论文部分内容阅读
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。
Eliminating visual jitter is the key to docking operations in robotics. Based on affine transformation, a recursive model of image is established. A KLT feature extraction algorithm based on gradient sub-region is designed. The relationship between gradient and gray level is analyzed. The feature points are matched by absolute error and optimality. The diamond search algorithm is used to improve the matching speed. An adaptive template algorithm is designed to solve the problem that the matching result is not unique. The least square method is used to solve the over-determined equations to obtain the motion parameters. The observation model of intentional motion parameters is deduced. Kalman filtering is used to remove unintentional motion. The filtered motion parameters are used to reconstruct the image to stabilize the video with jitter. Experiments on mobile robots in non-flat pavements. The results show that the relative parameter filtering is smoother than the absolute parameter filtering, and the algorithm has no mutual interference to the jitter compensation in the x and y directions. The video sequence processed by the algorithm is greatly improved compared with the original sequence to meet the accuracy requirements .