• Journal of KIISE:Software and Applications
• /
• v.36 no.4
• /
• pp.253-260
• /
• 2009

Many algorithms for computer vision and pattern recognition have recently been implemented on GPU (graphic processing unit) for faster computational times. However, the implementation has two problems. First, the programmer should master the fundamentals of the graphics shading languages that require the prior knowledge on computer graphics. Second, in a job that needs much cooperation between CPU and GPU, which is usual in image processing and pattern recognition contrary to the graphic area, CPU should generate raw feature data for GPU processing as much as possible to effectively utilize GPU performance. This paper proposes more quick and efficient implementation of neural networks on both GPU and multi-core CPU. We use CUDA (compute unified device architecture) that can be easily programmed due to its simple C language-like style instead of GPU to solve the first problem. Moreover, OpenMP (Open Multi-Processing) is used to concurrently process multiple data with single instruction on multi-core CPU, which results in effectively utilizing the memories of GPU. In the experiments, we implemented neural networks-based text extraction system using the proposed architecture, and the computational times showed about 15 times faster than implementation on only GPU without OpenMP.

• Journal of KIISE:Information Networking
• /
• v.28 no.4
• /
• pp.463-476
• /
• 2001

To support the mobility, mobile computers are generally equipped with lower capability than desktop PCs or workstations in terms of the size of a display, the processing power of CPU and so on. This may give a rise to limitation in mobile computers of supporting multimedia services such as World Wide Web which users would otherwise fully enjoy in desktop PCs. Approaches to reducing the limitations are distillating original multimedia data or converting them to text. Conventional proxy servers for mobile computer simply send distillated image files with the fixed size regardless of the display size of a target mobile computer. Since the cached data is kept separately for each user, they cannot be shared among users with the same display configuration and thus the proxy server could be overloaded. In this paper, we first classify various mobile computers based on their display capability in terms of display sizes and colors. We propose an enhanced proxy server called Class-based proxy that provides a mobile computer with distillated image files in proportion to its class display capacity. The proposed proxy server allows a mobile computer user to have a homepage view similar to that in PC or Workstation. Mobile computers with the same class share the cached image files, which are distillated appropriately for that class. This helps the proxy server to get higher cache hit ratio with improved efficiency and scalability.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.10
• /
• pp.377-382
• /
• 2018

This study deals with internal state and tissue density analysis methods for red ginseng grade determination. For internal measurement of red ginseng, there have been various studies on nondestructive testing methods since the 1990s, It was difficult to grasp the most important inner hole and inside whites in the grading. So in this study, we developed a closed capturing device for infra-red illumination environment, and developed an internal measurement system that can detect the presence and diameter of inner hole and inside whites. Made devices consisted of infrared lights with a high transmission rate of red ginseng in 920 nanometer wave band, a infra-red camera and a Y axis actuator with a red ginseng automatically controlled focus on the camera. The proposed algorithm performs an auto-focus system on the Y-axis actuator to automatically adjust the sharp focus of the object according to the size and thickness. Then red ginseng is rotated $360^{\circ}$ at $1^{\circ}$ intervals and 360 total images are acquired, and reconstructed as a sinogram through Radon transform and Back-projection algorithm was performed to acquire internal images of red ginseng. As a result of the algorithm, it was possible to acquire internal cross-sectional image regardless of the thickness and shape of red ginseng. In the future, if more than 10,000 different shapes and sizes of red ginseng internal cross-sectional image are acquired and the classification criterion is applied, it can be used as a reliable automated ginseng grade automatic measurement method.

• KIPS Transactions on Computer and Communication Systems
• /
• v.12 no.12
• /
• pp.363-370
• /
• 2023

Object detection technology that accurately recognizes the road and surrounding conditions is a key technology in the field of autonomous driving. In the field of autonomous driving, object detection technology requires real-time performance as well as accuracy of inference services. Task offloading technology should be utilized to apply object detection technology for accuracy and real-time on resource-constrained devices rather than high-performance machines. In this paper, experiments such as performance comparison of task offloading, performance comparison according to input image resolution, and performance comparison according to camera object resolution were conducted and the results were analyzed in relation to the application of task offloading for real-time object detection of autonomous driving in resource-constrained devices. In this experiment, the low-resolution image could derive performance improvement through the application of the task offloading structure, which met the real-time requirements of autonomous driving. The high-resolution image did not meet the real-time requirements for autonomous driving due to the increase in communication time, although there was an improvement in performance. Through these experiments, it was confirmed that object recognition in autonomous driving affects various conditions such as input images and communication environments along with the object recognition model used.

