• 센서학회지
• /
• 제33권3호
• /
• pp.119-124
• /
• 2024

Effective fire sensing is important to protect lives and property from the disaster. In this paper, we present an intelligent visual sensing method for detecting fires based on machine learning techniques. The proposed method involves a two-step process. In the first step, fire and non-fire images are used to train a convolutional neural network (CNN), and in the next step, feature vectors consisting of 256 values obtained from the CNN are used for the learning of a support vector machine (SVM). Linear and nonlinear SVMs with different parameters are intensively tested. We found that the proposed hybrid method using an SVM with a linear kernel effectively increased the recall rate of fire image detection without compromising detection accuracy when an imbalanced dataset was used for learning. This is a major contribution of this study because recall is important, particularly in the sensing of disaster situations such as fires. In our experiments, the proposed system exhibited an accuracy of 96.9% and a recall rate of 92.9% for test image data.

• 한국정밀공학회지
• /
• 제26권6호
• /
• pp.68-73
• /
• 2009

One of the traditional optical methods to monitor a tool is a CCD sensor-based vision system which captures an aspect of the tool in real time. In the case using the CCD sensor, specific lens-modules are necessary to monitor the tool with higher resolution than its pixel size, and a microprocessor is required to attain desired data from captured images. Thus theses additional devices make the entire measurement system complex. Another method is to use a pair of an optical source and a detector per measuring axis. Since the method is based on the intensity modulation, the structure of the measurement system is simper than the CCD sensor-based vision system. However, in the case measuring the three dimensional position of the tool, it is difficult to apply to micro machine-tools because there may not be space to integrate three pairs of an optical source and a detector. In this paper, in order to develop a tool-origin measurement system which is employed in micro machine-tools, the improved method to measure a tool origin in x, y and z axes is introduced. The method is based on the intensity modulation and employs one pair of an optical source radiating divergent beams and a quadrant photodiode to detect a three dimensional position of the tool. This paper presents the measurement models of the proposed tool-origin sensor. The models were verified experimentally The verification results show that the proposed method is possible and the induced models are available for design.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제55권4호
• /
• pp.161-164
• /
• 2006

In this thesis we designed the vision system to inspect the defect of a back light unit of plat panel display device. The vision system is divided into hardware and inspection algorithm of defect. Hardware components consist of illumination part, robot-arm controller part and image-acquisition part. Illumination part is made of acrylic panel for light diffusion and five 36W FPL's(Fluorescent Parallel Lamp) and electronic ballast with low frequency harmonics. The CCD(Charge-Coupled Device) camera of image-acquisition part is able to acquire the bright image by the light coming from lamp. The image-acquisition part is composed of CCD camera and frame grabber. The robot-arm controller part has a role to let the CCD camera move to the desired position. To take inspections of surface images of a flat panel display it can be controlled and located every nook and comer. Images obtained by robot-arm and image-acquisition board are saved on the hard-disk through windows programming and are tested whether there are defects by using the image processing algorithms.

• 한국공작기계학회:학술대회논문집
• /
• 한국공작기계학회 2004년도 춘계학술대회 논문집
• /
• pp.225-230
• /
• 2004

A cold ＆ hot rolling coil production line of iron nill consists of a kind of coherent automatic process, but an automatic labelling process still had technical difficulties in the automation of its process. The reason for difficulties in building an automatic process is that quantitative data for each rolled coil from every shipping is not easy to receive from the previous process. it is not possible to apply for a general and simple purpose robot that is actually worked through a taught position to the process because the size and direction of the coi1 has differed on every shipping. From these reasons. we introduce a robot vision system to accept an expected variable situation and to ensure the stability and flexibility of the process. This paper examines a study applied for similar cases and finds the position and direction of relied coil using the moment invariant algorithm proposed by Hu. In addition. the camera calibration and position error compensation algorithm is applied by the analysis of the relationship of transition in a space coordinate system. The construction of a robot vision system proposed by this paper is a more intellectual system than that of the automatic labelling system. which is already used to the Daihen steel nill of NEW JAPAN steel mill co. Ltd in Japan, and shows a better independent operation in the field of production.

• 한국정보통신학회논문지
• /
• 제5권6호
• /
• pp.1157-1162
• /
• 2001

본 논문에서는 영상 처리 기법을 이용하여 칼라 팔레트 외형상의 불량품을 식별하는 연구를 수행하고자 한다 먼저 기본적인 팔레트 자동선별시스템의 필요성에 대해 기술하며, 각 샘플링된 팔레트에 대해 영상처리기법을 이용한 불순물 검출 알고리즘을 제안하고자 한다. 또한 이를 상용화할 수 있도록 윈도우환경의 비전처리 프로그램을 제시하였다. 끝으로 본 연구에 대한 평가와 앞으로의 연구과제에 대해 기술하고자 한다.

