【作者】 朱昊； 黄源水； 付梦印；

【Author】 ZHU Hao,HUANG Yuan-shui,FU Meng-yin (School of Automation,Beijing Institute of Technology,Beijing 100081,China)

【机构】 北京理工大学自动化学院；

【摘要】 视觉导航是地面无人车辆导航的一种重要手段,对非结构化道路进行识别,可以刻画无人车辆的可通行区域,使其安全完成导航任务。通过特征提取算法,从图像中获得可高效区分不同区域的特征,进而采用分类算法完成识别的功能。将一种基于非负矩阵分解的特征提取算法应用于本领域中,利用非负矩阵分解提取非结构化道路图像特征,并通过支持向量机进行识别。实验结果表明,该算法提取的特征比传统的颜色、纹理特征分类效果更优。更多还原

【Abstract】 Visual navigation is an important means of navigation for unmanned ground vehicles(UGV).The identification of unstructured roads can help the unmanned vehicles describe the path that make it safe to complete navigation tasks.Through the feature extraction algorithm,obtain the features that can best distinguish different regions,then use classification algorithm to complete the identification function.Use a nonnegative matrix factorization (NMF) based feature extraction algorithm to select the features from unstructured road images,then use support vector machines for identification.Experimental results show that the extracted features display a better performance than the traditional color\texture features更多还原