平滑去噪分块K-means算法的机器人视觉图像处理

A Robot Vision Image Processing Method Based on Smoothing and Segment K-means Algorithm

针对机器人视觉目标图像信噪比低、背景噪声干扰大的特点,采用马尔科夫随机场(Markov Random Field,MRF)模型的平滑去噪方法对图像进行预处理。在此基础上,采用K-means聚类算法对图像进行聚类,将具有不同特征的目标区域分类,为进一步实现目标识别和跟踪提供基础。同时,为进一步克服移动机器人导航过程中视觉处理速度慢的缺陷,对图像进行分块划分,提取每个图像块的均值、方差和最大值作为特征值,从而提高算法的处理速度。

The MRF（Markov Random Field）-based smoothing is introduced into the pre-processing of robot vision target images which typically have strong background noise and low signal to noise ratio. The segmentation of the pre-processed image is performed by means of the K-means clustering algorithm, which prepares the image for target recognition and tracking. Besides, in order to accelerate the visual processing speed during mobile robot navigation, the image is divided into sections, and the variance and the maximum of each image section are extracted as the features for k-means clustering. The experiments with standard images and actual images are carried out, which prove that the algorithm has better recognition ability, higher detection accuracy and faster detection speed.