-
上述多种高光谱图谱系统所获取的是二维图像和一维高光谱信息,缺少距离探测深度信息。然而,在海洋应用中,物体目标距离信息探测也是海洋探测重要内容之一。因此笔者团队将高光谱成像技术和沙姆成像原理结合,研发了一种新型的连续光高光谱沙姆激光雷达系统,可应用于海洋生物的探测,在微藻和水母探测取得了一些进展;并且基于振镜凝视式高光谱技术,结合结构光三维探测技术,自主搭建了一套高精度四维高光谱探测系统,详细内容将在下文阐述。
-
沙姆激光雷达系统是基于沙姆成像原理(Scheimpflug Principle)的一种新型的连续光激光雷达。沙姆成像原理具体描述为通过倾斜像平面使得物平面、成像透镜平面和像平面相交于一条直线,以获取目标物体全面清晰的像,如图17所示。基于其形似书本翻页,这个规则也被称为合页规则(Hinge Rule)。满足沙姆原理的成像系统理论上可以在使用大口径的成像透镜时实现无穷远的景深,对长距离范围内的物体清晰成像,同时实现大口径和无限焦深。
图 17 沙姆原理示意图。O是坐标系的原点;O'是透镜的中心;O''是图像传感器的坐标原点;
$ \varphi $ 是像平面与透镜平面之间的夹角;$ \theta $ 是物平面与透镜平面之间的夹角;$ {v}_{0} $ 是像面中心与透镜中心的距离;$ d $ 是透镜中心到物平面的距离;$ {p}_{I} $ 是图像传感器上像点的像素位置;A、B为物平面上的物点,A'、B'为A、B对应的像点Figure 17. Schematic diagram of Scheimpflug principle. O is the origin of coordinate system; O' is the center of the lens; O'' is the coordinate origin of the image sensor;
$ \varphi $ is the included angle between image plane and lens plane;$ \theta $ is the included angle between object plane and lens plane;$ {v}_{0} $ is the length between the center of the image plane and the center of the lens; d is the length from the center of the lens to the object plane;$ {p}_{I} $ is is the pixel position of the image point on the image sensor; A and B are object points in the object plane, A' and B' are image points corresponding to A and B如图17所示,基于几何光学的原理,可以推导得到满足沙姆成像原理下,其图像传感器像素位置与成像物体距离之间的关系:
$$ {\textit{z}}=\frac{d[{p}_{I}\left({\rm sin}\varphi -{\rm cos}\varphi {\rm cot}\theta \right)+{v}_{0}]}{{p}_{I}\left({\rm cos}\varphi +{\rm sin}\varphi {\rm cot}\theta \right)+{v}_{0}{\rm cot}\theta } $$ (1) 式中:
$ {\textit{z}} $ 是物平面上的物点到坐标原点O的距离;$ d $ 是透镜中心到物平面的距离;$ \varphi $ 是像平面与透镜平面之间的夹角;$ \theta $ 是物平面与透镜平面之间的夹角;$ {v}_{0} $ 是像面中心与透镜中心的距离;$ {p}_{I} $ 是图像传感器上像点的像素位置,可以由公式(2)计算得到:$$ {p}_{I}=\left({N}_{p}/2-{n}_{p}\right){\omega }_{p} $$ (2) 式中:
$ {N}_{p} $ 是单列或单行的像素总个数;$ {n}_{p} $ 是每个像素对应图像传感器坐标原点O″的位置,O″为图像传感器的中心;$ {\omega }_{p} $ 是像素尺寸。$ {v}_{0} $ 可由成像公式计算可得:$$ {v}_{0}=fd/\left(d-f\mathrm{cos}\theta \right) $$ (3) 式中:
$ f $ 是成像透镜的焦距。$ \varphi $ 可以根据三角关系得到:$$ \varphi ={\rm arctan}\frac{f\mathrm{sin}\theta }{d-f\mathrm{cos}\theta } $$ (4) 当沙姆激光雷达系统应用于水下探测时,由于水的折射率与空气不同,光束在经过空气-水界面时会产生折射,这使系统中物点距离与像点像素位置的关系发生了变化,如果不加以矫正,所得到的距离信息将产生较大的误差,因此需要对空气-水界面的折射进行矫正以消除误差。如图18所示,沙姆成像原理在成像过程中,光束经过空气-水交界面后发生折射,使得真实的物点比未矫正的物点距离更远。根据折射定律和光束的几何关系,可以在
$ yo{\textit{z}}$ 面矫正z方向上产生的距离测量误差,校正后实际的物点距离为:图 18 距离矫正示意图。O是坐标系的原点;O'是透镜的中心;
$ {p}_{I} $ 是像平面上图像传感器每个像素的位置;A0是未考虑空气-水界面折射时物体的位置;A1是物体实际的位置;A1'是A1对应的像点;$ {\delta }_{1} $ 是入射光束与垂直方向的夹角;$ {\delta }_{2} $ 是折射光束与垂直方向的夹角;z0是坐标原点(O)与空气-水界面之间的距离;z是未矫正的距离(OA0);z1是矫正后实际的距离(OA1)Figure 18. Optical layout of distance correction. O is the origin of coordinate system; O' is the center of the lens;
