| 摘要: |
| 在无人机个体识别中，直接用双谱矩阵进行个体识别要计算复杂的二维模板，运算效率低。针对这一不足，提出了一种基于二维双树复小波变换的二次特征提取算法。该算法将双谱分解成多个方向子带图并计算其能量和能量偏差，将维度较高的双谱矩阵高效地转换为维数较低的图像纹理特征，再将其送入支持向量机实现无人机个体识别。采用实采的Phantom 3 Advanced与Mavic Pro图传信号对算法进行验证，结果表明，基于二维双树复小波变换比直接用双谱矩阵进行分类的运算效率高21倍，准确率相较于基于积分双谱、基于灰度共生矩阵、基于小波变换法有不同程度的提升，满足准确性和实时性的需求。 |
| 关键词: 无人机个体识别 双谱矩阵 二维双树复小波变换 特征提取 |
| DOI： |
|
| 基金项目:四川省科技厅重点研发项目(2020YFS0507，2020YFG0271）；成都市重点研发支撑计划项目（2019-YF05-02636-SN） |
|
| UAV individual recognition based on two-dimensional dual-tree complex wavelet transform |
| LUO Zhenghua,LI Xia,YANG Yaoru,XIANG Bo,LUO Xiaodi |
| (1.School of Electronic Information and Electrical Engineering,Chengdu University,Chengdu 610106,China;2.The Fifth Research Institute of Telecommunications Technology,Chengdu 610062,China) |
| Abstract: |
| In unmanned aerial vehicle（UAV） individual identification,directly using bispectral matrix for individual identification is inefficient because of the heavy computing cost of complex two-dimensional template.To overcome this shortcoming,a secondary feature extraction algorithm based on two-dimensional dual-tree complex wavelet transform is proposed.This algorithm decomposes the bispectrum into multiple directional subband maps and calculates their energy and energy deviation,so as to efficiently convert the bispectral matrix with higher dimensions into image texture features with lower dimensions,and then sends image texture features to the support vector machine（SVM） to realizes the individual identification of the UAVs.The algorithm is verified on actual Phantom 3 Advanced and Mavic Pro image transmission signals.Experiments show that the calculation efficiency based on the two-dimensional dual-tree complex wavelet transform is 21 times higher than that of the bispectral matrix.The accuracy is higher than applying methods based on the integral bispectrum,the gray-level co-occurrence matrix,and the wavelet transform.Improvements introduced by the proposed algorithm meet high accuracy and real-time needs. |
| Key words: UAV individual recognition bispectral matrix two-dimensional dual-tree complex wavelet transform feature extraction |
《电讯技术》编辑部 