Fluorescent Magnetic Particle Inspection Crack Object Extraction Method Study

Magnetic fluorescent-penetrant inspection is a NDT method that often used to test the surface cracks of ferromagnetic materials. It is widely used because of its high sensitivity, fast speed and its simple and dependable craft without the destruction of raw materials. Detection of fluorescent magnetic crack images is a rising research and practical subject in recent years. This paper researches the methods of target extraction on fluorescent magnetic crack images.In this paper, the defect characteristics of train wheelsets images are analyzed, the differences of crack defect and false defect are investigated. Several mainstream technologies of digital image segmentation are discussed, including edge-based segmentation, threshold processing and region-based segmentation. The algorithm of blocks extremums is presented on the base of logical assumptions to the crack characteristics. And in the segmentation images based on this algorithm, the crack target extraction algorithm based on histogram fitting curves division of multidimensional features is presented and realized. The advantages and disadvantages are analyzed after comparison of this algorithm and the global threshold segmentation.This paper focus on the target extraction algorithm of node characteristics. For the thin and reticulation characteristics of many cracks, a target extraction algorithm based on node characteristics is put forward. The core of the algorithm is detecting the crack nodes and filtering the pseudo nodes, then recovering the regions in the segmentation image based on double-blocks extremums and thinning. Experimental results show that the method has good effect, and easy to achieve.The experimental results show that the new methods are good both in terms of efficiency and accuracy, and have practical value.