基于LGMD网络的低成本视觉系统设计与实现

Design and realization of low-cost visual system based on LGMD network

为了进一步降低视觉系统在微机器人上应用的成本,提出了一种基于STM32F103的微型机器人嵌入式仿生视觉系统。视觉系统的灵感取自蝗虫的小叶巨型运动探测器(LGMD)视觉网络,研究表明,该网络对快速接近的物体有准确的响应,能够有效引导微机器人在动态环境下自主避碰。通过降低LGMD网络的灰度等级,减小输入图像大小,使其能够应用在计算资源有限的嵌入式微系统上。通过离线测试验证:该系统的可靠性和鲁棒性,可应用于群机器人的自主避碰。

In order to further reduce the application cost of vision system in micro-robot, an embedded bionic vision system of micro-robot based on STM32 F103 is proposed.The vision system is inspired by the lobula giant movement detector(LGMD)vision network of locusts.Research shows that this network has accurate response to rapidly approaching objects and can effectively guide the micro-robot to avoid collisions autonomally in dynamic environment.By reducing the gray level of the LGMD network and reducing the size of the input image, it can be applied to the embedded microsystem with limited computing resources.The reliability and robustness of the system are verified by off-line test, and it can be applied to autonomous collision avoidance of swarm robotics.