Concealed Object Detection Of Terahertz Images Based On Image Fusion Of Multi-angle

Terahertz radiation has good penetrability for clothing,wrapping and other similar materials,and its photon energy is only 1/106 of X-ray which is widely used in security inspection at present.Therefore,terahertz radiation will not cause ionization damage to human body while used in human screening.So it is promising that terahertz radiation will replace X-ray or millimeter wave to detect whether human carries dangerous objects or not.(1)In view of the low signal-to-noise ratio and low resolution of terahertz security image,a gray-scale enhancement and non-local mean(NLM)filtering algorithm are introduced.Selecting a reasonable threshold of gray-scale enhancement can remove the artifacts in images,which can make the image denoising has a better result.NLM filter not only reduces the noise of terahertz image,but also preserves the details of dangerous objects in the image,so that the images will not be blured.By introducing an integral image,the neighborhood similarity comparison in the same direction does not need repeated calculation,which greatly reduces the complexity of the algorithm,and makes the running time of the NLM filtering algorithm decreased from the original 180 seconds to the 2.561 seconds,which means it speeds up nearly 70 seconds.Therefore,the NLM filter can be used in real-time terahertz security inspection.(2)Then,threshold segmentation combined with top-hat transform are used to enhance and highlight the dangerous objects in terahertz image by designing reasonable structural elements,and the edge of terahertz image can be extracted based on Canny operator.(3)After the above process,the terahertz image does not increase the effective information,so the idea of image registration is introduced.The image of better quality is selected as the reference image,and the input image is weakened.Through the above-mentioned image processing,the step size and the number of iteration of registration are controlled,and the registration effect is optimized.Chose and merge the proper terahertz images captured by the same sensor but in different angles,the original blurred part of the reference image is enhanced,and dangerous objects is clearer and more complete to be obtained through the edge extraction.After fusing the terahertz image which was originally carried the gun but lacked in the buttstock,it can be restored,and offset the most of the information of human torso,and achieved good detection results.