• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.2
• /
• pp.109-121
• /
• 2020

Depth calculation of objects in a scene from images is one of the most studied processes in the fields of image processing, computer vision, and photogrammetry. Conventionally, depth is calculated using a pair of overlapped images captured at different view points. However, there have been studies to calculate depths from a single image. Theoretically, it is known to be possible to calculate depth using the diameter of CoC (Circle of Confusion) caused by defocus under the assumption of a thin lens model. Thus, this study aims to verify the validity of the thin lens model to calculate depth from edge blur amount which corresponds to the radius of CoC. For this study, a commercially available DSLR (Digital Single Lens Reflex) camera was used to capture a set of target sheets which had different edge contrasts. In order to find out the pattern of the variations of edge blur against varying combination of FD (Focusing Distance) and OD (Object Distance), the camera was set to varying FD and target sheet images were captured at varying OD under each FD. Then, the edge blur and edge displacement were estimated from edge slope profiles using a brute-force method. The experimental results show that the pattern of the variations of edge blur observed in the target images was apart from their corresponding theoretical amounts derived under the thin lens assumption but can still be utilized to calculate depth from a single image for the cases similar to the limited conditions experimented under which the tendency between FD and OD is manifest.

• Proceedings of the KSRS Conference
• /
• 2002.10a
• /
• pp.516-516
• /
• 2002

With the wide usage of LiDAR data and high-resolution satellite image, 3D modeling of buildings in urban areas has become an important research topic in the photogrammetry and computer vision field for many years. However the previous modeling has its limitations of merely texturing the image to the DSM surface of the study area and does not represent the relief of building surfaces. This study is focused on presenting a system of realistic 3D building modeling from consecutive stereo image sequences using digital camera. Generally when acquiring images through camera, various parameters such as zooming, focus, and attitude are necessary to extract accurate results, which in certain cases, some parameters have to be rectified. It is, however, not always possible or practical to precisely estimate or rectify the information of camera positions or attitudes. In this research, we constructed the collinearity condition of stereo images through extracting the distinctive points from stereo image sequence. In addition, we executed image matching with Graph Cut method, which has a very high accuracy. This system successfully performed the realistic modeling of building with a good visual quality. From the study, we concluded that 3D building modeling of city area could be acquired more realistically.

• 한국HCI학회:학술대회논문집
• /
• 2006.02a
• /
• pp.711-716
• /
• 2006

Since the introduction of digital camera to the mass market, the number of digital photos owned by an individual is growing at an alarming rate. This phenomenon naturally leads to the issues of difficulties while searching and browsing in the personal digital photo archive. Traditional approach typically involves content-based image retrieval using computer vision algorithms. However, due to the performance limitations of these algorithms, at least on the casual digital photos taken by non-professional photographers, more recent approaches are centered on time-based clustering algorithms, analyzing the shot times of photos. These time-based clustering algorithms are based on the insight that when these photos are clustered according to the shot-time similarity, we have "event clusters" that will help the user browse through her photo archive. It is also reported that one of the remaining problems with the time-based approach is that people perceive events in different scales. In this paper, we present an adaptive time-based clustering algorithm that exploits the usage history of digital photos in order to infer the user's preference on the event granularity. Experiments show significant performance improvements in the clustering accuracy.

• Journal of Korea Multimedia Society
• /
• v.13 no.4
• /
• pp.512-525
• /
• 2010

A PDP(Plasma Display Panel) Frame is critical part of PDP and also produces couple hundred thousand every month. In the process of mass production, product inspection is very important process. Also to increase the reliance, inspection each part and every final product is asked quite often. Purpose of this paper is to use computer vision system to inspect the PDP parts which is Automated visual process inspection. This paper contains the system design for inspecting defects of hole, tab, stud, rivet of PDP Frame. The system also can inspect various kinds of PDP frames. Quick and accurate 100% inspection of all shapes can improve the manufacturing productivity. Inspection results can be stored in a database and analyzed to find the cause of defects. After applying the system to the industry, the result shows the possibility of fast and accuracy of the inspection.

