基于视觉图像边缘特征的灌木丛或高草识别

灌木丛或高草的枝叶外扩,使得其灰度图像边缘呈"锯齿"状,故边缘点曲率值变化频率及变化范围较大,可选择边缘曲率值的变化特征描述该类障碍物的边界特征.首先,利用Fisher 线性分类器对原始灰度图像进行分割,对二值化图像去除孤立点、进行形态学膨胀处理和空穴区域填充;其次,利用LOG算子及细化算法提取出障碍物的单像素边缘;再...

Shrub or herb identification based on visual image edge feature

The edge of shrub or herb image is irregular since their leaves and branches are always protuberant.In this case,the obstacle′s edge will illustrate toothed after segmenting their monochrome images.It leads that the edge′s curvature has greatly fluctuant range in principle.Shrub or herb is identified based on edge-shape-factor,which is measured with edge′s curvature.Firstly,based on Fisher rules monochrome image is segmented applying morphological dilation and region growing separately after noise removing.Secondly,using LOG operator and thinning algorithm single-pixel edge of obstacle is obtained.Thirdly,after acquiring the chain code and curvature of obstacle′s every pixel,edge-shape-factor can be computed according to formula.Lastly,large amount of samples′training illustrates that the edge-shape-factor of shrub or herb ranges from 0.049 6to 0.076 2.Based on this method,shrub or herb is identified,which has good real-time ability and robustness.