Objective: In this study, we established Sandwich enzyme-linked immunosorbent assay (sELISA) methods to detect major soybean allergens β-conglycinin and glycinin. Methods: A single-factor experiment was conducted to optimize the method for isolating and enriching β-Conglycinin and glycinin from defatted soy protein powder, determining the addition amounts of the precipitants CaCl2 and NaHSO3, and increasing the separation time and speed to enhance the separation effect. The checkerboard method was used to determine the optimal working concentrations of the capture and detection antibodies, the reaction time for each step, and the dilution solvent for the target protein. An ELISA detection method was established and its performance was evaluated. Results: Before centrifugal separation of β-Conglycinin and glycinin, 10 mmol/L CaCl2 and 0.01 mmol/L NaHSO3 were added as precipitants and reducing agents, respectively. The precipitation was left overnight and the centrifugation speed was increased to 10,000 rpm, resulting in extraction rates of 41.81% and 54.06% for β-Conglycinin and glycinin, respectively, with clear protein bands. For the sandwich ELISA of β-Conglycinin, 0.5 μg/mL 3A9 mAb was used as the capture antibody, and 2E2mAb-HRP was diluted 1:4000 as the detection antibody. The antigen and detection antibody were incubated for 90 min each, and the color development was for 15 min. A four-parameter fit and linear fit were used to draw the standard curve. The LOD of this method was 1.36 ng/mL, and the LOQ was 5 ng/mL. In the actual sample spiked recovery experiment, the recovery rate was between 96% and 108%, and the intra-batch and inter-batch coefficient of variation were both less than 15%. For the sandwich ELISA of glycinin, 0.5 μg/mL 7D3 mAb was used as the capture antibody, and 4G4mAb-HRP was diluted 1:1000 as the detection antibody. The antigen and detection antibody were incubated for 120 min each, and the color development was for 15 min. A linear fit was used to draw the standard curve. The LOD of this method was 28.75 ng/mL, and the LOQ was 32.5 ng/mL. In the actual sample spiked recovery experiment, the recovery rate was between 90.3% and 107.2%, and the intra-batch and inter-batch coefficient of variation were both less than 15%, both two detection methods had no cross-reaction with other allergens, and the specificity was good. Conclusion: The sandwich ELISA methods established are sensitive, rapid, accurate, and stable, and can be used for the detection of β-Conglycinin and glycinin in food.