• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.5
• /
• pp.69-73
• /
• 2022

In this paper, we propose a new method for measuring fine dust through a LoG(Laplacian of Gaussian)-based edge detection technique. CCTV-based images in a video are collected for fine dust measurement, and image ranges are designated through RoI(Region of Interest). After clustering by applying the GMM(Gaussian Mix Model) to the specified area, we detect edge through the LoG algorithm and measure the detected edge strength. The concentration of fine dust is determined based on the measured intensity data of the edge. In this paper, we propose algorithm as the effectiveness of experiment. As a result of collecting and applying CCTV image in the video installed around the laboratory of this school for a month from June to July, the measured result value was proved through this experiment to be sufficient to calculate the concentration and range of fine dust.

• Journal of Broadcast Engineering
• /
• v.26 no.6
• /
• pp.748-757
• /
• 2021

In this paper, we propose a gaze estimation network in which eye landmark position detection and gaze direction vector estimation are integrated into one deep learning network. The proposed network uses the Stacked Hourglass Network as a backbone structure and is largely composed of three parts: a landmark detector, a feature map extractor, and a gaze direction estimator. The landmark detector estimates the coordinates of 50 eye landmarks, and the feature map extractor generates a feature map of the eye image for estimating the gaze direction. And the gaze direction estimator estimates the final gaze direction vector by combining each output result. The proposed network was trained using virtual synthetic eye images and landmark coordinate data generated through the UnityEyes dataset, and the MPIIGaze dataset consisting of real human eye images was used for performance evaluation. Through the experiment, the gaze estimation error showed a performance of 3.9, and the estimation speed of the network was 42 FPS (Frames per second).

• Proceedings of the IEEK Conference
• /
• 2005.11a
• /
• pp.383-386
• /
• 2005

본 논문에서는 Hough 변환을 이용하여 도로 차선의 핵심 정보를 추출하고 차선을 인식하는 방법을 제안하고 실시간으로 차선 인식이 용이 하도록 차선 검출의 고속화 방법을 제안한다. 고속화를 위해 이미지를 작은 영역(Interest Zone)으로 분할하고 분할된 영역에 대해 Hough 변환을 수행하여 영역내의 차선을 검출한다. 검출된 차선의 패턴 정보를 이용하여 다음 Step의 Interest Zone을 결정하고 Hough 변환의 수행을 반복하여 차선 검출을 시도 하였다. 또한 실험 영상을 대상으로 시뮬레이션 수행한 결과를 제시하고 제안 방법의 유효성을 검증하였다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2005.11a
• /
• pp.195-198
• /
• 2005

본 논문에서는 Hough 변환을 이용하여 도로 차선의 핵심 정보를 추출하고 차선을 인식하는 방법을 제안하고 실시간으로 차선 인식이 용이 하도록 차선 검출의 고속화 방법을 제안한다. 고속화를 위해 이미지를 작은 영역(Interest Zone)으로 분할하고 분할된 영역에 대해 Hough 변환을 수행하여 영역내의 차선을 검출한다. 검출된 차선의 패턴 정보를 이용하여 다음 Step의 Interest Zone을 결정하고 Hough 변환의 수행을 반하여 차선 검출을 시도 하였다. 또한 실험 영상을 대상으로 시뮬레이션 수행한 결과를 제시하고 제안 방법의 유효성을 검증하였다.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.2
• /
• pp.33-44
• /
• 2010

This paper propose the structure of real-time face detection hardware architecture for robot vision processing applications. The proposed architecture is robust against illumination changes and operates at no less than 60 frames per second. It uses Modified Census Transform to obtain face characteristics robust against illumination changes. And the AdaBoost algorithm is adopted to learn and generate the characteristics of the face data, and finally detected the face using this data. This paper describes the face detection hardware structure composed of Memory Interface, Image Scaler, MCT Generator, Candidate Detector, Confidence Comparator, Position Resizer, Data Grouper, and Detected Result Display, and verification Result of Hardware Implementation with using Virtex5 LX330 FPGA of Xilinx. Verification result with using the images from a camera showed that maximum 32 faces per one frame can be detected at the speed of maximum 149 frame per second.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.47 no.5
• /
• pp.1-9
• /
• 2010

Generally, dynamic hand gesture recognition is formed through preprocessing step, hand tracking step and hand shape detection step. In this paper, we present advanced dynamic hand gesture recognizing method that improves performance in preprocessing step and hand shape detection step. In preprocessing step, we remove noise fast by using dynamic table and detect skin color exactly on complex background for controling skin color range in skin color detection method using YCbCr color space. Especially, we increase recognizing speed in hand shape detection step through detecting Start Image and Stop Image, that are elements of dynamic hand gesture recognizing, using Guideline. Guideline is edge of input hand image and hand shape for comparing. We perform various experiments with nine web-cam video clips that are separated to complex background and simple background for dynamic hand gesture recognition method in the paper. The result of experiment shows similar recognition ratio but high recognition speed, low cpu usage, low memory usage than recognition method using learning exercise.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10d
• /
• pp.553-555
• /
• 2002