• Journal of the Korea Convergence Society
• /
• v.11 no.6
• /
• pp.1-6
• /
• 2020

In this study, research on character recognition, which is one of the fields of computer vision, was conducted. Optical character recognition, which is one of the most widely used character recognition techniques, suffers from decreasing recognition rate if the recognition target deviates from a certain standard and format. Hence, this study aimed to address this limitation by applying deep learning techniques to character recognition. In addition, as most character recognition studies have been limited to English or number recognition, the recognition range has been expanded through additional data training on Korean text. As a result, this study derived a deep learning-based character recognition algorithm for Korean text recognition. The algorithm obtained a score of 0.841 on the 1-NED evaluation method, which is a similar result to that of English recognition. Further, based on the analysis of the results, major issues with Korean text recognition and possible future study tasks are introduced.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.11 no.9
• /
• pp.901-908
• /
• 2016

Edge is the basic characteristic of image, edge detection is very important in image processing applications and computer vision area. Many studies are being performed to detect these edges by domestic and foreign researchers. The conventional edge detection methods such as Roberts, Sobel, Prewitt, and Laplacian etc, which are using a fixed value of mask are widely used and morphological gradient which uses dilation and erosion among morphology process techniques is also widely used. But these methods does not detect edges well in the diagonal direction or gradually changing image parts. Accordingly, in this paper, the modified top-hat and bottom-hat transform algorithms which are detecting edges well in the parts of diagonal direction or gradually changing image are proposed. The proposed algorithms present the detected edge images compared with the conventional methods and are evaluated performance by using cosine similarity.

• Journal of Electrical Engineering and Technology
• /
• v.6 no.6
• /
• pp.806-816
• /
• 2011

Interest about social security has recently increased in favor of safety for infrastructure. In addition, advances in computer vision and pattern recognition research are leading to video-based surveillance systems with improved scene analysis capabilities. However, such video surveillance systems, which are controlled by human operators, cannot actively cope with dynamic and anomalous events, such as having an invader in the corporate, commercial, or public sectors. For this reason, intelligent surveillance systems are increasingly needed to provide active social security services. In this study, we propose a core technique for intelligent surveillance system that is based on swarm robot technology. We present techniques for invader enclosing using swarm robots based on multiple distributed object environment. The proposed methods are composed of three main stages: location estimation of the object, specified object tracking, and decision of the cooperative behavior of the swarm robots. By using particle filter, object tracking and location estimation procedures are performed and a specified enclosing point for the swarm robots is located on the interactive positions in their coordinate system. Furthermore, the cooperative behaviors of the swarm robots are determined via the result of path navigation based on the combination of potential field and wall-following methods. The results of each stage are combined into the swarm robot-based invader-enclosing technique on multiple distributed object environments. Finally, several simulation results are provided to further discuss and verify the accuracy and effectiveness of the proposed techniques.

• Journal of Korea Multimedia Society
• /
• v.12 no.4
• /
• pp.473-482
• /
• 2009

Reconstructed 3D data from computer vision includes necessarily a noise or an error. When these data goes through a mesh process, the different 3D mesh data from original shape comes to make by a noise or an error. This paper proposed the method that smooths a noise effectively by noise analysis in reconstructed 3D data. Because the proposed method is smooths a noise using the area ratio of the mesh, the pre-processing of unusable mesh is necessary in 3D mesh data. This study detects a peak noise and Gaussian noise using the ratio of 3D volume and 2D area of mesh and smooths the noise with respect of its characteristics. The experimental results using synthetic and real data demonstrated the efficacy and performance of proposed algorithm.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.1
• /
• pp.142-152
• /
• 2016

GPS is not able to be used for indoor positioning and currently most of techniques emerging to overcome the limit of GPS utilize private wireless networks. However, these methods require high costs for installation and maintenance, and they are inappropriate to be used in the place where precise positioning is needed as in indoor parking lots. This paper proposes a vehicular indoor positioning method based on QR-code recognition. The method gets an absolute coordinate through QR-code scanning, and obtain the location (an relative coordinate) of a black-box camera using the tilt and roll angle correction through affine transformation, scale transformation, and trigonometric function. Using these information of an absolute coordinate and an relative one, the precise position of a car is estimated. As a result, average error of 13.79cm is achieved and it corresponds to just 27.6% error rate in contrast to 50cm error of the recent technique based on wireless networks.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.12
• /
• pp.2844-2851
• /
• 2015

How to recover from the multiple 2D images into 3D object has been widely studied in the field of computer vision. In order to improve the accuracy of the recovered 3D shape, it is more important that noise must be minimized and the number of image frames must be guaranteed. However, potential noise is implied when tracking feature points. And the number of image frames which is consisted of an observation matrix usually decrease because of tracking failure, occlusions, or low image resolution, and so on. Therefore, it is obviously essential that the number of image frames must be secured by recovering the missing feature points under noise. Thus, we propose the analytic approach which can control directly the error distance and orientation of missing feature point by the geometrical properties under noise distribution. The superiority of proposed method is demonstrated through experimental results for synthetic and real object.