The various mechanical indicators of foundation rock and soil are the core basis for the design of building foundation schemes, bearing capacity determination, and settlement prediction. Indoor sampling tests can easily disrupt the original stress state of the soil, while in-situ tests can maintain the natural structure and stress level of the soil layer and obtain true mechanical parameters. When conducting in-situ detection on site, the soil layer's own structure, the operation process of the test, on-site construction disturbances, and the supporting conditions of the instruments will continuously interfere with the final collected parameters. Systematic deviations in the indicators will directly mislead the design values of the foundation. This article focuses on three fundamental theories: maintaining the in-situ stress state of rock and soil, coupling changes of in-situ test parameters, and closed-loop control of on-site testing quality in geotechnical engineering. It sorts out various parameter interference factors existing throughout the entire process of on-site in-situ testing. A complete implementation control path is established from four aspects: pre-treatment of soil layers, standardized in-situ test operations, multi-parameter mutual verification mechanisms, and a full-process detection quality control system. This approach weakens the parameter distortion caused by external conditions, enhances the stability of in-situ test data for foundation rock and soil, and provides reliable technical support for the design and safety review of building foundation engineering.