• 한국측량학회지
• /
• 제20권3호
• /
• pp.243-253
• /
• 2002

최근 들어 수치사진측량은 산업분야를 비롯한 여러 분야에 그 활용범위가 확대되면서 측정정확도의 향상을 위한 다각적인 연구가 진행되고 있으며 그중 다중영상을 이용한 해석방법은 많은 연구자들로 하여금 관심을 모으고 있다. 그러나 이 방법은 영상획득의 용이성과 높은 해석정확도의 확보가 가능한 반면 처리데이터양의 증가로 많은 처리시간과 노력을 요하게 되므로 처리과정의 자동화를 통한 작업의 효율성 향상을 기하여야 한다. 이에 본 연구에서는 정밀시공이 요구되는 전파망원경에 대해 다중영상을 획득하여 이를 해석하기 위한 영상해석과정의 자동화를 구현하고 대상물에 대한 정밀 측정 및 정확도 향상을 꾀하고자 하였다. 연구결과 정밀한 측정을 요하는 대상물에 대해 다중영상해석을 통한 정확도 향상을 도모할 수 있었으며 이를 토대로 전파망원경의 정밀 측정 및 곡률산출을 수행하여 대상물의 시공 후 조립정확도를 검사할 수 있었다.

• 대한의용생체공학회:의공학회지
• /
• 제40권2호
• /
• pp.48-54
• /
• 2019

This paper proposed a method of validation data augmentation for improving the grading accuracy of diabetic macular edema (DME) using deep learning. The data augmentation technique is basically applied in order to secure diversity of data by transforming one image to several images through random translation, rotation, scaling and reflection in preparation of input data of the deep neural network (DNN). In this paper, we apply this technique in the validation process of the trained DNN, and improve the grading accuracy by combining the classification results of the augmented images. To verify the effectiveness, 1,200 retinal images of Messidor dataset was divided into training and validation data at the ratio 7:3. By applying random augmentation to 359 validation data, $1.61{\pm}0.55%$ accuracy improvement was achieved in the case of six times augmentation (N=6). This simple method has shown that the accuracy can be improved in the N range from 2 to 6 with the correlation coefficient of 0.5667. Therefore, it is expected to help improve the diagnostic accuracy of DME with the grading information provided by the proposed DNN.

• 한국측량학회:학술대회논문집
• /
• 한국측량학회 2003년도 추계학술발표회 논문집
• /
• pp.61-64
• /
• 2003

This study attempts to analyze both the mis-closures between triangulation points of 3 and 4 glade and permanent sites data, and accuracy of according as change the number of utilized GPS permanent sites and base lines for improvement of the accuracy in GPS surveying using permanent sites data. The result of this study show that the mis-closure between the two points and Jeonju/Chungju/Sangju/Daegu stations of NGI(National Geography Institute) are 0.0051 m, 0.0361 m, 0.0039 m and 0.0198 m respectively. It indicated that the mis-closures were less than the allowed values in the primary/secondary control point specification for GPS surveying, a mis-closure less than 30 mm for the distance less than 30 km and a mis-closure less than 1 PPM${\times}$D(km) for the distance greater than 30 km. Jinan 11 of actual surveying point for the base line 21.4911 km in Jeonju permanent site and 87.8156 km in Sangju permanent site, northing and easting for planimetric errors of Jinan 11 are 0.0120 m and 0.0113 m, northing and easting for planimetric errors of Jinan 12 are 0.0122 m and 0.0115 m.

• 한국공작기계학회논문집
• /
• 제13권3호
• /
• pp.8-15
• /
• 2004

A significant error in the end milling processes is generated due to using slender tools of which the strengths are not sufficient. In order to obtain the desired machining accuracy, therefore, it is general that at first the rough cut is implemented, then the finish cut is followed. The rough cut eliminates large volume and the finish cut does the remained part. This remaining portion after the rough cut is called as the finish allowance. Larger finish allowances make it hard to get precise dimensions at a following finish cut. Smaller finish allowances are helpful for good dimension, but it sometimes is responsible for inferior surface qualities and over cuts. This study suggests a guidance for the optimum finish allowance for machining accuracy improvement, in which the rough cuts are regulated to remain the desired margins without any over cuts. Some experiments were carried out with various cutting conditions including the change of tool strengths and depth of cuts, and also extended to up millings as well as down millings.

• 한국산업융합학회 논문집
• /
• 제23권2_2호
• /
• pp.209-218
• /
• 2020

Various studies and business investments have been performed on UAV in the field of spatial information industry, and it is judged that this industry has being evolved into an expansion stage as a legalization progresses. In addition, public institutions such as Korea Land and Geospatial Information Corporation, Korea Expressway Corporation, and Korea Land and Housing Corporation which have relatively much utilized spatial information work have entered into the stage of settling with active introduction for reasons of work efficiency and business management. However, surveying drones are still classified as expensive equipment, which is a burden on general business application and technology popularization. Moreover, the stabilization of reliability of various location information acquired from UAV is a part of ongoing research and supplementation. Therefore, in this study, to use image information acquired from low-cost UAV as reliable spatial information data, the flight altitude was changed and compared with the result of double transverse flight with conventional photographing technique. As a result, there was no change in the result value at low altitude, but the result showed more than 30% accuracy and accuracy improvement for the X, Y value at the altitude of 130m or higher than the conventional method.

