红外双波段图像融合系统关键技术研究

发布时间：2018-03-28 00:10

本文选题：图像增强　切入点：图像融合　出处：《中国科学院大学(中国科学院上海技术物理研究所)》2017年博士论文

【摘要】：经过多年的努力,红外成像探测系统已发展到第三代,即红外双/多波段图像融合系统。其核心在于利用景物在不同波段的响应特性,融合后使场景能有更加客观完整的信息描述,提高系统性能。本文针对红外双波段图像融合系统开展了相关关键技术研究。主要研究内容包括以下几个方面:构建了一个实时红外双波段融合信号处理平台。其主要分为三个子系统:高速数据传输系统、多核DSP实时信号处理系统及红外视频实时显示系统。高速数据传输系统运用RocketIO技术,提高了数据传输的带宽,实现了系统中数据高速可靠的传输。多核DSP实时信号处理系统采用多核DSP加FPGA的架构,提高了信号实时处理的能力。红外视频实时显示系统基于单板实现,通过视频接口可实时显示高清分辨率的红外图像。提出了一种基于自适应引导滤波的子带分解多尺度retinex红外图像增强算法。该方法将自适应引导滤波与子带分解多尺度retinex算法相结合,利用自适应引导滤波更加准确地估计了光照分量。实验证明,该算法能有效消除光晕现象及增强图像细节。提出了一种基于视觉显著性及NSCT的红外图像融合算法。首先对待融合图像进行NSCT分解,获取多尺度多方向的子带图像。然后利用视觉显著性检测结果指导各图像子带的融合,最后对融合后的子带图像进行NSCT反变换得到融合后图像。实验表明,该算法可以更好地保留目标区域信息,突出目标特征,融合后的图像较传统方法图像融合评价指标更好,纹理细节更加清晰。提出了一种基于背景抑制及图像融合的红外双波段小目标检测算法。首先分别对各个波段图像进行背景抑制,接着利用形态学膨胀降低图像配准误差的影响,最后对双波段图像进行叠加。实验证明,该算法可以有效提高小目标提取能力。
[Abstract]:After years of efforts, infrared imaging detection system has developed to the third generation, that is, infrared dual-band / multi-band image fusion system. The core of the system is to utilize the response characteristics of the scene in different bands. After fusion, the scene can be described with more objective and complete information. In order to improve the performance of the system, the key technologies of infrared dual-band image fusion system are studied in this paper. The main research contents are as follows: a real-time dual-band infrared image fusion signal processing platform is constructed. It is mainly divided into three subsystems: high speed data transmission system, Multi-core DSP real-time signal processing system and infrared video real-time display system. High-speed data transmission system uses RocketIO technology to improve the bandwidth of data transmission. The multi-core DSP real-time signal processing system adopts the structure of multi-core DSP and FPGA to improve the ability of real-time signal processing. The infrared video real-time display system is realized based on single board. Infrared images with high resolution can be displayed in real time by video interface. A sub-band decomposition multi-scale retinex infrared image enhancement algorithm based on adaptive bootstrapping filter is proposed. This method combines adaptive bootstrapping filtering with sub-band decomposition. Multi-scale retinex algorithm combined, The illumination component is estimated more accurately by using adaptive bootstrapping filter. This algorithm can effectively eliminate halo phenomenon and enhance image details. An infrared image fusion algorithm based on visual salience and NSCT is proposed. Firstly, the fused image is decomposed by NSCT. The multi-scale and multi-directional sub-band images are obtained, and the fusion of each sub-band is guided by visual salience detection results. Finally, the fused sub-band images are obtained by NSCT inverse transformation. The experimental results show that, The algorithm can better retain the information of the target area and highlight the target features. The fused image is better than the traditional image fusion evaluation index. The texture details are clearer. An infrared dual-band small target detection algorithm based on background suppression and image fusion is proposed. Then the effect of image registration error is reduced by morphological expansion, and finally the dual-band image is superposed. Experiments show that the algorithm can effectively improve the ability of small target extraction.
【学位授予单位】：中国科学院大学(中国科学院上海技术物理研究所)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】