컴퓨터지원진단(Computer Aided Diagnosis; CAD) 시스템은 방사선 의사들이 흉부 X-ray 영상에서 결절을 탐지하는데 있어 실제적으로 발생할 수 있는 오진율을 줄이고, 폐 결절이 존재하는 폐야에서 결절의 존재 유무를 판단하여 검출을 표시함으로써 진단율을 개선시킬 수 있도록 하였다. 본 논문은 흉부 X-ray 영상에서의 폐 결절을 추출하는데 유전자 알고리즘(Genetic Algorithm)을 이용한 템플릿 매칭(Template Matching) 방법을 제안한다. 제안한 방법은 흉부 X-ray 영상에 존재하는 결절과 레퍼런스 이미지를 매칭시켜 적합도를 계산한 후, 그 값을 통하여 수치가 낮은 개체를 선택하여 높은 개체와 교차시킨다. 그리고 레퍼런스 이미지는 결절이 존재하는 환자 X-ray 영상에서 샘플 노듈을 추출한 후 가우시안 분포를 갖는 512개의 레퍼런스 이미지를 생성하였다. 본 논문에서 사용된 영상은 결절 50개, 비결절 30개와 흉부 X-ray 영상에서 육안으로 판별이 가능한 결절 영상을 20개를 포함하여 총 100개 영상을 사용하였다. 실험 결과 83%의 결절을 자동 추출 하였으며, 가장 적절한 레퍼런스 이미지를 발견하고 이를 흉부영상에 매칭시켜 정확한 결절의 위치를 확인하였다.

• Journal of the Institute of Convergence Signal Processing
• /
• v.14 no.1
• /
• pp.27-32
• /
• 2013

It plays an important role to detect the shape of an ellipse in many application areas of image processing. But it is very difficult to detect the ellipse in the real image because the noise was involved in the image, other objects obscured the ellipse or the ellipses were overlap with each other. In this paper, we extract the boundary (edge) to detect ellipse in the image and perform the grouping process in order to reduce amount of information. As a result, the speed of the ellipse detection was improved. Also in order to the ellipse detection, we selected the five ellipse parameters at random And then to select the optimal parameters of the ellipse, the linear least-squares approximation is applied. To verify the ellipse detection, RANSAC algorithm is applied. After the algorithm proposed in this study was implemented, the results applied to the real images showed an aocuracy of 75% and speed was very fast to compared with other researches. It mean that the proposed algorithm was valuable to detect the ellipses in the image.

• Journal of Digital Contents Society
• /
• v.16 no.3
• /
• pp.389-395
• /
• 2015

Digital image provides many conveniences at the internet environment recently. A great number of applications, like Digital Library, Stock Image, Personal Image and Important Information, require the use of digital image. However it has fatal defect which is easy to be modified because digital image is only electronic file. Numerous digital image forgeries have become a serious problem due to the sophistication and accessibility of image editing software. Copy-Move forgery is the simplest type of forgery that involves copying portion of an image and paste it on different location within the image. There are many approaches to detect Copy-Move forgery, but all of them have their own limitations. In this paper, visual and invisible feature based forgery detection techniques are tested and analyzed. The analysis shows that pros and cons of these two techniques compensate each other. Therefore, a hybrid of visual based and invisible feature based forgery detection that combine the merits of both techniques is proposed. The experimental results show that the proposed algorithm has enhanced performance compared to individual techniques. Moreover, it provides more information about the forgery, like identifying copy and duplicate regions.

• Journal of the Microelectronics and Packaging Society
• /
• v.30 no.4
• /
• pp.69-78
• /
• 2023

With the rapid growth of artificial intelligence, the demand for semiconductors is enormously increasing everywhere. To ensure the manufacturing quality and quantity simultaneously, the importance of automatic defect detection during the packaging process has been re-visited by adapting various deep learning-based methodologies into automatic packaging defect inspection. Deep learning (DL) models require a large amount of data for training, but due to the nature of the semiconductor industry where security is important, sharing and labeling of relevant data is challenging, making it difficult for model training. In this study, we propose a new framework for securing sufficient data for DL models with fewer computing resources through a divide-and-conquer approach. The proposed method divides high-resolution images into pre-defined sub-regions and assigns conditional labels to each region, then trains individual sub-regions and boundaries with boundary loss inducing the globally coherent and seamless images. Afterwards, full-size image is reconstructed by combining divided sub-regions. The experimental results show that the images obtained through this research have high efficiency, consistency, quality, and generality.