$ {p}_{I} $ is the position of each pixel of the image sensor on the image plane; A0 is the position of the object without considering the refraction of the air-water interface; A1 is the actual position of the object; A1' is the image point corresponding to A1;$ {\delta }_{1} $ is the angle between the incident light and the vertical direction;$ {\delta }_{2} $ is the angle between the refracted light and the vertical direction; z0 is the distance between the origin of the coordinates (O) and the air-water interface; z is the uncorrected distance (OA0); z1 is the actual distance (OA1) after correction$$ {{\textit{z}}}_{1}={{\textit{z}}}_{0}+\left( {{\textit{z}}}_{0}\right)\frac{\mathrm{tan}{\delta }_{2}}{\mathrm{tan}{\delta }_{1}} $$ (5) 式中:z是未经矫正的距离(OA0);z1是矫正后的距离(OA1);z0是透镜平面与空气-水界面之间的垂直距离(OM)。
$ {\delta }_{1} $ 和$ {\delta }_{2} $ 分别是入射光束和折射光束与空气-水界面的夹角。 -
非弹性高光谱沙姆激光雷达系统是将一维沙姆成像原理与高光谱成像技术结合的一种新型的连续光激光雷达系统,可以用于探测物质荧光光谱。满足沙姆成像原理的沙姆激光雷达系统理论上具有无穷景深,可以灵敏的探测到激光照射路径上物体所发出的荧光信号。图19(a)所示是笔者团队搭建的新型非弹性高光谱沙姆激光雷达系统的实验样机,主要包括激光发射模块和接收模块,图19(b)是其成像原理图。发射模块使用的是中心波长为446 nm的商用蓝色激光二极管,其最大输出功率为1.5 W,可根据应用场景进行更换合适波长和功率的半导体激光器。接收模块由成像透镜、倾斜的狭缝、准直透镜、长通滤光片、棱镜-光栅-棱镜(PGP)结构、聚焦透镜和CMOS传感器组成。物体在激光激发下产生的荧光首先经过焦距为50 mm的佳能成像镜头聚焦到缝宽为50 μm的狭缝,然后通过焦距为75 mm的消色差双胶合的准直透镜准直,随后通过PGP结构进行分光,最终经过与准直透镜规格完全一致的聚焦透镜聚焦成像在面阵CMOS相机上。PGP分光结构中的闪耀光栅每毫米300刻槽,楔形棱镜的转向角为6°。可根据激光器波长和物质荧光光谱选择合适的长通滤波片滤除背景杂散光,以提高图像的信噪比,方便实验数据的处理和分析。
图 19 (a)非弹性高光谱沙姆激光雷达系统样机;(b)非弹性高光谱沙姆激光雷达成像原理图:L1和L2是准直透镜,OF是一种长通滤光片。P1和P2是两个对称的楔形棱镜,G是每毫米300个刻槽的透射光栅
Figure 19. (a) A prototype of inelastic hyperspectral Scheimpflug lidar system; (b) Inelastic hyperspectral Scheimpflug lidar imaging schematic: L1 and L2 are collimated lenses, and OF is a long-pass optical filter. P1 and P2 are two symmetrical wedge prisms, and G is a transmission grating with 300 grooves per mm
为调试方便,将成像透镜平面与物平面的夹角
$ \theta $ 设置为90°。经过计算可得成像透镜平面与像平面夹角$ \varphi $ 为14°。因此,狭缝和面阵CMOS相机需要分别倾斜14°,以满足共轭成像关系。此外,狭缝方向必须严格与光栅分光方向垂直,以保证距离和光谱定标准确。面阵相机所获取的图像,一个方向记录光谱信息,一个方向记录空间信息(即距离)。在光谱数据处理时,在原始光谱图像中选择感兴趣距离范围以及光谱范围,将对应的像素点强度按照空间方向进行叠加,并根据公式(5)归一化,就可以得到测量物体的归一化荧光光谱数据。$$ {I_{norm}}\left( \lambda \right) = \frac{{{I_{origin}}\left( \lambda \right) - {I_{\min }}}}{{{I_{\max }} - {I_{\min }}}} $$ (6) 将系统应用于实验前,需要先对系统的距离和光谱进行标定。如图20(a)所示,未进行折射矫正的数据与矫正后的真实数据误差较大,这体现了折射矫正的必要性和准确性。从图中可以看到,非弹性高光谱沙姆激光雷达系统距离空气-水界面的距离约为2.4 m。光谱标定采用的是标准的汞灯光源,标定结果如图20(b)所示,可以区分开576.960 nm和579.066 nm的谱线,系统的光谱分辨率在2.1 nm左右。
图 20 (a)距离定标结果;(b)光谱标定结果
Figure 20. (a) Distance calibration result; (b) Spectral calibration result
笔者利用自主搭建的非弹性高光谱沙姆激光雷达实验样机,分别在实验室环境和近岸实地环境下进行了多次水中生物探测实验。
在实验室环境下,笔者团队对蓝色安朵仙水母、褐色安朵仙水母和巴布亚硝水母进行了荧光光谱探测,图21给出了这三种水母的实物照片。将水母依次置于方形玻璃器皿中后(玻璃器皿在446 nm和488 nm激光激发下不会产生荧光),放置于规格为240 cm×40 cm×30 cm的水箱内,使用最大功率为1.5 W、中心波长分别为446 nm和488 nm的半导体激光二极管作为激发光先后对这些水母进行探测,都观察到了明显的红色荧光信号。这些红色荧光主要由于这三种水母体内有单细胞虫黄藻共生,而虫黄藻体内的叶绿素在446 nm和488 nm的激光激发下会产生680 nm附近的红色荧光。对探测到的水母荧光光谱信号进行处理后的光谱图如图22所示,可以看出三种水母荧光光谱曲线的趋势大致相同。
图 21 (a)蓝色安朵仙水母样品;(b)褐色安朵仙水母样品;(c)巴布亚硝水母样品
