Research On Blind Pixel Compensation And Non-uniformity Correction Technology Of Sea Surface Infrared Imaging System

With the " the Belt and Road " strategy put forward,my country’s marine industry has developed rapidly,related maritime activities have continued to increase,and maritime traffic accidents have also occurred frequently.Infrared imaging system is widely used in search and rescue ships due to its long detection distance,strong anti-interference ability and all-weather characteristics.Affected by factors such as infrared detector materials and corrosion processes,the output response of each unit of the infrared detector has a large error,which will seriously affect the output image quality of the infrared imaging system.Therefore,it is necessary to perform blind pixel detection,blind pixel compensation and non-uniform correction on infrared images.In this thesis,the blackbody response characteristics,blind pixel noise characteristics and their generation mechanism are described in detail.Secondly,for the blind pixel detection method,the principle of the double-reference source detection method and the 3 σ detection method is deeply analyzed.This experiment can be analyzed on the time axis,which can eliminate the influence of random noise on the detection results and improve the accuracy of blind pixel detection.Finally,starting from the blind element compensation method based on single frame image,this thesis focuses on the linear interpolation method.On the basis of this method,an extended neighborhood compensation method is proposed to solve the problems of continuous blind pixel and blind pixel clusters,and the blind pixel compensation results of blackbody image and sea surface infrared image are analyzed in detail.Thesis expounds the characteristics of sea surface infrared images and the generation mechanism of non-uniform noise.The One-point,two-point,and multi-point non-uniform correction methods among the non-uniform correction methods based on blackbody calibration are studied.On the basis of these correction methods,a multi-point non-uniform correction method based on probability least squares is given.As a preliminary correction method,it is used to eliminate fixed pattern noise in non-uniform noise.This method uses multiple blackbody calibration temperature points for fitting,so that the output response characteristic curve can more accurately approximate the response of the detector in the linear range curve,which overcomes the contradiction between the dynamic range of temperature measurement and the correction accuracy.Although the preliminary correction method can remove a large amount of non-uniform noise,there will still be residual non-uniformity and streak noise and it can’t do anything about the random noise generated by the infrared imaging system when it is working.Aiming at the above problems,thesis studies the time-domain high-pass filter correction method and the neural network correction method in the nonuniform correction method based on multi-frame scenes.These two methods can automatically adjust the correction parameters with the change of the environment,and has good applicability.Since these two correction methods are based on the time domain,the calculation of continuous multi-frame video images will be very large,and the influence of factors such as the processing speed and storage capacity of the hardware system needs to be considered.Not suitable for real-time non-uniform correction of sea surface infrared images.In this thesis,a non-uniform correction method based on Gaussian high-pass filter variance compensation moment matching is presented as a fine correction method.After the initial correction,starting from the scene image based on a single frame,this method deeply studies the non-uniform correction method based on variance compensation moment matching.In this method,there will be serious shadow noise problems in the high-frequency area of the image after correction.Two-dimensional Gaussian high-pass filtering is used to extract the high-frequency sensitive area of the image to eliminate the influence of the high-frequency area on the non-uniform correction and improve the accuracy of the non-uniform correction.In order to effectively verify the method given in this thesis,several infrared image acquisition experiments were carried out for the laboratory blackbody environment and the actual sea surface environment,and the reliability was verified by using the real blackbody image and sea surface image.After subjective observation and objective index evaluation The experimental results show that the method given in this thesis can eliminate blind pixel noise and non-uniform noise well,improve the output quality of infrared images,and has important research significance in the field of engineering applications.