• 대한전자공학회논문지SP
• /
• 제43권1호
• /
• pp.85-90
• /
• 2006

본 논문에서는 빠르고 안정적이면서도 높은 검색 성공률을 보장하는 내용기반 음악 정보 검색 시스템을 구축하였다. 시스템 질의 구간이나 질의 길이에 따른 시스템 불안정성 문제를 해결할 수 있는 DB 구축 방법인 MFC기법과 각 Superclass별로 특징 벡터의 차수를 차등 적용하여 시스템의 검색 속도를 향상시킬 수 있는 기법을 적용하였다. Superclass를 적용한 시스템은 SuperClass를 적용하지 않은 시스템과의 검색 성공률, 검색 속도 그리고 검색 Precision 비교 실험에서 대등한 성능을 유지하면서 검색 속도를 $20\%\~40\%$ 향상시켰다.

• 한국정밀공학회지
• /
• 제19권7호
• /
• pp.97-105
• /
• 2002

The use of rapid prototyping (RP) has reduced time to market, cut total costs and improved product quality by giving design and manufacturing teams the opportunity to verify and fine tune designs before committing them to expensive tooling and fabrication. In order to improve their unique characteristics according to the working principles, Variable Lamination Manufacturing process (VLM-ST) and corresponding CAD/CAM software (VLM-Slicer) is developed. The objective of this study is to improve the accuracy of VLM-ST process, and it can be done by offset fur cutting error correction, cutting path overrun fur sharp edge and reference shape generation for off-line stacking. It has been shown that, through the verification experiments for given practical shapes, the proposed algorithms are effective for diverse categories of three-dimensional shapes.

• Journal of Computational Design and Engineering
• /
• 제2권2호
• /
• pp.96-104
• /
• 2015

Aluminum die casting is an important manufacturing process for mechanical components. Die casting is known to be more accurate than other types of casting; however, post-machining is usually necessary to achieve the required accuracy. The goal of this investigation is to develop machining- free aluminum die casting. Improvement of the accuracy of planar and cylindrical parts is expected by correcting metal molds. In the proposed method, the shape of cast aluminum made with the initial metal molds is measured by 3D scanning. The 3D scan data includes information about deformations that occur during casting. Therefore, it is possible to estimate the deformation and correction amounts by comparing 3D scan data with product computer-aided design (CAD) data. We corrected planar and cylindrical parts of the CAD data for the mold. In addition, we corrected the planar part of the metal mold using the corrected mold data. The effectiveness of the proposed method is demonstrated by evaluating the accuracy improvement of the cast aluminum made with the corrected mold.

• 한국군사과학기술학회지
• /
• 제12권1호
• /
• pp.114-122
• /
• 2009

An adaptive Kalman filter is designed as a post-navigation filter to improve the accuracy of GPS receiver's navigation performance in high dynamic environments. Not only the adaptive Kalman filter reduces the large noise error of navigation data which is obtained by least square method, but also the filter is not degraded as normal Kalman filter in high acceleration movements because the system noise is estimated. Also an initialization structure of the filter is desisted in consideration for irregular output condition of navigation data by least squared method such as reacquisition status in GPS receiver. The filter performance is verified by GPS simulator which has the simulation capability of high velocity and acceleration. Finally, a vehicle test including DGPS is executed to conform the real improvement of that filter performance. This filter can be applied to various data measurement systems to improve accuracy in high dynamic conditions besides GPS receiver.

• 전기전자학회논문지
• /
• 제27권4호
• /
• pp.561-566
• /
• 2023

딥러닝의 발전으로 다양한 컴퓨터 비전 연구를 수행할 수 있게 됐다. 딥러닝은 컴퓨터 비전 연구 중 이미지 처리에서 높은 정확도와 성능을 보여줬다. 하지만 대부분의 이미지 처리 방식은 이미지의 시각 정보만을 이용해 이미지를 처리하는 경우가 대부분이다. 이미지-텍스트 쌍을 활용할 경우 이미지와 관련된 설명, 주석 등의 텍스트 데이터가 이미지 자체에서는 얻기 힘든 추가적인 맥락과 시각 정보를 제공할 수 있다. 본 논문에서는 이미지-텍스트 쌍을 활용하여 이미지와 텍스트를 분석하는 딥러닝 모델 제안한다. 제안 모델은 이미지 정보만을 사용한 딥러닝 모델보다 약 11% 향상된 분류 정확도 결과를 보였다.