• 한국융합학회논문지
• /
• 제12권1호
• /
• pp.29-35
• /
• 2021

자동차 번호판 인식은 지능형 교통시스템에서 핵심적인 역할을 담당한다. 따라서 효율적으로 자동차 번호판의 숫자 및 문자영역을 검출하는 것은 매우 중요한 과정이다. 본 연구에서는 딥러닝과 의미론적 영상분할 알고리즘을 적용하여 효과적으로 자동차 번호판의 번호영역을 검출하는 방법을 제안한다. 제안된 방법은 화소 투영과 같은 전처리과정 없이 번호판 영상에서 바로 숫자 및 문자영역을 검출하는 알고리즘이다. 번호판 영상은 도로 위에 설치된 고정 카메라로 부터 획득한 영상으로 날씨 및 조명변화 등을 모두 포함한 다양한 실제 상황에서 촬영된 것을 사용하였다. 입력 영상은 색상변화를 줄이기 위해 정규화하고 실험에 사용된 딥러닝 신경망 모델은 Vgg16, Vgg19, ResNet18 및 ResNet50이다. 제안방법의 성능을 검토하기 위해 번호판 영상 500장으로 실험하였다. 학습을 위해 300장을 할당하였으며 테스트용으로 200장을 사용하였다. 컴퓨터모의 실험결과 ResNet50을 사용할 때 가장 우수하였으며 95.77% 정확도를 얻었다.

• 수산경영론집
• /
• 제33권1호
• /
• pp.69-85
• /
• 2002

Reduction of fishing grounds, which was caused by the recent fisheries agreements between Korea, Japan, and China, Is Increasing the important of coastal fisheries. Korea's coastal fisheries, however, need special management measures owing to the following problems. First, there are too many fishing fleets even though fisheries resources are continuously decreasing. Despite the fact that some of the fishing vessels have been scrapped, there still remain too many vessels and especially approximately 27,000 vessels increased during 1997-1998. Second, as the condition of fisheries resources changed radically, many fishermen are fishing by methods which they were not permitted to. These fishing methods cannot be legally supported and so there is a huge gap between the system and the reality, Third, two or three licenses are given to each coastal fishing vessel because a single license cannot give sufficient income, but some of these are formally acquired. So under such circumstances, efficient management of fisheries is impossible. Fourth, absence of demarcation among regions and industries is causing frequent conflicts and there are concerns about the decreasing fisheries resources due to competitive fishing practices. Therefore, considering the above mentioned problems Korea's coastal fisheries management should be developed as the following: First, new licenses should be limited while expanding the buy-back program. The government is currently planning to limit new licenses by introducing the fixed license number system in coastal fisheries but is somewhat passive about the buy-back program. Second, fisheries management which is based on self-regulation should be established. In order to increase the effect of fisheries management, the fishermen should decide by themselves the fish and fishing methods they would be exploiting and directly regulate them. Third, it is necessary to integrate the licenses of coastal fisheries. Since coastal fisheries management through the license system has distinct limitations, it is preferable to unify risking licenses and let the fishermen decide specific matters on their own. Finally, it is necessary to establish boundaries among the regions and industries. Joint fishing areas among regions(cities and provinces) should be established and fishing in other areas should be permitted on condition of paying the required fees. On the other hand, it is also necessary to permit coastal fishing only within certain distances.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2009년도 정보 및 제어 심포지움 논문집
• /
• pp.284-286
• /
• 2009

In this paper, we consider a cognitive radio system operating as secondary user. It uses an empty channel that is not currently used by primary users having the license to the channel. In the previous works, secondary user looks for an empty channel by choosing any channel in order or randomly and by sensing the channel to distinguish whether primary users are using. But if primary user is fixed type, we will find an empty channel faster than the mentioned channel selecting methods by using a method considering prior information about cases that primary user used the channel, since it is possible to analogize the channel access possibility of primary user according to regular time and position. Therefore, we propose a channel searching method based on the channel list for the purpose of reducing the channel searching time and improving throughput of secondary users. Firstly, we determine a weighting value of each channel based on the history of channel activities of primary users. This value is added to current channel state buffer and we search an empty channel from channel with smallest value to one with the biggest value. Finally, we compare the performances of the proposed method with those of the sequential channel searching and the random channel searching methods in terms of the average channel searching time and the average number of transmissions of secondary user.