欢迎访问《图学学报》 分享到：
图学学报

图学学报

• 图像与视频处理 • 上一篇    下一篇

深度残差网络的无人机多目标识别

  

  1. 1. 南昌航空大学信息工程学院，江西 南昌 330063；
    2. 南昌航空大学无损检测技术教育部重点实验室，江西 南昌 330063
  • 出版日期:2019-02-28 发布日期:2019-02-27
  • 基金资助:
    国家自然科学基金项目(61663030，61663032)

Multi-Objective Identification of UAV Based on Deep Residual Network

  1. 1. School of Information Engineering, Nanchang Hangkong University, Nanchang Jiangxi 330063, China;
    2. Key Laboratory of Nondestructive Testing, Nanchang Hangkong University, Ministry of Education, Nanchang Jiangxi 330063, China
  • Online:2019-02-28 Published:2019-02-27

PDF (PC)

292

摘要： 传统目标识别算法中，经典的区域建议网络(RPN)在提取目标候选区域时计算量 大，时间复杂度较高，因此提出一种级联区域建议网络(CRPN)的搜索模式对其进行改善。此外， 深层次的卷积神经网络训练中易产生退化现象，而引入残差学习的深度残差网络(ResNet)，能 够有效抑制该现象。对多种不同深度以及不同参数的网络模型进行研究，将两层残差学习模块 与三层残差学习模块结合使用，设计出一种占用内存更小、时间复杂度更低的新型多捷联式残 差网络模型(Mu-ResNet)。采用 Mu-ResNet 与 CRPN 结合的网络模型在无人机目标数据集以及 PASCAL VOC 数据集上进行多目标识别测试，较使用 ResNet 与 RPN 结合的网络模型，识别准 确率提升了近 2 个百分点。

关键词: 无人机, 残差网络, 级联区域建议网络, 目标识别

Abstract:  In traditional target recognition algorithms, the classical region proposal net (RPN) has large amount of computation and high complexity of time at extracting the target candidate region. Cascade region proposal network (CRPN) is proposed as a new search method for improving the performance of RPN, in which residual learning based deep residual network (ResNet) is also used effectively to suppress the degradation phenomenon in deep-level convolution neural networks. Aimed at the network models with different depths and parameters, a novel multi-strapdown residual network (Mu-ResNet) model, which is of less memory and lower time complexity, is designed by combining two-layer and three-layer residual learning modules. The combination model of Mu-ResNet and CRPN is used for multi-target recognition test by using the unmanned aerial vehicle (UAV) target data and PASCAL VOC data. The results have shown that nearly 2% of recognition accuracy is increased compared with the combination model of ResNet and RPN.

Key words: unmanned aerial vehicle (UAV), residual network, cascade region proposal network, target recognition