At present，aluminum-plastic blister packaging is the main method of pharmaceutical packaging，whose imaging changes greatly as the light or shooting angle varies on its surface. Therefore，the visual line-inspection technology，instead of traditional manual inspection，is adopted in this paper to solve the grain-leaking problem during the packaging process. An segmentation algorithm of target region is designed to precisely extractthe target region of aluminum-plastic blister packaging with different sizes or shapes. At the same time，an image segmentation algorithm is designed to distinguish the position of tablets accurately under the unstable light andcomplicated surface reflections in the actual detection. And then，the visual inspection system based on Open CV and Qt has been developed. Using 200 million pixels 30-frame camera，an experiment platform has been set up，which can adapt to a variety of non-white pharmaceutical packaging of aluminum-plastic blister detection，as its image detection speed must be within 150m/s. Besides，the system has lower requirement for illumination and can replace manual inspection with its great adaptability.