Abstract: A sensing and control system, which includes coaxial vision sensor, DSP processor and open NC system, was designed for CO₂ laser cutting.By coaxially detecting the images of the sparks jet, it is found that the maximal sparks length of the coaxial image, which can be determined by the highest pixels of different brightness ranges of the sparks image, corresponds to the best roughness of cutting face.Experiments for self-optimizing control of laser cutting quality were performed by utilizing the images of sparks jet. The results show that the optimal cutting speed for the best roughness of the bottom edge could be found automatically and quickly.