• 대한기계학회논문집A
• /
• 제27권2호
• /
• pp.310-316
• /
• 2003

One of the significant subject in the automatic arc welding is to establish control system of the welding parameters for controlling bead geometry as a criterion to evaluate the quality of arc welding. This paper proposes an inference algorithm for bead geometry in CMA Welding using Neuro-Fuzzy algorithm. The characteristic welding parameters are measured by the circuit composed of hall sensor, voltage divider tachometer, etc. and then the bead geometry of each weld pool is calculated and detected by an image processing with CCD camera and a measuring with microscope. The relationships between the characteristic welding parameters and the bead geometry have been arranged empirically. From the result of experiments, membership functions and fuzzy rules are tuned and determined by the learning of neural network, and then the relationship between actual bead geometry and inferred bead geometry are concluded by fuzzy logic controller. In the applied inference system of bead geometry using Neuro-Fuzzy algorithm, the inference error percent is within -5%∼+4% in case of bead width, -10%∼+10% in bead height, -5%∼+6% in bead area, -10%∼+10% in penetration. Use of the Neuro-Fuzzy algorithm allows the CMA Welding system to evaluate the quality in bead geometry in real time as the welding parameters change.

• Journal of Mechanical Science and Technology
• /
• 제18권7호
• /
• pp.1222-1235
• /
• 2004

This study examines the influence of geometry on the internal flow and macroscopic behavior of the spray in Diesel nozzles. For this investigation, two bi-orifice nozzles were employed: one cylindrical and one conical. The first step is to use a non-destructive characterization method which is based on the production of silicone moulds so that the precise internal geometry of the two nozzles can be measured. At this stage the nozzles have been characterized dimensionally and therefore the internal flow can be studied using CFD calculations. The results gained from this experiment make it possible also to ascertain the critical cavitation conditions. Once the critical cavitation conditions have been identified, the macroscopic parameters of the spray can be studied in both cavitating and non-cavitating conditions using a test rig pressurized with nitrogen and with the help of a image acquisition system and image processing software. Consequently, research can be carried out to determine the influence that cavitation has on macroscopic spray behavior. From the point of view of the spray macroscopic behavior, the main conclusion of the paper is that cavitation leads to an increment of the spray cone angle. On the other hand, from the point of view of the internal flow, the hole outlet velocity increases when cavitation appears. This phenomenon can be explained by the reduction in the cross section of the liquid phase in the outlet section of the hole.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권6호
• /
• pp.2044-2059
• /
• 2022

Recent works have shown that the geometric constraint can be harnessed to boost the performance of CNN-based camera localization. However, the existing strategies are limited to imposing image-level constraint between pose pairs, which is weak and coarse-gained. In this paper, we introduce a pixel-level epipolar geometry constraint to vanilla localization framework without the ground-truth 3D information. Dubbed EpiLoc, our method establishes the geometric relationship between pixels in different images by utilizing the epipolar geometry thus forcing the network to regress more accurate poses. We also propose a variant called EpiSingle to cope with non-sequential training images, which can construct the epipolar geometry constraint based on a single image in a self-supervised manner. Extensive experiments on the public indoor 7Scenes and outdoor RobotCar datasets show that the proposed pixel-level constraint is valuable, and helps our EpiLoc achieve state-of-the-art results in the end-to-end camera localization task.

• 소성∙가공
• /
• 제9권4호
• /
• pp.404-412
• /
• 2000

It is desirable to use the square grid analysis with the aid of the stereo vision and image processing techniques in order to automatically measure the surface-strain distribution over a stamped part. But this method has some inherent problems such as the difficulty in enhancement of bad images, the measurement error due to the digital image resolution and the limit of the area that can be measured at a time. Therefore, it is still hard to measure the strain distribution over the entire surface of a medium-or large-sized stamped part even by using an automated strain measurement system. In this study, several methods which enable to solve these problems considerably without losing accuracy and precision In measurement are suggested. The superposition of images that have different high-lightened or damaged part from each other gives much enhanced image. A new algorithm for constructing of the element connectivity from the line-thinned image helps recognize up to 1,000 elements. And the geometry assembling algorithm including the global error minimization makes it possible to measure a large specimen with reliability and efficiency.

