The microstructure of high-strength steel (HSS), such as grain size and phase fraction, is a key factor determining its final mechanical properties, including toughness plasticity, and fatigue performance. Traditional metallographic analysis methods suffer from limitations such as strong subjectivity and insufficient quantification. To this end, it is necessary to explore and evaluate the application potential and impact of machine vision technology in the automatic recognition of HSS microstructures, quantitative analysis, and the study of its correlation with mechanical properties. A deep learning-based image segmentation and feature extraction algorithm was developed and optimized, achieving high-precision, automated recognition and quantitative characterization of complex microstructures. Furthermore, it focused on establishing machine learning models to predict key mechanical properties of HSS (tensile strength, elongation) using extracted microstructural feature parameters (e.g., phase fraction, phase content). Through the implementation of these technologies, an analytical approach for influencing microstructure and property regulation is provided.