| Single-pixel imaging,as the name suggests,is a technique for imaging using a single-pixel detector.The history of single-pixel imaging can be traced back to mechanical point-scan imaging and the Hadamard transform imaging spectrometer popular in the 1960s and 1970s.Later,due to the advent of the focal plane array and its excellent performance in the visible light spectrum imaging,the single-pixel imaging technology gradually faded out of the mainstream imaging field.In the past two decades,inspired by quantum imaging and compressed sensing algorithms,single-pixel imaging technology has re-emerged and has become a research hotspot that has been widely concerned by scholars around the world.Today’s single-pixel imaging mainly refers to the use of a single-pixel detector without spatial resolution,combined with light field modulation technology or structured light coding technology,using iterative correlation algorithms,compressed sensing algorithms or deep learning algorithms reconstruct scene image.The single-pixel imaging system has a simple structure and mainly includes a single-pixel detection module and a structured light modulation module.The performance of these two directly determines the indicators of single-pixel imaging.With the rapid development of semiconductor materials and processing technology,the existing single-pixel detectors have excellent performance,generally possessing extremely strong weak light sensing capability,extremely high response speed,extremely large spectral sensing range and extremely low production cost.With these significant advantages,single-pixel imaging is expected to become an imaging technology that breaks through the barriers of traditional focal plane array cameras.So far,single-pixel imaging has been excavated and verified to have excellent performance in non-visible light and even non-optical wavelength range imaging,imaging in extreme environments,super-resolution imaging,3D imaging and other fields.However,as another element of single-pixel imaging,the structured light modulation module has not been able to match the performance of single-pixel detectors in all aspects,resulting in that the advantages of single-pixel imaging have not been fully utilized.The most prominent point is that the modulation speed of structured light is not fast enough,and due to the unique imaging mechanism of single-pixel imaging,the speed of single-pixel imaging is generally slow,which is not suitable for capturing dynamic scenes,which is not suitable for capturing dynamic scenes,greatly limiting the trend of single-pixel imaging to more widely applied scenes.In order to improve the single-pixel imaging speed and give full play to the high-bandwidth advantages of single-pixel detectors,this paper starts with the analysis of the single-pixel imaging mechanism,and proposes innovative solutions in structured light encoding forms and structured light encoding devices,which not only has realized the single pixel image capture of high-speed motion target,and make the single pixel imaging speed reached the ultrafast imaging level.In addition,this thesis combines the single pixel imaging model with the traditional camera imaging technology,and proposes a single exposure compression imaging scheme to improve the frame rate of ordinary cameras.In general,the main research contents and innovations of this thesis are as follows:1.Benefiting from the full understanding of the single-pixel imaging mechanism,a singlepixel imaging scheme for high-speed moving targets is proposed,which breaks.the previous encoding form of single-pixel imaging.When imaging a high-speed moving target,the coding pattern remains stationary,and relying on the displacement property of the moving target allows it to autonomously complete a series of spatial codes necessary for single-pixel imaging on the stationary pattern.This scheme not only makes single-pixel imaging get rid of the need for dynamic encoding devices,but also the encoding speed is completely determined by the target moving speed.Therefore,as long as the single-pixel detection rate meets the requirements of Nyquist sampling law,the moving target image can be reconstructed.We demonstrate the feasibility of this scheme through physical experiments,and compare and analyze the reconstruction effect using the compressive sensing algorithm at different sampling rates.In order to further improve the scheme,a set of convolutional neural network model was built independently for reconstruction at low sampling rate,and finally high-quality imaging at 5%sampling rate was achieved.2.A new mask is designed and fabricated as a structured light encoding device in single pixel imaging.The outer ring of the mask is printed with a cleverly designed coding pattern,and a complete single-pixel imaging coding can be completed by driving the mask to rotate once.Using the mask as a structured light encoding device for single-pixel imaging has two significant advantages.First,the motor used to drive the mask usually has a very high speed,so it can achieve a high imaging speed.Secondly,the spectral response range of the mask is much larger than that of the commonly used spatial light modulator,so it can be used for wide spectral imaging.Through three experiments,we prove that the mask can not only achieve high quality single-pixel imaging of static targets,but also achieve simultaneous imaging in ultraviolet,visible and infrared bands at 100fps imaging speed.3.A time-resolved single-pixel imaging scheme is proposed and verified.In this scheme,time dimension is introduced into the signal acquisition process of each coded pattern for periodic dynamic scenes.After a group of coding is completed,all signals are grouped according to the relative time in the cycle,and then images representing different moments of the scene are reconstructed successively.In this scheme,the time resolution of the single-pixel imaging system depends only on the working bandwidth of the single-pixel detector.In the verification experiment,we first completed time-resolved 2D imaging using a single pixel detector in conventional linear mode,and then realized time-resolved 3D imaging by combining Fourier transform contouring.The imaging speed of 2D and 3D is up to 2,000,000fps.We then used a single-photon detector operating in geiger mode to image the flying laser pulse at a speed of 1012fps.More importantly,in the laser pulse imaging results,we not onUy clearly observed the flight attitude of the laser pulse,but also found the phenomenon of "temporal Airy disk" in ultrafast imaging.4.Based on the principle of computational temporal ghost imaging,a snapshot compressive imaging scheme is proposed.By introducing a spatial light modulator into the traditional camera imaging structure,the scene during the exposure period of the camera is divided into several sub-periods with a high-speed refresh pattern and spatially encodes them respectively.The image recorded by the camera is the encoded scene of these sub-periods.accumulation.During reconstruction,the single image recorded by the camera and the coding matrix used during exposure are used to reconstruct the time signals of all spatial pixel positions according to the computational temporal ghost imaging algorithm,and then these time signals are combined to form several images at different times.image.This solution can effectively improve the frame rate and time resolution capability of the camera,and the improved frame rate multiples far exceed the snapshot compressive imaging technology based on compressed sensing.We first verified the feasibility of the scheme through simulation.Then,in order to improve the conversion efficiency from spatial resolution to temporal resolution,we proposed three encoding and reconstruction strategies,and verified them and analyzed the imaging results. |