• 한국산학기술학회논문지
• /
• 제13권9호
• /
• pp.3856-3863
• /
• 2012

스테레오 영상으로부터 3차원 형상을 구현함에 있어 사용자의 개입을 최소로 필요로 하는 기법을 개발하였다. 형상구현은 특정 기하학 그룹을 평가하는 3단계로 이루어진다. 1단계는 영상에 존재하는 epipolar 기하 평가로 각 영상에서의 특정점들을 일치시킨다. 2단계는 소실점 방법을 이용하여 투영공간에서 특정평면을 찾는 affine 기하 평가이다. 3단계에서는 카메라의 자기보정을 포함하며 3차원 모델이 얻어질 수 있는 계량 기하 변수를 구한다. 이 방법의 장점은 형상구현을 위해 스테레오 영상을 보정할 필요가 없는 것으로, 그 구현가능성을 실증하였다.

• 한국BIM학회 논문집
• /
• 제8권1호
• /
• pp.15-23
• /
• 2018

Geometry control of precast members is the most important technology for modular construction. In this paper, image-based modeling and rendering (IBMR) technology was adopted for 3D modeling of precast elements. It is necessary to use match-casting method for precast post-tensioned column assembly. Preassembly using 3D models created by image processing can minimize construction error. Augmented reality devices are used to check the geometry of the segment. Laboratory-scale tests were performed. The proposed process has been applied to the real precast bridge pier segments.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2001년도 ICCAS
• /
• pp.164.3-164
• /
• 2001

In this paper, we present a method for obtaining the actual modeling data through reconstructing a 3D data from an image and a method of estimating the geometrical information and the camera location of architectural objects from a photograph containing a virtual environment are introduced. Our approach combines both geometry-based and image-based modeling techniques. The modeling system is effective and robust because it exploits constrains that are characteristic of architectural scenes. Our approach can recover models for use in either geometry-based rendering systems. We present results that demonstrate out approach´s ability to create realistic renderings of architectural scenes from viewpoints far from the original photographs.

• 한국소성가공학회:학술대회논문집
• /
• 한국소성가공학회 1999년도 춘계학술대회논문집
• /
• pp.269-272
• /
• 1999

It is still hard to measure the strain distribution over entire surface of a medium or large-size stamped part even by using an automated strain measurement system. Several methods which enable to enlarge the measurement range without losing accuracy and precision are suggested in this study, The superposition of images having different high-lightened or damaged part each other results in an enhanced image. A new method for constructing the element connectivity from a line-thinned image makes it possible to identify up to 1,000 elements. And the geometry assembling algorithm is proved very efficient in which the whole area to be measured is divided into several parts ; the coordinate transformation between every two adjacent parts is obtained from the concept of the least square error ; and the 3-D shape or strain distribution over the whole surface is assembled,

• International Journal of Control, Automation, and Systems
• /
• 제5권6호
• /
• pp.663-673
• /
• 2007

This paper presents a new moving obstacle detection method using an optical flow in mobile robot with an omnidirectional camera. Because an omnidirectional camera consists of a nonlinear mirror and CCD camera, the optical flow pattern in omnidirectional image is different from the pattern in perspective camera. The geometry characteristic of an omnidirectional camera has influence on the optical flow in omnidirectional image. When a mobile robot with an omnidirectional camera moves, the optical flow is not only theoretically calculated in omnidirectional image, but also investigated in omnidirectional and panoramic images. In this paper, the panoramic image is generalized from an omnidirectional image using the geometry of an omnidirectional camera. In particular, Focus of expansion (FOE) and focus of contraction (FOC) vectors are defined from the estimated optical flow in omnidirectional and panoramic images. FOE and FOC vectors are used as reference vectors for the relative evaluation of optical flow. The moving obstacle is turned out through the relative evaluation of optical flows. The proposed algorithm is tested in four motions of a mobile robot including straight forward, left turn, right turn and rotation. The effectiveness of the proposed method is shown by the experimental results.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2004년도 Proceedings of ISRS 2004
• /
• pp.125-127
• /
• 2004

High-resolution satellite images offer abundance information of the earth surface for remote sensing applications. The information includes geometry, texture and attribute characteristic. The pixel-based image classification can't satisfy high-resolution satellite image's classification precision and produce large data redundancy. Object-oriented information extraction not only depends on spectrum character, but also use geometry and structure information. It can provide an accessible and truly revolutionary approach. Using Beijing Spot 5 high-resolution image and object-oriented classification with the eCognition software, we accomplish the cultures' precise classification. The test areas have five culture types including water, vegetation, road, building and bare lands. We use nearest neighbor classification and appraise the overall classification accuracy. The average of five species reaches 0.90. All of maximum is 1. The standard deviation is less than 0.11. The overall accuracy can reach $95.47\%.$ This method offers a new technology for high-resolution satellite images' available applications in remote sensing culture classification.