• Journal of Korea Multimedia Society
• /
• v.20 no.5
• /
• pp.769-781
• /
• 2017

Deep learning networks like Convolutional Neural Networks (CNNs) show successful performances in many computer vision applications such as image classification, object detection, and so on. For implementation of deep learning networks in embedded system with limited processing power and memory, deep learning network may need to be simplified. However, simplified deep learning network cannot learn every possible scene. One realistic strategy for embedded deep learning network is to construct a simplified deep learning network model optimized for the scene images of the installation place. Then, automatic training will be necessitated for commercialization. In this paper, as an intermediate step toward automatic training under fisheye camera environments, we study more precise human localization in fisheye images, and propose an accurate human localization method, Automatic Ground-Truth Labelling Method (AGTLM). AGTLM first localizes candidate human object bounding boxes by utilizing GoogLeNet-LSTM approach, and after reassurance process by GoogLeNet-based CNN network, finally refines them more correctly and precisely(tightly) by applying saliency object detection technique. The performance improvement of the proposed human localization method, AGTLM with respect to accuracy and tightness is shown through several experiments.

• Proceedings of the KIEE Conference
• /
• 2005.10b
• /
• pp.64-66
• /
• 2005

In this paper, some automatic inspection algorithms for LCD module are suggested. Human eyes are very good for the inspection in many industrial areas. However, very bright LCD back lighting may cause permanent damage to the human eyes. Also, the growing size of the LCD make it more difficult for the human inspectors. Therefore, using camera set, automatic inspection process becomes very essential for the future LCD industry.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2000.10a
• /
• pp.51-56
• /
• 2000

This study is to implement automatic inspection system of motor body panel by visual system. since it is inefficient to examine adhesion condition of components with the naked eye, image processing algorithm is presented to replace the existing manual inspection process with an automatic inspection process the developed automatic inspection system presents tile examination method of the adhesion condition of components with a pixed camera, which leads to an increase of the productivity and a reduction of manufacturing cost

• Journal of the Ergonomics Society of Korea
• /
• v.24 no.2
• /
• pp.35-43
• /
• 2005

The purpose of this study is to suggest computer assisted neck-base's landmark identification algorithms and measurement extraction methods from three-dimensional human scan data. So we developed the algorithms for automatic identification of landmarks related to the neck-base types. The subjects were 58 women $18{\sim}24$ years of age. Their body were measured directly and indirectly by using camera and three-dimensional body scanner. They were measured during the months of October in 2001. Based on the characters of classified neck-base types, algorithms for the automatic identification of landmarks and methods of automatic measurement are developed. The three-dimensional automatic measuring program is made by $C^{++}$ language. Using this program, 4 landmarks are identified and 6 items are measured. In the verifying the precision of automatic measurement, the height measurements(cervicale, side neck point, front neck point) were relatively accurate, but neck-base width measurement was measured wide.

• Journal of the Korean Society of Food Science and Nutrition
• /
• v.36 no.12
• /
• pp.1583-1588
• /
• 2007

PC-camera based automatic titration was executed in the acidity analyses of vinegar, milk and Takju. The average hue value (Havg) of 144 pixels in the image of the sample solution being titrated was computed and followed up at regular time intervals during titration in order to detect the titration end point. The Havg increase of 5 degrees from the first Havg was regarded as reaching at the end point in the cases of vinegar and milk. The Havg increase set up to detect the end point was 70 degrees in the case of Takju. In the case of vinegar, the volume of added titrant (0.1 N NaOH) was $21.409{\pm}0.066mL$ in manual titration and $21.403{\pm}0.055mL$ in automatic titration (p=0.841). In the case of milk, it was $1.390{\pm}0.025mL$ in manual titration and $1.388{\pm}0.027mL$ in automatic titration (p=0.907). In the case of Takju, it was $4.738{\pm}0.028mL$ in manual titration and $4.752{\pm}0.037mL$ in automatic titration (p=0.518). The high p values suggested that there were good agreements between manual and automatic titration data in all three food samples. The automatic method proposed in this article was considered to be applicable not only to acidity titrations but also to most titrations in which the end points can be detected by color change.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.38 no.5
• /
• pp.77-85
• /
• 2001

