Research On Nonuniformity Correction Technique Based On Domestic Infrared Focal Plane Array

Infrared focal plane array (IRFPA) imaging system is one of the most rapidly developing IR technology. It is widely used in different fields for its advantages. Due to western countries’ blocking, our country’s research progress in this field is relatively slow. Nowadays, we have been able to develop the IRFPA detector independently, but due to the limitations of manufacturing process, material level and other technical conditions, there is a big gap between our country and foreign country in the performance of the same IRFPA products. Especially, the nonuniformity of our detectors’ response is very seriously.In this paper, we take GWIR 03 02 X1A IRFPA detector as a research object of the domestic detector. Studying the performance and imaging characteristics of this detector, introducing both calibration-based nonuniformity correction algorithms and scene-based nonuniformity correction algorithms, which are widely used in thermal imaging system. And then, we proposed improved algorithm for both algorithms in this paper.For calibration-based non-uniformity correction, to the problems that the ROM-based transmission system of On-Chip Calibration(OCC) data will consume logical resources and the data numbers dependent on the original test,we proposed a new adaptive OCC transmission technology based on FLASH+SDRAM framework. This technology can not only improves the transmission rate of OCC datas, but also adaptively get the OCC datas from the optimal output of the detector. To the problem of blind visual effect in traditional calibration-based non-uniformity correction, we proposed a new calibration-based non-uniformity correction with image compensation. This method doesn’t need periodic block correction, and can guarantee the image quality for a long time without degradation.For the scene-based non-uniformity correction algorithm, there exist three problems:the model is complex; scene priori requirements are high; ghosting. To solve these problems, we proposed a new non-uniformity model based on full variation theory, and use Split Bregman method to minimize the equation iteratively. This model can correction nonuniformity based on single frame with the advantages of low computational complexity and wide scope of application. We also design a GPU platform based on CUDA+VS2010, which make the algorithm be applied to infrared video stream in real time.