Figure 21. (a) A sample of blue Cassiopea andromedah; (b) A sample of brown Cassiopea andromeda; (c) A sample of Mastigias papua
图 22 高光谱沙姆激光雷达测得的三种水母的荧光高光谱图。(a)激发光波长为446 nm;(b)激发光波长为488 nm
Figure 22. Fluorescence hyperspectra of three kinds of jellyfish measured by hyperspectral Scheimpflug lidar: (a) The wavelength of excitation light is 446 nm; (b) The wavelength of excitation light is 488 nm
根据实验室环境下的实验结果可以看出,光源为446 nm或488 nm的非弹性高光谱沙姆激光雷达都可以探测到水母的荧光信号,可以应用于体内有虫黄藻共生的水母的探测。
笔者团队还使用非弹性高光谱沙姆激光雷达系统实地对棕囊藻和水母等生物进行了探测。2021年1月下旬,中国深圳大鹏金沙湾-南澳近岸海域爆发了球形棕囊藻赤潮。笔者团队在岸边捞取了不少球形棕囊藻囊体样品,较大的囊体的直径可达15~16 mm,较小的囊体的直径也在几个毫米,如图23(a)所示。随后,笔者团队将这些球形棕囊藻囊体样品置于500 mm×100 mm×50 mm规格的亚克力板透明水箱内,并置于距离高光谱沙姆激光雷达约3 m的地方。此时激光雷达的光源采用的是最大功率1.5 W的中心波长为446 nm的蓝色半导体激光二极管。
图 23 (a)球形棕囊藻实物图,较大的囊体的直径可达15~16 mm;(b)岸上水箱实验测得的球形棕囊藻囊体的归一化荧光光谱;(c)在近岸渔排现场测量的照片;(d) 在近岸渔排现场测的球形棕囊藻囊体的归一化荧光光谱
Figure 23. (a) A photograph of phaeocystis globosa showed that the diameter of large cysts could reach 15-16 mm; (b) Normalized fluorescence spectra of phaeocystis globosa cysts measured in the onshore tank experiment; (c) A photograph of on-site measurement in the nearshore fishing ground; (d) Normalized fluorescence spectra of phaeocystis globosa cysts measured in the nearshore fishing ground
对探测到的球形棕囊藻囊体的荧光进行了归一化处理。如图23(b)所示,在500~550 nm之间的是微弱的水拉曼的峰,在650~800 nm之间是叶绿素荧光峰。随后,笔者团队在2021年1月24日晚上,在近岸的渔排上利用非弹性高光谱沙姆激光雷达系统直接对近岸的棕囊藻囊体进行探测。如图23(c)所示,红圈内的小红点就是海水中的球形棕囊藻囊体在446 nm的激光激发下产生的红色荧光信号。图23(d)是对现场测量的球形棕囊藻囊体荧光光谱的处理结果,从图中可以看到水的拉曼非常强。其次,对比在岸上水箱测量的球形棕囊藻囊体,现场测量的囊体的荧光在750 nm附近的小肩峰消失了,这是由于水对长波的吸收较大,这部分的荧光被水吸收了。
此外,笔者团队还于中国深圳大鹏金沙湾-南澳海域捕捞了一些该海域常见的水母(多管水母、双生水母),如图24所示。在实验室环境下,提取活性较高的水母置于装有纯净海水的透明培养皿内,作为实验前的预处理。
将非弹性高光谱沙姆激光雷达系统的光源换成最大功率为1.5 W、中心波长为488 nm的半导体激光器,作为水母荧光的激发光源。装有纯净海水的透明比色皿在该波长下不会产生荧光。水母实验样品被置于距离雷达约3 m的位置。对获取的水母荧光光谱进行归一化处理,图25(a)和(b)分别给出了多管水母和双生水母在488 nm激发下产生的归一化荧光光谱,其荧光峰峰值都在515 nm左右,双生水母的光谱范围更广,在550~600 nm之间仍有很明显的荧光。实验结果证实了光源为488 nm的非弹性高光谱沙姆激光雷达探测系统还可应用于多管水母和双生水母的探测。
-
为融合二维空间、一维光谱、一维深度等四维信息于一体,笔者团队基于振镜凝视式高光谱技术,结合结构光三维探测技术,搭建了一套能实现高光谱分辨率(3 nm)、高空间分辨率、高深度精度(27.5 μm)的四维高光谱探测系统[32],其示意图和系统参数指标如图26及表1所示。其光谱范围可覆盖400~800 nm;光谱分辨率可达3 nm;深度精度为27.5 μm;四维探测时间小于80 s。系统四维探测分为两部分:第一部分是结构光三维重建及高光谱立方体数据采集;第二部分是三维空间数据与高光谱数据融合形成四维数据集,四维融合探测结果如图27所示。
表 1 四维高光谱探测系统参数
Table 1. Four dimensional hyperspectral detection system parameters
Specifications of the HSDA system Characteristic parameters Spectral range/nm 400-800 Spectral resolution/nm <3 Depth resolution/mm 0.0275 Plane fit standard deviation/mm 0.0269 Num of 4D points <800×800 Acquisition time/s <80 FOV/(°) 30 Measurement field/mm2 96×64 Working distance/mm 300-600 图 27 系统流程图。(a)原始4D模型;(b)原点模型中A、B、C、D点的光谱曲线;(c)基于光谱信息的点云分割4D模型;(d)不同波长(450~750 nm)的单色图像,用于不同的3D透视[32]
