ABSTRACT
In the process of mineral deposits formation, elements often show abnormal enrichment. Therefore, the inversion of element contents (the prediction of element contents in mineral deposits through hyperspectral data) has some indicative significance to the exploration of mineral resources. In order to solve the problem of nonlinear copper contents inversion using hyperspectral remote sensing, this paper put forward an improved Gradient Boosting Regression Tree (IGBRT) method to invert copper contents. The main innovations of this paper are: (1) replacing the simple average function with the k-nearest-neighbour weighted average function as the node prediction function to improve the accuracy; (2) using the adaptive reduction step instead of the fixed reduction step to improve the efficiency. At the end of the paper, taking the Altun region, Xinjiang province, China as the experimental area, the iron element which was high correlated with the copper was chosen as the intermediate variable, and the contents relationship between them was established to solve the problem of the small number of copper samples in the study area. Using the improved Gradient Boosting Regression Tree method to predict the iron contents, then the copper contents could be predicted through the relationship between them. The results showed that the IGBRT algorithm had the coefficient of determination (R2) of 0.744, which were 0.185 more than the traditional one, and the learning efficiency was increased by 39.4%. The results of this study can provide a reference for remote sensing inversion of copper contents in mineral deposits of uninhabited areas, and have some important significance for the delineation of copper prospects.
Acknowledgements
This work was supported in part by the second Tibetan Plateau Scientific Expedition and Research (STEP) (2019QZKK0806), in part by the National Natural Science Foundation of China (42071312), in part by the Hainan Hundred Special Project, in part by the Hainan Provincial Department of Science and Technology (ZDKJ2019006), and in part by the Strategic Priority Science Program of Chinese Academy of Sciences (Class A) (XDA19070202).
We give our thanks to Linhai Jin, Haifeng Ding and Hongying Zhou for their data support.
Data Availability Statement
Data supporting reported results can be found through the author’s e-mail, [email protected].
Disclosure
No potential conflict of interest was reported by the author(s).