. Slave process is always polling the input queue to get recognition request. If there are some requests from master process in the input queue, slave process converts the image in the input queue to device ID character string, gas usage amount character string and position information of the strings, returns the information to output queue, and switch to idle mode to poll the input queue. Master process gets final information form the output queue and delivers the information to the mobile device. We used total 27,120 gasometer images for training, validation and testing of 3 types of deep neural network. 22,985 images were used for training and validation, 4,135 images were used for testing. We randomly splitted 22,985 images with 8:2 ratio for training and validation respectively for each training epoch. 4,135 test image were categorized into 5 types (Normal, noise, reflex, scale and slant). Normal data is clean image data, noise means image with noise signal, relfex means image with light reflection in gasometer region, scale means images with small object size due to long-distance capturing and slant means images which is not horizontally flat. Final character string recognition accuracies for device ID and gas usage amount of normal data are 0.960 and 0.864 respectively.

• The Transactions of the Korea Information Processing Society
• /
• v.5 no.11
• /
• pp.2845-2854
• /
• 1998

Median filter can be implemented in the binary domain based on threshold decomposition, stacking property, and linear separability. In this paper, we develop one-dimensional and two-dimensional parallel algorithms for the median filter on a reconfigurable mesh with buses(RMESH) which is suitable for VLSI implementation. And we evaluate their performance by comparing the time complexities of RMESH algorithms with those of algorithms on mesh-connected computer. When the length of M-valued 1-D signal is N and w is the window width, the RMESH algorithm is done in O(Mw) time and mesh algorithm is done in $O(Mw^2)$ time. Beside, when the size of M-valued 2-D image is $N{\times}N$ and the window size is $w{\times}w$, our algorithm on $N{\times}N$ RMESH can be computed in O(Mw) time which is a significant improvement over the $O(Mw^2)$ complexity on $N{\times}N$ mesh.

• Design & Manufacturing
• /
• v.14 no.2
• /
• pp.14-20
• /
• 2020

In OLED panel production process, it is necessary to cut a part of protective film as a preprocess for lighting inspection. The current method is to recognize only the fiducial mark of the cut-out panel. Bare Glass Cutting does not compensate for machining cumulative tolerances. Even though process defects still occur, it is necessary to develop technology to solve this problem because only the Align Mark of the panel that has already been cut is used as the reference point for alignment. There is a lot of defective lighting during panel lighting test because the correct protective film is not cut on the panel power and signal application pad position. In laser cutting process to remove the polarizing film / protective film / TSP film of OLED panel, laser processing is not performed immediately after the panel alignment based on the alignment mark only. Therefore, in this paper, we performed real time inspection which minimizes the mechanism tolerance by correcting the laser cutting path of the protective film in real time using Machine Vision. We have studied calibration algorithm of Vision Software coordinate system and real image coordinate system to minimize inspection resolution and position detection error and edge detection algorithm to accurately measure edge of panel.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.5
• /
• pp.14-21
• /
• 2020

In order to provide confidential services between two communicating parties, block data encryption using a symmetric secret key is applied. A power analysis attack on a cryptosystem is a side channel-analysis method that can extract a secret key by measuring the power consumption traces of the crypto device. In this paper, we propose an attack model that can recover the secret key using a power analysis attack based on a deep learning convolutional neural network (CNN) algorithm. Considering that the CNN algorithm is suitable for image analysis, we particularly adopt the recurrence plot (RP) signal processing method, which transforms the one-dimensional power trace into two-dimensional data. As a result of executing the proposed CNN attack model on an XMEGA128 experimental board that implemented the AES-128 encryption algorithm, we recovered the secret key with 22.23% accuracy using raw power consumption traces, and obtained 97.93% accuracy using power traces on which we applied the RP processing method.

• Proceedings of the KSRS Conference
• /
• v.1
• /
• pp.90-93
• /
• 2006

The first Geostationary Ocean Color Imager (GOCI) onboard its Communication Ocean and Meteorological Satellite (COMS) is scheduled for launch in 2008. GOCI includes the eight visible-to-near-infrared (NIR) bands, 0.5km pixel resolution, and a coverage region of 2500 ${\times}$ 2500km centered at 36N and 130E. GOCI has had the scope of its objectives broadened to understand the role of the oceans and ocean productivity in the climate system, biogeochemical variables, geological and biological response to physical dynamics and to detect and monitor toxic algal blooms of notable extension through observations of ocean color. The special feature with GOCI is that like MODIS, MERIS and GLI, it will include the band triplets 660-680-745 for the measurements of sun-induced chlorophyll-a fluorescence signal from the ocean. The GOCI will provide SeaWiFS quality observations with frequencies of image acquisition 8 times during daytime and 2 times during nighttime. With all the above features, GOCI is considered to be a remote sensing tool with great potential to contribute to better understanding of coastal oceanic ecosystem dynamics and processes by addressing environmental features in a multidisciplinary way. To achieve the objectives of the GOCI mission, we develop the GOCI Data Processing System (GDPS) which integrates all necessary basic and advanced techniques to process the GOCI data and deliver the desired biological and geophysical products to its user community. Several useful ocean parameters estimated by in-water and other optical algorithms included in the GDPS will be used for monitoring the ocean environment of Korea and neighbouring countries and input into the models for climate change prediction.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2015.05a
• /
• pp.367-370
• /
• 2015

