With the development of industrial robots and modern industrial, the more performance requirements for industrial robots are needed. To improve production efficiency and product quality, intelligent, high speed and high precision are essential requirements for industrial robots. In summary of domestic intelligent beer bottle mouth defect detection method based on machine vision, highspeed and highaccuracy is still a problem to be solved. This paper presents the fourcircle positioning method based on circle fitting assessment method, which greatly improves the accuracy of bottle mouth detection area, and the smart bottle mouth defects detection method based on subregion hysteresis thresholding segmentation of projection features. Collected 488 image tests, the detection accuracy is 99.4%, the average speed of detection is 38 ms. The algorithm proposed in this paper has high detection speed and high detection precision, it can be well applied in the modern industrial robot with high speed and high precision.