A novel object detection algorithm is proposed for moving object detection under weak lighting. Firstly, background model is trained for obtaining steady background image. Definition of the background and current frame image are improved by image enhancement processing. Secondly, foreground objects is got by background subtracting method and the difference image is de-noised and mended. Finally, shadow in difference image is eliminated and exact moving object is got by different brightnessly between foreground, shadow and background. The effectiveness of the proposed method is demonstrated experimentally with several video clips collected in a indoor night environment.