http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
SHOMY: Detection of Small Hazardous Objects using the You Only Look Once Algorithm
김은찬,Jinyoung Lee,Hyunjik Jo,Kwangtek Na,Eunsook Moon,Gahgene Gweon,Byungjoon Yoo,경연웅 한국인터넷정보학회 2022 KSII Transactions on Internet and Information Syst Vol.16 No.8
Research on the advanced detection of harmful objects in airport cargo for passenger safety against terrorism has increased recently. However, because associated studies are primarily focused on the detection of relatively large objects, research on the detection of small objects is lacking, and the detection performance for small objects has remained considerably low. Here, we verified the limitations of existing research on object detection and developed a new model called the Small Hazardous Object detection enhanced and reconstructed Model based on the You Only Look Once version 5 (YOLOv5) algorithm to overcome these limitations. We also examined the performance of the proposed model through different experiments based on YOLOv5, a recently launched object detection model. The detection performance of our model was found to be enhanced by 0.3 in terms of the mean average precision (mAP) index and 1.1 in terms of mAP (.5:.95) with respect to the YOLOv5 model. The proposed model is especially useful for the detection of small objects of different types in overlapping environments where objects of different sizes are densely packed. The contributions of the study are reconstructed layers for the Small Hazardous Object detection enhanced and reconstructed Model based on YOLOv5 and the non-requirement of data preprocessing for immediate industrial application without any performance degradation.
김은찬,강국진,Kim, Eun-Chan,Gang, Guk-Jin 한국기계연구원 1987 기계연구원소보 Vol.17 No.-
This report describes the study result of the hydrodynamics characteristics for the wake duct. Two wake ducts for the DWT 25,000 Product Carrier were designed and manufactured. The resistance and propulsion performance of the model with them was evaluated by model tests. The object of the present research for the wake duct is to establish the foundation of the design technique and the performance prediction ability of the one, and furthermore, to prepare the base of the development of new type energy saving device
김은찬,강국진,Kim, Eun-Chan,Gang, Guk-Jin 한국기계연구원 1984 기계연구원소보 Vol.12 No.-
A new kind of bulb called Stern-End-Bulb(SEB) for the improvement of the after part of fine hull forms was developed. The reduction of wave resistance and the improvement of the powering performance for the ship with SEB were shown by the ship model tests, At the same time, the characteristics of wave in the vicinity of the stern and the mechanism of the resistance reduction by SEB were investigated. By the systematical variation of the SEB size, the optimum size of SEB was obtained.
김은찬,양승일,Kim, Eun-Chan,Yang, Seung-Il 한국기계연구원 1987 기계연구원소보 Vol.17 No.-
Since the towing tank was operated from early 1979, the test and analysis methods have been established and applied for the performance prediction of ships. Especially the analysis programs for the resistance test ('EHP') and self-propulsion test ('DHP') based on the 1978 ITTC performance prediction method was modified as a name of 'PPTT' in order to include the form factor calculation, two-dimensional analysis method, the prediction on multi-screw ship and the organization of data filing system. Recently the program 'PPTT' was improved to cover the procedure of data fairing, the analysis of propeller-open-water test results carried out at low and high Reynolds numbers, etc. This paper describes the newly improved analysis program 'PTI'.
Deep-Learning Based Honeycomb Pattern Removal in Fiber Bundle Imaging
김은찬,서태원,양성욱 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.4
Fiber bundle endomicroscopy has potential in in-vivo or in-situ optical diagnostics because of its compact size and flexible structure. However, acquired images through fiber bundles suffer from artifact such as a honeycomb pattern. The fiber bundle composed of thousands of single optical fibers inevitably includes void imaging space blocked by cladding layers, while cores are subject to delivering light and imaging. Therefore, restoration of images from the fiber bundle imaging is necessary for improving the quality of imaging and the accuracy of diagnosis. Herein, we introduce a honeycomb pattern removal method based on a deep-learning approach. A training dataset is created by overlaying an initially captured white-reference-image with intact images from ImageNet, while simply taking ground truth from the intact images without supplementary optical instruments. A total number of training images is 276,480, and Adam optimizer is used with 1e-4 learning rate. Evaluation metrics, such as variance-based smoothness and Rayleigh-based line separation criteria were used to verify the pattern removal with preservation of detail. As a result, the quality measure of the proposed method is 0.819, and the other methods, such as a median filter, Gaussian filter, and a barycentric interpolation method, are 0.744, 0.758, and 0.702, respectively.