Infrared imaging guidance is the the major trends of guidance in weapons development. Taking weaponry preliminary research fund projects as foundation, the paper makes in-depth analysis and research in nonuniformity correction of infrared focal plane array (IRFPA),which is one of important issues of the infrared image preprocessing.The main work and features are as follows:1, In order to make nonuniformity correction be dead against, the paper starts from the composition of the infrared thermal imaging systems.And then the causes of non-uniformity of infrared focal plane and the main characteristics of infrared images is introduced. In the latter part of Chapter 2, the response characteristic curve of the detector cells is generated in accordance with the measured data, which makes preparation for the next work.2, On the basis of serious research on the domestic and international developments in this regard, two kind algorithms are analysised .One is calibration based correction, including one point NUC algorithm, two points NUC algorithm and the "S" curve correction algorithm;One is scene based NUC algorithm, including NUC algorithm based on Kalman filtering and the NUC algorithm based on BP neural network.3, Based on the work above,two integrated algorithms are put forward.One combines one point NUC and wavelet transform algorithm, which increases the dynamic range of temperature of the algorithm applies; One combines two point NUC and the "S" curve based correction algorithm, which not only increases dynamic range but also improves the calibration accuracy of the algorithm. At the same time, the amount of computation is greatly reduced.4, In order to meet the needs of project practice, the improved multi-point NUC algorithm, Kalman filtering algorithm and neural network algorithm based on on the classification of the region are introduced. Compared with the traditional multi-point NUC ,the improved multi-point NUC reduce the need for storage of data significantly ;Kalman filtering algorithm meet the needs of real-time correction through the off-line calculation of correction parameters; neural network algorithm based on on the classification of the region reduced the fuzzy on the target and background that is made by traditional neural network algorithm. This algorithm especially can be used in detection, tracking and identification of small target.5, Making the process of the non-uniformity correction complete, in Chapter 4 of this article the blind-pixel compensation algorithm is introduced. Whether the blind pixel is the edge of goal or not , deal the different situations in different ways. A motion analysis based blind-pixel compensation algorithm is introduced. The compensation algorithm has high accuracy and yields a good result.All of the algorithm involved in the paper have been simulated by Matlab . And all have obtained experimental results. The conclusion of the paper has been proofed powfully through the large number of experimental data. |