The purpose of this study is to suggest Vergence Information Extracting Algorithm(VIEA) which enables quick and accurate vergence information achievement for automation of vergence and focus control of horizontal moving stereoscopic camera. Firstly, for this purpose, the geometric structure of horizontal moving stereoscopic camera device was analyzed and linear relation between the vergence and the focus control. Then stereoscopic camera was designed and produced with the application of vergence and focus relation formula. Finally, VIEA that uses Cepstrum filter was employed to implement Automatic Vergence and Focus Controlling Stereoscopic Camera System(AVFCSCS). VIEA showed lower vergence achievement time and error ratio in comparison with existing algorithms. The suggested system in this study substantially reduced the controlling time and error-ratio as to make it possible to achieve natural and clear images. It also simplified the handling of stereoscopic camera for the convenience of end-users.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.11 no.5
• /
• pp.615-622
• /
• 2018

Recent advancements in camera communication (CamCom) technology using visible light exploited to use display as an luminance source to modulate the data for visible light data communication. The existing display-CamCom techniques uses the selected region of interest based camera capturing approach to detect and decode the 2D color coded data on display screen. This is not effective way to do communicate when the user on mobility. This paper propose the automatic display detection using Lambertian color segmentation combined with canny edge detection algorithms for CamCom in order to avoid manual region of interest selection to establish communication link between display and camera. The automatic display detection methods fails using conventional edge detection algorithms when content changes dynamically in displays. In order to solve this problem lambertian color segmentation combined with canny edge detection algorithms are proposed to detect display automatically. This research analysed different algorithms on display edge recognition and measured the performance on rendering dynamically changing content with color code on display. The display detection rate is achieved around 96% using this proposed solutions.

• Journal of the Korean Society for Precision Engineering
• /
• v.21 no.4
• /
• pp.194-201
• /
• 2004

Optical fibers are indispensable fer optical communication systems that transmit large volumes of data at high speed. But the aligning technology under the sub-micron accuracy is required for the precise axis adjustment and connection. For the purpose of precise alignment of the optical arrays, in this research, we have developed the 12-axis(with 8 automated axis and 4 manual axis) automatic optical fiber alignment system including the image processing-based searching system, the automatic loading system using the robot and the suction toot and the automatic UV bonding system. In order to obtain the sub-micron alignment accuracy, two 4-axis PC-based motion controllers and the two 50nm resolution 6-aixs micro-stage actuated by micro stepping motors are adopted. The fiber aligning procedure consists of two steps. Firstly, the optical wave guide and an input optical array are aligned by the 6-axis input micro-stage with the IR camera. The image processing technique is introduced to reduce primary manual aligning time and result in achieving the 50% decrease of aligning time. Secondly, the IR camera is replaced by the output micro-stage and a wave guide and two optical arrays are aligned simultaneously before the laser power intensity delivered to the optical powermeter reached the threshold value. When the aligning procedure is finished, the wave guide and arrays are W bonded. The automatic loading/unloading system is also introduced and the entire wave guide handing time is reduced significantly compared to the former commercial aligning system.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.25 no.6
• /
• pp.536-541
• /
• 2015

In this paper, we propose a the image-based automatic flight control system for the mini drone. Automatic flight system with a camera on the ceiling and markers on the floor and landing position is designed in an indoor environment. Images from the ceiling camera is used not only to recognize the makers and landing position but also to track the drone motion. PC sever identifies the location of the drone and sends control commands to the mini drone. Flight controller of the mini drone is designed using state-machine algorithm, PID control and way-point position control method. From the, The proposed automatic flight control system is verified through the experiments of the mini drone. We see that known makers in environment are recognized and the drone can follows the trajectories with the specific ㄱ, ㄷ and ㅁ shapes. Also, experimental results show that the drone can approach and correctly land on the target positions which are set at different height.

• Journal of Korea Society of Industrial Information Systems
• /
• v.14 no.5
• /
• pp.139-148
• /
• 2009

In this paper, I propose an automatic analysis method of golf swine from single-camera video sequences. I define necessary swing features for automatic swing analysis in 2-dimensional environment and present efficient swing analysis methods using various image processing techniques including line and edge detection. The proposed method has two characteristics compared with previous swing analysis systems and related studies. First, the proposed method enables an automatic swing analysis in 2-dimension while previous systems require 3-dimensional environment which is relatively complex and expensive to run. Second, swing analysis is done automatically without human intervention while other 2-dimensional systems necessarily need analysis by a golf expert. I tested the method on 20 swing video sequences and found the proposed method works effective for automatic analysis of golf swing.