An x-ray vision is a useful method to measure or inspect objects which can not be done by conventional camera vision method. Our problem is to identify the pose of an object from an x-ray projection image. It is assumed here that the x-ray imaging coordinates, such as the locations of the x-ray source and the image plane, are predetermined and that the object geometry is known. For simplicity, only polyhedral objects are considered whose image features consist of corner points and edge lines. By detecting the corner points or lines in the image, the pose of an object can be estimated iteratively by using the feature matching between an object and an x-ray projection image. To extract the features of objects in x-ray images, we propose a reliable and efficient image processing algorithm. The method uses the typical characteristics of the x-ray image for the polyhedral object that the intensity distribution also can be represented by a polyhedron which consists of planes. By this approximation, edges in the image can be determined by extraction of the edge lins between the adjacent planes in the intensity distribution. Then, by applying Hough Transform to the edge images and analyzing the lines, the vertices can be determined.
In the final manuscript, the performance of the proposed image processing method is compared with that of the conventional ones such as sobel, laplacian operators. Also, performance of the algorithms proposed here will be discussed in detail including the limitations of the method.
