| Due to defects of manufacturing process and materials of the infrared focal plane array device in the infrared imaging system,the unprocessed original image will show obvious streaks or block noise,which seriously affects the quality of the infrared image.Therefore,non-uniformity correction is the key to obtaining high quality infrared images.In order to expand the working dynamic range of the infrared system,it is usually necessary to change the system integration time.Since the pixel response is related to the integration time and exhibits a certain non-uniformity,the change of the integration time will lead to the degradation of the image uniformity.In this thesis proposed a non-uniformity correction method based on calibration data.This method uses the neural network principle to construct a correction model that integrates the integration time term.The black body gray image and corresponding integration time are used as inputs,and the gray average value is used as the expected value.The trained correction network can effectively adapt to the non-uniformity caused by the change of integration time.It is found through experiments that the average non-uniformity after correction for blackbody images with different integration time and temperature has decreased by about 80.8% compared with the conventional TPC method.For the phenomenon that the output gray value changes with the integration time,based on the black body radiation theory and the detector response model,this thesis derives a correction model with the target radiation exitance of camera response as the output value to build a network,and still uses the black body gray image and the corresponding integration time as the input,calculate the corresponding exitance of black body radiation as the output expected value,and the trained correction network can maintain the stability of the output for the infrared image whose integration time changes.After verification,the same infrared scene image of 800,900 and 1000after corrected,the relative change of the output value is only 0.27% and 0.36%,which is far lower than the relative change of 5.1% and 8.4% after corrected by the gray-scale correction method,and still keep the non-uniformity correction effect of the image.For the response shift phenomenon caused by the long-term operation of the infrared system or the change in ambient temperature,an online modifying correction coefficients method is proposed.This method adds a uniform shading cover in front of the lens or adjusts the lens to align the uniform scene to the current frame image and integration time as input.The bias term is corrected online after about twenty iterations,the non-uniformity can be reduced by 87.9% compared with the output image before correction coefficients modified. |
