基于改进YOLOv4深度学习的有雾海面船只识别

Ship identification of foggy sea surface based on improved YOLOv4 deep learning algorithm

为了提高有雾场景下船只检测识别的准确率,本文运用四分法计算求解大气光值实现对暗通道先验去雾算法的优化,引入空洞卷积方法和K-means++聚类算法改进YOLOv4算法,提出改进的暗通道先验去雾算法和改进YOLOv4深度学习的船只检测方法。通过与不同去雾算法和船只识别算法进行对比实验分析,改进后的方法更好地实现了海面船只的实时检测及分类识别。实验结果表明该方法解决了原去雾算法中去雾图像亮度偏暗等问题,提高了船只识别的准确率与实时性,对海上有雾环境条件下的船只实时检测研究具有一定的理论指导意义。

In order to improve the accuracy of ship detection and recognition in foggy scenes,the quartering method is proposed to calculate the atmospheric light value,so as to optimize the dark channel prior dehazing algorithm.And at the same time,the hole convolution method and K-means++clustering algorithm are introduced to improve the YOLOv4 algorithm.An improved dark channel prior dehazing algorithm and improved YOLOv4 deep learning ship detection method are proposed.Through comparative experiments and analysis with different defogging algorithms and ship recognition algorithms,it proves that the improved method better can realize the real-time detection and classification recognition of sea ships.The experimental results show that this method solves the problem of dark image brightness in the original defogging algorithm,improves the accuracy and real-time performance of ship recognition,having certain theoretical guiding significance to the real-time detection of ships in the foggy environment at sea.