Initial-pressed juice is an important intermediate product in cane sugar industry, and sugar brix is a key indicator for evaluating sugar quality. Real-time evaluation of sugar quality requires determining the content of sugar brix in all steps of the cane sugar process. Near-infrared (NIR) spectroscopy is simple, rapid and non-destructive technologies on the analysis of material contents. In this study, the chemometric algorithm of parameter-combined tuning of Savitzky-Golay (SG) smoother and Partial Least Squares (PLS) regression was utilized for NIR analysis of sugar brix contents in sugarcane initial-pressure juice. The algorithms of combined optimization of SG smoother and PLS regression was achieved and the calibration models were optimally established by screening the expanded 540 SG smoothing modes and the 1-30 latent valuables (LV). The optimized models have high predictive accuracy. These results confirm that the combined optimization of SG smoothing modes and PLS LVs is effective in the quantitative determination of sugar brix contents in sugarcane initial-pressure juice, and that the NIR spectroscopic technology with its chemometric algorithms have the potential in the analysis of cane sugar intermediates.