基于磁梯度张量的磁目标模式识别方法
|
摘要
针对目前地下小型磁目标形状识别局限于磁测数据的反演,受测量精度影响大,识别效果不理想的问题,提出了基于磁梯度张量和支持向量机的地下磁目标模式识别方法。该方法将机器学习的方法引入地下磁目标识别领域,利用量子粒子群改进的支持向量机(QPSO-SVM)识别地下小目标的形状。同时从样本信号中计算并分离出基于磁梯度张量矩阵的9个特征量联合识别磁目标,并对磁异常数据进行化极和延拓处理,提高了数据质量,使数据特征更突出。仿真和实验结果证明,本方法克服了重磁数据正、反演过程中大量的公式推导和计算,降低了对磁测数据精度的依赖,提高了识别正确率。
In view of the fact that the shape recognition of underground small magnetic targets is limited to the inversion of magnetic data, and the influence of measurement accuracy is large, and the recognition effect is not ideal, a magnetic gradient tensor and support vector machine based underground magnetic target pattern recognition method was proposed. This method introduced the machine learning method into the field of underground magnetic target recognition, and used the Quantum Particle Swarm Optimization Support Vector Machine(QPSO-SVM) to identify the shape of the underground small target. At the same time, the nine feature quantities based on the magnetic gradient tensor matrix were calculated and separated from the sample signal to jointly identify the magnetic target, and the magnetic anomaly data was subjected to the polarization and extension processing, which improved the data quality and made the data features more prominent. The simulation and experimental results showed that the method overcome a large number of formula derivation and calculation in the process of gravity and magnetic data inversion and inversion, which reduced the dependence on the accuracy of magnetic measurement data and improved the recognition accuracy.
In view of the fact that the shape recognition of underground small magnetic targets is limited to the inversion of magnetic data, and the influence of measurement accuracy is large, and the recognition effect is not ideal, a magnetic gradient tensor and support vector machine based underground magnetic target pattern recognition method was proposed. This method introduced the machine learning method into the field of underground magnetic target recognition, and used the Quantum Particle Swarm Optimization Support Vector Machine(QPSO-SVM) to identify the shape of the underground small target. At the same time, the nine feature quantities based on the magnetic gradient tensor matrix were calculated and separated from the sample signal to jointly identify the magnetic target, and the magnetic anomaly data was subjected to the polarization and extension processing, which improved the data quality and made the data features more prominent. The simulation and experimental results showed that the method overcome a large number of formula derivation and calculation in the process of gravity and magnetic data inversion and inversion, which reduced the dependence on the accuracy of magnetic measurement data and improved the recognition accuracy.
引文
[1]王林飞,郭灿灿,薛典军,等.磁梯度张量解析信号分析法及其在场源位置识别中的应用[J].地球物理学进展,2016,31(3):1164-1172.
[2]孙刃.磁异常正反演技术在水下磁性目标探测中的应用研究[D].北京:中国地质大学,2014.
[3]吴国超.基于磁异常的目标体定位反演方法研究[D].吉林:吉林大学,2015.
[4]徐熠.磁梯度张量异常反演算法与算例[D].北京:中国地质大学,2015.
[5]李金朋,张英堂,范红波,等.基于磁梯度张量的地下小目标相关成像方法[J].探测与控制学报,2016,38(3):75-78.
[6]尹刚,张英堂,范红波,等.基于磁传感器阵列的磁性目标跟踪方法[J].上海交通大学学报,2015,49(12):1748-1752.
[7]Luthi S M.Textural segmentation of digital rock images into bedding units using texture energy and cluster labels[J].Mathematical Geology,1994,26(2):181-196.
[8]Rosati I,Cardarelli E.Statistical pattern recognition technique to enhance anomalies in magnetic surveys[J].Journal of Applied Geophysics,1997,37(2):55-66.
[9]Brown M P,Poulton M M.Locating buried objects for environmental site investigations using neural networks[J].Journal of Environmental & Engineering Geophysics,2008,1(3).
[10]Calderón-Macías C,Sen M K,Stoffa P L.Artificial neural networks for parameter estimation in geophysics[J].Geophysical Prospecting,2010,48(1):21-47.
[11]Ehret B.Pattern recognition of geophysical data[J].Geoderma,2010,160(1):111-125.
