Research On Computational Ghost Imaging Target Recognition Technology

Computational ghost imaging?CGI?is a new imaging method,which is different from traditional imaging and uses intensity fluctuation correlation between two light fields to reconstruct the image.CGI can break some limitations of traditional imaging such as the lens-less imaging,anti-turbulence imaging and anti-scattering imaging.Although CGI has many advantages,its imaging quality is susceptible to interference.In the process of imaging,insufficient sampling times and unknown the depth of object will lead to the degradation of image quality.Object identification technology is an important research field of artificial intelligence,and it plays an important role in many situations that are not suitable for manual operation and artificial vision.Aiming at the key problems in CCI,we carry out the research on object recognition technology in CGI in this paper.The main researches are as follows:?1?A CGI experimental system using speckle was constructed.In this system,the lookup table of spatial light modulator and the deep Fresnel region of speckle propagation are calibrated,and the virtual diffraction light field pattern is calculated using the light field propagation theory based on angular spectrum.?2?In order to identify the object with certain errors in depth measurement or estimation,a CGI object identification method combining SIFT?Scale Invariant Feature Transform?algorithm and support vector machine algorithm?SVM?was proposed.We demonstrate that a high identification rate above 85%can also be achieved when the defocus degree is in the range of|?|?5?denotes the axial coherence length in given system?.?3?Aiming at the identification of targets at completely unknown depths in deep Fresnel region,a object search algorithm in deep Fresnel region is proposed in this paper.The proposed algorithm uses the image quality evaluation function DBC to estimate the defocusing degree of the reconstructed image,and a depth search algorithm and the adaptive focusing window selection algorithm are designed to search for the depth position of the object.Experiments show that when the search range is divided into N depth position,the images and depth of object can be obtained after log₂iterative by the proposed approach.Finally,a method combining object depth search algorithm with CGI identification object method is presented.Experiments show that the method can identify the object after the depth of object is roughly searched.This paper applies object identification to the CGI system to improve the practicability and flexibility of CGI.