Aiming at remote sensing image target detection objects with large aspect ratio and arbitrary orientation,the general detector high accuracy detection is not high,this paper proposes a YOLOv5-based remote sensing image rotating target detection,using rotational regression Loss KLD loss,converting the rotating rectangular envelope frame into a Gaus-sian distribution,which provides a more accurate regression prediction.The realization results show that the proposed algo-rithm achieves a detection accuracy(mAP50)of 77.97%in the DOTA dataset,which is an improvement of 5.75%compared to the base model YOLOv5m.Experiments demonstrate the effectiveness of the algorithm in this paper in detecting rotating targets in remote sensing images with high accuracy.
关键词
深度学习/遥感图像/高斯分布/旋转目标检测
Key words
deep learning/remote sensing images/Gaussian distribution/rotating target detection
维普
万方数据