• Journal of the Institute of Convergence Signal Processing
• /
• v.22 no.4
• /
• pp.149-155
• /
• 2021

In this study, we intend to develop a defective road surface object recognition model that automatically detects road surface defects that restrict the movement of the transportation handicapped using electric mobile devices with deep learning. For this purpose, road surface information was collected from the pedestrian and running routes where the electric mobility aid device is expected to move in five areas within the city of Busan. For data, images were collected by dividing the road surface and surroundings into objects constituting the surroundings. A series of recognition items such as the detection of breakage levels of sidewalk blocks were defined by classifying according to the degree of impeding the movement of the transportation handicapped in traffic from the collected data. A road surface object recognition deep learning model was implemented. In the final stage of the study, the performance verification process of a deep learning model that automatically detects defective road surface objects through model learning and validation after processing, refining, and annotation of image data separated and collected in units of objects through actual driving. proceeded.

• Journal of Animal Science and Technology
• /
• v.49 no.5
• /
• pp.677-688
• /
• 2007

돼지를 포함한 대부분의 동물은 일정한 발정주기를 가지고 일정한 시기에 배란을 하는 자연배란동물이지만, 토끼, 고양이, 밍크 등의 암놈은 교미자극에 의해 배란이 일어나는 유기배란동물이다. 또한 1년에 한 번만 발정하는 단발정동물과 1년에 수차례 발정하는 다발정동물이 있다. 이 중에서 모돈은 1년에 수차례 발정하는 다발정 동물로서 발정기에 들면 비발정기와는 다른 행동을 나타낸다(Diehl 등, 2001). 양돈가의 수익을 최대화하기 위해서는 비생산일수를 최소로 줄여야 한다. 모돈의 비생산일수를 줄일 수 있는 한 가지 방법은 성공적으로 교배를 시키는 것이다. 이처럼 성공적으로 교배를 시키기 위해서는 수정적기를 정확히 예측해야 한다. 만약 수정적기를 정확히 판단하지 못하여 수태가 되지 않으면, 비생산일수가 늘어나 손실을 입게 된다. 따라서 수정적기를 정확히 판단하는 것은 모돈의 성공적인 인공수정에 있어서 중요한 요소이다. 수정적기는 배란이 일어나기 전 10시간에서 12시간 사이이며, 발정이 시작되는 시점을 기준으로 하였을 때 경산돈의 경우 26시간에서 34시간 사이이고 미경산돈의 경우는 18시간에서 26시간 사이이다(Evans 등, 2001). 현재 하루에 두 번 모돈의 발정을 확인하는 것이 일반화되어 있으며, 이 때 웅돈을 접촉시키거나 육안관찰을 통하여 발정 유무를 판단한다. 이러한 방법에는 숙련된 기술과 풍부한 경험이 요구될 뿐만 아니라 총 소요노동력의 30% 정도가 요구된다(Perez 등, 1986). 하루에 두 번밖에 발정을 감지하지 않기 때문에 발정이 언제 시작되었는지를 정확히 알 수 없으며, 또한 발정의 대부분이 새벽에 시작되므로 수정적기를 정확히 판단하기란 매우 어렵다. 만약 발정을 감지했더라도 적기에 인공수정을 하지 못한다면, 수태율이 낮아지므로 경제적 손실이 초래된다. 현재 이러한 문제점 때문에 2회에서 3회에 걸쳐 인공수정을 하고 있으나 이에 따른 소요비용과 소요노동력 등은 양돈가의 부담을 가중시키는 요인이 되고 있다. 돼지는 발정기가 되면 비발정기에 나타내지 않던 외음부의 냄새를 맡는 행동, 귀를 세우는 행동 및 승가허용 행동 등을 나타낸다(Diehl 등, 2001). 또한 돼지는 비발정기에 비하여 발정기에 더 많은 활동량을 나타낸다(Altman, 1941; Erez and Hartsock, 1990). Freson 등(1998)은 스톨에서 개별적으로 사육되고 있는 모돈의 활동량을 적외선센서를 이용하여 측정함으로써 발정을 86%까지 감지하였다고 보고하였다. 그러나 이 연구는 단지 모돈의 발정을 감지하였을 뿐 번식관리에 있어서 가장 중요한 수정적기의 판단 기준을 제시하지 못하였다. 따라서, 본 연구는 스톨에서 사육되는 모돈의 활동량을 측정함으로써 발정시작시각을 감지하고 이를 기준으로 인공수정적기를 예측할 수 있는 인공수정적기 예측 장치를 개발한 후 이의 성능을 농장실증실험을 통하여 시험하고자 수행되었다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.12 no.5
• /
• pp.865-872
• /
• 2017

