• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.10a
• /
• pp.485-488
• /
• 2013

최근 자동차 산업의 활성화로 인해 교통사고 급증이 사회 문제화 되면서 사고를 미연에 방지할 수 있는 운전자 보조 시스템 연구가 활발하게 이루어지고 있다. 일반적으로 자동차 사고 원인의 70% 이상이 운전자 과실에 의해서 발생되고 전체 추돌사고의 75%가 시속 29km 이하의 속도에서 발생한다. 이를 예방하기 위해서 운전자의 인지 판단을 보조하는 시스템의 개발이 많이 이루어지고 있는데, 예를 들어 자동 주차 시스템, AVM(Around View Monitoring) 시스템 등이 있다. 본 논문에서는 AVM 시스템 중 원근 왜곡을 보정하는 단계에서 직선 및 교점을 검출할 때, NMS(Non-Maximum Suppression)를 적용한 허프 변환 방법을 사용할 것이다. 또한 기존의 Sub-Pixel을 이용한 직선 및 교점 검출 방법과 NMS을 적용한 허프 변환 방법을 사용한 직선 및 교점을 검출하는 방법을 비교 분석함으로써 제안하는 NMS를 적용한 허프변환을 이용한 직선 및 교점을 검출하는 방법을 사용하여 보다 효율적인 AVM 시스템의 구현 가능성을 검증한다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2016.10a
• /
• pp.382-383
• /
• 2016

Canny edge detection of the image is composed of four kinds of Gaussian filter, gradient calculation, Non-maximum suppression, and Hypothesis Thresholding. Feature is the ratio between the vehicle body, the windows, and the wheels obtained from the edge image. Features that make the proportion of these vehicles are different for each respective model. We have identified by application of this algorithm where only a small vehicle.

• IEIE Transactions on Smart Processing and Computing
• /
• v.6 no.4
• /
• pp.253-261
• /
• 2017

In this paper, a corner detection method based on a new non-cornerness measure is presented. Rather than evaluating local gradients or surface curvatures, as done in previous approaches, a non-cornerness function is developed that can identify stable corners by testing an image region against a set of desirable corner criteria. The non-cornerness function is comprised of two steps: 1) eliminate any pixel located in a flat region and 2) remove any pixel that is positioned along an edge in any orientation. A pixel that passes the non-cornerness test is considered a reliable corner. The proposed method also adopts the idea of non-maximum suppression to remove multiple corners from the results of the non-cornerness function. The proposed method is compared with previous popular methods and is tested with an artificial test image covering several corner forms and three real-world images that are universally used by the community to evaluate the accuracy of corner detectors. The experimental results show that the proposed method outperforms previous corner detectors with respect to accuracy, and that it is suitable for real-time processing.

• Smart Media Journal
• /
• v.7 no.3
• /
• pp.16-21
• /
• 2018

This paper presents the design and implementation of a deep learning-based automated weed detector on onion fields. The system is based on a Convolutional Neural Network that specifically selects proposed regions. The detector initiates training with a dataset taken from agricultural onion fields, after which candidate regions with very high probability of suspicion are considered weeds. Non-maximum suppression helps preserving the less overlapped bounding boxes. The dataset collected from different onion farms is evaluated with the proposed classifier. Classification accuracy is about 99% for the dataset, indicating the proposed method's superior performance with regard to weed detection on the onion fields.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.1795-1811
• /
• 2019

In order to achieve rapid and accurate detection of vehicle objects in complex traffic conditions, we propose a novel vehicle detection method. Firstly, more contextual and small-object vehicle information can be obtained by our Joint Feature Network (JFN). Secondly, our Evolved Region Proposal Network (EPRN) generates initial anchor boxes by adding an improved version of the region proposal network in this network, and at the same time filters out a large number of false vehicle boxes by soft-Non Maximum Suppression (NMS). Then, our Mask Network (MaskN) generates an example that includes the vehicle occlusion, the generator and discriminator can learn from each other in order to further improve the vehicle object detection capability. Finally, these candidate vehicle detection boxes are optimized to obtain the final vehicle detection boxes by the Fine-Tuning Network(FTN). Through the evaluation experiment on the DETRAC benchmark dataset, we find that in terms of mAP, our method exceeds Faster-RCNN by 11.15%, YOLO by 11.88%, and EB by 1.64%. Besides, our algorithm also has achieved top2 comaring with MS-CNN, YOLO-v3, RefineNet, RetinaNet, Faster-rcnn, DSSD and YOLO-v2 of vehicle category in KITTI dataset.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.3
• /
• pp.840-860
• /
• 2023

Aiming at the serious occlusion and slow tracking speed in pedestrian target tracking and recognition in complex scenes, a target tracking method based on improved YOLO v5 combined with Deep SORT is proposed. By merging the attention mechanism ECA-Net with the Neck part of the YOLO v5 network, using the CIoU loss function and the method of CIoU non-maximum value suppression, connecting the Deep SORT model using Shuffle Net V2 as the appearance feature extraction network to achieve lightweight and fast speed tracking and the purpose of improving tracking under occlusion. A large number of experiments show that the improved YOLO v5 increases the average precision by 1.3% compared with other algorithms. The improved tracking model, MOTA reaches 54.3% on the MOT17 pedestrian tracking data, and the tracking accuracy is 3.7% higher than the related algorithms and The model presented in this paper improves the FPS by nearly 5 on the fps indicator.

