| The space infrared remote sensing imaging system has the advantages of good concealment and around-the-clock detection,and is widely used in the fields of ocean,environment,resources,defense,military and so on.The space infrared earth observation video camera can be in the form of "gazing at" imaging for continuous monitoring of specific situations or specific target,it can get the dynamic information in the scene in ways similar to the "video".Due to the complex environment,platform jitter and attitude adjustment,the video camera obtains the infrared video sequence contains jitter,fuzzy,even seriously affected the scene monitoring and dynamic information retrieval in the process of imaging.The space stabilization system can effectively suppress the random vibration of the imaging system and guarantee the detector outputs stable video sequence.Electronic stability is one of the main methods of stabilizing image technology.In this paper,from the point of electronic image stabilization of infrared video camera,firstly,the design of the hardware electronics is introduced,which is mainly the design of low noise imaging system;Secondly,the infrared image preprocessing is studied,including the infrared image dynamic noise and non-uniformity noise suppression and the infrared image super-resolution reconstruction.Then,the real-time electronic image stabilization algorithm for infrared video camera based on image local feature is studied,which is mainly used as second image stabilization system in the observe video camera image stabilization system;Meanwhile,the tracking of dynamic information or specific targets in the scene is also one of the main tasks of the infrared video camera,at last,the infrared vision target tracking algorithm based on kernel correlation filtering is studied.The main research contents and innovation points of this thesis are as follows:Firstly,The system composition and image stabilization method of space infrared video camera are studied.And then the electronic hardware design of infrared imaging system is studied.Mainly studied the low noise design of the circuit about the imaging system.It is mainly including power supply module,analog signal conditioning module,analog-digital conversion,control circuit based on FPGA.After the test,thre circuit noise is about 0.1 8)(1,the imaging system noise is about 0.38)(1.Secondly,study on dynamic noise and non-uniformity noise suppression of infrared images.Infrared image usually contains more noise,and infrared image noise suppression is the primary task of other more advanced tasks such as electronic image stabilization,target detection and target tracking.Therefore,the traditional image denoising and scene based nonuniformity correction method are studied,and the improved non-local mean filtering algorithm is proposed.The improved algorithm is superior to the original algorithm whether it is subjective visual effect or MSE,PSNR and SSIM three objective evaluation indexes.And puts forward the improved neural network based on infrared image nonuniformity correction algorithm,the improved algorithm not only can effectively inhibit the nonuniformity noise,also can effectively restrain the generation of "ghost" compared with the original algorithm.Thirdly,aiming at the features of low resolution,vague details and low signal-tonoise ratio of infrared image,the infrared image super-resolution reconstruction algorithm based on convolution neural network is studied.To improve the traditional SRCNN algorithm,a new network structure is proposed based on SRCNN network structure,and the activation function ReLU is replaced by RReLU.Compared to the original algorithm,the improved algorithm can obtain a higher PSNR and better image quality after the image reconstruction.Fourth,according to the infrared video camera stabilization system requirements,an electronic image stabilization algorithm based on image local characteristics is studied.In the order to satisfy the stability and speed of the system at the same time,the precision of the stability image is sacrificed to meet the steady image speed,an optimized and improved real-time electronic image stabilization algorithm based on SURF and ORB features is proposed.It is mainly based on two aspects: one is to reduce the number of pyramid layers in the extraction of SURF feature points.The reduction of the number of layers leads to the robustness of the extracted feature points,although it reduces the stability image accuracy but improves the steady image speed;Secondly,the nearest neighbor search FLANN is used to replace Hamming distance matching,which not only improves the matching speed,but also improves the matching precision,thus improving the stability image accuracy.The improved algorithm is aimed at 288 x 384 surface array detector imaging system,the image stabilization accuracy is 0.1978 pixel,and the single frame time is 0.0311 seconds.The improved electronic image stabilization algorithm has certain reference significance for the following video camera stabilization system engineering.Fifth,aiming at the demand of infrared video camera for specific target tracking in the scene,the infrared vision target tracking algorithm based on nuclear correlation filtering is studied.In view of the traditional nuclear related filter tracking algorithm robustness is poor,do not have scale adaptive and the poor ability to resist shade,the improved nuclear related filtering algorithm is proposed for infrared visual target tracking,It mainly improves the traditional KCF algorithm in two aspects: First,the idea of target tracking algorithm proposed by Danellian is adopted to add a scale filter in KCF,which is specially used to predict target scale changes.Second,the idea of longterm target tracking algorithm proposed by Z.Hong was used for reference to add an occlusion judgment mechanism based on image feature point matching in the KCF algorithm.The improved algorithm is not only adaptive to the infrared vision target tracking,but also can effectively deal with the occlusion of the target and realize the target long time tracking.The optimize and improved algorithm can be used for reference in the follow-up infrared video camera tracking scene. |
PDF Full Text Request