Recently, in response to the Internet age, the demand for hardware devices has been increasing, centering on the rapidly growing smart home field, due to the growth and management of sensor and control technology, mobile application, network traffic, big data management and cloud computing. In order to maintain the sustainable development of the hardware system, it is necessary to update the system, and the hardware device is absolutely necessary in real time processing of complex data (voice, image, etc.) as well as data collection. In this paper, we propose a method to simplify the control and communication method by integrating the hardware devices in two operating systems in a unified structure to solve the simultaneous control and communication method of hardware under different operating systems. The performance evaluation results of the proposed integrated hardware and the cloud control system connected to the cloud server are described and the main directions to be studied in the field of internet smart home are described.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.5
• /
• pp.745-750
• /
• 2013

In many recent studies, a variety of environmental control system for practical gardening facilities such as facility house and plant factory have been proposed. However, the plants have been exposed to growth disorder and disease and pest injury because the temperature and humidity have not properly controlled so far. Therefore, a lot of damage of farmers have been reported. The air circulation fan and industrial dehumidifier have been currently utilized as the countermeasures, but they do not meet the expectation. In this study, the growth phase of each plant is recognized by using cycle-by-cycle plants growth recogniztion algorithm to provide optimal environment according to the growth phases of each plant.he productivity can be raised by using cycle-by-cycle plant growth recognition monitoring system because it optimally controls the environment by cycle that is required for plant growth.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.3
• /
• pp.693-698
• /
• 2018

This paper proposes a vehicle detection method for parking areas using unmanned aerial vehicles (UAVs) and using YOLOv2, which is a recent, known, fast, object-detection real-time algorithm. The YOLOv2 convolutional network algorithm can calculate the probability of each class in an entire image with a one-pass evaluation, and can also predict the location of bounding boxes. It has the advantage of very fast, easy, and optimized-at-detection performance, because the object detection process has a single network. The sliding windows methods and region-based convolutional neural network series detection algorithms use a lot of region proposals and take too much calculation time for each class. So these algorithms have a disadvantage in real-time applications. This research uses the YOLOv2 algorithm to overcome the disadvantage that previous algorithms have in real-time processing problems. Using Darknet, OpenCV, and the Compute Unified Device Architecture as open sources for object detection. a deep learning server is used for the learning and detecting process with each car. In the experiment results, the algorithm could detect cars in a dense area using UAVs, and reduced overhead for object detection. It could be applied in real time.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.18 no.6
• /
• pp.535-544
• /
• 2016

Special tunnel design and construction methods have been suggested due to developments of subway and tunnel. Collapse accidents of tunnel bring enormous damage. So, observation and analysis for the safety of tunnelling and behaviour of surrounding ground are important. But, it is not economical to implement the field test in every time. Therefore, this study has measured ground behaviour due to excavation of 2-arch tunnel station according to offset between grouped pile and tunnel by laboratory model test. For the model test, trapdoor device was adopted. Tunnelling is simulated by volume loss of 2-arch tunnel. Ground displacements are observed by close range photogrammetric method and image processing. In addition, these data are compared with numerical analysis.