Uncertainty Measurement And Application Of LSF FWHM For Satellite Cameras

It is very important to measure Point Spread Function(PSF) and Modulation Transfer Function(MTF) of satellite cameras, as they are of great significance in assessment of imaging system performance and restoration of remote sensing images. Considering that satellite cameras work in an obvious different environment in space compared to the static statement in lab, on-orbit PSF and MTF cannot be replaced by static PSF and MTF measured in lab. In other words, it is necessary to extract PSF and MTF from actual remote sensing images. Point method, periodic-targets method, pulse method and knife-edge method are four typical ways to do that. Among these methods, knife-edge method having relative high accuracy for response of high resolution satellite cameras and less complicated requires for targets than point method, becomes the most common way to gain on-orbit PSF and MTF inland and abroad. However, in order to measure PSF and MTF precisely, artificial targets with specific demands are needed. Thus the monitoring of imaging performance of satellite cameras become not a continuous result but a discrete one limited by the time when the targets are set and the location where the targets are set. Even though the discrete result is widely used in performance monitoring and image restoration, continuous one is also in desperate need. Therefor natural targets such as parking lots、long straight highroad and great bridge across water are usually made use of, with relative low accuracy compared to artificial targets. But it is an effective way to make continuous assessment for PSF and MTF, which is meaningful for long term on-orbit performance of satellite cameras and real-time image restoration.Sub-pixel edge detection and edge spread function extraction are two key steps of knife-edge method. Researches of both two steps are made since a long time. Sub-pixel edge detection using Fermi Function is a consistent result of researches, however, there is no unified conclusion about edge spread function extraction. An self-adaptive method is developed on the basis of modified SG method. Simulation experiments are conducted to verify its effectiveness. Actual remote sensing images are also used to analyze the advantages and disadvantages between various ESF extraction methods. Spline method and spline-SG method have almost the same results, a little better than MSG and SASG method. On the other hand, SASG has better performance when stripe noise exists, and has simpler algorithm than spline method. Thus, SASG method has its effectiveness and practicability in knife-edge method for PSF and MTF measuring.After the research of key steps of knife-edge method, image factors which affect the veracity of knife-edge method are concerned. Artificial simulation targets are used to find out the limitations for knife-edge method, including knife-edge angle、knife-edge size and SNR, et al. Analyses of these factors aim at getting exact quantitative ranges for each factor, in which knife-edge method can make precise calculations for PSF and MTF. Use FWHM(Full Width at Half Maximum) of PSF and MTF value at normalized Nyquist frequency(NyquistMTF) as two characteristic parameters of PSF and MTF. Simulations show that if the target, either natural or artificial, has an angel between 6°~10°, calculated FWHM will have ±1% accuracy and uncertainty below 2%, while calculated MTF will have ±5% accuracy and uncertainty also below 2%. Taking both accuracy and uncertainty into account, the best knife-edge angle will be 8°. The valid size of knife-edge image changes as the angle changes. When the target has an angle of 8°in the center of the image to be calculated, at least 20 pixels are needed in width and 15 pixels in height. To ensure the accuracy, SNR of the image should at least be 40 dB. Researches also show that Nyquist MTF is very sensitive depending on the change of FWHM, and thus the performance of satellite cameras can be analyzed only by FWHM when precision are not rigidly needed.Besides, because of the large scroll angle of satellites ranging from 0°to 40°, GSD(Ground Sampling Distance) will sure be affected. For the knife-edge method, the change of GSD will have an effect on sampling of the gray level distribution of the edge, which also affects the accuracy of the method. Simulation experiments are conducted to explore the law how the results change as the GSD changes along with the scroll angle. The study shows that the description will be less particular when GSD largens, while image size decreases, FWHM minishes and MTF curve increases. However, image restoration using PSFs measured at large scroll angle has poor quality. PSFs measured at small scroll angle will bring better restoration quality.