• Journal of computational fluids engineering
• /
• v.21 no.4
• /
• pp.90-95
• /
• 2016

Leidenfrost drops with ellipsoidal shaping can control the bouncing height by adjusting the aspect ratio(AR) of the shape at the moment of impact. In this work, we focus on the effect of the AR and the impact Weber number(We) on the non-axisymmetrical spreading dynamics of the drop, which plays an important role in the control of bouncing. To understand the impact dynamics, the numerical simulation is conducted for the ellipsoidal drop impact upon the perfect non-wetting solid surface by using volume of fluid method, which shows the characteristics of the spreading behavior in each principal axis. As the AR increases, the drop has a high degree of the alignment into one principal axis, which leads to the consequent suppression of bouncing height with shape oscillation. As the We increases, the maximum spreading diameters in the principal axes both increase whereas the contact time on the solid surface rarely depends on the impact velocity at the same AR. The comprehensive understanding of the ellipsoidal drop impact upon non-wetting surface will provide the way to control of drop deposition in applications, such as surface cleaning and spray cooling.

• Clinical and Experimental Reproductive Medicine
• /
• v.25 no.2
• /
• pp.217-231
• /
• 1998

The effectiveness of intrauterine insemination (IUI) combined with controlled ovanan hyperstimulation (COH) in the treatment of infertility with various etiologies was compared in a total of 152 cycles. Patients received a maximum of three IUI cycles for the treatment. Severe male ($<2\times10^6$ motile sperm) or age factor (> 39 y) patients were excluded in this study. Pregnancy was classified as clinical if a gestational sac was seen on ultrasound. The overall clinical pregnancy rate was 7.9% per cycle (12/152) and 9.7% per patient (12/124). The pregnancy rates were 0% in unstimulated natural (0/18), 7.5% in CC (3/40), 8.2% in CC+hMG (4/49), 5.9% in GnRH-a ultrashort (1/17), 5.9% in GnRH-a long (1/17) and 27.3% in dual suppression cycles (3/11), respectively. The pregnancy rate was higher in dual suppression cycle than other stimulated cycles, but this was not significant. The multiple pregnancy rates were 25.0% (2 twins and 1 triplet). No patient developed ovarian hyperstimulation. Abortion rates were 66.7% in CC (2/3) and 100% in ultrashort cycles (1/1). The livebirth rate was 5.9% per cycle (9/152) and 7.3% per patient (9/124). There were no differences in age, duration of infertility, follicle size, total ampules of gonadotropins and days of stimulation between pregnant and non-pregnant groups. However, significant(P<0.05) differences were observed in the level of estradiol $(E_2)$ on the day of hCG injection ($3,266.6{\pm}214.2$ vs $2,202.7{\pm}139.4$ pg/ml) and total motile sperm count ($212.1{\pm}63.4$ vs $105.1{\pm}9.9\times10^6$) between pregnant group and non-pregnant group. These results suggest that IUI combined with successful ovarian stimulation tends to improve the chance of pregnancy as compared to IUI without COH and a total motile sperm count may be considered predictive of the success for pregnancy.

• ETRI Journal
• /
• v.29 no.5
• /
• pp.679-681
• /
• 2007

In this letter, we demonstrate a technique for suppression of transients in output bursts of an erbium-doped fiber amplifier (EDFA) in an optical burst network. To suppress the transients, the EDFA is forward-fed by non-fluctuating input utilizing a power-modulated burst control packet channel. Using the technique, we obtained a maximum 1.7 dB reduction in gain transient in the EDFA output, and we transmitted 9.953 Gbps data bursts and 2.488 Gbps burst control packets stably.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.06a
• /
• pp.80-81
• /
• 2019

본 논문에서는 자연스러운 장면 영상에서 임의의 방향성을 가진 텍스트를 검출하기 위한 기법을 제안한다. 텍스트 검출을 위한 기본적인 프레임 워크는 Faster R-CNN[1]을 기반으로 한다. 먼저 RPN(Region Proposal Network)을 통해 다른 방향성을 가진 텍스트를 포함하는 bounding box를 생성한다. 이어서 RPN에서 생성한 각각의 bounding box에 대해 세 가지의 서로 다른 크기로 pooling된 특징지도를 추출하고 병합한다. 병합한 특징지도에서 텍스트와 텍스트가 아닌 대상에 대한 score, 정렬된 bounding box 좌표, 기울어진 bounding box 좌표를 모두 예측한다. 마지막으로 NMS(Non-Maximum Suppression)을 이용하여 검출 결과를 획득한다. COCO Text 2017 dataset[2]을 이용하여 학습 및 테스트를 진행하였으며 주관적으로 평가한 결과 기울어진 텍스트에 적합하게 회전된 영역을 얻을 수 있음을 확인하였다.