Figure 27. Flow chart of system.(a) Original 4D model; (b) Spectrum curves of point A, B, C and D in the origin model; (c) 4D model with point cloud segmentation based on spectral information; (d) Monochrome images at different wavelengths (450-750 nm) for different 3D perspectives[32]
基于自主搭建的四维高光谱探测系统,对真实植株与塑料植株的四维高光谱进行了探测实验,其探测结果如图28~图29所示。从图29(d)光谱曲线可以看出,绿色植物的反射光谱在450 nm和670 nm处有吸收峰,这是由叶绿素的光吸收引起的,而塑料植物的反射光谱在440 nm和470 nm处有吸收峰。提取上述几处峰值点,A和B的光谱之间的差异可实现最大化。因此,尽管真实植株与塑料植株的三维点云模型具有非常相似的外观,但两者可以通过反射光谱很容易地进行区分。也就是说,每个3D点对应一个唯一的光谱数据,可以作为区分真植物和塑料植物的主要特征。上述人脸及植株探测实验,验证了文中四维探测系统具有高光谱分辨率、高空间分辨率、高深度精度的探测能力,也体现了该套系统在水下原位探测海洋生物的四维高光谱信息的巨大潜力。
图 28 (a)绿色植物;(b)被测绿色植物的高光谱立方体;(c)绿色植物的三维点云模型;(d)在不同角度和不同波长(450~675 nm)下观察到的绿色植物的四维数据;(e)放大叶片数据;(f)叶片中蓝色矩形区域点云的分布;(g)点云的曲面面片[32]
Figure 28. (a) The green plant; (b) The hyperspectral cube of the measured green plant; (c)3D point cloud model of the green plant; (d) The 4D data of the green plant observed at different angles and at different wavelengths (450-675 nm); (e) Enlargement of the leaf blade margin; (f) The distribution of the point cloud of blue rectangular region in the leaf blade; (g) The surface patch of the point cloud[32]
Application of hyperspectral imager and lidar in marine biological detection
-
摘要:
海洋是涵盖超过70%地球表面的连续海水,发展先进的海洋生物光学监测手段对海洋生态系统的保护至关重要。文中综述了笔者团队在小型高光谱图谱仪与激光雷达系统搭建及其在海洋生物检测等应用上的部分近期工作。图谱仪方面介绍了不同空间扫描方式的高光谱图谱仪在透射、反射及荧光等不同工作模式下,对数种藻类、斑马鱼等海洋生物进行了图谱检测,并且基于图谱数据结合机器学习算法实现了微藻种类的精准分类和藻类生长周期的准确预测;在激光雷达方面详述了使用非弹性高光谱沙姆激光雷达系统在实验室和近岸实地环境进行了多次水生生物的测量实验,成功获取其荧光高光谱,证明了非弹性高光谱沙姆激光雷达系统在海洋生物监测应用中的巨大潜力。此外,笔者团队还搭建了一套四维凝视成像探测系统能实现高光谱分辨率(3 nm)、高空间分辨率、高深度精度(27.5 μm)的精准探测。
Abstract:Oceans are continuous waters that cover more than 70% of the earth's surface. The optical monitoring of marine life is very important for the protection of marine ecosystem. In this paper, a review on our recent work in the construction of compact hyperspectral spectrometers and lidar systems and their applications in e.g. marine biological detection was given. Hyperspectral imagers with different spatial scanning methods were demonstrated, which were used to detect several kinds of algae, zebrafish and other marine organisms under different modes, such as transmission, reflection and fluorescence modes. In addition, based on some machine learning algorithm, accurate classification of microalgae and accurate prediction of algae growth cycle were achieved. In the aspect of lidar, an inelastic hyperspectral Scheimpflug lidar system has been used to measure aquatic organisms in laboratory and inshore field environment and their fluorescence hyperspectra have been captured successfully, which demonstrated the great potential of the inelastic hyperspectral Scheimpflug lidar system in the application of marine biological monitoring. A four-dimensional detection system which could achieve high spectral resolution (3 nm), high spatial resolution and high depth precision (27.5 μm) was also presented.
