海面双波段红外图像配准与融合方法研究

发布时间：2018-11-22 09:55

【摘要】：随着我国经济的迅猛发展,海洋新兴产业快速崛起,海上运输业和旅游业日益活跃。但是随之而来的海上遇险事故频频发生,导致人员伤亡和巨大的财产损失。海上遇险事故大多发生在雾天或大风浪等恶劣天气下,当救助直升机对遇险目标进行搜寻时,红外成像是最适用于全天候的海面遇险目标搜寻的实用系统。目前我国在救助直升机上的搜寻系统中,绝大多数还没有采用红外搜寻系统,只有极少数救助直升机上采用了单波段红外成像系统,但单波段的红外成像系统所采集的图像可用信息量受限,导致海面遇险目标的识别率受到影响。随着现代科学技术的发展,利用双波段红外成像系统进行数据采集,利用两个波段信息的互补性,可提高系统的目标检测能力和抗干扰能力。本文主要研究海面双波段红外图像配准与融合方法,结合红外成像技术自身特点,对同一海面场景的双波段红外图像进行融合处理,实现信息互补。图像配准是图像融合的关键步骤,未能精确配准的图像会造成融合后的图像出现模糊,但在本项目背景下由于海面遇险目标较小,海浪结构极为相似,特征点较少,常用的配准方法,如尺度不变特征转换(SIFT)算法都达不到很好的效果,本文结合硬件系统内中红外相机和远红外相机位置相互固定这一特点,提出了一种改进的基于仿射变换的图像配准方法。在完成精确配准后,本文采用了基于小波变换的图像融合方法。经过深入研究发现,随着海面目标场景温度的变化,在某些特定的时间段内,目标在中红外图像和远红外图像中的亮度相对于背景亮度会出现有的会亮,有的会暗,灰度值反转的现象,这时常用的加权平均法会降低目标的对比度。本文针对该现象提出了一种改进的低频分量融合规则,即利用局部峰值法检测疑似目标区域,通过判定疑似目标区域与背景的亮度关系来制定融合策略。本文用融合质量评价指标对融合后的图像进行了评价。实验结果表明,使用本文的方法融合后的图像效果较好,更有利于后续图像处理算法对海面遇险目标进行检测与识别。
[Abstract]:With the rapid development of China's economy and the rapid rise of marine emerging industries, marine transportation and tourism are becoming more and more active. However, the subsequent maritime distress accidents occurred frequently, resulting in casualties and huge property losses. Most marine distress accidents occur in severe weather such as fog or storm. When rescue helicopters search for distress targets, infrared imaging is the most suitable system for the all-weather sea distress target search. At present, most of the search systems on rescue helicopters in our country have not adopted infrared search systems. Only a very few rescue helicopters have adopted single-band infrared imaging systems. However, the available information of the images collected by the single band infrared imaging system is limited, which results in the influence of the recognition rate of the targets in distress on the sea surface. With the development of modern science and technology, the dual-band infrared imaging system is used to collect data, and the complementary information between the two bands can improve the ability of target detection and anti-jamming of the system. This paper mainly studies the registration and fusion method of sea surface dual-band infrared image. Combining the characteristics of infrared imaging technology, the dual-band infrared image of the same sea surface scene is fused to realize the complementary information. Image registration is the key step of image fusion. The image that can not be accurately registered will result in blurred image after fusion. But in the background of this project, because of the small target of sea distress, the wave structure is very similar, and the feature points are less. The common registration methods, such as scale invariant feature conversion (SIFT) algorithm, can not achieve good results. This paper combines the fixed position of mid-infrared camera and far-infrared camera in hardware system. An improved image registration method based on affine transformation is proposed. After accurate registration, the image fusion method based on wavelet transform is adopted in this paper. After in-depth study, it is found that, with the change of the temperature of the sea surface target scene, the brightness of the target in the mid-infrared image and far-infrared image will appear some light and some dark relative to the background brightness in some specific time periods. When the gray value is reversed, the weighted average method can reduce the contrast of the target. In this paper, an improved low frequency component fusion rule is proposed, that is, the local peak value method is used to detect the suspected target region, and the fusion strategy is formulated by judging the brightness relationship between the suspected target region and the background. In this paper, the fusion quality evaluation index is used to evaluate the fused image. The experimental results show that the fusion method proposed in this paper has a better effect and is more beneficial to the subsequent image processing algorithm for the detection and recognition of the target in distress on the sea surface.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】