The Study Of Pushbroom Pixel-by-Pixel Multi-Gain Imaging Spectroscopy

Imaging spectroscopy is an advanced optical remote sensing technology for the earth,which is mainly reflected in the ability to obtain the spatial and spectral information of ground objects at the same time.It has important applications in fine land classification,water quality inversion,mineral exploration,environmental protection and so on.Pushbroom is the mainstream imaging mode of the current spaceborne imaging spectrometer,which has the characteristics of no moving parts,instantaneous wide coverage and high signal-to-noise ratio.Due to the limitation of large dynamic range and high signal-to-noise ratio,the current load design can only meet the monitoring needs of a single field such as ocean,resources and environment or atmosphere.The application of advanced microelectronics technology will break through this limitation,so that the designed system has the advantages of high performance and diversified application scenarios,and can well meet the research needs of multi factor and multi process comprehensive earth observation.To meet this requirement,this paper presents a pixel-by-pixel multi-gain self-switching spectral imaging solution,and studies the key technologies,system design optimization and comprehensive testing.The main research work and innovations of this paper are as follows:1)The contradiction between dynamic range and signal-to-noise ratio is analyzed from the engineering principle,and a feasible solution to the limitation of pushbroom imaging mode is proposed,which is pushbroom pixel-by-pixel multi-gain self-switching spectral imaging.Based on the 3-level TG technology of the custom CMOS detector,the optimal gain selection at the pixel level is achieved by one exposure and multiple readouts,taking advantage of the non-destructive photocharge reading.Downloaded as“2 bit gain code+14 bit DN value”,the ground deduces the calibration factor from the gain code to obtain the true pupil radiance.The system can switch the pixel level of four gains to meet the high performance requirements of comprehensive remote sensing monitoring.At present,no such satellite imaging spectrometer and related applications have been reported at home and abroad,and the system scheme has some originality.2)Based on MODTRAN atmospheric radiation transfer model and spectral data of multi types of ground objects such as water,vegetation and concrete,the pupil radiance simulation under different conditions was carried out,and the effects of solar altitude angle,water vapor content and other factors were analyzed.According to the parameters of the imaging system,the corresponding signal electrons for different radiance are calculated,which can provide reference for the previous system design.The results show that the upper limit of the scene dynamic range is about 2.5 Me~-.From the monitoring type,the dynamic range can be roughly set to four blocks,namely,cumulus and snow(7.62×10~4～2.01×10~6e~-),concrete and sand(2.13×10~4～8.03×10~5e~-),vegetation(1.68×10~4～2.92×10~5e~-),beach and water(1.0×10~4～2.3×10~5e~-).In addition,the signal intensity is most affected by the solar zenith angle.The changes of water vapor content and visibility increase the dynamic range of weak signals such as water.Pixel-by-pixel multi-gain imaging can well deal with the impact of these imaging conditions.3)Constraints for gain number and scale design are constructed.The effects of full well charge design on SNR and dynamic range of the system are analyzed from the perspectives of overall dynamic range,load SNR index,influence of gain switching on SNR,and engineering difficulty.After simulation,it can be seen that for the same dynamic range,multi gain imaging SNR is better than single gain,and the change is not obvious when the number of gains exceeds 3～4.The SNR gain is greatly affected by the change of ground object type.When the number of gains is 2～4 and the ratio is2～5.5,the design meets the imaging requirements.It is determined that the full-well charge of the four gains is 2.5 Me~-(ultra low gain),600 Ke~-(low gain),120 Ke~-(medium gain),and 24 Ke~-(high gain),respectively.4)Starting from the pixel topology and readout link,the specificity of multi-gain imaging is discussed,and the influence of multi level ping-pong reading and saturated DN value judgment on image noise is analyzed.A Poisson-gaussian noise model in spectral mode and channel mode with variable multi-gain is established by analyzing the type of time-domain noise.Based on this model,the probability that pixels will be read out with low gain due to noise is calculated.The sources and effects of system spatial noise are analyzed from optical,electronic and other perspectives.The influence of noise introduced by manufacturing process errors into readout links is studied.Effective noise suppression methods such as adjusting parity line sampling sequence and AB frame distinguishing readout are presented.Finally,the four gains pushbroom spectral imaging was simulated using MWI data as the input pupil radiance.The intrinsic characteristics of the multi-gain spectral image and the effect of noise on the target texture characteristics were analyzed.5)Based on the engineering principle prototype of pushbroom progressive multi-gain imaging spectrometer,the performance of dark field noise,pixel conversion gain,dynamic range,signal-to-noise ratio and other system parameters are tested.The effects of PGA and multi line merging on signal-to-noise ratio,the effect of field of view position on spectral resolution,and the small dynamic range of a single gain are analyzed.The experimental results show that the total dynamic range of the system is107.9d B,the signal-to-noise ratio under typical radiance of each band is higher than the index requirements,the minimum spectral sampling step is better than 2nm,and the overall performance is superior.For the fixed gain pushbroom image,the non-uniformity correction effects of two-point method,moment matching and other methods are compared and analyzed.Considering that the laboratory absolute calibration can convert the DN value into radiance and eliminate the influence of spatial non-uniformity response on multi-gain image fusion,multi-gain fusion and non-uniformity correction are selected by pixel-by-pixel calibration coefficient fitting.