基于全景视觉的光伏可调支架清扫机器人导航方法  

Navigation Method of Photovoltaic Adjustable Support Cleaning Robot Based on Panoramic Vision

作　　者：刘国能 王旭花 刘拓 原坤 雷江 LIU Guoneng;WANG Xuhua;LIU Tuo;YUAN Kun;LEI Jiang(Weinan Branch of State Power Investment Shaanxi New Energy Co.,Ltd.,Weinan 714000,China)

机构地区：[1]国家电投集团陕西新能源有限公司渭南分公司,陕西渭南714000

出　　处：《微型电脑应用》2025年第2期162-166,170,共6页Microcomputer Applications

摘　　要：针对光伏可调支架清扫机器人导航能力不足的问题,提出一种基于全景视觉的光伏可调支架清扫机器人导航方法。通过模型预测控制(MPC)技术设计防偏移器,通过分析清扫车实际运行过程中的状态构建方程,建立模型并计算相对应的预测值,从而引导清扫车移动,减少导航偏移现象发生,提高定位精度。在尺度不变特征变换(SIFT)算法的基础上进行改进,引入最小核值相似区优化算法,计算各点的响应值,优化全景视觉图像,提高清扫车导航定位的效率。实验结果验证,所提方法导航定位时间仅为传统SIFT算法的18.38%,对偏移角度的修正效率提高了39.29个百分点。所提方法使清扫车导航定位时的运动状态更加稳定,提升了导航精度,有效提高了清扫机器人的导航能力。To solve the problem of insufficient navigation ability of photovoltaic adjustable support cleaning robot,a navigation method of photovoltaic adjustable support cleaning robot based on panoramic vision is proposed.The anti-drift device is designed by model-based predictive control(MPC)technology,equations are constructed by analyzing the state of the sweeper during its actual operation,a model is built and the corresponding predicted values are calculated,thereby guiding the sweeper to move,reducing the occurrence of navigation drift,and improving the positioning accuracy.Based on scale-invariant feature transform(SIFT),the minimum kernel similar region optimization algorithm is introduced to calculate the response value of each point,optimize the panoramic vision image,and improve the efficiency of the navigation and positioning of the sweeper.Experimental results show that the navigation and positioning time of the proposed method is only 18.38%of that of the traditional SIFT algorithm.The correction efficiency of drift angle is increased by 39.29 percentage points.The proposed method makes the moving state of the sweeper more stable,improves the navigation accuracy,and effectively improves the navigation ability of the cleaning robot.

关 键 词：全景视觉 光伏组件 清扫机器人 MPC SIFT 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]