Rail Surface Defect Recognition Method Based on AdaBoost Multi-classifier Combination

Rail surface defects have the characteristics of various types and complex morphological characteristics. It is difficult to obtain accurate classification results only by using a single classification method. Therefore, this paper presents a rail surface defect recognition method based on AdaBoost multi-classifier combination. Firstly, defect attributes are described by extracting geometric shape and gray level features of defect area, and Relief algorithm is used to select defect features and filter out features unrelated to classification. Then by using AdaBoost multi-classifier combination method and taking CART decision tree as a weak classification algorithm to design a combined classifier, rail surface defects classification is realized. The results show that this method can effectively identify three common types of defects: rail surface peeling block, tread crack and fish scale peeling crack. Key Words: Rail surface defects, Feature extraction, AdaBoost multi-classifier combination, CART decision tree