The weight percent SiO₂ of rocks is a basic parameter for the classification of geological bodies and the analysis of genetic evolution, therefore, lots of rocks are collected in the field and analyzed to obtain their weight percent SiO₂ in the laboratory. This method has always been one of the main working means for basic geological research and mineral exploration. Aiming at the disadvantages of long time and high cost of chemical analysis in the laboratory, the paper took thirty-one rocks with different weight percent SiO₂ collected in the field as experimental samples, and carried out prediction method research of weight percent SiO₂ of rocks based on the technology of thermal infrared remote sensing, then constructed three quantitative prediction models. By comprehensive accuracy evaluation of three models, it is concluded that the average accuracy of each model is higher than 80%, but the accuracy for the samples whose real weight percent SiO₂ are higher than 60% is higher than 90%, and the accuracy is up to 99.98%. The results could provide a new technical means for geologists to obtain the weight percent SiO₂ of rocks in the field quickly.

岩石的SiO ₂重量百分比是进行地质体分类和遗传演化分析的基本参数，因此，在实验室中采集了大量岩石并进行分析，以获得其SiO ₂的重量百分比。这种方法一直是基础地质研究和矿产勘探的主要工作手段之一。针对实验室化学分析时间长，成本高的弊端，以田间采集的31份不同SiO ₂重量百分比的岩石为实验样品，进行了SiO ₂重量百分比的预测方法研究。基于热红外遥感技术对岩石进行了定量分析，建立了三种定量预测模型。通过对三个模型的综合精度评估，可以得出每个模型的平均精度均高于80％，但实际重量百分比SiO ₂大于60％的样品的精度高于90％，并且精度高达99.98％。该结果可为地质学家快速获得野外岩石中SiO ₂的重量百分比提供一种新的技术手段。