摘要
为提高机器人自主避障的稳定性和可靠性,通过对蝗虫神经系统中具有碰撞预警能力的小叶巨大运动检测器(LGMD)神经网络进行优化处理,构建适用于嵌入式微型机器人的仿生视觉避障系统。针对LGMD网络在黑暗环境中碰撞感知性能较差,将传统图像处理算法与仿生网络相结合,通过融合拉普拉斯锐化和高斯模糊的激励来增强碰撞对象的扩展边缘,提出基于图像增强的碰撞检测神经网络(LGMD-LS)。利用MATLAB软件对模型进行视频仿真分析,结果表明:相较于LGMD模型,改进算法在黑暗环境中能有效识别迫近障碍物,具有较好的鲁棒性。在自制微型机器人上进行实物验证,结果表明:机器人在黑暗场景中能够有效避障,验证了算法的可靠性。为应用于实际场景下机器人碰撞检测提供参考依据。
In order to improve the stability and reliability of autonomous obstacle avoidance of robots,a bionic visual obstacle avoidance system suitable for embedded micro robots is constructed by optimizing the lobula giant movement detector(LGMD)neural network of locust with collision warning ability in the grasshopper neural system.Aiming at the poor collision perception performance of LGMD network in dark environment,traditional image processing algorithm is combined with the bionic network.By fusing the excitation of Laplace sharpening and Gaussian blur to enhance the extended edge of the collision object,a collision detection neural network based on image enhancement(LGMD-LS)is proposed.The video simulation analysis of the model using MATLAB software shows that compared with LGMD model,the improved algorithm can effectively identify approaching obstacles in dark environment and has better robustness.Physical verification is carried out on the self-made micro robot.The results show that the robot can effectively avoid obstacles in dark scenes,which verifies the reliability of the algorithm.It provides a reference basis for robot collision detection in practical scenes.
作者
汪杰
雷斌
蒋林
李港
苏冲
WANG Jie;LEI Bin;JIANG Lin;LI Gang;SU Chong(College of Machinery and Automation,Wuhan University of Science and Technology,Wuhan 430081,China;Institute of Robotics and Intelligent Systems,Wuhan University of Science and Technology,Wuhan 430081,China)
出处
《传感器与微系统》
CSCD
北大核心
2024年第4期53-56,60,共5页
Transducer and Microsystem Technologies
基金
国家重点研发计划资助项目(2019YFB1310000)
湖北省自然科学基金资助项目(2018CFB626)。
关键词
移动机器人
蝗虫视觉神经网络
动态避障
碰撞检测
mobile robot
locust visual neural network
dynamic obstacle avoidance
collision detection
