基于YOLOv8的卫星遥感图像快速目标检测方法

A Fast Target Detection Method for Satellite Remote Sensing Images Based on YOLOv8

目标检测相关技术已经广泛应用于空间目标监视、卫星自动寻轨等领域,也是计算机视觉领域最重要和最具有挑战性的研究分支之一,逐渐成为国内外军事领域的应用热点.在现代空天对抗中,通过卫星遥感图像获取近地飞行器目标,借以快速判断敌方部队的有生力量,将使我方部队占据战略优势.针对卫星观测距离过远、遥感图像背景复杂等问题,研究基于一阶段轻量化网络YOLOv8的小样本目标检测算法.通过图像翻转、马赛克数据增强及mixup数据增强等图像增强手段提高了模型的泛化性能;通过多次调整优化函数、降低类别损失增益及降低掩模比等参数调整策略提高了模型的平均精度;通过使用参数预设及加载原优化函数导出的模型提高了模型的运算效率.提出的方法在公开的飞行器数据集进行了验证,验证指标包括查准率-查全率(precision-recall)、平均精度(mAP)和画面每秒传输帧数(FPS).结果表明本文提出的改进型网络模型能满足卫星遥感图像的快速目标检测需要.

Target detection related technology has been widely used in space target surveillance,satellite automatic orbit finding and other fields.It is also one of the most important and challenging research branches in the field of computer vision and has gradually become a hot research area in the military field at home and abroad.In the mod-ern air-space confrontation,the acquisition of near-earth vehicle targets through satellite remote sensing images can quickly judge the effective strength of the enemy forces,which enables our troops to occupy a strategic advantage.Aiming at the problems such as too long satellite observation distance and complex background of remote sensing image,the small sample target detection algorithm is studied based on one-stage light weighted network YOLOv8.The research work of this paper mainly includes three aspects.Firstly,the generalization performance of the model is improved by image enhancement methods such as image flipping,Mosaic data enhancement and mix-up data en-hancement.Secondly,the average accuracy of the model is improved by adjusting the optimization function,reduc-ing the class loss gain and reducing the mask ratio.Thirdly,the computational efficiency of the model is improved via the preset parameter and loading the model derived from the original optimization function.The method pro-posed in this paper is verified on the public aircraft data set,and the verification indexes include precise-recall(P-R),average accuracy(mAP)and the number of frames per second(FPS).The results show that the improved network model proposed in this paper can meet the needs of fast target detection in satellite remote sensing images.