移动机器人视觉抖动的Kalman滤波补偿

消除视觉抖动是机器人移动中对接作业的关键。基于仿射变换,建立了图像的递推运动模型;设计了基于梯度的分区域的KLT特征提取算法,分析了梯度与灰度变化的关系;利用绝对误差和最优进行特征点的匹配,并利用菱形搜索算法来提高匹配速度,设计自适应模板算法来解决匹配结果不惟一的问题;利用最小二乘法求解超定运动方程组,得到运动参数。推导得到有意运动参数的观测模型;利用Kalman滤波去除无意运动;利用滤波后的运动参数重构图像,对含抖动的视频进行稳像补偿。在非平整路面内移动机器人上开展实验。结果表明,相对参数滤波比绝对参数滤波更平滑,且算法对x和y方向的抖动补偿无相互干扰,经过该算法处理后的视频序列与原序列相比结果得到较大改善,满足准确性要求。

Vision Jitter Compensation Scheme Using Kalman Filter Method for Mobile Robot

Elimination of visual jitter plays a key role in docking operation of mobile robots.Based on affine transformation,image kinematics and recurrence relations were established.The sub-region KLT feature extraction algorithm was designed based on the pixel gradient,and relationships between gradient with pixel intensity were analyzed.Optimization of sum of absolute difference was used to match feature points.Diamond search template was adopted to improve the matching speed.To the question of matching results is not the unique,adaptive varying scale template algorithm was proposed.To obtain the motion parameters the least squares was used to solve the over-determined equation of motion.According to the observation model of intended motion,unintended motion was removed using Kalman filter to synthetic parameters including jitter.Then filtered parameters were used to reconstruct image.On the autonomous robot experiment was completed.Results show that relative parameters filter can get smoother image than absolute parameters,x and y direction filters were interference-free.Scheme proposed meet the requirement of accuracy.