-
Key words:
- hyperspectral imager /
- Scheimpflug lidar system /
- jelly fish /
- phaeocystis /
- 4D detection
-
图 1 显微高光谱成像仪原理图及系统标定结果。(a)原型显微高光谱成像仪的照片;(b)光学元件示意图:1.物镜 2.成像镜头 3.狭缝 4.准直透镜 5.棱镜 6.光栅 7.棱镜 8.管透镜 9.CMOS;(c)校准光源原始光谱图像;(d)系统测量的定标光谱,光谱分辨率约为3 nm;(e)波长和像素之间的校准结果;(f)商业光谱仪测量的校准光源光谱;(g)分辨率测试板的重建高光谱图像,分辨线(4.39 μm);(h)整个工作系统的原型示意图[27]
Figure 1. Schematic diagram of microscopic hyperspectral imager and the results of system calibration. (a) A photo of the prototype microscopic hyperspectral imager; (b) Schematic illustration of the optical elements: 1. Objective, 2. Imaging lens, 3. Slit, 4. Collimator lens, 5. Prism, 6. Grating, 7. Prism, 8. Tube lens, 9. CMOS; (c) Original spectral image of a calibration source; (d) Spectrum of the calibration source measured by our system. The spectral resolution is about 3 nm; (e) Calibration result between the wavelength and pixel index; (f) Spectrum of the calibration source measured by a commercial spectrometer; (g) Reconstructed hyperspectral image of the resolution test target. Resolvable lines in element 6 of group 6 (4.39 μm); (h) Prototype of the whole working system[27]
图 2 (a)侧面光照条件下阴影生成示意图;(b)多模式检测示意图。各部件示意图:1.单模光纤,2.光纤准直器,3.透镜,4.分束器,5.长通滤波片,6.物镜,7.样品和运动平台,8.高光谱成像系统[27]
Figure 2. (a) Schematic diagram of the shadow generation under lateral illumination condition; (b) Illustration of multi-mode detections. Reflection mode and fluorescence mode use epi-illumination, while transmission mode employs trans-illumination. Schematic diagram of each component: 1. Single mode fiber, 2. Fiber collimator, 3. Doublet lens, 4. Beam splitter, 5. Long pass filter, 6. Objective, 7. Sample and motion stage, 8. Infinity-corrected hyperspectral imaging system, consisting of the imaging module[27]
图 3 透射模式下斑马鱼高光谱检测结果。(a)水下斑马鱼的光学显微图像,比例尺:100 μm;(b)高光谱重建斑马鱼高光谱图像。比例尺:100 μm;(c)斑马鱼局部放大图像和三个标注点,包括透明鳍(标记为红点)、斑点(标记为蓝星)和卵黄囊(标记为黄色三角形);(d)三个标注点对应的透射光谱[27]
Figure 3. Experimental results the hyperspectral detection of a zebrafish under the transmission mode. (a) Optical microscopic image of a zebrafish underwater. Scale bar: 100 μm; (b) Hyperspectral image of a zebrafish reconstructed from our system. Scale bar: 100 μm; (c) Partial enlarged image of the zebrafish and three points of interest, including the transparent fin (marked as red dot), speckle (marked as blue star) and yolk sac (marked as yellow triangle); (d) Corresponding transmittance spectrum at the three points of interest[27]
图 4 (a)用MMHI系统测定了5种致灾微藻的荧光光谱;(b)归一化光谱;(c)用商用光谱仪测定了5种致灾微藻的荧光光谱;(d)归一化光谱[27]
Figure 4. (a) Measured fluorescence spectra of five disaster-causing microalgae by the MMHI system; (b) Normalized spectra of (a); (c) Measured fluorescence spectra of five disaster-causing microalgae by a commercial spectrometer; (d) Normalized spectra of (c)[27]
图 6 (a)旋转高光谱图谱仪实物图;(b)原理图和光学元件(1.成像镜头,2.线缝,3.准直透镜,4.光栅,5.棱镜,6.成像透镜);(c)系统安装于旋转支架上[29]
Figure 6. (a) The photo of our rotational hyperspectral scanner; (b) Schematic diagram and optical elements (1. imaging lens; 2. line slit; 3. collimating lens; 4. grating; 5. prism; 6. imaging lens); (c) Installed on a rotational mount[29]
图 13 三种微藻吸收特性验证。(a)原始LED光源和三种微藻的透射光谱;(b)归一化光谱;(c)微藻相对于LED光源的透射率;(d)商用分光光度计所测量的微藻吸光度[31]
Figure 13. Verification of absorption characteristics of three microalgae. (a) Original transmission spectra of the LED light source and three microalgae; (b) Normalized spectra of (a); (c) Transmittance of the microalgae relative to the LED light source; (d) Absorbance of the microalgae measured by a commercial spectrophotometer[31]
图 14 (a)使用线性支持向量机分类后的三种微藻PC2和PC1散点图;(b) 180个样品(包括60个角毛藻、60个衣藻和60个棕囊藻)透射光谱的混淆矩阵;(c) ROC曲线对应于(a)中的SVM分类器[31]
Figure 14. (a) Scatter plot of the scores of PC2 versus PC1 of three microalgae along with the linear SVM classifiers. The entire two-dimensional space is divided into three regions as labeled; (b) Confusion matrix of all transmission spectra for 180 samples (including 60 chaetoceros, 60 chlamydomonas and 60 phaeocystis); (c) The ROC curves corresponding to the SVM classifiers in (a)[31]
