Based on image recognition technology of the automatic feature extraction sequence images

In continuous casting process, it is difficult to directly observe the phase change of the crystallizer melt during the melting and crystallization of mold fluxes. To improve the drawbacks of SHTT II melt and crystallization temperature tester, this paper develops automatic feature extraction of sequence images and modeling analysis of sequence images based on image recognition technology. Image segmentation and recognition techniques are used to automatically extract the two temperature data in the upper left corner of each image. The images are pre-processed to obtain binarized images with only black and white colors. The pixels are inverted, and the figures are located and segmented using grayscale histograms of horizontal and vertical projections. After obtaining the individual digits, the digits from 0 to 9 are artificially labeled. The SSIM values of the other digits are compared and identified by calculating the SSIM values of the other digits with the labeled digits. Two temperature curves were drawn based on the obtained temperature results, and the analysis yielded accurate test results for 1#wire and inaccurate results for 2#wire.