Image segmentation of argon blowing based on improved Otsu algorithm

摘要

Argon blowing at the bottom of ladle is a process for smelting high-quality steel. For realizing the argon blowing automatic control of flow based computer vision, the argon blowing interface is monitored by a camera. Image segmentation of argon blowing is the key to get clear exposed area of molten steel. In order to ensure the accuracy and real-time of image segmentation, an improved Otsu algorithm is proposed. Firstly, the simulated annealing algorithm is adopted to optimum the new individual selection process of genetic Algorithm. Then use the proportion of the target area pixels to the image as the weight to modify the formula of maximum interclass variance. Finally, the improved Otsu formula is used as the fitness function of the genetic algorithm for image thresholding. Experiments show that the proposed algorithm can achieve greater F-score than other algorithms, and the requirement of real-time image processing can be satisfied, which is valuable in engineering application.