基于改进Yolov5n的无人机对地面军事目标识别算法

针对目前主流的目标检测算法在真实航拍战场数据背景下识别精度低,误检率与漏检率高等问题,对Yolo目标识别算法进行了研究,提出一种基于改进Yolov5n的轻量化航拍军事目标检测模型；首先,采用ECA注意力机制与主干网络C3模块融合,以解决航拍图像背景复杂且存在相似目标干扰问题；其次,引入归一化高斯瓦萨斯坦距离(NWD)代替CIoU损失函数,提高对模糊小目标的检测识别；最后,采用GSConv轻量化卷积代替标准卷积,减轻模型重量；经过实验测试,改进后的算法模型平均检测精度达到81.5%,提升0.9个百分点,模型大小为3.4MB,减轻0.4MB,识别速度为每秒113帧；实验结果表明该模型在轻量化的同时保持着高精度的航拍军事目标检测；

Uav ground military target recognition algorithm based on improved Yolov5n

Aiming at the problems of low recognition accuracy and high false detection rate and missing rate of the current mainstream target detection algorithms in the background of real aerial battlefield data, the Yolo target recognition algorithm is studied,A lightweight aerial military target detection model based on improved Yolov5n is proposed,Firstly, ECA attention mechanism is integrated with C3 module of backbone network to solve the problem of complex background and similar object interference in aerial images.Secondly, the CIoU loss function is replaced by A Normalized Gaussian Wasserstein Distance (NWD) to improve the detection and recognition of fuzzy small targets.Finally, GSConv lightweight convolution is used instead of standard convolution to reduce the weight of the model. Through experimental tests, the average detection accuracy of the improved algorithm model reaches 81.5%, an increase of 0.9%, the model size is 3.4MB, reduced 0.4MB, the recognition speed is 113 frames per second.The experimental results show that the model is lightweight while maintaining high precision aerial military target detection.