图 15 一组混合微藻的物种分类示意图。(a)混合微藻在可见波段的高光谱图像;(b)混合微藻单波段(680 nm)的高光谱图像;(c)加上图像掩模后图(a)的种属鉴定结果,蓝色和红色区域分别代表莱茵衣藻和棕囊藻;(d)模板图像掩模后由图(b)上的阈值分割所生成图像[31]
Figure 15. Demonstration for the species identification from a group of mixed microalgae. (a) Hyperspectral image of the mixed microalgae at the broad visible bands; (b) Hyperspectral image of the mixed microalgae at the single spectral band (680 nm); (c) The result of species identification on (a) after adding the image mask, the blue and red regions represent the chlamydomonas and phaeocystis, respectively; (d) The template image mask, generated by threshold segmentation on (b)[31]
图 16 (a)棕囊藻25天以上的生长曲线 1:滞后期 2:指数增长期 3:稳定期 4:下降期;(b)棕囊藻的归一化透射光谱;(c)训练集对生长阶段的预测结果;(d)测试集对生长阶段的预测结果[31]
Figure 16. (a) Growth curve of phaeocystis over 25 days. 1: lag phase, 2: exponential growth phase, 3: stable phase, 4: decline phase; (b) Normalized transmission spectra of phaeocystis; (c) Predicted results of the growth stage by the training set; (d) Predicted results of the growth stage by the test set[31]
图 17 沙姆原理示意图。O是坐标系的原点;O'是透镜的中心;O''是图像传感器的坐标原点;
$ \varphi $ 是像平面与透镜平面之间的夹角;$ \theta $ 是物平面与透镜平面之间的夹角;$ {v}_{0} $ 是像面中心与透镜中心的距离;$ d $ 是透镜中心到物平面的距离;$ {p}_{I} $ 是图像传感器上像点的像素位置;A、B为物平面上的物点,A'、B'为A、B对应的像点Figure 17. Schematic diagram of Scheimpflug principle. O is the origin of coordinate system; O' is the center of the lens; O'' is the coordinate origin of the image sensor;
$ \varphi $ is the included angle between image plane and lens plane;$ \theta $ is the included angle between object plane and lens plane;$ {v}_{0} $ is the length between the center of the image plane and the center of the lens; d is the length from the center of the lens to the object plane;$ {p}_{I} $ is is the pixel position of the image point on the image sensor; A and B are object points in the object plane, A' and B' are image points corresponding to A and B图 18 距离矫正示意图。O是坐标系的原点;O'是透镜的中心;
$ {p}_{I} $ 是像平面上图像传感器每个像素的位置;A0是未考虑空气-水界面折射时物体的位置;A1是物体实际的位置;A1'是A1对应的像点;$ {\delta }_{1} $ 是入射光束与垂直方向的夹角;$ {\delta }_{2} $ 是折射光束与垂直方向的夹角;z0是坐标原点(O)与空气-水界面之间的距离;z是未矫正的距离(OA0);z1是矫正后实际的距离(OA1)Figure 18. Optical layout of distance correction. O is the origin of coordinate system; O' is the center of the lens;
$ {p}_{I} $ is the position of each pixel of the image sensor on the image plane; A0 is the position of the object without considering the refraction of the air-water interface; A1 is the actual position of the object; A1' is the image point corresponding to A1;$ {\delta }_{1} $ is the angle between the incident light and the vertical direction;$ {\delta }_{2} $ is the angle between the refracted light and the vertical direction; z0 is the distance between the origin of the coordinates (O) and the air-water interface; z is the uncorrected distance (OA0); z1 is the actual distance (OA1) after correction图 19 (a)非弹性高光谱沙姆激光雷达系统样机;(b)非弹性高光谱沙姆激光雷达成像原理图:L1和L2是准直透镜,OF是一种长通滤光片。P1和P2是两个对称的楔形棱镜,G是每毫米300个刻槽的透射光栅
Figure 19. (a) A prototype of inelastic hyperspectral Scheimpflug lidar system; (b) Inelastic hyperspectral Scheimpflug lidar imaging schematic: L1 and L2 are collimated lenses, and OF is a long-pass optical filter. P1 and P2 are two symmetrical wedge prisms, and G is a transmission grating with 300 grooves per mm
图 23 (a)球形棕囊藻实物图,较大的囊体的直径可达15~16 mm;(b)岸上水箱实验测得的球形棕囊藻囊体的归一化荧光光谱;(c)在近岸渔排现场测量的照片;(d) 在近岸渔排现场测的球形棕囊藻囊体的归一化荧光光谱
Figure 23. (a) A photograph of phaeocystis globosa showed that the diameter of large cysts could reach 15-16 mm; (b) Normalized fluorescence spectra of phaeocystis globosa cysts measured in the onshore tank experiment; (c) A photograph of on-site measurement in the nearshore fishing ground; (d) Normalized fluorescence spectra of phaeocystis globosa cysts measured in the nearshore fishing ground
图 27 系统流程图。(a)原始4D模型;(b)原点模型中A、B、C、D点的光谱曲线;(c)基于光谱信息的点云分割4D模型;(d)不同波长(450~750 nm)的单色图像,用于不同的3D透视[32]
