Based on structured light vision, this paper presents an interactive projection system. The imaging characteristics in projection systems are analyzed and the gesture figures in the projection area are captured by a camera. Image processing and feature extraction algorithms are presented for gesture recognition. The gesture figures and their shades are segmented, and the centroid and area features are extracted to recognize the gesture movement, thus the gesture instructions are recognized by the projection system. The system adjusts the projection contents according to the gesture instructions, to realize non-contact control and human-machine interaction. Experimental results verify the presented system and method.