高光谱图像配准模块开发毕业论文

 2021-04-07 12:04

摘 要

本文实现了两套高光谱成像系统所成图像之间的配准,对影响高光谱图像配准后效果因素进行了多组实验,并开发了一套能够实现高光谱图像自动配准的模块。首先利用可见近红外高光谱相机和短波近红外高光谱相机以外部推扫的形式采集了多组标定板的图像,在手动配准中研究了选点数目、拟合函数、重采样方法对配准精度的影响,经过比较,标记点的数目会直接影响配准的准确度,使用多项式拟合函数时的配准效果比较好,三种不同的重采样方法对本次配准实验的效果并无显著性差异,还有精度评价中发现配准的精度误差能够控制在0-1.3毫米之间,并且通过Matlab程序的图像处理、阈值分割、腐蚀、标点与排序,利用仿射变换矩阵对图像进行配准,完成了高光谱图像的自动配准模块的开发,测试显示能够满足高光谱图像配准的要求。

关键词:高光谱图像配准;模块开发;重采样方法

Hyperspectral image registration module development

ABSTRACT

In this paper, the registration between the images of two sets of hyperspectral imaging systems is realized. Several sets of experiments are carried out on the factors affecting the registration of hyperspectral images, and a module capable of realizing automatic registration of hyperspectral images is developed. Firstly, the images of multiple sets of calibration plates were collected by external near-infrared hyperspectral camera and short-wave near-infrared hyperspectral camera. The number of selected points, fitting function and resampling method were studied in manual registration. The effect of quasi-precision, after comparison, the number of marked points will directly affect the accuracy of the registration. The registration effect is better when using the polynomial fitting function, and the effects of three different resampling methods on the registration experiment are There is no significant difference, and the accuracy error of the registration found in the accuracy evaluation can be controlled between 0-1.3 mm, and the image is processed by Matlab program, threshold segmentation, corrosion, punctuation and sorting, and the image is imaged by affine transformation matrix. Registration was carried out to complete the development of an automatic registration module for hyperspectral images, which was shown to meet the requirements for hyperspectral image registration.

Key words:Hyperspectral image registration;Module development;Resampling method

目 录

1 绪论………………………………………………………………………………………1

1.1 研究背景…………………………………………………………………………1

1.2 研究内容与技术路线………………………………………………………………3

2 实验数据获取………………………………………………………………………………5

2.1 实验系统………………………………………………………………………5

2.2 实验设计………………………………………………………………………6

2.2.1 相同系统先后数据采集…………………………………………………7

2.2.2 不同系统先后数据采集…………………………………………………7

3 图像配准………………………………………………………………………………9

3.1 手动配准…………………………………………………………………………9

3.1.1 配准操作…………………………………………………………………9

3.1.2 数据处理…………………………………………………………………9

3.1.3 结果比较…………………………………………………………………17

3.2 自动配准………………………………………………………………………17

3.2.1 自动配准算法……………………………………………………………17

3.2.2 自动配准程序开发………………………………………………………18

3.3 精度评价…………………………………………………………………………25

3.3.1 评价方法…………………………………………………………………25

3.3.2 自动评价算法与程序开发………………………………………………26

3.3.3 精度对比…………………………………………………………………28

4 结论……………………………………………………………………………………43

4.1 总结………………………………………………………………………………43

4.2 结论………………………………………………………………………………43

致谢 ……………………………………………………………………………………44

参考文献 ………………………………………………………………………………45

附录A ………………………………………………………………………………47

附录B ………………………………………………………………………………50

附录C ………………………………………………………………………………51

附录D ………………………………………………………………………………52

1 绪论

1.1 研究背景

高光谱图像主要的分类方法分为监督分类方法和非监督分类方法,其中监督分类方法主要介绍了平行多面体分类方法、最大似然分类方法、人工神经元分类方法 ;非监督分类方法主要介绍了K-means分类方法、ISDATA分类方法、谱聚类分类方法。同时还综述了支持向量机分类方法、最小二乘支持向量机分类方法、决策树分类方法等新型分类方法[1]。同时,成像光谱仪的光谱分辨力能够达到纳米级,分光方式也趋于多样化,既有高光谱分辨率又能够保证清晰空间图像的探测技术推动人们加强了其在资源勘探、环境监测和食品检测方面的应用[2]

高光谱成像技术在各种果蔬产品的快速和非破坏性检测中和农产品的检测中得到了充分的应用。高光谱成像技术对果蔬的检测分为外部质量检测和内部质量检测,通过检测可以对果蔬进行好坏鉴定以及分级处理,比如利用红外线高光谱成像技术可以对苹果内水分含量进行测量。对于农产品的检测而言,主要是杂质(谷壳和稻草)检测,品种辨识,作物生长形态辨识病虫害监测。不仅如此,也可以将高光谱成像技术应用于大面积成品率的估算中[3-6]

您需要先支付 80元 才能查看全部内容!立即支付

课题毕业论文、开题报告、任务书、外文翻译、程序设计、图纸设计等资料可联系客服协助查找,优先添加企业微信。