Abstract: Taking into account the relatively poor versatility and effectiveness of the general robot visual navigation filtering algorithm, and taking advantage of the existing the median filter and multi-scale adaptive fusion filter, a combinatorial optimization filter based on the median and multi-scale was proposed. Firstly, the algorithm uses the median filter, and then uses the multi-scale filter. Finally the filtered results are fused according to smallest mean absolute error criterion(MAE). Experimental results show that the proposed algorithm can better filter out a variety of common noise for the robot road images, which enhances the versatility and effectiveness of the algorithm.

Key words: Mean Absolute Error (MAE), median filter, multi-scale, combinational optimization

摘要： 考虑到一般机器人视觉导航滤波算法的通用性和有效性比较差的缺点，结合中值滤波器和多尺度自适应融合滤波器这两种算法的优点，提出一种基于中值和多尺度的组合优化滤波器。算法首先应用中值滤波，然后进行多尺度滤波处理,最后根据最小平均绝对误差准则，将滤波后结果进行融合处理。实验结果表明：该算法能很好地滤除机器人道路图像各种常见的噪声，从而提高算法的有效性和通用性。

关键词: 平均绝对误差, 中值滤波器, 多尺度, 组合优化