相控阵仿生激光三维成像方法研究

With the advantages of rich information, good anti-interference, etc., three-dimensional (3D) Light Detection and Ranging (LiDAR) has been widely used in the military and civil fields, such as remoting sensing, visual tracking, geomorphic mapping, unmanned vehicle and artificial intelligence. However, current 3D LiDAR systems, including scan and non-scan types, the performances cannot simultaneously meet the requirements from different applications. The issues are embodied in two aspects, namely, (i) the trade-off between high resolution and real-time, and (ii) ranging accuracy. Therefore, it is necessary to develop highly performances of 3D LiDAR. .For issue (i), we propose a novel 3D imaging mechanism based on the features of a bionic eye and the advantages of optical phased array (OPA). The remarkable features of retina-like imaging are space-variant sampling and visual compressed transmission. Combined the advantages of flexible beam direction with high efficiency of OPA, we propose a retina-like 3D LiDAR by the use of liquid OPA. We achieve foveated 3D image with high resolution, but low resolution for the periphery, so the merits of both high resolution and real-time are simultaneously achieved. For issue (ii), low ranging accuracy exists in conventional 3D LiDAR, because current ranging discriminating methods are based on unipolar pulsed echoes which are easily affected by the factors such as noise and pulses broadening and distortion, etc. For example, the peak discrimination in unipolar pulsed echoes is low sensitivity, i.e. slowing changing rate around the peak, which leads to low ranging accuracy because the pulse broadening and fluctuation. Here, we propose a ranging discriminating method to obtain bipolar pulsed echoes by the use of differential-optical path method. The common noise including background noise and low-frequency noise is suppressed. Moreover, high sensitivity is obtained by changing peak discrimination into discriminating zero-cross point, which leads to high ranging accuracy...The prototype is developed by studying the contents above. We carry out comparative experiments to test theoretical analysis and demonstrate the advantages of the proposed method, which paves the way for designing a high performance of 3D LiDAR.

三维成像激光雷达因获取信息丰富、抗干扰性好等优势广泛应用于地貌勘测、无人驾驶、人工智能等军民领域。现有三维成像方法受成像体制、噪声、探测器件等因素制约，难以满足诸多领域的应用需求。集中体现在两方面：（1）高分辨与实时性难以兼顾；（2）测距精度亟待提高。.针对问题（1），本项目从成像体制源头创新，将相控阵光束指向灵活、高效的优势与视网膜非均匀采样相结合，提出相控阵仿生三维成像方法，通过研究并优化变分辨率扫描三维成像模型，达到对感兴趣目标高分辨成像，周边压缩的目的，从而兼顾高分辨与实时性；针对问题（2），提出双光路差分测距方法，将单极性回波峰值检测，转变为双极性过零点检测，大幅提高了时刻鉴别灵敏度，有效解决传统测距方法因信噪比低下而导致难以提高测距精度的问题。.构建相控阵仿生三维成像系统，形成从理论分析到试验测试的闭环验证。为实现高分辨、高精度实时的三维成像激光雷达奠定理论基础与技术支撑。

随着科学技术的飞速发展，激光三维成像由于获取目标信息更为丰富、分辨 率高等优势，广泛应用于遥感测量、生物医疗、人工智能等军用与民用领域。目前，三维成像激光雷达正朝着远距离、高分辨率、高实时性与轻量化的方向发展。但现有三维成像激光雷达受成像体制的制约，导致分辨率、实时性和测距精度成为制约激光三维成像综合性能难以提高的关键瓶颈。.本项目通过研究人眼视网膜传输机理，从仿人眼出发，建立基于仿人眼成像机理的激光三维成像模型。提出了一种基于液晶相控阵的光束扫描方法，实现变分辨率成像；同时，提出基于液体透镜的双光路差分高精度测距方法，通过优化光路差分结构，实现对回波的高精度时刻鉴别。解决了传统激光三维成像无法同时实现大视场和高分辨率的高精度成像问题。通过理论与试验验证，构建了相控阵仿生激光三维成像及优化设计理论，实现了集大视场、高分辨率、高精度于一体的相控阵仿生成像。通过完成本项目，不仅为相控阵仿生激光三维成像系统的设计提供了理论基础与试验依据，同时为解决诸多基础学科对光电成像综合性能不佳的共性问题提供一种设计借鉴。