Surface images of aluminum alloy after laser cleaning are obtained in real-time with a high speed coupled device and light emitting diode light sources. We design an on-line detection system based on machine vision and propose a dynamic threshold fast position coupling algorithm for laser cleaning aluminum alloy. The proposed algorithm solves the problem of uneven light in laser high-speed cleaning process, realizes the accurate segmentation and the quick positioning of qualified and unqualified areas. The proposed system can detect the quality of the laser cleaning aluminum alloy in real-time. The detection time is reduced, and the recognition accuracy is improved. The system can ensure the cleaning quality.