선박 운항에서 GPS 위치 정보의 신뢰성은 안전 항해의 핵심 요소이다. 그러나 GPS 신호 교란(재밍, 스푸핑 등)이 증가함에 따 라 대체 위치 결정 시스템의 필요성이 대두되고 있다. 본 연구는 딥러닝 기반 객체 인식 기술인 YOLO v2 네트워크를 활용하여 선박에 설 치된 카메라 영상을 분석하고, 이를 기반으로 선박 위치를 추정하는 보조 시스템을 개발하였다. 구체적으로 광양항 입항 영상에서 좌현의 멧돌초 고립장애표지와 우현의 광양제철 사일로를 객체로 선정하여 인식하였으며, 바운딩 박스 중심점과 카메라 위치를 연결한 위치선 (Line of Position)을 생성하고 기준선과의 교각을 측정하여 2D 해도 상에 위치를 작도하였다. 학습 결과 객체 인식 정확도는 94% 이상으로 나타났으며, GPS 위치와 비교 시 최소 오차 3m, 최대 오차 33m, 평균 RMSE 18.51m를 기록하였다. 이는 GPS 신호 교란 시 대안으로 활용 가능하며, 협수로나 항만 등 정밀 위치가 요구되는 환경에서 유용함을 검증하였다. 향후 자율운항선박 시스템과의 연계를 통해 GNSS 이 중화 장비로 발전할 수 있을 것으로 기대된다.

With advancements in high-resolution scanners and high-performance computers, the use of whole slide imaging (WSI) in digital pathology has increased. WSI scans glass slides and stores them in digital format, making them immune to damage or discoloration, and enabling remote pathology review and peer review. Additionally, with the development of artificial intelligence, research using deep learning models in pathology has become more widespread. In this study, the You Only Look Once (YOLO)v8 model was used to train artificial intelligence to detect apoptotic bodies commonly observed in rodent livers. A total of 1,558 rat liver images containing apoptotic bodies were collected and followed by labeling and data augmentation using flipping and rotation techniques to expand the dataset to 3,738 images. The dataset was then divided into training, validation, and test sets to develop and evaluate a model for object recognition. The training was conducted with an epoch set to 300. The YOLOv8 model detected apoptotic bodies with a mean average precision at 50% value of 0.882. Although the model’s accuracy for detecting individual apoptotic bodies may not seem extremely high, it is important to note that the size of apoptotic bodies is very small compared to hepatocytes, making them harder to detect. However, the model’s overall performance is expected to improves significantly with a larger dataset. The YOLOv8 model successfully detected apoptotic bodies with high accuracy. This can help reduce the workload of toxicologic pathologists and decrease the time and cost involved in pathology review. Furthermore, with an increased dataset, even higher accuracy can be expected in the future.

The lane designation and the bus-only lane system for traffic speed and road safety are difficult to crack down on, and for this purpose, crackdown methods using image recognition technologies are being studied. Existing studies require continuous learning or additional equipment, and it is difficult to classify combined vehicles such as vans and pickup trucks. Therefore, in this study, YOLO and EasyOCR were mixed to classify combined vehicles through vehicle type symbols. For combined vehicles, higher accuracy was shown than classification using YOLO. Due to the nature of Hangul, the accuracy was slightly lowered because the OCR was not accurately recognized, but if it is used with the existing YOLO classification, high accuracy of crackdown will be possible.

자율운항선박이 상용화되어 연안을 항해하기 위해서는 해상의 장애물을 탐지할 수 있어야 한다. 연안에서 가장 많이 볼 수 있 는 장애물 중의 하나는 양식장의 부표이다. 이에 본 연구에서는 YOLO 알고리즘을 이용하여 해상의 부표를 탐지하고, 카메라 영상의 기하 학적 해석을 통해 선박으로부터 떨어진 부표의 거리와 방위를 계산하여 장애물을 시각화하는 해상물체탐지시스템을 개발하였다. 1,224장 의 양식장 부표 사진으로 해양물체탐지모델을 훈련시킨 결과, 모델의 Precision은 89.0 %, Recall은 95.0 % 그리고 F1-score는 92.0 %이었다. 얻 어진 영상좌표를 이용하여 카메라로부터 떨어진 물체의 거리와 방위를 계산하기 위해 카메라 캘리브레이션을 실시하고 해상물체탐지시 스템의 성능을 검증하기 위해 Experiment A, B를 설계하였다. 해상물체탐지시스템의 성능을 검증한 결과 해상물체탐지시스템이 레이더보 다 근거리 탐지 능력이 뛰어나서 레이더와 더불어 항행보조장비로 사용이 가능할 것으로 판단된다.

Future autonomous vehicles need to recognize the ego lanes required for lane change and the side left and right lanes differently. Therefore, multi-lane recognition is needed. In this study, using the YOLO network, mainly used for object recognition, the proposed method recognizes the ego, left and right side lanes as different objects and identifies the correct lanes. As a result of the performance evaluation on the TuSimple test data, the proposed method recognized the ego lanes and the left and right side lanes differently. It showed very stable lane recognition results. And by detecting lanes that do not exist in the ground truth of TuSimple data, the proposed method is very robust in lanes detection. Nevertheless, studies related to learning data reinforcement in which lanes are located in the center or at the left and right edges of the image and accurate network learning for lanes are needed.

We need data such as the number of lanes for lane change on the road as well as environmental and object recognition of the road for the autonomous vehicle of the future. This study proposed an algorithm that recognizes the left and right lanes and the center lane while driving differently from the black box image taken from a car. In general, deep learning does not recognize lanes individually but recognizes all lanes as only one lane. Therefore, using YOLO's object recognition function, the left and right lanes and the center lane were detected as different lanes, and a heuristic method was applied to recognize multi-lanes as more correct lanes. As a result of the performance evaluation, we confirmed that the proposed method detects the lane more accurately than Fast R-CNN and only YOLOv2.