Accurate Real-time Ship Target detection Using Yolov4
Bingde Wang, Bing Han, Liutao Yang
Abstract
With the development of artificial intelligence, surface ships have been developing toward intelligence and autonomy. Fast and accurate ship target detection is an important source of visual perception information of intelligent ship. In this paper, a fast and accurate ship target detection algorithm based on YOLOv4 is proposed for the problems of low accuracy and poor real-time performance of existing ship detection algorithms. YOLOv4 is currently one of the fastest and most accurate target detection algorithms with a wide range of applications in various fields. For ship targets, in order to achieve better detection results, some methods are used to improve the algorithm, such as K-means clustering, model structure modification and Mixup. In order to verify the effectiveness of the method, a ship target detection dataset is constructed by collecting various ship targets. And the training and tests of the algorithm model are completed in a GPU cloud server. The performance of the algorithm is compared to that of Faster R-CNN, SSD and YOLOv3 through experiments. The results show that this method has satisfactory accuracy and real-time detection of ship targets.