论文部分内容阅读
非结构化道路环境复杂多变,但是路两旁的植被较为显著,可用于限定路面的不可通行区域.在复杂的室外环境中,植被区域容易受到天气、阴影、路况等多种因素干扰而产生误检.为此本文提出了一种基于高斯核SVM(支持向量机)的植被检测方法,通过基于超像素的稀疏表示法来分析并学习样本多维色彩空间特征,进而构造分类准则有效获取植被信息,并采用栅格概率滤波来优化检测结果,提高检测精度.实验表明,该方法很好地解决了非结构化道路环境中的植被检测问题,对光照、路况等变化也具有较强的抗干扰能力,且具备较好的实时性和可靠性.在实际应用中,有效地限制了路面的不可通行区域,保障了移动智能机器人在复杂道路环境中的安全行驶区域.
The unstructured road environment is complicated and changeable, but the vegetation on both sides of the road is more obvious and can be used to restrict the unacceptable area of the road surface. In complicated outdoor environment, the vegetation area is easily disturbed by many factors such as the weather, shadow and road conditions In this paper, we propose a vegetation detection method based on Gaussian kernel SVM (Support Vector Machine), analyze and learn the multidimensional color space features by means of superpixel-based sparse representation, and then construct classification rules to effectively obtain vegetation information, And raster probability filter is used to optimize the detection results and improve the detection accuracy.The experiment shows that this method can well solve the problem of vegetation detection in unstructured road environment and also has strong anti-interference ability to light, road conditions and other changes , And has good real-time and reliability.In practical application, the unacceptable area of the road surface is effectively limited, and the safe driving area of the mobile intelligent robot in the complex road environment is ensured.