-
海洋宏生物原位观测需要兼顾二维图像的“观”和三维图像的“测”。传统水下激光扫描雷达和相机复合的技术可以利用激光扫描雷达获得三维点云数据,利用相机获得高分辨率静态二维图像,通过异源数据融合实现目标二维图像和三维点云数据获取[17]。但是,该复合技术中激光扫描雷达的机械扫描工作方式往往导致空间分辨率低、实时性差,不利于海洋生物等小尺寸运动目标的探测。因此,如何抑制水体散射噪声实现高对比度二维图像和高分辨率三维图像的实时获取,成为海洋宏生物原位观测的关键。针对此,提出了水下激光雷达相机LiRAI,技术上兼顾并超越激光雷达和水下摄像机复合的技术方案,实时获得百万像素的二维强度图像和三维点云数据。
如图2顶部所示,传统的水下光学成像是全景深图像,成像链路中的水体散射、感兴趣目标和背景都被成像,且三维空间降维为二维平面图像,因此,图像呈现低对比度的特点,存在LOST问题。不同于传统水下光学成像,LiRAI可以抑制水体后向散射噪声并过滤背景,仅对感兴趣的采样区进行成像,同时获得高对比度的二维图像和高分辨率三维图像,其成像效果如图2底部所示。
LiRAI的工作原理如图3所示,典型的LiRAI系统主要由脉冲激光器、门控成像器件、时序控制器和图像处理模块构成。其中,脉冲激光器发射纳秒级激光脉冲,经照明镜头整形后以泛光形式照明水体;门控成像器件可实现纳秒级快门,采集目标的回波信号并转化为图像;时序控制器主要是输出同步触发信号,触发脉冲激光器和门控成像器件按照设计的编码时序工作。工作过程中,在时序控制器编码时序控制下,脉冲激光器向水体发射激光脉冲,激光脉冲在水中传播时会被水体吸收和散射,遇到目标时会被目标反射或散射形成后向传输的回波信号,在设定的延时τA下,门控成像器件开启快门接收回波信号,持续开启时间tGate后快门关闭,获得A帧;在设定的延时τB下,快门以同样的门宽开启工作,获得B帧。延时τA和τB满足τB=τA+tGate,其中tGate为快门门宽;快门门宽与激光脉宽满足tGate = tLaser。当激光脉冲和选通脉冲均为矩形脉冲时,在卷积作用下,A帧和B帧的距离能量包络均为三角形。其中,三角形包络的上升沿称为头信号区,下降沿称为尾信号区。如图3所示,A帧的尾信号与B帧的头信号交叠,该交叠区域为LiRAI的有效光立体采样区。
光立体采样区内的三维空间信息是基于笔者提出的三角形距离能量相关三维成像技术实现的[16, 18],通过A帧和B帧交叠区的能量灰度比关系可获得目标的距离信息r:
$$ r=\frac{{\tau }_{\rm A}c}{2n}+\frac{{I}_{\rm head,B}}{{I}_{\rm head,B}+{I}_{\rm tail,A}}D $$ (1) 式中:Itail,A和Ihead,B分别为A帧的尾信号区灰度值和B帧的头信号区灰度值;n为水的折射率;c为光在真空中传播的速度;D 为光立体采样区的景深,其大小为:
$$ D=\frac{{t}_{\rm Laser}c}{2n} $$ (2) 通过公式(1)获得目标距离信息后,利用摄像机模型便可重建目标三维空间信息。
A帧和B帧图像交集叠加便可获得光立体采样区的二维强度图像:
$$ {I}_{\rm sampling}={I}_{\rm head,B}+{I}_{\rm tail,A} $$ (3) 由公式(1)和(3),LiRAI便同时获得了光立体采样区的二维图像和三维空间信息。
-
针对交叠鱼群和底栖生物伪装存在的LOST问题,图4给出了基于蒙特卡洛仿真的LiRAI探测结果。仿真中,水体衰减系数为0.1/m,吸收系数为0.02/m,散射系数为0.08/m,不对称因子为0.9,光子数为100亿,典型工作距离为5 m。图4(a)为鱼群LiRAI仿真结果,包括二维强度图像(左侧)和三维图像(右侧);图4(b)是底栖伪装的比目鱼LiRAI仿真结果,也包括二维强度图像(左侧)和三维图像(右侧)。由图4中可以看出,当反应目标纹理等辐射特性的二维图像中目标对比度低时,利用三维图像中目标的空间信息差异可有效发现低对比度目标。
在梳理LiRAI技术特点前,先分析一下水下视频剖面仪UVP5如何解决海洋微生物原位观测中的低对比度目标探测、采样体积量化和目标尺寸测量的问题。如图5所示,UVP5采用两组波长625 nm的LED对射照明水体,相机光轴垂直于LED对射方向采集照明水体内的目标图像,实现暗场成像,从而提高浮游动物图像的对比度,而采样体积的大小则近似认为是LED对射形成的空间立体采样区,同时在已知相机与采样区距离信息的条件下,基于摄像机模型便可量化测量目标的尺寸[4]。但是,由于UVP5的采样区是LED对射形成的,采样体积受限于LED的机械安装布局而难以灵活调节。
对于LiRAI技术,若门控成像器件的CCD或CMOS的靶面宽度为w,高度为h,成像镜头的焦距为f,如图6所示,则LiRAI的光立体采样区的体积V的大小为:
$$ V=\frac{whD({R}_{\rm A}^{2}+{R}_{\rm B}^{2}+{R}_{\rm A}{R}_{\rm B})}{3{f}^{2}} $$ (4) 公式(4)中,D已由公式(2)给出,RA和RB由延时τA和τB决定,其大小分别为:
$$ {R}_{\rm A}=\frac{{\tau }_{\rm A}c}{2n} $$ (5) $$ {R}_{\rm B}=\frac{{\tau }_{\rm B}c}{2n}=\frac{({\tau }_{\rm A}+{t}_{\rm Gate})c}{2n} $$ (6) 从公式(4)~(6)可以看出,通过调节系统的焦距、延时和快门宽度可以调节光立体采样区的大小,通过调节延时可以调节光立体采样区的位置。因此,不同于UVP5,LiRAI的光立体采样区体积是可调的,理论上不受机械限制,这样就可以满足不同尺寸海洋宏生物的观测。观测大尺寸海洋宏生物时,可以调大光立体采样区;观测小尺寸海洋宏生物时,可以调小光立体采样区。