Figure 27. Flow chart of system.(a) Original 4D model; (b) Spectrum curves of point A, B, C and D in the origin model; (c) 4D model with point cloud segmentation based on spectral information; (d) Monochrome images at different wavelengths (450-750 nm) for different 3D perspectives[32]
图 28 (a)绿色植物;(b)被测绿色植物的高光谱立方体;(c)绿色植物的三维点云模型;(d)在不同角度和不同波长(450~675 nm)下观察到的绿色植物的四维数据;(e)放大叶片数据;(f)叶片中蓝色矩形区域点云的分布;(g)点云的曲面面片[32]
Figure 28. (a) The green plant; (b) The hyperspectral cube of the measured green plant; (c)3D point cloud model of the green plant; (d) The 4D data of the green plant observed at different angles and at different wavelengths (450-675 nm); (e) Enlargement of the leaf blade margin; (f) The distribution of the point cloud of blue rectangular region in the leaf blade; (g) The surface patch of the point cloud[32]
表 1 四维高光谱探测系统参数
Table 1. Four dimensional hyperspectral detection system parameters
Specifications of the HSDA system Characteristic parameters Spectral range/nm 400-800 Spectral resolution/nm <3 Depth resolution/mm 0.0275 Plane fit standard deviation/mm 0.0269 Num of 4D points <800×800 Acquisition time/s <80 FOV/(°) 30 Measurement field/mm2 96×64 Working distance/mm 300-600 -
[1] Vahtmäe E, Paavel B, Kutser T. How much benthic information can be retrieved with hyperspectral sensor from the optically complex coastal waters? [J]. J Appl Remote Sens, 2020, 14(1): 016504. doi: 10.1117/1.JRS.14.016504 [2] Klemas V V. Coastal and environmental remote sensing from unmanned aerial vehicles:An overview [J]. Journal of Coastal Research, 2015, 31(5): 1260-1267. doi: 10.2112/JCOASTRES-D-15-00005.1 [3] Lou Xiulin, Hu Chuanmin. Diurnal changes of a harmful algal bloom in the East China Sea: Observations from GOCI [J]. Remote Sensing of Environment, 2014, 140: 562-572. doi: 10.1016/j.rse.2013.09.031 [4] Pettersen R, Johnsen G, Bruheim P, et al. Development of hyperspectral imaging as a bio-optical taxonomic tool for pigmented marine organisms [J]. Organisms Diversity & Evolution, 2014, 14(2): 237-246. [5] Bansod B, Singh R, Thakur R. Analysis of water quality parameters by hyperspectral imaging in Ganges River [J]. Spatial Information Research, 2018, 26: 203-211. doi: 10.1007/s41324-018-0164-4 [6] Jia Beibei, Yoon Seung-Chul, Zhuang Hong, et al. Prediction of pH of fresh chicken breast fillets by VNIR hyperspectral imaging [J]. Journal of Food Engineering, 2017, 208: 57-65. doi: 10.1016/j.jfoodeng.2017.03.023 [7] Cai F, Lu W, Shi W, et al. A mobile device-based imaging spectrometer for environmental monitoring by attaching a lightweight small module to a commercial digital camera [J]. Scientific Reports, 2017, 7(1): 15602. doi: 10.1038/s41598-017-15848-x [8] Gardner B, Reddy R, Mayerich D, et al. Application of vibrational spectroscopy and imaging to point-of-care medicine: A review [J]. Appl Spectrosc, 2018, 72(S1): 52-84. [9] Cai F, Dan W, Min Z, et al. Pencil-like imaging spectrometer for bio-samples sensing [J]. Biomedical Optics Express, 2017, 8(12): 5427. doi: 10.1364/BOE.8.005427 [10] Yao Xinli, Li Shuo, He Sailing. Dual-mode hyperspectral bio-imager with a conjugated camera for quick object-selection and focusing [J]. Progress in Electromagnetics Research, 2020, 168: 133-143. doi: 10.2528/PIER20080308 [11] Wei Lin, Su Kang, Zhu Siqi, et al. Identification of microalgae by hyperspectral microscopic imaging system [J]. Spectroscopy Letters, 2017, 50(1): 59-63. doi: 10.1080/00387010.2017.1287094 [12] Chang Chih-Chung, Lin Chih-Jen. Libsvm: A library for support vector machines [J]. ACM Transactions on Intelligent Systems and Technology, 2011, 2(3): 1-27. doi: 10.1145/1961189.1961199 [13] Kong Z, Liu Z, Zhang L, et al. Atmospheric pollution monitoring in urban area by employing a 450-nm Lidar system [J]. Sensors (Basel), 2018, 18(6): 1880. doi: 10.3390/s18061880 [14] Kong Z, Ma T, Cheng Y, et al. Feasibility investigation of a monostatic imaging