• Journal of information and communication convergence engineering
• /
• v.16 no.3
• /
• pp.173-178
• /
• 2018

Accurate classification of cloud images is a challenging task. Almost all the existing methods rely on hand-crafted feature extraction. Their limitation is low discriminative power. In the recent years, deep learning with convolution neural networks (CNNs), which can auto extract features, has achieved promising results in many computer vision and image understanding fields. However, deep learning approaches usually need large datasets. This paper proposes a deep learning approach for classification of cloud image patches on small datasets. First, we design a suitable deep learning model for small datasets using a CNN, and then we apply data augmentation and dropout regularization techniques to increase the generalization of the model. The experiments for the proposed approach were performed on SWIMCAT small dataset with k-fold cross-validation. The experimental results demonstrated perfect classification accuracy for most classes on every fold, and confirmed both the high accuracy and the robustness of the proposed model.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.63.1-63
• /
• 2001

At 3-D vision measuring, the camera calibration is necessary to calculate parameters accurately. Camera calibration was developed widely in two categories. One is that establishes reference points in space, and the other is that uses the grid type frame and statistical method. But, the former has difficult to setup reference points and the latter has low accuracy. In this paper we present an algorithm for camera calibration using perspective ratio of the grid type frame with different line widths. The advantage of this algorithm is that it can estimate position, pose and distance between camera and object ...

• Journal of Institute of Control, Robotics and Systems
• /
• v.16 no.4
• /
• pp.368-373
• /
• 2010

A new indoor mobile robot localization method is presented. Robot recognizes well designed single color landmarks on the ceiling by vision system, as reference to compute its precise position. The proposed likelihood prediction based method enables the robot to estimate its position based only on the orientation of landmark.The use of single color landmarks helps to reduce the complexity of the landmark structure and makes it easily detectable. Edge based optical flow is further used to compensate for some landmark recognition error. This technique is applicable for navigation in an unlimited sized indoor space. Prediction scheme and localization algorithm are proposed, and edge based optical flow and data fusing are presented. Experimental results show that the proposed method provides accurate estimation of the robot position with a localization error within a range of 5 cm and directional error less than 4 degrees.

• Electronics and Telecommunications Trends
• /
• v.13 no.6 s.54
• /
• pp.13-33
• /
• 1998

최근 2, 3년 사이에 국내에서도 많은 병렬 머신이 도입되면서 병렬처리에 대한 관심이 높아지고 있다. 본 고에서는 미국에서 최근 고성능 컴퓨팅 기술 개발 사업의 일환으로 추진하고 있는 Grand Challenge Problems에 속하지만 다른 과학계산 응용과는 특성이 다른 컴퓨터 비전 태스크를 병렬화 하는 여러 가지 방법에 대해 살펴본다. 먼저 컴퓨터 비전 태스크와 이를 병렬화 할 때 일반적인 특징에 대해서 설명한다. 그리고 하위 레벨(low-level), 중간 레벨(intermediate-level), 상위 레벨(high-level) 태스크 각각을 예로 들면서 병렬처리 방법에 대해 설명한 후, 여러 레벨의 비전 태스크를 종합적으로 병렬화 할 때 제기되는 문제로서 태스크 병렬성(task parallelism) 및 이질적 처리(heterogeneous processing)에 대해서 알아본다. 마지막으로 이러한 컴퓨터 비전 태스크에서의 여러가지 병렬처리에 대한 벤치마크에 대하여 살펴본다.

• Journal of Advanced Marine Engineering and Technology
• /
• v.32 no.1
• /
• pp.184-191
• /
• 2008

In order for a mobile robot to move under unknown or uncertain environment, it must have an environmental information. In collecting environmental information, the mobile robot can use various sensors. In case of using ultrasonic sensors to collect an environmental information, it is able to comprise a low-cost environmental recognition system compared with using other sensors such as vision and laser range-finder. This paper proposes a map building algorithm which can collect environmental information using ultrasonic sensors. And also this paper suggests a traveling algorithm using environmental information which leads to the map building algorithm. In order to accomplish the proposed traveling algorithm, this paper additionally discusses a position revision algorithm.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• v.19 no.4
• /
• pp.16-23
• /
• 2020

An electric vehicle operates at high currents and requires real-time monitoring of the entire system for ensuring efficiency and safety of the vehicle. Current sensors are applied to drive the motors, inverters, and battery control systems, and are the key components to ensure constant monitoring of the magnitude and waveforms of the operating current. In this study, a heat treatment process condition to influence the performance of Permalloy current sensors was developed; the correlation between the output capacity, low-temperature characteristics, and high-temperature characteristics of the current sensor was studied; and the process was optimized to meet the required output accuracy and temperature characteristics.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.27 no.6
• /
• pp.972-981
• /
• 1990

Many researchers have been interested in whether complex computational problems can be solved by the neural net or not. Especially, problems of early vision are integrated by Tikhonov's regularization theory. Regularization technique can be realized in resistive net. In this paper, we suggest the resistive net with upper and lower thresholder to be able to compute shape from shading and to solve its discontinuous problem. We simulate three algorithms-Horn's algorithm, resistive net and up-low thrwsholding net -with respect to three cases-fully boundary, boundary losing partly and noisy image. As being able to cope with crease and discontinuous parts, we get the good 3D shape from shading.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.1
• /
• pp.224-241
• /
• 2015

Multiview plus depth (MVD) videos are widely used in free-viewpoint TV systems. The best-known technique to determine depth information is based on stereo vision. In this paper, we propose a novel local stereo matching algorithm which is radiometric invariant. The key idea is to use a combined matching cost of intensity and gradient based similarity measure. In addition, we realize an adaptive cost aggregation scheme by constructing an adaptive support window for each pixel, which can solve the boundary and low texture problems. In the disparity refinement process, we propose a four-step post-processing technique to handle outliers and occlusions. Moreover, we conduct stereo reconstruction tests to verify the performance of the algorithm more intuitively. Experimental results show that the proposed method is effective and robust against local radiometric distortion. It has an average error of 5.93% on the Middlebury benchmark and is compatible to the state-of-art local methods.

• Proceedings of the KSME Conference
• /
• 2008.11a
• /
• pp.1255-1258
• /
• 2008

Roll to roll (R2R) system, known as 'web processing', is the process of producing these electronic devices on a roll of flexible plastic. With the need of improved performance and productivity in R2R industry, effective control and on-line supervision for web quality is essential. In this report, we present a system for on-line measurement of wrinkles, one of defects incurring due to compressive stresses developed in the web. This system is able to capture an image generated when a well defined line shape laser beam passes through a transparent web. The system calculates 3D shape information, including the height of the wrinkle on the web, and displays the images for the shape information of the web in real time. By using area scan camera and machine vision laser, this system takes more advantages of setting up as a simple and low cost system compared to the line scan camera systems that widely used in web manufacturing. Specific calibration method and analysis on the achievable accuracy will be discussed.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.148-153
• /
• 2009

Few methods have dealt with segmenting multiple images with analogous content. Concurrent images of a scene and gathered images of a similar foreground are examples of these images, which we term consistent scene images. In this paper, we present a method to segment these images based on manual segmentation of one image, by iteratively propagating information via multi-level cues with adaptive confidence. The cues are classified as low-, mid-, and high- levels based on whether they pertain to pixels, patches, and shapes. Propagated cues are used to compute potentials in an MRF framework, and segmentation is done by energy minimization. Through this process, the proposed method attempts to maximize the amount of extracted information and maximize the consistency of segmentation. We demonstrate the effectiveness of the proposed method on several sets of consistent scene images and provide a comparison with results based only on mid-level cues [1].