对于UVP5,由于三维空间投影为二维图像,采样体积内的距离信息仍是丢失的,只能以光立体采样区的单一粗距离信息作为目标的距离。然而,LiRAI则可重建光立体采样区内的三维空间信息,从而获得每个目标的精细距离信息和特征尺寸信息。
此外,在低对比度目标探测方面,LiRAI具有与UVP5类似的光切片成像的特点,可以过滤水体噪声和背景干扰的特点。不同的是:UVP5是通过两组LED对射形成光切片,而LiRAI则是通过激光选通的方式形成光切片,具有更灵活的调节能力[19]。
综上所述,LiRAI用于海洋宏生物探测具有以下技术特点:
(1)单一系统同时获得二维图像和三维图像,二维像素与三维体素一一对应,无需异源数据融合;
(2)高对比度二维成像,仅对感兴趣光立体采样区成像,直接过滤水体噪声和背景干扰;
(3)无机械扫描高分辨率三维成像;
(4)光立体采样体积可灵活调节,满足不同尺寸海洋生物观测需求。
LiRAI的这些技术特点有利于海洋宏生物量化分析:二维图像可用于行为分析和种群识别;三维图像可用于获得目标特征尺寸信息,进行粒径谱分析;利用三维图像中目标和目标之间、目标和背景间的空间差异可提高交叠目标数量统计的准确度;在获得生物数量以及采样体积信息后可反演生物丰度信息。此外,利用LiRAI可以获得时间序列图像,增加时间维度,用于生物量时空分布分析。
Underwater light ranging and imaging for macro marine life in-situ observation and measurement
-
摘要:
海洋宏生物原位“观”和“测”对于海洋生态环境、海洋生物资源和海底矿产资源的研究和评估具有重要的意义。目前用于海洋宏生物原位观察的传统水下摄像机存在目标辐射特性、水体光散射、距离信息丢失等导致的低对比度目标探测难的问题。针对此,提出了水下激光雷达相机,可以兼顾并超越传统激光扫描雷达与摄像机复合的技术方案,利用单一系统同时获得百万像素高对比度的二维强度图像和高分辨率的三维图像,且二维图像中的像素和三维图像中的体素一一对应,并介绍了基于该技术研制的“凤眼”系统,其光立体采样区体积可调,距离分辨率优于1 cm,像素数为1360×1024。自2018年起,“凤眼”在我国南海海域进行了4个航次的海上试验,获取了海底宏生物及微地形地貌图像,最大工作深度达到3 291 m。
Abstract:Marine macro life in-situ observation and measurement is of great significance to research and evaluate marine ecological environment, marine biological resources and seabed mineral resources. Traditional underwater cameras for in-situ observation of marine macro life have problems with low-contrast target (LOST) caused by target radiation characteristics, water light scattering, and loss of distance information in 2D images. Light ranging and imaging (LiRAI) technique was proposed, which could take into account and surpass the traditional lidar and camera composite technical solution, using a single system to simultaneously obtain high-contrast 2D intensity images and high-resolution 3D images with mega-pixels, and pixels in 2D images correspond to voxels in 3D images one by one. “Fengyan” systems were established based on LiRAI for marine macro life in-situ observation and measurement. The optical sampling volume was adjustable, the range resolution was better than 1 cm, and the number of pixels was 1 360×1 024. Since 2018, four voyages of sea trials had been conducted in the South China Sea, and images of marine life and sea floor had been obtained successfully, and the maximum working depth of “Fengyan” was 3 291 m.