[12]谢永茂.基于模板匹配的局部磁异常识别方法[D].北京:中国地质大学,2012.
[13]尹刚,张英堂,李志宁,等.磁偶极子梯度张量的几何不变量及其应用[J].地球物理学报,2016,59(2):749-756.
[14]张昌达.航空磁力梯度张量测量——航空磁测技术的最新进展[J].工程地球物理学报,2006,3(5):354-361.
[15]张光辉,程昱.干扰识别的量子粒子群和支持向量机算法[J].中国电子科学研究院学报,2011,6(5):490-493.
[16]王小川.MATLAB神经网络43个案例分析[M].北京:北京航空航天大学出版社,2013.
[17]孙瑶琴.改进粒子群算法优化支持向量机在故障诊断中的应用研究[J].计算机测量与控制,2017,25(3):48-50.
[18]李旭芳,王士同.基于QPSO训练支持向量机的网络入侵检测[J].计算机工程与设计,2008,29(1):34-36.
[19]荆磊,杨亚斌,陈亮,等.改进的阻尼法低纬度磁异常化极方法[J].地球物理学报,2017,60(2):843-850.
[20]Keating P.An improved technique forreduction to the pole at low latitudes[J].Geophysics,1996,61(1):131-137
[21]孟慧.磁梯度张量正演、延拓、数据解释方法研究[D].吉林:吉林大学,2012.
[2]孙刃.磁异常正反演技术在水下磁性目标探测中的应用研究[D].北京:中国地质大学,2014.
[3]吴国超.基于磁异常的目标体定位反演方法研究[D].吉林:吉林大学,2015.
[4]徐熠.磁梯度张量异常反演算法与算例[D].北京:中国地质大学,2015.
[5]李金朋,张英堂,范红波,等.基于磁梯度张量的地下小目标相关成像方法[J].探测与控制学报,2016,38(3):75-78.
[6]尹刚,张英堂,范红波,等.基于磁传感器阵列的磁性目标跟踪方法[J].上海交通大学学报,2015,49(12):1748-1752.
[7]Luthi S M.Textural segmentation of digital rock images into bedding units using texture energy and cluster labels[J].Mathematical Geology,1994,26(2):181-196.
[8]Rosati I,Cardarelli E.Statistical pattern recognition technique to enhance anomalies in magnetic surveys[J].Journal of Applied Geophysics,1997,37(2):55-66.
[9]Brown M P,Poulton M M.Locating buried objects for environmental site investigations using neural networks[J].Journal of Environmental & Engineering Geophysics,2008,1(3).
[10]Calderón-Macías C,Sen M K,Stoffa P L.Artificial neural networks for parameter estimation in geophysics[J].Geophysical Prospecting,2010,48(1):21-47.
[11]Ehret B.Pattern recognition of geophysical data[J].Geoderma,2010,160(1):111-125.
[12]谢永茂.基于模板匹配的局部磁异常识别方法[D].北京:中国地质大学,2012.
[13]尹刚,张英堂,李志宁,等.磁偶极子梯度张量的几何不变量及其应用[J].地球物理学报,2016,59(2):749-756.
[14]张昌达.航空磁力梯度张量测量——航空磁测技术的最新进展[J].工程地球物理学报,2006,3(5):354-361.
[15]张光辉,程昱.干扰识别的量子粒子群和支持向量机算法[J].中国电子科学研究院学报,2011,6(5):490-493.
[16]王小川.MATLAB神经网络43个案例分析[M].北京:北京航空航天大学出版社,2013.
[17]孙瑶琴.改进粒子群算法优化支持向量机在故障诊断中的应用研究[J].计算机测量与控制,2017,25(3):48-50.
[18]李旭芳,王士同.基于QPSO训练支持向量机的网络入侵检测[J].计算机工程与设计,2008,29(1):34-36.
[19]荆磊,杨亚斌,陈亮,等.改进的阻尼法低纬度磁异常化极方法[J].地球物理学报,2017,60(2):843-850.
[20]Keating P.An improved technique forreduction to the pole at low latitudes[J].Geophysics,1996,61(1):131-137
[21]孟慧.磁梯度张量正演、延拓、数据解释方法研究[D].吉林:吉林大学,2012.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |