In recent years, the development of “green tire” has attracted much attention. Many kinds of rubber additives are needed in the manufacturing process of green tires, and the content of rubber additives is closely related to whether green tires can meet the standards. Therefore, it is important to quantitatively detect the rubber additives in tire rubber. THz-TDS technology has been successfully applied in the field of quantitative analysis of substances. However, when the quantitative analysis object is a multi-component mixture, the results of quantitative analysis will not be satisfactory due to the overlap and distortion of the mixture spectrum. In order to solve this problem, Zernike moment is introduced as a spectral pretreatment technology into terahertz spectral quantitative analysis of multi-component mixtures of rubber additives. A quantitative analysis method of terahertz spectrum based on Zernike moment and support vector regression (ZM-SVR) is proposed. Firstly, three rubber additives, zinc oxide, silica and 2-Mercaptobenzothiazole (MBT), which affect the quality of green tires, were used as quantitative detection objects. Three rubber additives and nitrile-butadiene rubber were prepared as multi-mixture experimental samples, and the terahertz spectra of samples were measured by terahertz time-domain spectroscopy system. Then, terahertz spectroscopy was analyzed and processed. After obtaining the three optical parameters of absorption coefficient, extinction coefficient and refractive index, the three optical parameters were constructed into the THz three-dimensional spectrum of the sample, and the characteristic information of the THz three-dimensional spectral gray-scale image was extracted by Zernike moment. Finally, the quantitative model between the characteristic information of the THz three-dimensional spectral gray-scale image of the sample and the content of the target component was established by using support vector regression. The target component content in the mixture sample was analyzed. The correlation coefficients of the forecasting set of the quantitative model obtained by this method were greater than or equal to 0952 2, and the root mean square error was less than or equal to 2267 2%. To further verify the validity of this method, the results of quantitative analysis were compared with those of PLS and SVR. Compared with the quantitative analysis results obtained by conventional methods, the accuracy and stability of the results obtained by Zernike moment combined with support vector regression method have been significantly improved. Therefore, Zernike moment combined with support vector regression provides a new method for terahertz spectroscopy quantitative detection of multi-component mixture of rubber additives, and has broad application prospects in the field of quality detection of green tires and rubber.