• 한국콘텐츠학회논문지
• /
• 제16권8호
• /
• pp.395-403
• /
• 2016

최근 들어 영상처리는 여러 분야에서 사용되어지고 있다. 영상처리에서 많이 연구되어지고 있는 기술은 실시간으로 객체를 추적하는 기술이다. 객체를 추적하는 방법은 보행자를 추적하는 HOG(Histogram of Oriented Gradients), 전경과 배경 분리 방법을 사용하는 Codebook 같은 방법 들이 많이 알려져 있다. 그러나 객체가 움직이거나 동적인 배경, 조명변화가 심할 경우 객체 추출이 어려워진다. 본 논문에서는 ROI(Region of Interest)기반 깊이영상과 컬러영상의 특징을 이용해 객체를 추출하는 방법을 제안한다. 첫 번째, 깊이 영상에서 배경분리를 통해 객체의 위치를 찾아 ROI로 설정해준다. 두 번째, 컬러영상을 이용하여 영상의 특징점을 찾는다. 세 번째, 특징점과 객체의 볼록헐(convex hull) 구성점들을 이용하여 새로운 윤곽을 만들어 더 정확한 객체를 추출하도록 한다. 마지막으로 본 논문에서 제안한 방법과 기존 방법과의 비교를 통해 제안한 방법의 결과가 좀 더 정확한 객체를 추출하고 있음을 검증하였다.

• 전기학회논문지
• /
• 제68권1호
• /
• pp.182-188
• /
• 2019

Autonomous mobile robots need SLAM (simultaneous localization and mapping) to look for the location and simultaneously to make the map around the location. In order to achieve visual SLAM, it is necessary to form an algorithm that detects and extracts feature points from camera images, and gets the camera pose and 3D points of the features. In this paper, we propose MPROSAC algorithm which combines MSAC and PROSAC, and compare the performance of optimization method and the filtering method for feature-based monocular visual SLAM. Sparse Bundle Adjustment (SBA) is used for the optimization method and the extended Kalman filter is used for the filtering method.

• 전기학회논문지
• /
• 제59권12호
• /
• pp.2314-2318
• /
• 2010

For human-robot interaction, a robot should recognize the meaning of human behavior. In the case of static behavior such as face expression and sign language, the information contained in a single image is sufficient to deliver the meaning to the robot. In the case of dynamic behavior such as gestures, however, the information of sequential images is required. This paper proposes behavior classification by using fuzzy classifier to deliver the meaning of dynamic behavior to the robot. The proposed method extracts feature points from input images by a skeleton model, generates a vector space from a differential image of the extracted feature points, and uses this information as the learning data for fuzzy classifier. Finally, we show the effectiveness and the feasibility of the proposed method through experiments.

• 한국측량학회지
• /
• 제35권1호
• /
• pp.41-53
• /
• 2017

A thermal camera is used to obtain thermal information of a certain area. However, it is difficult to depict all the information of an area in an individual thermal image. To form a high-resolution panoramic thermal image, we propose an optimized FAST (feature from accelerated segment test) algorithm to combine two or more images of the same scene. The FAST is an accurate and fast algorithm that yields good positional accuracy and high point reliability; however, the major limitation of a FAST detector is that multiple features are detected adjacent to one another and the interest points cannot be obtained under no significant difference in thermal images. Our proposed algorithm not only detects the features in thermal images easily, but also takes advantage of the speed of the FAST algorithm. Quantitative evaluation shows that our proposed technique is time-efficient and accurate. Finally, we create a mosaic of the video to analyze a comprehensive view of the scene.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2003년도 하계종합학술대회 논문집 Ⅳ
• /
• pp.1887-1890
• /
• 2003

This paper presents a three-dimensional (3D) head pose estimation algorithm using the stereo image. Given a pair of stereo image, we automatically extract several important facial feature points using the disparity map, the gabor filter and the canny edge detector. To detect the facial feature region , we propose a region dividing method using the disparity map. On the indoor head & shoulder stereo image, a face region has a larger disparity than a background. So we separate a face region from a background by a divergence of disparity. To estimate 3D head pose, we propose a 2D-3D Error Compensated-SVD (EC-SVD) algorithm. We estimate the 3D coordinates of the facial features using the correspondence of a stereo image. We can estimate the head pose of an input image using Error Compensated-SVD (EC-SVD) method. Experimental results show that the proposed method is capable of estimating pose accurately.

