Prochloraz and imazalil are commonly used preservative fungicides for citrus fruits. The mixture of the two pesticides can effectively reduce pathogenic bacteria drug resistance and achieve better preservative effects. However, the high concentration of pesticide residues on the surface of fruits and vegetables will affect consumers health. In this paper, a rapid and accurate method for detecting pesticide residues in the citrus epidermis was established based on the combination of surface-enhanced Raman spectroscopy and chemometrics methods by taking ugly orange as matrix, prochloraz and imazalil mixed pesticides as the research object. In order to compare the enhancement effects of the gold sol and silver sol, they were respectively used on prochloraz standard solution, imazalil standard solution and mixed pesticide solution of prochloraz and imazalil based on orange peel extract at the concentration of 10 mg·L-1 respectively. Then the Raman spectra of the prepared sample solutions were collected. The results showed that the enhancement effect of gold sol in prochloraz standard solution orimazalil standard solution was better, while silver sol in prochloraz and imazalil mixed pesticides was good. Additionally, to obtain the best reinforcing effect of gold sol substrate, a comparative test was carried out, which determined that the volume ratio of gold sol reinforced substrate to two standard pesticide solutions is 1∶1, and the concentration of agglomerating agent NaCl is 1 mol·L-1. According to the direction from high to low, the spectra of prochloraz standard solutions and imazalil standard solutions at different concentrations were collected. The detection limits were lower than 1 mg·L-1 and 0.5 mol·L-1 respectively, within the maximum pesticide residue limit of 5 mg·L-1 for citrus crops stipulated by the state. In the quantitative analysis experiment of mixed pesticides with prochloraz and imazalil, taking orange epidermis extract as matrix and silver sol with better enhancement effect as enhancement substrate, the surface enhanced Raman spectra of mixed pesticides including prochloraz and imazalil with gradient concentration (5~42 mg·L-1) were collected. Then, various pretreatment methods were used to optimize the spectral data, and four regression models containing support vector regression (SVR), support vector regression optimized by the Grey Wolf (GWO-SVR)algorithm, support vector regression optimized by particle swarm optimization (PSO-SVR)and support vector regression optimized by genetic (GA-SVR) algorithm were compared to establish anaccurate and reliable quantitative model. The results showed that the best prediction effect was achieved on the regression model which was established by support vector regression optimized by Grey Wolf (GWO-SVR) algorithm with the characteristic peak intensities of 829 and 1 168 cm-1 as input after the first-order difference preprocessing. The corrected correlation coefficient (RC), the root mean square error (RMSEC) of the correction set, the predicted correlation coefficient (RP), and the root mean square error (RMSEP) of the correction predicted were 0.978, 1.655 mg·L-1, 0.967 and 2.227 mg·L-1 respectively. In conclusion, the proposed method was proved to be effectively applied for the qualitative and quantitative detection of mixed pesticides with prochloraz and imazalil in the citrus epidermis. It could provide a new approach for detecting pesticide residues in citrus.