摘要
现如今无人机技术在民用和军用领域得到了广泛应用,但是由于无人机技术的易于获取和操作,导致无人机的“黑飞”与“滥飞”问题严重。传统的反无人机检测方法在复杂的飞行环境下常面临误检、漏检等问题,导致对无人机的检测不够准确。因此,文章提出了一种基于超分辨率重建与YOLOv5s融合的检测方法。该方法使用改进的增强型超分辨率重建生成对抗网络(Real-ESRGAN)提升图像分辨率,使得检测网络能够提取到小目标的更多特征信息;使用优化的YOLOv5s对重建后的图像进行检测。实验结果表明,该方法在反无人机检测方面表现出色,精度高达90.3%,相较SSD、YOLOv7等经典目标检测模型效果更好。
Nowadays,drone technology is extensively utilized in civilian and military sectors.However,due to the ease of access and operation of drone technology,the problem of“black flight”and“indiscriminate flight”of drones is serious.Traditional anti drone detection methods often face issues such as false detections and missed detections in complex flight environments,resulting in inaccurate detection of drones.Therefore,a detection approach is introduced,which combines super-resolution reconstruction with the YOLOv5s algorithm.Firstly,an improved enhanced super-resolution reconstruction generative adversarial network(Real ESRGAN)is used to enhance image resolution,enabling the detection network to extract more feature information of small targets.Then an optimized YOLOv5s is used to detect the reconstructed image.Finally,the experimental results show that this method performs well in anti drone detection,with the accuracy up to 90.3%,which is better than classical object detection models such as SSD and YOLOv7.
作者
田帅
李盛
王露曼
邱博之
Tian Shuai;Li Sheng;Wang Luman;Qiu Bozhi(Xijing University,Xi’an 710123,China)
出处
《无线互联科技》
2024年第8期103-105,共3页
Wireless Internet Technology
基金
国家自然科学基金,项目名称:男女性语音产生机制差异的空气动力学建模对比研究,项目编号:11974289。
关键词
超分辨率重建
注意力机制
小目标检测
复杂背景
反无人机算法
super resolution reconstruction
attention mechanism
small object detection
complex background
anti drone algorithm