• Annual Conference of KIPS
• /
• 2019.05a
• /
• pp.515-518
• /
• 2019

Semantic segmentation has remained as a challenging problem in the field of computer vision. Given the immense power of Convolution Neural Network (CNN) models, many complex problems have been solved in computer vision. Semantic segmentation is the challenge of classifying several pixels of an image into one category. With the help of convolution neural networks, we have witnessed prolific results over the time. We propose a convolutional neural network model which uses Fully CNN with deconvolutional pixel layers. The goal is to create a hierarchy of features while the fully convolutional model does the primary learning and later deconvolutional model visually segments the target image. The proposed approach creates a direct link among the several adjacent pixels in the resulting feature maps. It also preserves the spatial features such as corners and edges in images and hence adding more accuracy to the resulting outputs. We test our algorithm on Karlsruhe Institute of Technology and Toyota Technologies Institute (KITTI) street view data set. Our method achieves an mIoU accuracy of 92.04 %.

• Journal of Information Processing Systems
• /
• v.18 no.3
• /
• pp.319-331
• /
• 2022

The capability to examine an X-ray image is so far a challenging task. In this work, we suggest a practical and novel algorithm based on image fusion to inspect the issues such as background noise, blurriness, or sharpness, which curbs the quality of dual-energy X-ray images. The current technology exercised for the examination of bags and baggage is "X-ray"; however, the results of the incumbent technology used show blurred and low contrast level images. This paper aims to improve the quality of X-ray images for a clearer vision of illegitimate or volatile substances. A dataset of 40 images was taken for the experiment, but for clarity, the results of only 13 images have been shown. The results were evaluated using MSE and PSNR metrics, where the average PSNR value of the proposed system compared to single X-ray images was increased by 19.3%, and the MSE value decreased by 17.3%. The results show that the proposed framework will help discern threats and the entire scanning process.

• Annual Conference of KIPS
• /
• 2010.04a
• /
• pp.365-368
• /
• 2010

LCD 모니터의 백라이트로 CCFL 형광체를 많이 사용하고 있으나 그 불량여부는 육안에 의존하고 있다. 그러나 육안 검사는 CCFL 현광체의 불량에 대한 일관성 있는 기준이 결여되고, 노동집약적인 검사로 인해 산업적 재해가 발생할 수 있다. CCFL 불량유무를 자동 판별하기 위해서는 물리적 촬영환경과 자동검출 알고리즘이 필수적이다. 본 논문에서는 CCFL 형광체의 불량을 자동으로 검사하기 위한 촬영환경을 설명하고, 그 촬영환경에 알맞은 대표적인 불량검출 알고리즘을 제안한다. 실험결과에 따르면, 알고리즘은 불량 판별율 98.86%와 과검율 3.34%의 성능을 보였다.

• Annual Conference of KIPS
• /
• 2011.04a
• /
• pp.353-356
• /
• 2011

스마트 폰의 등장과 모바일 인터넷을 제공함에 따라 휴대폰은 음성통신 수단이 아닌 웹을 통하여 서비스를 제공받는 도구 또는 각종 게임 및 응용 어플리케이션을 제공하는 놀이수단으로도 발전하였고, 이로인하여 사용량도 증가하였다. 사용량의 급증으로 인하여 모바일 광고에 대한 업계의 관심도 증가 하였지만, 한정적인 출력화면에 의하여 제한적일 수밖에 없다. 이를 보완하기 위해, 본 논문에서는 기업의 로고 광고의 효과를 극대화 할 수 있는 안드로이드 기반 로고를 인식하는 증강현실 시스템을 제안 하였고, 이를 구현 하여 실 제폰에 탑재한 후 다양한 성능 분석을 하였다. 실험결과, 그 가능성은 확인하였지만 현하드웨어 성능으로는 실시간으로 지원하기에는 역부족임을 알 수 있었다.

• Journal of Biosystems Engineering
• /
• v.28 no.6
• /
• pp.511-516
• /
• 2003

Computer vision technology has been utilized as one of the most powerful tools to automate various agricultural operations. Though it has demonstrated successful results in various applications, the current status of technology is still for behind the human's capability typically for the unstructured and variable task environment. In this paper, a man-machine interactive hybrid decision-making system which utilized a concept of tole-operation was proposed to overcome limitations of computer image processing and cognitive capability. Tasks of greenhouse watermelon cultivation such as pruning, watering, pesticide application, and harvest require identification of target object. Identifying water-melons including position data from the field image is very difficult because of the ambiguity among stems, leaves, shades. and fruits, especially when watermelon is covered partly by leaves or stems. Watermelon identification from the cultivation field image transmitted by wireless was selected to realize the proposed concept. The system was designed such that operator(farmer), computer, and machinery share their roles utilizing their maximum merits to accomplish given tasks successfully. And the developed system was composed of the image monitoring and task control module, wireless remote image acquisition and data transmission module, and man-machine interface module. Once task was selected from the task control and monitoring module, the analog signal of the color image of the field was captured and transmitted to the host computer using R.F. module by wireless. Operator communicated with computer through touch screen interface. And then a sequence of algorithms to identify the location and size of the watermelon was performed based on the local image processing. And the system showed practical and feasible way of automation for the volatile bio-production process.