• 한국공작기계학회논문집
• /
• 제16권1호
• /
• pp.60-68
• /
• 2007

It has been conducted by laser vision sensor for weldability estimation of front-bead after doing high speed butt laser welding of any condition. It has been developed a real time GUI(Graphic User Interface) system for weldability application in the basis of texts and field qualify levels. In the reference of bead imperfections, defects absolute position and defects intensity index of front-bead in the basis of formability reference, it has been produced a weldability estimation and defects intensity index of back-bead by back propagation neural network. In the result of by comparing measuring data by laser vision sensor of back-bead and data by back propagation neural network of one, it has been shown the similar results. Finally, under knowledge of welding condition in production line, it has been conducted a weldability estimation of back-bead only in knowledge of informations of front-bead data without using laser vision sensor or welding inspection experts and furthermore it can be used data for final inspection results of back-bead.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제8권11호
• /
• pp.4103-4117
• /
• 2014

Image processing and computer vision algorithms are gaining larger concern in a variety of application areas such as robotics and man-machine interaction. Vision allows the development of flexible, intelligent, and less intrusive approaches than most of the other sensor systems. In this work, we determine the location and orientation of a mobile robot which is crucial for performing its tasks. In order to be able to operate in real time there is a need to speed up different vision routines. Therefore, we present and evaluate a method for introducing parallelism into the multiple non-overlapping camera pose estimation algorithm proposed in [1]. In this algorithm the problem has been solved in real time using multiple non-overlapping cameras and the Extended Kalman Filter (EKF). Four cameras arranged in two back-to-back pairs are put on the platform of a moving robot. An important benefit of using multiple cameras for robot pose estimation is the capability of resolving vision uncertainties such as the bas-relief ambiguity. The proposed method is based on algorithmic skeletons for low, medium and high levels of parallelization. The analysis shows that the use of a multiprocessor system enhances the system performance by about 87%. In addition, the proposed design is scalable, which is necaccery in this application where the number of features changes repeatedly.

• 디지털융복합연구
• /
• 제12권11호
• /
• pp.349-355
• /
• 2014

본 논문은 인간의 생물학적 시각정보처리과정 특성과 웨이블릿을 이용한 인공시각시스템을 제안한다. 인공시각시스템은 인간의 생물학적 시각정보 처리과정을 이용하며 맹인의 인공시각시스템 제작 및 일반적인 인식시스템에 적용할 수 있다. 본 논문에서는 인간의 망막에서 신경절 세포까지 정보 처리과정을 모델링하여 구성하였고 신경절 세포에서 후두엽 초기시각피질까지 전달되는 정보 복원과정을 모델링하여 초기시각피질에 표현되는 영상정보를 구성하였다. 실험과정은 인간의 시각정보 처리과정 특성인 저주파, 고주파 분해를 웨이블릿 변환으로 시각 정보를 구현하였고 이를 이용하여 인식 시스템에 적용하였다. 실험에 사용한 데이터는 AT&T 얼굴데이터베이스를 사용하였다. 그리고 제안하는 인간의 시각정보처리 과정 특성을 이용한 방법이 영상인식 시스템의 정확성을 향상시킬 수 있음을 실험을 통하여 증명하고자 한다.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 2001년도 추계종합학술대회
• /
• pp.587-591
• /
• 2001

본 연구에서는 영상 처리를 이용하여 외형적인 칼라 팔레트내의 불량품을 식별하는 실험을 수행하고자 한다. 기본적인 팔레트 자동선별시스템에 대해 기술하며, 각 샘플링된 팔레트에 대해 영상처리기법을 이용한 불순물 검출 알고리듬을 제안하고자 한다. 또한 이를 상용화할 수 있도록 윈도우환경의 비전처리 프로그램을 제시하였다. 끝으로 본 연구에 대한 평가와 앞으로의 연구과제에 대해 기술하고자 한다.

• 전자공학회논문지 IE
• /
• 제48권1호
• /
• pp.16-24
• /
• 2011

본 논문에서는 트림 T형 용접너트 생산 시스템 중 불량품을 자동으로 선별할 수 있는 자동 선별기를 비전 검사기를 사용하여 개발하였다. 카메라로부터 입력되는 영상 신호에 대해 히스토그램을 활용한 경계 판별 및 나사산 검출, 이진 모폴로지 연산(Binary morphology operation)을 활용한 얼룩 검출 등의 기법을 활용하였다. 기존의 검사기나 육안 검사에서 불가능하였던 수치 검사를 0.1mm의 정밀도로 검사할 수 있도록 하였으며, 이를 통해 제조단가를 25% 절감하고 생산성을 330% 이상 향상시킬 수 있었다.