• Journal of Electrical Engineering and Technology
• /
• 제11권6호
• /
• pp.1846-1856
• /
• 2016

This study proposes a novel feature point initialization method in order to improve the accuracy of feature point positions by fusing a vision sensor and a lidar. The initialization is a process that determines three dimensional positions of feature points through two dimensional image data, which has a direct influence on performance of a 6-DoF bearing-only SLAM. Prior to the initialization, an extrinsic calibration method which estimates rotational and translational relationships between a vision sensor and lidar using multiple calibration tools was employed, then the feature point initialization method based on the estimated extrinsic calibration parameters was presented. In this process, in order to improve performance of the accuracy of the initialized feature points, an iterative automatic scaling parameter tuning technique was presented. The validity of the proposed feature point initialization method was verified in a 6-DoF bearing-only SLAM framework through an indoor and outdoor tests that compare estimation performance with the previous initialization method.

• 반도체디스플레이기술학회지
• /
• 제23권1호
• /
• pp.56-60
• /
• 2024

In this paper, we propose a SIFT feature-based dynamic stitching algorithm for image calibration and correction of a 360-degree surround view system. The existing surround view system requires a lot of processing time and money because in the process of image calibration and correction. The traditional marker patterns are placed around the vehicle and correction is performed manually. Therefore, in this study, images captured with four fisheye cameras mounted on the surround view system were distorted and then matched with the same feature points in adjacent images through SIFT-based feature point extraction to enable image stitching without a fixed marker pattern.

• 대한전자공학회논문지SP
• /
• 제37권6호
• /
• pp.25-37
• /
• 2000

특징점의 궤적은 인접한 프레임에 존재하는 특정점 사이의 대응관계로 정의할 수 있다. 실제 영상열에서 존재할 수 있는 잘못된 특징점(false positive, false negative)들은 특징점의 대응관계를 결정할 때 많은 문제를 야기하기 때문에 특징점의 대응관계를 찾는 문제는 어려운 문제로 알려져 있다. 본 논문에서는 새로운 궤적의 나타남, 사라짐 등 불완전한 궤적을 갖는 특징점들을 고려하는 특징점 추적기법을 제안한다. 정합 척도로서 가중치가 부여된 유클리디언 거리를 사용하고 특징점의 운동특성을 잘 반영할 수 있도록 그 가중치를 자동으로 조정한다. 대응점 탐색과정에서 치명적인 영향을 줄 수 있는 애매한 특징점이 존재하는 경우를 고려하여 인접한 프레임 사이의 정합점 결정을 그래프에 의한 최적 대응점 탐색문제로 해결한다. 제안하는 대응점 탐색 알고리즘은 실제 영상열에서 나타날 수 있는 잘못된 특징점들이 대응관계를 결정할 때 주는 영향을 최소화하기 위하여 국부 최적(local optimal)을 찾게되며, 인접한 두 프레임에 m, n개의 특징점이 주어졌을 경우, 최선의 경우 O(mn), 최악의 경우 O($m^2n$)의 계산량을 필요로 한다. 제안하는 알고리즘은 정합과정에서 잘못된 특징점을 고려하고, 특징점의 운동특성을 잘 반영함으로써 대량의 특징점을 추적하는데도 충분히 적용할 수 있음을 실험을 통해 확인하였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.793-797
• /
• 2003

In this paper, we present a self-localization method for mobile robots using vertical line features of indoor environment. When a 2D map including feature points and color information is given, a mobile robot moves to the destination, and acquires images by one camera from the surroundings having vertical line edges. From the image, vertical line edges are detected, and pattern vectors meaning averaged color values of the left and right region of each line segment are computed. The pattern vectors are matched with the feature points of the map using the color information and the geometrical relationship of the points. From the perspective transformation of the corresponded points, nonlinear equations are derived. Localization is carried out from solving the equations by using Newton's method. Experimental results show that the proposed method using mono view is simple and applicable to indoor environment.

• International Journal of Contents
• /
• 제13권4호
• /
• pp.70-79
• /
• 2017

Images are an important element in patents and many experts use images to analyze a patent or to check differences between patents. However, there is little research on image analysis for patents partly because image processing is an advanced technology and typically patent images consist of visual parts as well as of text and numbers. This study suggests two methods for using image processing; the Scale Invariant Feature Transform(SIFT) algorithm and Optical Character Recognition(OCR). The first method which works with SIFT uses image feature points. Through feature matching, it can be applied to calculate the similarity between documents containing these images. And in the second method, OCR is used to extract text from the images. By using numbers which are extracted from an image, it is possible to extract the corresponding related text within the text passages. Subsequently, document similarity can be calculated based on the extracted text. Through comparing the suggested methods and an existing method based only on text for calculating the similarity, the feasibility is achieved. Additionally, the correlation between both the similarity measures is low which shows that they capture different aspects of the patent content.