Coarse-to-fine Brinell hardness indentation diameter measurement

Computer vision based method becomes an efficient way to measure the Brinell hardness indentation diameter in recent years. Improving the measurement accuracy and the algorithm speed in such low contrast images is the most important topic. In this paper, we proposed a coarse-to-fine measurement method. In the coarse stage, we get a low-resolution image which is down-sampled from the original high-resolution image, and locate the initial position of the coarse indentation circle C0 by using existence probability map. In the fine stage, we projected circle C0 points back into the original image space, and get a series of seed points. The seed points and its neighborhoods determined the precise points of the indentation circle with gaussian fitting method. Finally, we carry out least square method to fit the precise points, and calculate the indentation diameter. Experiment results show the promising performances of our proposed method, which meet the requirement of accuracy, speed, and consistency.