基于特征聚合的空中加油锥套检测算法  

Aerial refueling cone sleeve detection algorithm based on feature aggregation

作　　者：陶承阳 袁杰[1] 回天 加尔肯别克 TAO Chengyang;YUAN Jie;HUI Tian;Jiaerkenbieke(School of Electrical Engineering,Xinjiang University,Urumqi 830046,China;Unmanned System Research Institute,Northwestern Polytechnical University,Xi’an 710072,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830046 [2]西北工业大学无人系统技术研究院,陕西西安710072

出　　处：《现代电子技术》2022年第13期152-158,共7页Modern Electronics Technique

基　　金：国家自然科学基金项目(61863033);新疆维吾尔自治区“天山青年计划”:优秀青年科技人才培养项目(2019Q018)。

摘　　要：针对自主空中加油过程中受油机成像端由于运动模糊、失焦、遮挡等出现退化帧的情况,造成检测算法性能下降的问题,将Yolov3网络与FlowNet 2.0网络相结合,通过光流估计网络获得含有物体运动信息的光流场,根据光流场信息将临近帧提取的特征图对齐到当前帧并进行特征聚合,从而增强当前帧的特征质量,增强算法在退化帧上的检测效果,提高对加油锥套的平均检测精度。在测试集上的实验结果表明,该算法的平均检测精度达到85.81%,召回率达到97.33%,相较于Yolov3算法在平均检测精度方面提升9.87%,在召回率方面提升8.06%。根据所提的加油锥套检测稳定性指标,评估结果表明该算法的稳定性相较于Yolov3算法提高了20.14%,且在视频退化帧上的检测效果得到了明显的提升。In the process of autonomous aerial refueling,the frame degradation may appear at the imaging end of the refueled aircraft due to blurring motion,out⁃of⁃focus and occlusion,which can cause the performance reduction of the detection algorithm.In this paper,Yolov3 network and FlowNet 2.0 network are combined,and the optical flow field containing the motion information of the object is obtained by the optical flow estimation network.According to the information of the optical flow field,the feature images extracted from the adjacent frames are aligned to the current frame and feature aggregation is carried out,so that the feature quality of the current frame is enhanced,the detection effect of the algorithm on degraded frames is enhanced,and the average detection accuracy for the refueling cone sleeve is improved.The experimental results on the test set show that the average detection accuracy of the proposed algorithm is 85.81%and its recall rate is 97.33%.In comparison with the Yolov3 algorithm,its average detection accuracy is improved by 9.87%and its recall rate is improved by 8.06%.According to the proposed stability index of detecting the refueling cone sleeve,the evaluation results show that the stability of the proposed algorithm is improved by 20.14%in comparison with that of the Yolov3 algorithm,and the detection effect on video degradation frames has been significantly improved.

关 键 词：加油锥套检测 特征聚合 自主空中加油 光流估计 光流场获取 检测算法 实验结果分析 

分 类 号：TN99-34[电子电信—信号与信息处理] TP391.4[电子电信—信息与通信工程]