Research On Non-uniformity Correction Technology Of Aerospace CMOS Camera

Aerospace optical remote sensing imaging is one of the effective means to quickly and accurately obtain earth observation data,and it is also a research hotspot in recent years.Space remote sensing imaging system with high resolution,wide range,multifunction,small volume,light weight and low power consumption has become the development trend.CMOS image sensors are widely used in aerospace optical remote sensing cameras for their advantages of simple drive,high integration,low power consumption,and high frame rate.However,the non-uniform noise of space CMOS camera causes poor image quality,low signal-to-noise ratio and lack of information.Therefore,in this paper,in view of the non-uniformity problem of aerospace CMOS cameras,systematic research work has been carried out in terms of noise source,response nonlinearity,correction parameter storage size,correction parameter update,etc.The details are as follows:1.The causes and characteristics of non-uniformity noise in space CMOS camera are analyzed,the correction model of non-uniformity noise is established,and the subjective and objective evaluation criteria of non-uniformity correction effect are defined.2.Starting from practical engineering applications,based on the calibration method,a CMOS image non-uniformity correction method based on an adaptive moving window is proposed.The threshold range is estimated by selecting windows of different widths,and the image column variance is compared with the threshold to complete the image block processing,and then perform non-uniformity correction according to the image block.The method can effectively correct the non-uniformity noise caused by the nonlinear response of CMOS image sensor.At the same time,multiple pixels share a set of correction parameters,and introduce the second-order fitting response curve of least square method,which greatly reduces the amount of correction parameters,and saves the storage space of the non-uniformity correction parameters.3.On the basis of scene method,a neural network non-uniformity correction method based on the regional attributes of neighboring pixels is proposed.In this method,the correction parameters are updated iteratively by the real scene information,the initial value of iteration is obtained by the calibration method,and the iteration step is adjusted adaptively according to the gradient information.The problems such as the correction parameters cannot be updated,the pixel correction parameters in the same column are the same,and the iteration time is long are solved,so as to improve the accuracy of non-uniformity correction.Finally,from the peak signal-to-noise ratio,flatness and root mean square error and other objective indicators,after being corrected by the adaptive moving window method,its most advantageous non-uniformity is reduced to less than 2%,and the amount of correction parameters is reduced by 62.8% compared with column-by-twopoint correction.The peak signal-to-noise ratio is 41.37 d B,and the root mean square error is 0.6835,flatness is 0.0606.After correction by the regional attribute method,the peak signal-to-noise ratio is 47.71 d B,the root mean square error is 0.1835,and the flatness is 0.0111.The proposed method has achieved good correction effect both from subjective observation of simulation image and experimental data verification,and can play a better role in suppressing non-uniform fringe noise.