Abstract: A new nondestructive testing method based on the principle of magnetic-optical imaging was investigated in this paper. Fractal dimension detection was applied to extract features and estimate the optimal scale on the magnetic-optical images of high strength steel that has micro weld defects on the surface. The extracted features of weld defects were trained and tested by using classification algorithm Adabost to build weld defect characteristics and realize automatic identification of magnetic optical images for high strength steel surface defects. The experimental results show that the characteristics of high strength steel weld defects can be detected by using magnetic-optical imaging. Furthermore, the position, shape and category of weld defects can be identified by fractal dimension analysis based on the magnetic-optical images.