-
-
[1] Sun Song, Sun Xiaoxia. Future ocean and our research strategy [J]. Bulletin of the Chinese Academy of Sciences, 2016, 31(12): 1285-1292. (in Chinese) [2] Wang Pinxian. Developing deep-sea science and technology: Perspectives and pitfall [J]. Scinece & Technology Review, 2021, 39(3): 71-79. (in Chinese) [3] Davis C S, Thwaites F T, Gallager S M, et al. A three-axis fast-tow digital Video Plankton Recorder for rapid surveys of plankton taxa and hydrography [J]. Limnology and Oceanography: Methods, 2005, 3(2): 59-74. doi: 10.4319/lom.2005.3.59 [4] Picheral M, Guidi L, Stemmann L, et al. The underwater vision profiler 5: an advanced instrument for high spatial resolution studies of particle size spectra and zooplankton [J]. Limnology and Oceanography: Methods, 2010, 8(9): 462-473. doi: 10.4319/lom.2010.8.462 [5] Tristan Biard, Lars Stemmann, Marc Picheral, et al. In situ imaging reveals the biomass of giant protists in the global ocean [J]. Nature, 2016, 532(7600): 504-519. doi: 10.1038/nature17652 [6] Feng Jingchun, Liang Jianzhen, Zhang Si, et al. Development of deep-sea biological resources exploitation equipment [J]. Strategic Study of CAE, 2020, 22(6): 67-75. (in Chinese) [7] Simone Marini, Emanuela Fanelli, Valerio Sbragaglia, et al. Tracking fish abundance by underwater image recognition [J]. Scientific Reports, 2018, 8(1): 13748. doi: 10.1038/s41598-018-32089-8 [8] Maureen Mellody. Robust Methods for the Analysis of Images and Videos for Fisheries Stock Assessment: Summary of a Workshop [M]. Washington DC: The National Academies Press, 2014. [9] Kresimir Williams, Alex De Robertis, Zachary Berkowitz, et al. An underwater stereo-camera trap [J]. Methods in Oceanography, 2014, 11: 1-12. doi: 10.1016/j.mio.2015.01.003 [10] Jaffe J S. Underwater optical imaging: the past, the present, and the prospects [J]. IEEE Journal of Oceanic Engineering, 2015, 40(3): 683-700. doi: 10.1109/JOE.2014.2350751 [11] Lythgoe J N. Polarized light and underwater vision [J]. Nature, 1967, 213(5079): 893-894. doi: 10.1038/213893a0 [12] Keil T, Immarco A, Kerpchar M. Recnet underwater range-gated measurements [C]//SPIE, 1968, 12: 33-43. [13] Ines Dumke, Autun Purser, Yann Marcon, et al. Underwater hyperspectral imaging as an in situ taxonomic tool for deep-sea megafauna [J]. Scientific Reports, 2018, 8(1): 12860. doi: 10.1038/s41598-018-31261-4 [14] Miquel Massot-Campos, Gabriel Oliver-Codina. Optical sensors and methods for underwater 3D reconstruction [J]. Sensors, 2015, 15(112): 31525-31557. doi: 10.3390/s151229864 [15] Mariani P, Quincoces I, Haugholt K. Range-gated imaging system for underwater monitoring in ocean environment [J]. Sustainability, 2019, 11(1): 162. [16] Wang Xinwei, Sun Liang, Lei Pingshun, et al. Underwater 3D triangular range-intensity correlation imaging beyond visibility range (invited) [J]. Infrared and Laser Engineering, 2018, 47(9): 0903001. (in Chinese) [17] Oceaneering. Geoscience and AUV Surveys. [EB/OL]. [2021-04-15] https://www.oceaneering.com/survey-and-mapping/geoscience-and-auv-surveys/. [18] Wang Xinwei, Li Youfu, Zhou Yan. Triangular-range-intensity profile spatial -correlation method for 3D supper-resolution range-gated imaging [J]. Applied Optics, 2013, 52(30): 7399-7046. doi: 10.1364/AO.52.007399 [19] Wang Xinwei, Li Youfu, Zhou Yan. Multi-pulse time delay integration method for flexible 3D super-resolution range-gated imaging [J]. Optics Express, 2015, 23(6): 7820-7831. doi: 10.1364/OE.23.007820 [20] Sun Liang, Wang Xinwei, Liu Xiaoquan, et al. Lower-upper-threshold correlation for underwater range-gated imaging self-adaptive enhancement [J]. Applied Optics, 2016, 55(29): 8248-8255. doi: 10.1364/AO.55.008248 [21] Wang Minmin, Wang Xinwei, Zhang Yue, et al. Range-intensity-profile prior dehazing method for underwater range-gated imaging [J]. Optics Express, 2021, 29(5): 7630-7640. doi: 10.1364/OE.417131 [22] Wang Minmin, Wang Xinwei, Sun Liang, et al. Underwater 3D deblurring-gated range-intensity correlation imaging [J]. Optics Letters, 2020, 45(6): 1455-1458. doi: 10.1364/OL.385983 [23] Yang Yuqing, Wang Xinwei, Sun Liang, et al. Binning-based local-threshold filtering for enhancement of underwater 3D gated range-intensity correlation imaging [J]. Optics Express, 2021, 29(6): 9385-9395. doi: 10.1364/OE.420359
计量
- 文章访问数: 616
- HTML全文浏览量: 167
- 被引次数: 0