• ETRI Journal
• /
• 제40권3호
• /
• pp.376-388
• /
• 2018

In through-the-wall radar imaging (TWRI), the presence of front and side walls causes multipath propagation, which creates fake targets called multipath ghosts. They populate the scene and reduce the probability of correct target detection, classification, and localization. In modern TWRI, specular multipath exploitation has received considerable attention for reducing the effects of multipath ghosts. However, this exploitation is challenged by the requirements of the reflecting geometry, which is not always available. Currently, the demand for a high radar image resolution dictates the use of a large aperture and wide bandwidth. This results in a large amount of data. To tackle this problem, compressive sensing (CS) is applied to TWRI. With CS, only a fraction of the data are used to produce a high-quality image, provided that the scene is sparse. However, owing to multipath ghosts, the scene sparsity is highly deteriorated; hence, the performance of the CS algorithms is compromised. This paper presents and discusses the adverse effects of multipath ghosts in TWRI. It describes the physical formation of ghosts, their challenges, and existing suppression techniques.

• 한국위성정보통신학회논문지
• /
• 제9권4호
• /
• pp.77-81
• /
• 2014

본 논문은 영상에서 실시간으로 움직임 물체와 물체의 위치를 검출하는 방법을 제안한다. 첫째로 영상으로부터 2개의 연속된 프레임 차분을 통해 움직이는 물체를 추출하는 방법을 제안한다. 만약 두 프레임이 캡쳐되는 사이의 간격이 길다면, 실제 움직이는 물체의 꼬리 같은 거짓 움직임 물체를 생성한다. 두번째로 본 논문은 도플러 효과와 HSV 색상 모델을 사용하여 이 문제들을 해결하는 방법을 제안한다. 마지막으로 물체의 분할과 위치 설정은 상기의 단계에서 얻은 결과가 조합되어 완료된다. 제안된 방법은 99.2%의 검출율을 갖고, 과거에 제안된 다른 비슷한 방법들 보다는 비교적 빠른 속도를 갖는다. 알고리즘의 복잡성은 시스템의 속도에 직접적인 영향을 끼치기 때문에, 제안된 방법은 낮은 복잡성을 가져 실시간 움직임 검출을 위해 사용 될 수 있다.

• 응용통계연구
• /
• 제34권3호
• /
• pp.449-460
• /
• 2021

본 논문에서는 음성위조공격탐지(Voice spoofing detection) 문제에 데이터 증강을 적용한다. ASVspoof 2017은 리플레이 공격 탐지에 대해 다루며 진짜 사람의 음성과 환경이나 녹음·재생 장치의 조건들을 다르게 하여 위조한 가짜 음성을 분류하는 것을 목적으로 한다. 지금까지 이미지 데이터에 대한 데이터 증강 연구가 활발히 이루어졌으며 음성에도 데이터 증강을 시도하는 여러 연구가 진행되어왔다. 하지만 음성 리플레이 공격에 대한 데이터 증강시도는 이루어지지 않아 본 논문에서는 데이터 증강기법을 통한 오디오 변형이 리플레이 공격 탐지에 어떠한 영향을 미치는지에 대해 탐구해본다. 총 7가지의 데이터 증강기법을 적용해보았으며 그 중 DVC, Pitch 음성 증강기법이 성능향상에 도움되었다. DVC와 Pitch는 기본 모델 EER의 약 8% 개선을 보여주었으며, 특히 DVC는 57개의 환경변수 중 일부 환경에서 눈에 띄는 정확도 향상이 있었다. 가장 큰 폭으로 증가한 RC53의 경우 DVC가 기본 모델 정확도의 약 45% 향상을 이끌어내며 기존에 탐지하기 어려웠던 고사양의 녹음·재생 장치를 잘 구분해냈다. 본 연구를 토대로 기존에 증강기법의 효과에 대한 연구가 이루어지지 않았던 음성 위조 탐지 문제에서 DVC, Pitch 데이터 증강기법이 성능 향상에 도움이 된다는 것을 알아내었다.

• 한국산업융합학회 논문집
• /
• 제24권6_2호
• /
• pp.699-706
• /
• 2021

Existing license plate recognition system is used as an optical character recognition method, but a method of using deep learning has been proposed in recent studies because it has problems with image quality and Korean misrecognition. This requires a lot of data collection, but the collection of license plates is not easy to collect due to the problem of the Personal Information Protection Act, and labeling work to designate the location of individual license plates is required, but it also requires a lot of time. Therefore, in this paper, to solve this problem, five types of license plates were created using a virtual Korean license plate generation program according to the notice of the Ministry of Land, Infrastructure and Transport. And the generated license plate is synthesized in the license plate part of collectable vehicle images to construct 10,147 learning data to be used in deep learning. The learning data classifies license plates, Korean, and numbers into individual classes and learn using YOLOv5. Since the proposed method recognizes letters and numbers individually, if the font does not change, it can be recognized even if the license plate standard changes or the number of characters increases. As a result of the experiment, an accuracy of 96.82% was obtained, and it can be applied not only to the learned license plate but also to new types of license plates such as new license plates and eco-friendly license plates.

• 천문학회보
• /
• 제43권1호
• /
• pp.41.1-41.1
• /
• 2018

We introduce a current status and future plans of Korea Microlensing Telescope Network (KMTNet) microlensing experiments, which include an observational strategy, pipeline, event-finder, and collaborations with Spitzer. The KMTNet experiments were initiated in 2015. From 2016, KMTNet observes 27 fields including 6 main fields and 21 subfields. In 2017, we have finished the DIA photometry for all 2016 and 2017 data. Thus, it is possible to do a real-time DIA photometry from 2018. The DIA photometric data is used for finding events from the KMTNet event-finder. The KMTNet event-finder has been improved relative to the previous version, which already found 857 events in 4 main fields of 2015. We have applied the improved version to all 2016 data. As a result, we find that 2597 events are found, and out of them, 265 are found in KMTNet-K2C9 overlapping fields. For increasing the detection efficiency of event-finder, we are working on filtering false events out by machine-learning method. In 2018, we plan to measure event detection efficiency of KMTNet by injecting fake events into the pipeline near the image level. Thanks to high-cadence observations, KMTNet found fruitful interesting events including exoplanets and brown dwarfs, which were not found by other groups. Masses of such exoplanets and brown dwarfs are measured from collaborations with Spitzer and other groups. Especially, KMTNet has been closely cooperating with Spitzer from 2015. Thus, KMTNet observes Spitzer fields. As a result, we could measure the microlens parallaxes for many events. Also, the automated KMTNet PySIS pipeline was developed before the 2017 Spitzer season and it played a very important role in selecting the Spitzer target. For the 2018 Spitzer season, we will improve the PySIS pipeline to obtain better photometric results.