用于车辆自主驾驶的多线激光雷达成像机理及关键技术研究

Line detection laser radar is an essential 3-D information imaging sensor in the field of intelligent robot,unmanned vehicle and digital city. The number of detecting lines of this kind of laser radar developed by the foreign countries has reached up to 64, but there are no relative reports on the studies of line detection laser radar for unmanned vehicle in china. A new scheme was proposed to design the line detection laser radar, that the output of high-frequency all-solid-state lasers is reshaped with special optical system to generate the irradiating source, and apd-array is combined with the joint technology of field of view to realize the multi-line detection. Its outstanding advantage is that the number of detection lines can be set according to the requirement and the maximum is 128. We will develop the design,analysis and simulation system of line detection laser radar based on its imaging mechanism. An new multi-line detection laser radar used in umanned vehicle will be developed after solving following key technolgies,such as line-laser uniformly irradiation, apd-array's automatic gain control and balance,timing and synchronization with high precesion,3-D scene reconstruction based on multi-line detection. At last, we will do the outdoor experimets of the multi-line laser radar to provide a theoretical and technological basis for realizing the nationalization multi-line laser radar with proprietary intellectual property rights.

线探测激光雷达是智能机器人、自主驾驶以及数字化城市等必备的三维立体信息感知设备。国外该类激光雷达的多线探测数目已经达到了64线，国内尚无同类研究的报道，急需打破国外技术封锁。本项目提出以高重频新型全固态激光器的输出经特殊整形设计作为照明光源，多元线阵雪崩光电探测器结合视场拼接实现多线接收探测，形成了一种与国外现有车载激光雷达不同的新颖技术途径，其突出优势是可按需要设定探测线数，最多可扩展至128线。本研究将在深入分析多线探测激光雷达成像机理基础上，建立该类激光雷达设计、分析和仿真平台，通过突破高重频线激光照明的均匀化方法、多元线阵雪崩光电探测的自动增益控制和均衡方法、高精度计时与同步技术、以及基于多线探测的三维场景重构等关键技术，研制出能用于自主驾驶平台的新型多线探测激光雷达样机，并进行外场实验验证，为最终实现具有自主知识产权的多线探测激光雷达的国产化奠定坚实的理论和技术基础。

多线激光雷达是智能机器人视觉、自主驾驶导航以及数字化城市测绘等领域必备的三维立体信息感知设备。多年来该类雷达一直为国外产品所垄断，国内迫切需要开展多线探测激光雷达的成像机理研究，进而攻克各项关键技术，为实现具有自主知识产权的多线探测激光雷达的国产化奠定坚实的理论和技术基础。. 本项目在理论上取得的创新性成果包括：利用Cramer-Rao理论，推导了高斯脉冲、倒置抛物线脉冲和四次方脉冲模型下激光雷达测距精度极限的理论解；引入Monte Carlo方法，建立了多线激光雷达测距成像的数字化全波形仿真模型；以随机介质中的光传输理论为基础，得到了雾霾条件下激光雷达回波信号及测距数据分布的解析解；从统计光学出发，建立了光子计数激光雷达测距数据的概率模型，揭示了光脉冲宽度、回波光子数和死区时间等与测距数据统计分布之间的依赖关系；基于各向异性总变分和稀疏图像分解方法，设计了高速运动条件下图像运动模糊的复原算法；上述理论研究工作均得到了充分的实验验证，为多线激光雷达参数设计、性能评估及环境适应性分析提供了有效手段。. 本项目在执行过程中，还相继攻克了线阵激光照明、多通道高精度时间-数字转换、线阵雪崩探测器低噪声增益平衡驱动和多线探测三维场景重构等多项关键技术，研制了12线、16线和32线测距模块，以研制出的12线激光雷达原理样机做性能测试试验，实现作用距离＞100米，单线测频达到10kHz，测距精度优于5cm，且通过模块拼接，最大线数可扩展到64线。此外，将高重频测距与二维振镜扫描相结合，以单线扫描代替多线探测，形成了一种与现有车载激光雷达完全不同的创新方案，经实验验证，该方案的激光雷达不但可以达到128线以上的等效探测线数，而且其水平和俯仰视场以及探测线数均可实时电控，且具有低成本、凝视成像、可实现嵌入式车体安装等优势，为新一代车载激光雷达提供了一种极具前景的解决方案。