In order to establish an efficient calibration method for the microscopic parameters of soil-rock mixtures with different rock contents and improve the accuracy of the discrete element simulation of direct shear tests, numerical model samples of soil-rock mixtures were constructed based on the particle flow code in 2 dimensions (PFC2D) and Python programming to realize Monte Carlo random generation of rock blocks. A cluster particle aggregate was used to simulate irregular rock blocks, and a systematic study was conducted on the calibration of microscopic parameters between rock blocks and soil particles. A microscopic parameter grading calibration process based on uniaxial tension, uniaxial compression, biaxial compression, and direct shear tests was proposed. The influence of rock content on the shear strength characteristics of soil-rock mixtures was investigated. The results show that with increasing rock content, the cohesion of the soil-rock mixture initially rises and subsequently declines, peaking at 60% rock content, while the internal friction angle increases monotonically. The study has clarified the influence of key microscopic parameters in the contact model on macroscopic mechanical behavior, and the simulated internal friction angles and cohesion values are very close to the results of in-situ tests.