Research On Near Infrared Single Pixel Imaging System And Algorithm

With the rapid development of microelectronic integrated circuits in recent years,the processing technology of large-scale CCD or CMOS integrated sensor arrays has become increasingly mature and the cost is getting lower and lower.Thus,conventional image and video acquisition technology relying on the CCD or CMOS sensor arrays can acquire huge collection of raw data for imaging at low cost in visible band.However,due to the inherent sparsity of the signal,the huge raw data causes waste of the collection resources,storage resources and transmission resources.In addition,imaging systems based on large-scale CCD or CMOS integrated sensor arrays are very expensive in the near infrared range.The compressed sampling scheme facilitates the birth of a single-pixel camera by means of signal sparsity and recovery algorithms.A single-pixel camera encodes the reflected or transmitted light of the scene through an observation matrix on the spatial light modulator to obtain a series of compressed observations,and combines the recovery algorithm to achieve reconstructed image.Based on the compressed sensing sampling scheme,this paper designs a near-infrared single-pixel camera based on the off-axis Newtonian reflection telescope.Through Zemax modeling and parameter optimization,parameters of acquiring weak target near-infrared radiation are obtained.The observation matrix compresses and encodes the target scene to determine the quality and resolution of the image.In this paper,different kinds of observation matrices are designed to simulate the different resolutions of different scenes to obtain the optimal matrix.This paper explore three different recovery algorithms.Combined with the observation matrix for joint simulation,different recovery algorithms can be efficiently applied to static scenes or dynamic scenes respectively,achieving high time resolution and high spatial resolution reconstructed image in near infrared bands.To improve the accuracy and stability of the compression measurements,the high sampling rate and the flip rate of the spatial light modulator are traded off.In each observation period of the spatial light modulator,multiple sampling and equalization processing is performed,strongly suppressing the interference of background noise,and improving the imaging quality of the near-infrared single-pixel camera.Based on Zemax optimized optical system,improved Labview platform sampling scheme and preferred combination of observation matrix and recovery algorithm,the actual built near-infrared single-pixel camera achieves high quality reconstructed images in different resolutions of simple or complex scenes in visible and near-infrared bands.Based on the data collected by the actual platform,the reconstructed image quality is evaluated by a grayscale mean gradient standard without reference.Therefore,the off-axis Newtonian reflection telescope single-pixel imaging system based on the compressed sensing sampling scheme can achieve multi-band,high-quality target scene reconstruction under the condition of a small amount of effective target scene information values.This strategy provides a low-cost,simple and compact solution the camera for lack of cameras based on large-scale CCD or CMOS integrated sensor arrays in the infrared band.