同步扫描光场成像术及水下三维视觉传感应用

The underwater imaging suffers of the following problems: limited range visibility, low contrast, non uniform lighting, blurring, bright artifacts, color diminished and noise. In order to capture the 3D scene with high quality, this project explores the properties of the light in water medium and a new underwater optical method is proposed to obtain the 3D image of the scene. We call it synchronous scanner light field imaging, which projects the controlled laser with two scanners and capture all the light with multiplexing theorem. First, the property of the light field is analyzed when the projected laser varies in luminous intensity, wavelength, and polarization. Then, the multiplexing theorem of the light field is studied. A general model would be developed to depict the process of multiplexing the dimensions (such as time, space and frequency) of the plenoptic function onto an image. Based on this theory, the mechanisms of light field imaging and the principle of light field recovery will be studied respectively. After obtaining the light field, the light components are extracted, especially the direct light is identified with clustering method. Finally, the 3D imaging is reconstructed from the direct light of each laser point, and the best image quality and imaging distance can be achieved by changing the projection light source parameters to active adaptation various water environment. In a word, this project will provide a new perspective of underwater imaging and is the supplement of the traditional theory based on light scattering. A variety of innovative research results are expect and can be demonstrate superior application in related areas, such as underwater exploration, development of the oceans, and so on.

不同于以往基于二维成像的水下视觉，本项目针对水下光成像,尤其是恶劣水环境应用面临的诸多挑战，从成像区域物方所有光线构成的光场出发,研究投射光源参数改变（包括激光的光强、波长、偏振态等）对光场变化影响的规律和特点，建立光场复用和水下光场恢复的多模态方法，研究水下光场的时间、空间、频率及偏振属性和聚类分析方法。在这一基础上，为克服以往水下主动激光扫描三维成像受散射、吸收影响，导致图像质量差、成像距离短等问题，本项目研究将水下光场成像与同步扫描三维视觉传感相结合，提出并实施一种全新的同步扫描光场成像三维视觉传感技术和系统，通过对记录光线特征的聚类识别、重聚焦特征分析，实现将各种散射光与目标反射光相分离，从而提高水下三维视觉传感系统的性能。进一步，研究投射光源参数改变对水环境的主动适应机制，使得图像质量和成像距离最佳。项目研究成果可为水下探测、海洋开发利用、海洋装备升级等提供必要的基础技术支撑。

项目针对光在水下传播受到水介质及水中微粒的散射、吸收影响，导致图像质量下降等问题，将同步扫描三维成像系统与光场成像相结合，提出并实施一种全新的同步扫描光场成像三维视觉传感技术和系统。在系统构建上，针对传统的激光扫描三维成像系统，设备体积大、视场小、盲区大、功耗大等问题，提出一种同步扫描三角测距系统异形结构，构建了小型化激光同步扫描原型系统，并在这一实验系统基础上开展相关实验研究；在系统特性研究及优化上，开展同步扫描三角测距水下激光成像系统的蒙特卡洛仿真分析，探讨了目标距离、激光光强、基线距离、物镜口径等参数变化对图像对比度的影响规律及水下适应机制，并对水下激光同步扫描三角测距成像系统作了优化设计；在基于光场图像的恢复增强及清晰度提高方面，针对光在水下传输时的衰减以及后向散射带来的模糊问题，提出了水下图像暗通道恢复及剔除算法，改善了图像质量，针对图像对比度下降这一特性，提出一种基于光场前后两次重聚焦，再作差分处理的散射背景光分离算法及计算机自动判定流程，提高了成像系统接收到反射光点的对比度；在应用拓展方面，针对深海生物原位观测需要，利用光场成像相关理论，提出水下线激光动态扫描三维成像方法，进一步研究和解决一系列关键技术问题，主要包括：从光场理论出发对水下双目立体成像理论及激光扫描技术系统进行了重构优化；对由于受到海水颠簸带来的被测目标与扫描系统之间相对运动造成的图像重建困难，提出了一种动态条件下基于视场中场景特征点的误差补偿方法，实现水下激光扫描系统的高精度动态测量。经实验室精度试验，系统达到水下三维重建精度（标准差）为2.28mm；经南海真实海域实测试验，表明样机具有深海原位三维测量重建功能,在距离样机2m处,对标准球棒两球心距测量结果的标准差为 2.22mm。该项目的研究成果有望为深海生物学、生态学研究提供迫切需要的观测手段。