• 한국전자통신학회논문지
• /
• 제15권6호
• /
• pp.1209-1216
• /
• 2020

본 연구에서는 무인항공기 원격탐사 기법과 딥러닝 기반 객체인식 알고리즘을 활용한 해안표착폐기물 탐지기법을 제안한다. 항공영상 내에 존재하는 해안표착폐기물을 탐지하기 위해 심층신경망 기반 객체 인식 알고리즘을 제안하였다. PET, 스티로폼, 기타 플라스틱의 3가지 클래스의 이미지 데이터셋으로 심층신경망 모델을 훈련시켰으며, 각 클래스별 탐지 정확도를 Darknet-53과 비교하였다. 이를 통해 해안표착 폐기물을 무인항공기를 통해 성상별 모니터링할 수 있었으며, 향후 본 연구에서 제안하는 방법이 적용될 경우 해변 전체에 대한 성상별 전수조사가 가능하며, 이를 통해 해양환경 감시 분야의 효율성 증대에 기여할 수 있을 것으로 판단된다.

• 수산해양교육연구
• /
• 제25권1호
• /
• pp.78-86
• /
• 2013

In order to analyze the pollution extent of small plastic debris(SPD) ashore on the shoreline of coastal flow, 12 of survey was conducted at Song-Jung beach of Busan for several years. The sampled beach was divided into 9 sites with unit area($m^2$). Many of SPD were detected in the southern part of the beach and classified into 11 items as P1 to P11 according to the contents. Average densities of total items' weight were $2.955g/m^2$ and weights of P2, P3 item were composed of about 64% among them. And average densities of total items' quantity were $56.259ea/m^2$ and quantities of P6 only were composed of about 63%. Seeing the seasonal variation, fall season was abundant extremely whereas nearly nothing in spring. The correlation of weights and quantities have reliable coefficients to some extent on sites and season but nearly don't have reliances on item, tide, wind and precipitation. Many researching data were required if possible in order to discussing about the correlation.