1. College of Geo-exploration Science and Technology, Jilin University, Changchun 130026, China;
2. Xi'an Center of Mineral Resources Survey, China Geological Survey, Xi'an 710100, China;
3. Key Laboratory of Urban Land Resources Monitoring and Simulation, MNR, Shenzhen 518000, Guangdong, China

In the process of mineral exploration, it is vital to delineate polymetallic anomalies using geochemical data. To solve the problems of the traditional linear inversion model, which is highly complex, slow and with poor model effect, a remote sensing geochemical nonlinear inversion model is proposed based on stacked auto-encoder (SAE) and extreme learning machine (ELM). The geochemical contents and abnormal distribution of Cu, Pb, Zn, W and Mo in the soil from Chenzhou of Hunan Province were studied. Experiment results show that the inversion results of SAE-ELM are of high accuracy, and the average relative error of each element is 0.222, and there is a good spatial correspondence between anomaly distribution and polymetallic anomaly reference map.