• Structural Engineering and Mechanics
• /
• v.91 no.5
• /
• pp.443-457
• /
• 2024

At present, the traditional concrete surface inspection methods based on artificial vision have the problems of high cost and insecurity, while the computer vision methods rely on artificial selection features in the case of sensitive environmental changes and difficult promotion. In order to solve these problems, this paper introduces deep learning technology in the field of computer vision to achieve automatic feature extraction of structural damage, with excellent detection speed and strong generalization ability. The main contents of this study are as follows: (1) A method based on DeepLabV3+ convolutional neural network model is proposed for surface detection of post-earthquake structural damage, including surface damage such as concrete cracks, spaling and exposed steel bars. The key semantic information is extracted by different backbone networks, and the data sets containing various surface damage are trained, tested and evaluated. The intersection ratios of 54.4%, 44.2%, and 89.9% in the test set demonstrate the network's capability to accurately identify different types of structural surface damages in pixel-level segmentation, highlighting its effectiveness in varied testing scenarios. (2) A semantic segmentation model based on DeepLabV3+ convolutional neural network is proposed for the detection and evaluation of post-earthquake structural components. Using a dataset that includes building structural components and their damage degrees for training, testing, and evaluation, semantic segmentation detection accuracies were recorded at 98.5% and 56.9%. To provide a comprehensive assessment that considers both false positives and false negatives, the Mean Intersection over Union (Mean IoU) was employed as the primary evaluation metric. This choice ensures that the network's performance in detecting and evaluating pixel-level damage in post-earthquake structural components is evaluated uniformly across all experiments. By incorporating deep learning technology, this study not only offers an innovative solution for accurately identifying post-earthquake damage in civil engineering structures but also contributes significantly to empirical research in automated detection and evaluation within the field of structural health monitoring.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.3
• /
• pp.1284-1292
• /
• 2015

As visual surveillance systems become more and more common in human lives, approaches based on these systems to solve security problems in practice are boosted, especially in railway applications. In this paper, we first propose a robust snag detection algorithm and then present a railway security system by using a combination of multiple sensors and the vision based snag detection algorithm. The system aims safety at several repeatedly occurred situations including slope protection, inspection of the falling-object from bridges, and the detection of snags and foreign objects on the rail. Experiments demonstrate that the snag detection is relatively robust and the system could guarantee the security of the railway through these real-time protections and detections.

• Current Optics and Photonics
• /
• v.4 no.5
• /
• pp.411-420
• /
• 2020

Non-contact three-dimensional (3D) measuring technology is used to identify defects in miniature products, such as optics, polymers, and semiconductors. Hence, this technology has garnered significant attention in computer vision research. In this paper, we focus on shape from focus (SFF), which is an optical passive method for 3D shape recovery. In existing SFF techniques using interpolation, all datasets of the focus volume are approximated using one model. However, these methods cannot demonstrate how a predefined model fits all image points of an object. Moreover, it is not reasonable to explain various shapes of datasets using one model. Furthermore, if noise is present in the dataset, an error will be generated. Therefore, we propose an algorithm based on polynomial regression analysis to address these disadvantages. Our experimental results indicate that the proposed method is more accurate than existing methods.

• Smart Structures and Systems
• /
• v.15 no.6
• /
• pp.1583-1600
• /
• 2015

When natural disasters occur, including earthquakes, tsunamis, and debris flows, they are often accompanied by various types of damages such as the collapse of buildings, broken bridges and roads, and the destruction of natural scenery. Natural disaster detection and warning is an important issue which could help to reduce the incidence of serious damage to life and property as well as provide information for search and rescue afterwards. In this study, we propose a novel computer vision technique for debris flow detection which is feature-based that can be used to construct a debris flow event warning system. The landscape is composed of various elements, including trees, rocks, and buildings which are characterized by their features, shapes, positions, and colors. Unlike the traditional methods, our analysis relies on changes in the natural scenery which influence changes to the features. The "background module" and "monitoring module" procedures are designed and used to detect debris flows and construct an event warning system. The multi-criteria decision-making method used to construct an event warring system includes gradient information and the percentage of variation of the features. To prove the feasibility of the proposed method for detecting debris flows, some real cases of debris flows are analyzed. The natural environment is simulated and an event warning system is constructed to warn of debris flows. Debris flows are successfully detected using these two procedures, by analyzing the variation in the detected features and the matched feature. The feasibility of the event warning system is proven using the simulation method. Therefore, the feature based method is found to be useful for detecting debris flows and the event warning system is triggered when debris flows occur.

• Journal of the Ergonomics Society of Korea
• /
• v.29 no.6
• /
• pp.925-932
• /
• 2010

Most of the research on three-dimensional interaction field have showed different accuracy in terms of sensing, mode and method. Furthermore, implementation of interaction has been a lack of consistency in application field. Therefore, this study is to suggest research trends of three-dimensional interaction using meta-analysis. Searching relative keyword in database provided with 153 domestic papers and 188 international papers covering three-dimensional interaction. Analytical coding tables determined 18 domestic papers and 28 international papers for analysis. Frequency analysis was carried out on method of action, element, number, accuracy and then verified accuracy by effect size of the meta-analysis. As the results, the effect size of sensor-based was higher than vision-based, but the effect size was extracted to small as 0.02. The effect size of vision-based using hand motion was higher than sensor-based using hand motion. Therefore, implementation of three-dimensional sensor-based interaction and vision-based using hand motions more efficient. This study was significant to comprehensive analysis of three-dimensional motion recognition for interaction and suggest to application directions of three-dimensional interaction.