Research On X-ray Three-dimensional Stereo Imaging Method Based On Stereo Matching Technology

X-ray imaging can penetrate the object and see the internal structure of the object,has been widely used in airports,stations and other security industry.However,because its perspective is two-dimensional,it cannot distinguish the structural information of different levels,nor can it locate each object,which may cause security risks due to occlusion.Existing CT and CL imaging methods can obtain different depth information of images through reconstruction,but the complex calculation,long imaging time and high cost cannot meet the online detection requirements of security check.In view of the above problems,combined with the knowledge of binocular stereo vision,this paper carried out the research of dual-perspective X-ray 3D stereo imaging method.Difficult problems for X-ray image feature extraction and matching,the deficiency of traditional feature matching algorithm in feature point extraction was analyzed,and a regional feature matching method was studied.Using the principle of binocular stereo vision and combining the gray information of double view image to segment the object,the problem of difficult image matching is optimized to a certain extent.In order to improve the matching rate of the image,the attribute similarity of the segmentation result is used to match the image,and then based on the matching result and parallax,the layered reconstruction is carried out to realize the recognition of different depths of the object.But considering the image segmentation results are greatly influenced by image itself and gray,gray similar objects especially will affect the reconstruction could not be effective segmentation effect,according to the problem,in this paper,combining the theory of image registration,research field conversion algorithm based on markers,through the preset markers,calculate the Angle of view transformation model.According to the feature points after the field of view conversion model,the images from two perspectives are matched and the object is finally displayed in 3D.Experimental results show that the algorithm proposed in this paper can effectively recognize the blind area image,improve the recognition of the occluded object,and do not need to carry out multi-angle scanning of the object,and reduce the imaging time to a certain extent.