井下巷道地磁匹配特征的CEA卷积增强的分析

针对井下某些巷道地磁空间变化平缓，地磁匹配概率低的问题，构建了井下巷道地磁卷积增强算子（convolution enhancement algorithms, CEA），进行地磁匹配前的目标区域和匹配向量的卷积增强预处理，去除数据噪声和增强识别特征。以Laplace、高通滤波（High Pass）、索伯尔滤波（Sobel）图像卷积算子为基础，通过列向量特征的锐化处理，建立了井下巷道地磁卷积增强的Laplace、High Pass和Sobel卷积算子模板。选取某金矿4个巷道的地磁数据，开展了CEA算子卷积前后的均方差算法地磁匹配定位的仿真试验。试验结果表明，CEA算子卷积可以增强匹配序列和地磁图的地磁空间特征，降低了匹配数据中的噪声影响。在数据CEA卷积前后的地磁统计特征对比中发现，Laplace算子不仅保持了原有地磁图变化特征，还增大了数据空间变化的差异度，降低了相关性，效果明显。特别是600 nT的高噪声干扰匹配试验中，Laplace算子卷积能够降低噪声对地磁定位扰动影响，有效提高了地磁匹配定位的概率和精度，具有较强的鲁棒性，适合作为井下巷道地磁匹配的数据预处理模型。

Performance Analysis of Convolution Enhancement of CEA Operator for Underground Geomagnetic Matching

  Obiectives  Underground geomagnetic positioning is a new method for emergency refuge and rescue. Its geomagnetic matching probability and accuracy will be affected by the richness and stability of the matching area. There is a problem of low probability of geomagnetic matching in some areas where the geomagnetic space changes gently.  Methods  A convolution enhancement operator of the underground is constructed, named convolution enhancement algorithms(CEA), which is used to convolution enhancement processing of geomagnetic data in underground matching area and geomagnetic vector of the target moving track, so as to remove the noise of measured geomagnetic data and enhance the richness of geomagnetic information. CEA operator is a kind of convolution sharpening of geomagnetic features with column quantization, is the modification of Laplace, High Pass and Sobel operators of image convolution, which is suitable for underground strip area. The simulation test of mean-square-error algorithm(MSD) geomagnetic matching before and after the convolution of CEA operator is carried out, which data is the geomagnetic data of 4 tunnels in a gold mine.  Results and Conclusions  The test results show that the statistical parameters of geomagnetic data fluctuate obviously after convolution by three CEA operators. The variation of geomagnetic data features before and after convolution by the High Pass operator is small, the variation of geomagnetic standard deviation and geomagnetic roughness is small, and the correlation coefficient sometimes even weakens. The information entropy of geomagnetic data before and after convolution by the Sobel oper?ator is increased, which cannot keep the consistency of geomagnetic map trend, appears alienation. The Laplace operator has obvious enhance effect, which keeps the change of the original geomagnetic map, increases the difference degree of the data and reduces their correlation. Especially in the matching test of 600 ?nT noise, the convolution processing of Laplace operator can reduce the impact of noise on geomagnetic positioning disturbance, improve the probability and accuracy of geomagnetic matching positioning, and show strong robustness, which is suitable for data preprocessing model of underground geomagnetic matching.