摘要
深入研究了流行的目标识别方法YOLOv3,将Inception模块融入其特征提取网络darknet-53中,从而得到新网络darknet-139。相比YOLOv3特征提取网络,新网络具有更好的特征提取能力。采集并制作算法所需的数据集,分别在YOLOv3和本文算法上进行训练并测试。实验结果表明,相比YOLOv3,本文算法的平均识别率提升了约2%。
The popular target recognition method YOLOv3 is deeply studied,and the Inception module is integrated into the feature extraction network darknet-53 to get a new network darknet-139.Compared with YOLOv3,the new network has better ability in feature extraction.The data set required by the algorithm is collected and made,and were trained and tested on YOLOv3 and the proposed algorithm,respectively.The experimental results show that the average recognition rate of the proposed algorithm is about 2% higher than that of YOLOv3.
作者
吕攀飞
王曙光
Lü Panfei;Wang Shuguang(Army Artillery and Air Defense Forces Academy,Hefei,Anhui 230031,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2019年第7期108-114,共7页
Laser & Optoelectronics Progress
基金
陆军装备部"十三五"预研基金
关键词
图像处理
无人机作战平台
人工智能
目标识别
image processing
unmanned aerial vehicle combat platform
artificial intelligence
target recognition