As a popular defect of steel, corrosion had been a big challenge to industry safe and structural health. For atmosphere corrosion characterization and evaluation, a clustering by fast search and find of density peaks (CFSFDP) algorithm, combined with gap statistic (GS) method is utilized to corroded Q235 carbon steel tubes. With the proposed method, three natural atmosphere corroded samples are investigated and classified. The proposed method successfully identifies the samples with different service periods. The temperature gradient, which indicates the heat generation and conductivity, is used to analyze cluster center selection. The matching rate is presented as a feature to reflect the corrosion state difference.