The present invention discloses a train wheel tread extraction and splicing method based on image segmentation and image fusion. The method comprises the steps of: dividing a wheel tread primary areaaccording to the feature of wheel image collection device fixed point imaging, employing a gray-scale projection technology, further locating a tread area according to the size of the gray projectionvalue and the variation rate, on this basis, employing an edge detection method to detect a wheel flange wire, completing wheel tread extraction, and finally, calculating the minimum smooth width according to the gray difference of a wheel tread overlapping area, splicing the wheel tread through a pixel matching method based on a space, and generating a 360-degree wheel tread panoramic image. Thewheel tread extraction and splicing method based on image processing belongs to the pre-processing technology of a locomotive damaging detection system, can effectively and rapidly extract the locomotive tread area, can improve the preprocessing speed of the locomotive damaging detection system, and has high practicability and wide application prospect.