This paper proposes an improve method of the detection performance of traffic lights for autonomous driving cars. Earlier detection methods used to adopt color thresholding, template matching and based learning maching methods, but its have some problems such as recognition rate decreasing, slow processing time. The proposed method uses both detection mask and morphological preprocessing. Firstly, input color images are converted to YCbCr image in order to strengthen its illumination, and horizontal edge components are extracted in the Y Channel. Secondly, the region of interest is detected according to morphological characteristics of the traffic lights. Finally, the traffic signal is detected based on color distributions. The proposed method showed that the detection rate and processing time improved rather than the conventional algorithm about some surrounding environments.

• Journal of Digital Contents Society
• /
• v.10 no.4
• /
• pp.579-585
• /
• 2009

In [9], Tyrtshnikov proposed a preconditioned approach to derive a general solution from a Toeplitz linear system. Furthermore, the process of selecting a preconditioner matrix from symmetric Toeplitz matrix, which has been used in previous studies, is introduced. This research introduces a new method for finding the preconditioner in a Toeplitz system. Also, through analyzing these preconditioners, it is derived that eigenvalues of a symmetric Toeplitz are very close to eigenvalues of a new preconditioner for T. It is shown that if the spectrum of the preconditioned system $C_0^{-1}T$ is clustered around 1, then the convergence rate of the preconditioned system is superlinear. From these results, it is determined to get the superliner at the convergence rate by our good preconditioner $C_0$. Moreover, an advantage is driven by increasing various applications i. e. image processing, signal processing, etc. in this study from the proposed preconditioners for Toeplitz matrices. Another characteristic, which this research holds, is that the preconditioner retains the properties of the Toeplitz matrix.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.10
• /
• pp.3290-3299
• /
• 2000

Digital watermarking has similar concepts with channel coding thechnique for transferring data with minimizing error in noise environment, since it should be robust to various kinds of data manipulation for protecting copyrights of multimedia data. This paper proposes a digital watermarking technique which is robust to various kinds of data manipulation. Intellectual property rights information is encoded using a convolutional code, and block-interleaving technique is applied to prevent successive loss of encoded data. Encoded intelloctual property rithts informationis embedded using spread spectrum technique which is robust to cata manipulation. In order to reconstruct intellectual property rights information, watermark signalis detected by covariance between watermarked image and pseudo rando noise sequence which is used to einbed watermark. Embedded intellectual property rights information is obtaned by de-interleaving and cecoding previously detected wtermark signal. Experimental results show that block interleaving watermarking technique can detect embedded intellectial property right informationmore correctly against to attacks like Gaussian noise additon, filtering, and JPEG compression than general spread spectrum technique in the same PSNR.

• Korean Journal of Remote Sensing
• /
• v.11 no.3
• /
• pp.129-145
• /
• 1995

SeaBeam2000, a multibeam echo sounder, is a new generation seabed mapping system of which a single swath covers an angular range of -60.deg. to 60.deg. from the vertical direction with 121 beams. It provides high-density and high-quality bathymetric data along with sidescan acoustic data. The purpose of the research is to develop a system for processing multibeam underwater acoustic and bathymetric data using digital signal processing techniques. Recently obtained multibeam echo sounder data covering a survey area in the East Sea of Korea ($37{\circ}$.00'N to $37{\circ}$30'N and $129{\circ}$40'E to $130{\circ}$30'E) are preliminarily processed using the developed system and reproduced in the raster image format as well as three dimensionally visualized form.

• Journal of IKEEE
• /
• v.26 no.1
• /
• pp.27-35
• /
• 2022

This paper proposes an ultrasound vessel-pattern imaging algorithm with low computational complexity. The proposed imaging algorithm reconstructs blood-vessel patterns by only detecting blood flow, and can be applied to a real-time signal processing hardware that extracts an ultrasonic finger-vessel pattern. Unlike a blood-flow imaging mode of typical ultrasound medical imaging device, the proposed imaging algorithm only reconstructs a presence of blood flow as an image. That is, since the proposed algorithm does not use an I/Q demodulation and detects a presence of blood flow by accumulating an absolute value of the clutter-filter output, a structure of the algorithm is relatively simple. To verify a complexity of the proposed algorithm, a simulation model for finger vessel was implemented using Field-II program. Through the behavioral simulation, it was confirmed that the processing time of the proposed algorithm is around 54 times less than that of the typical color-flow mode. Considering the required main building blocks and the amount of computation, the proposed algorithm is simple to implement in hardware such as an FPGA and an ASIC.