中文摘要：

在层析成像中，介质的随机扰动将对反演精度有很大影响．本文在反演迭代过程中，通过自相关平滑算法，将先验约束由单纯的均值拓展为模型方差、相关系数等统计参数，实现了随机介质的分布特征对层析反演的先验约束，有利于压制高频干扰，提高层析成像的精度．岩溶区洞穴探测的工程实例表明，在传统层析反演算法无法获得合理解释的情况下，采用自相关平滑算法仍能获得合理的地质剖面．

英文摘要：

Tomography inversion was largely deteriorated by background media stochastic disturbance. Considering the stochastic disturb of the study area, an Autocorrelation Smooth Method（ASM） was introduced into computer tomography, which successfully extended pre-restriction from mean values to variances, mean values, and autocorrelation coefficients etc. The ASM stochastic computer tomography can suppress high frequency disturbance and improve inversion precision. Engineering examples of electromagnetic tomography exploration to karst cares show that when the traditional tomography method could not get logical interpretation , ASM could still get reasonable geological profiles.