Refractories are essential in high-temperature industries such as metallurgy and construction, where they are subjected to chemical corrosion and mechanical damage under thermal-chemical-mechanical coupling conditions during service, which can lead to the deterioration of furnace linings and impact product quality. It is difficult for traditional methods to directly observe the damage process of refractories in-situ under high temperature and complex environment, and postmortem analysis is often used in this method, which lacks the process information under thermal-chemical-mechanical coupling conditions, leading to the deviation of analysis results. This study investigates the influence of varying contents of molten slag and different basicity (C/S ratios) on the corrosion behavior of corundum based refractories by combining high-temperature machine vision technology with Digital Image Correlation (DIC). The results indicate that the 3–1 mm aggregates can enhance the slag resistance of refractories. High basicity molten slag and excessive slag quantities will disrupt the speckle patterns and affect the accuracy of data. With the quantification and identification of molten slag penetration and corrosion area using the full-field average strain curve and strain nephogram on the refractory surface, a promissing method for precise research and real-time evaluation of corrosion behavior of refractory under complex environment is established.