The work aims to quickly detect the liquid level, impurities and bottle cracks, etc. after the wine filling. Machine vision technology was used and an appropriate detection device was chosen to build a machine vision inspection system for wine filling quality. The algorithm procedures were designed by using Matlab R2015a software to detect the liquid level, impurities and cracks. Sobel operator in the vertical direction was utilized to detect liquid level. Background subtraction method was adopted to detect impurities and cracks. The recognition rate of liquid level measurement was 97% by means of Sobel operator edge detection after wine filling. After the detection of impurities in the bottle in background subtraction method, the recognition rate was 98%. The experiment has certain operability in detecting the liquid level and impurities after wine filling.