基于场景的红外焦平面非均匀性校正算法及FPGA实现

发布时间：2018-01-24 01:00

本文关键词： 红外焦平面阵列非均匀性校正 FPGA 鬼影盲元处理图像增强　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：近些年,红外热成像技术发展速度迅猛,红外焦平面阵列作为现今主流红外成像系统的核心器件,其品质的优劣直接影响红外图像成像质量。因生产原料、加工工艺等问题红外焦平面阵列存在着固有的非均匀性,需通过信号处理算法来解决。国内外对焦平面非均匀性校正算法做了大量的研究工作,基于场景的非均匀性校正算法因其自身优势已然成为近年来的研究热点。当前基于场景的校正算法硬件实现主要面临两个难题:算法复杂性大,硬件实现受制约;易产生鬼影。本文主要针对以上两点对基于场景的非均匀性校正算法进行了研究,并实现了一种基于FPGA硬件平台的红外焦平面非均匀性校正系统。本论文设计了基于场景的红外焦平面非均匀性校正算法的红外成像硬件系统,包括主处理芯片FPGA的选型、红外焦平面阵列的硬件和软件驱动、AD和DA转换模块的设计。研究并实现了红外成像系统必备的盲元处理和图像增强模块,盲元处理模块采用双参考辐射源检测算法和邻域替代算法,去除了大部分的盲元点噪声,有效提高图像成像质量。图像增强模块选用平台直方图均衡化算法,大幅度提高图像对比度,增强图像可视性。分析了基于场景的经典神经网络算法及不足,提出改进型基于场景的非均匀性校正算法。并从图像清晰度、非均匀性以及鬼影的抑制等多个角度,对比分析了经典神经网络校正算法和本文改进型校正算法的优劣。在FPGA平台上实现了改进型基于场景的非均匀性校正算法移植,校正算法模块包括采样矩阵窗口模块、校正参数实时读写模块和校正算法模块等部分。经测试,该算法在文中搭建的红外焦平面阵列成像系统的硬件平台上实现了对实时红外视频的校正,能够快速去除红外图像中的非均匀性噪声,大幅改善红外图像成像效果,同时有效抑制鬼影的产生。
[Abstract]:In recent years, infrared thermal imaging technology is developing rapidly. Infrared focal plane array (IRFPA), as the core device of mainstream infrared imaging system, its quality directly affects the imaging quality of infrared image. There is inherent inhomogeneity in infrared focal plane array (IRFPA), which needs to be solved by signal processing algorithm. A lot of research work has been done on focal plane nonuniformity correction algorithm at home and abroad. The scene based nonuniformity correction algorithm has become a research hotspot in recent years because of its own advantages. At present, the hardware implementation of scene based correction algorithm mainly faces two problems: the complexity of the algorithm. Hardware implementation is restricted; It is easy to produce ghost image. In this paper, the scene based nonuniformity correction algorithm is mainly studied for the above two points. An infrared focal plane nonuniformity correction system based on FPGA hardware platform is implemented. This paper designs an infrared imaging hardware system based on scene based infrared focal plane nonuniformity correction algorithm. Including the main processing chip FPGA selection, infrared focal plane array hardware and software driver. The design of AD and DA conversion module. The blind element processing and image enhancement module of infrared imaging system are studied and implemented. The blind element processing module adopts dual-reference emitter detection algorithm and neighborhood substitution algorithm. Most of the blind point noise is removed and the image quality is improved effectively. The image enhancement module uses the platform histogram equalization algorithm to greatly improve the image contrast. Enhance the visibility of images. Analyze the classical neural network algorithm based on scene and its shortcomings, and propose an improved non-uniformity correction algorithm based on scene. Inhomogeneity and the suppression of ghosts and other angles. The advantages and disadvantages of the classical neural network correction algorithm and the improved correction algorithm in this paper are compared and analyzed. The improved scene based nonuniformity correction algorithm is transplanted on the FPGA platform. The correction algorithm module includes sampling matrix window module, correction parameter real-time reading and writing module, correction algorithm module and so on. The algorithm realizes the correction of real-time infrared video on the hardware platform of the infrared focal plane array imaging system in this paper, and it can quickly remove the non-uniform noise in the infrared image. Greatly improve the imaging effect of infrared images, and effectively suppress the generation of ghost images.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】