桥梁巡检机器人实时稳像算法研究  

Research on real-time image stabilization algorithm for bridge inspection robots

作　　者：张浩然 ZHANG Haoran(Research Institute of Highway,Beijing 100088 China)

机构地区：[1]交通运输部公路科学研究院,北京100088

出　　处：《山东交通科技》2024年第6期70-75,共6页

摘　　要：针对桥梁巡检机器人在桥梁上自主移动过程中受桥梁振动影响机器视觉稳定问题,提出一种融合阈值滤波和RANSAC的全局运动估计快速算法。首先,检测图像中Harris特征点并将其稀疏化,跟踪特征点坐标并计算位移矢量,通过融合算法估计最佳运动参数,最后利用Kalman算法对全局运动中的高频抖动进行补偿完成稳像。试验结果表明,该算法能快速处理包含大量局部运动的桥面场景,在分辨率为640×360情况下可实现30 fps实时稳像。A global motion estimation algorithm combining threshold filtering and RANSAC is proposed for bridge inspection robots.Firstly,detect Harris feature points in the image and sparsify them.Track the coordinates of feature points and calculate the displacement vector.The optimal motion parameters are estimated through fusion algorithms.Finally,the Kalman algorithm is used to compensate for high-frequency jitter in global motion and complete image stabilization.The experimental results show that the algorithm can quickly process bridge deck scenes containing a large amount of local motion,and can achieve real-time image stabilization at 30 fps with a resolution of 640×360.

关 键 词：电子稳像 运动估计 HARRIS角点 桥梁巡检机器人 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]