距离选通成像在无人机识别分类中的应用北大核心CSCD

针对无人机夜间难以识别的问题,采用YOLOv5算法对无人机激光图像进行识别分类。首先对距离选通成像原理进行了说明,其次介绍了YOLOv5网络结构,然后建立了五种无人机激光图像数据集,最后使用YOLOv5的四种网络模型对数据进行训练。实验结果得出,四种模型的准确率都是99.73%,损失率都在1%以内,同时,在YOLOv5s网络下五种无人机识别分类的平均识别率为96%、94.6%、95.2%、95.7%、95.4%,识别准确率较高。

Application of range gated imaging in UAV recognition and classification

Aiming at the problem that UAVs are difficult to identify at night,the YOLOv5 algorithm is used to identify and classify UAV laser images.Firstly,the principle of distance gated imaging is explained,and then the YOLOv5 network structure is introduced,then five types of UAV laser image datasets are established.Finally,the data are trained using four network models of YOLOv5.The experimental results show that the accuracy rate of the four models is 99.73%,and the loss rate is within 1%.Meanwhile,the average recognition rate of the five UAV recognition classifications under the YOLOv5s network is 96%,94.6%,and 95.2%,95.7% and 95.4%,with high recognition accuracy.