Regarding the characteristics of images of train wheel tread defects

defect extraction techniques were studied

for the purpose the train wheel tread region segmentation algorithm based on the stationary wavelet transform adaptive threshold was developed

the burning defect region extraction algorithm that the approximate location based on the block segmentation combined with the accurate location was proposed

the peeling defect extraction method based on searching way of tread edge lines was also proposed based on the searching way of tread edge lines. The proposed algorithms were validated by two images of tread defects. The system was tested by samples

the results show that the missing recognition rates of the burning and peeling were 8.3% and 5.3% respectively and false acceptance rate was 5.1%.