With the wide development of injection molding process, the embedded monitoring system based on machine vision has been developed to automatically monitoring abnormality of injection molding processing. First, the construction of hardware system and embedded software system were designed. Then camera calibration was carried on to establish the accurate model of the camera to correct distortion. Next the segmentation algorithm was applied to extract the monitored objects of the injection molding process system. The realization procedure of system included the initialization, process monitoring and product detail detection. Finally the experiment results were analyzed including the detection rate of kinds of the abnormality. The system could realize the multi-zone monitoring and product detail detection of injection molding process with high accuracy and good stability.