一种基于深度学习的禁飞区无人机目标识别方法  

A UAV Target Recognition Method for No Flying Zone Based on Deep Learning

作　　者：虞晓霞 刘智[1] 耿振野[1] 陈思锐 YU Xiaoxia;LIU Zhi;GENG Zhenye;CHEN Sirui(School of Electronics and Information Engineering,Changchun University of Science and Technology,Changchun 130022)

机构地区：[1]长春理工大学电子信息工程学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2018年第3期95-101,共7页Journal of Changchun University of Science and Technology(Natural Science Edition)

摘　　要：随着科学技术的迅速发展,中小型无人机逐步走向民用市场。但由于监管技术的缺失,无人机擅闯禁飞区事件屡见不鲜,严重扰乱了空域交通安全。针对传统无人机探测方法成本高、效率低和适应性差等问题,引入深度学习这一新技术,提出了一种基于深度学习的禁飞区无人机目标识别方法。通过对Le Net-5模型进行结构改进,构建一个无人机特征学习网络,经过训练后得到效果良好的模型,实现无人机目标的自主识别。其识别结果可以为禁飞区的监控预警系统提供重要信息,进一步有效保障重要空域的交通安全。实验结果表明,该方法可以有效实现禁飞区无人机目标识别,且误差率比直接应用经典Le Net-5模型减小0.36%。With the rapid development of science and technology, small and medium-sized unmanned aerial vehicle（UAV） gradually move towards civilian market. However,because of the lack of regulatory technology,it is not uncommon for UAV to engage in no-fly zones,seriously disrupting the traffic safety in airspace. Aimed at the problems of high cost,low efficiency and poor adaptability of traditional UAV detection methods,in this paper,a new idea that deep learning and proposes a method of UAV target recognition based on deep learning was introduced. By changing the structure of the Le Net-5 model,a UAV feature learning network was constructed,and a better model was obtained after training to realize the autonomous recognition of the UAV target. The result of the identification can provide important information for the monitoring and early warning system in the no-fly zone,and further effectively protect traffic safety in important airspace. The experimental results show that the method can effectively realize the target recognition in no-fly zone, and the error rate is reduced by 0.36% than the direct application of classical Le Net-5 model.

关 键 词：无人机 禁飞区 深度学习 自主识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]