本项目属机器视觉和多普勒雷达技术的交叉领域，拟研究一种基于电磁波多普勒效应、仅对运动目标敏感的新概念人工“多普勒视觉”。本项目以获取适用于人工智能应用场景和多样性需求的无线运动感知为研究目标，研究基于单频连续波体制的微型多普勒雷达传感器分布式组网架构，以可扩展、无连接的硬件系统和实时、集成的软件算法，实现近距离、三维、多个独立运动目标的运动分离、定位、追踪和运动模式识别，并研制原理样机，使之具备实时输出坐标空间内多维、多目标运动数字信息的能力。与基于计算视频和图像技术的传统机器视觉相比，基于电磁波多普勒效应的“多普勒视觉”在运动感知灵敏度、运动分离、算法简便和计算容量等方面具有内禀的优势，可以不受光照和气象条件影响“全天候”工作，具有重要的科学研究价值和工程实用意义。

This project belongs to an interdisciplinary area of machine vision and Doppler radar technique, aiming to investigate a new-concept artificial "Doppler vision" that is only sensitive to moving objects based on Doppler effect of electromagnetic waves. Targeting to wireless motion sensing applicable to application scenarios and diverse requirements of artificial intelligence, this research focuses on the separation, localization, tracing and pattern recognition of short-range, three-dimensional, multiple independent motions based on extendable, wireless connected, distributed network architecture consisting of miniaturized continuous-wave Doppler radar sensors and real-time, integrated algorithms. Experimental prototype will be developed to demonstrate the capability of real-time output of the digital information of multi-dimensional, multiple motions in specified coordinate space. Compared with conventional machine vision, the Doppler effect based "Doppler vision" has specific advantages in motion sensitivity, separation, algorithms and computing volume, and can work in "all-weather" without being impacted by illumination and weather conditions. These make this research have important values in both scientific research and engineering applications.