lidar with a parallel-placed image sensor for atmospheric remote sensing [J]. Journal of Quantitative Spectroscopy and Radiative Transfer, 2020, 254: 107212. doi: 10.1016/j.jqsrt.2020.107212 [15] Mei L, Brydegaard M. Atmospheric aerosol monitoring by an elastic Scheimpflug lidar system [J]. Opt Express, 2015, 23(24): A1613-A1628. doi: 10.1364/OE.23.0A1613 [16] Mei L, Guan P, Yang Y, et al. Atmospheric extinction coefficient retrieval and validation for the single-band Mie-scattering Scheimpflug lidar technique [J]. Opt Express, 2017, 25(16): A628-A638. doi: 10.1364/OE.25.00A628 [17] Gao Fei, Li Jingwei, Lin Hongze, et al. Oil pollution discrimination by an inelastic hyperspectral Scheimpflug lidar system [J]. Opt Express, 2017, 25(21): 25515-25522. doi: 10.1364/OE.25.025515 [18] Malmqvist E, Brydegaard M, Alden M, et al. Scheimpflug lidar for combustion diagnostics [J]. Opt Express, 2018, 26(12): 14842-14858. doi: 10.1364/OE.26.014842 [19] Duan Z, Yuan Y, Lu J C, et al. Underwater spatially, spectrally, and temporally resolved optical monitoring of aquatic fauna [J]. Opt Express, 2020, 28(2): 2600-2610. doi: 10.1364/OE.383061 [20] Zhao G, Malmqvist E, Rydhmer K, et al. Inelastic hyperspectral lidar for aquatic ecosystems monitoring and landscape plant scanning test[C]//The 28th International Laser Radar Conference (ILRC 28), EPJ Web of Conferences, 2018, 176: 01003. [21] Zhao G, Ljungholm M, Malmqvist E, et al. Inelastic hyperspectral lidar for profiling aquatic ecosystems [J]. Laser & Photonics Reviews, 2016, 10(5): 807-813. doi: 10.1002/lpor.201600093 [22] Chen Kun, Gao Fei, Chen Xiang, et al. Overwater light-sheet Scheimpflug lidar system for an underwater three-dimensional profile bathymetry [J]. Appl Opt, 2019, 58(27): 7643-7648. doi: 10.1364/AO.58.007643 [23] Gao Fei, Lin Hongze, Chen Kun, et al. Light-sheet based two-dimensional Scheimpflug lidar system for profile measurements [J]. Opt Express, 2018, 26(21): 27179-27188. doi: 10.1364/OE.26.027179 [24] He Sailing, Chen Xiang, Li Shuo, et al. Small hyperspectral imagers and lidars and their marine applications [J]. Infrared and Laser Engineering, 2020, 49(2): 0203001. (in Chinese) doi: 10.3788/IRLA202049.0203001 [25] Lin Hongze, Zhang Yao, Mei Liang. Fluorescence Scheimpflug LiDAR developed for the three-dimension profiling of plants [J]. Opt Express, 2020, 28(7): 9269-9279. doi: 10.1364/OE.389043 [26] Luo Longqiang, Chen Xiang, Xu Zhanpeng, et al. A parameter-free calibration process for a Scheimpflug LIDAR for volumetric profiling [J]. Progress in Electromagnetics Research, 2020, 169: 117-127. doi: 10.2528/PIER20120701 [27] Xu Zhanpeng, Jiang Yiming, He Sailing. Multi-mode microscopic hyperspectral imager for the sensing of biological samples [J]. Applied Sciences, 2020, 10(14): 4876. doi: 10.3390/app10144876 [28] Cai Fuhong, Chen Jie, Xie Xiaofeng, et al. The design and implementation of portable rotational scanning imaging spectrometer [J]. Opt Commun, 2020, 459: 125016. doi: 10.1016/j.optcom.2019.125016 [29] Luo Longqiang, Li Shuo, Yao Xinli, et al. Rotational hyperspectral scanner and related image reconstruction algorithm [J]. Sci Rep, 2021, 11: 3296. doi: 10.1038/s41598-021-82819-8 [30] Xu Zhanpeng, Jiang Yiming, Ji Jiali, et al. Classification, identification, and growth stage estimation of microalgae based on transmission hyperspectral microscopic imaging and machine learning [J]. Opt Express, 2020, 28(21): 30686-30700. doi: 10.1364/OE.406036 [31] Kürüm U, Wiecha P R, French R, et al. Deep learning enabled real time speckle recognition and hyperspectral imaging using a multimode fiber array [J]. Opt Express, 2019, 27(15): 20965-20979. doi: 10.1364/OE.27.020965 [32] Luo Jing, Li Shuo, Forsberg Erik, et al. 4D surface shape measurement system with high spectral resolution and great depth accuracy [J]. Opt Express, 2021, 29(9): 13048-13070. doi: 10.1364/OE.423755
计量
- 文章访问数: 736
- HTML全文浏览量: 304
- 被引次数: 0