摘 要

近年来高光谱成像技术在农林产品检测中的应用不断增加，但多基于单一波段的成像观测。本文提出了一种双波段高光谱成像系统配准方法，该方法可以将不同波段的高光谱图像进行融合叠加，以获得目标物体表面每个像素在两个波段的光谱信息。首先采集了标定板和木块的双波段高光谱图像，然后对比选择了多项式、仿射变换、局部三角网等图像配准的校正算法，同时通过图像处理手段自动提取标定板的标记点坐标，提高了配准过程的自动化，即通过控制点自动采集来代替纯手动配准，最终获得了可见近红外到短波近红外融合的高光谱木块图像。本方法面向农林产品的检测，农林产品上人为的或天然的平面都可以作为配准的对象，并且对水平面和非水平面都是适用的。

关键词：高光谱图像；配准；短波近红外；可见近红外；图像融合

Dual-band hyperspectral imaging system registration method for planar targets

ABSTRACT

In recent years, the application of hyperspectral imaging technology in the detection of agricultural and forestry products has been increasing, but more based on single-band imaging observations. In this paper, a dual-band hyperspectral imaging system registration method is proposed, which can superimpose and fusion hyperspectral images of different bands to obtain spectral information of two pixels of each pixel on the surface of the target object. Firstly, the dual-band hyperspectral image of the calibration plate and the wooden block is collected, and then the correction algorithm of image registration such as polynomial, affine transformation and local triangulation is selected, and the coordinates of the calibration point of the calibration plate are automatically extracted by image processing means. The automation of the registration process is improved, that is, the automatic acquisition by the control point is used instead of the pure manual registration, and finally the high-spectral wood block image of the visible near-infrared to short-wave near-infrared fusion is obtained. The method is applicable to the detection of agricultural and forestry products, and artificial or natural planes on agricultural and forestry products can be used as registration objects, and are applicable to both horizontal and non-horizontal planes.

Keywords: hyper-spectral image; registration; short-wave near-infrared; visible near-infrared; image fusion

目 录

1. 绪论……………………………………………………………………………………………………1
   1. 研究背景…………………………………………………………………………………………1
   2. 国内外发展状况…………………………………………………………………………………2
      1. 高光谱成像技术………………………………………………………………………… 2
      2. 图像配准技术…………………………………………………………………………… 3
   3. 研究内容与技术路线……………………………………………………………………………4
2. 数据获取实验…………………………………………………………………………………………7
   1. 设备简介…………………………………………………………………………………………7
   2. 数据获取实验设计………………………………………………………………………………8
   3. 主要数据…………………………………………………………………………………………9
   4. 小结………………………………………………………………………………………………9
3. 配准处理………………………………………………………………………………………………11
   1. 配准流程…………………………………………………………………………………………11
   2. 软件选择…………………………………………………………………………………………12
   3. 配准处理与分析…………………………………………………………………………………13
   4. 配准结果评估……………………………………………………………………………………17
      1. 人工感官评价…………………………………………………………………………… 17
      2. 数值量化对比…………………………………………………………………………… 17
   5. 小结………………………………………………………………………………………………17
4. 配准改进………………………………………………………………………………………………18
   1. 需求分析…………………………………………………………………………………………18
   2. 自动配准控制点提取……………………………………………………………………………18
      1. 算法设计………………………………………………………………………………… 19
      2. 利用Matlab编程实现算法……………………………………………………………… 20
   3. 自动配准效果评估………………………………………………………………………………24
      1. 人工感官评估…………………………………………………………………………… 24
      2. 数值量化对比…………………………………………………………………………… 25
   4. 光谱衔接测试……………………………………………………………………………………27
   5. 小结………………………………………………………………………………………………30
5. 结论……………………………………………………………………………………………………31

致谢…………………………………………………………………………………………………………32

参考文献……………………………………………………………………………………………………33

附录A………………………………………………………………………………………………………35