摘要
道路检测是智能车辆研究的重要内容.道路检测往往受到阴影遮挡、噪声、道路边界或标志线不连续等影响,这些问题导致传统的道路检测算法鲁棒性较差.针对此类问题,本文提出了一种基于变形模板的道路检测方法.首先,对道路边界或标志线进行边缘提取,然后构造了参数可变的道路模型及其同道路边缘轮廓之间的似然函数,最后通过遗传算法搜寻该函数的全局最大值,从而获得参数可变道路模型的最佳参数,仿真实验表明,本文提出的算法对阴影遮挡、噪声、道路边界或标志线不连续等问题均有较好的检测效果.
Road recognition is an important task in intelligent vehicle research. Road image is usually influenced by
shadows, noise and discontinuity of road edge or marked line, and this induces the traditional edge-based algorithm's
robustness decreases greatly. To avoid negative influence from threshold selection, a deformable template based
road recognition algorithm is presented in this paper. Firstly, preprocess road image with edge operator to get
the edge information, then construct a deformable template medel of road contour and its likelihood function
which define the fitting degree for a given group of deformable template parameters, finally genetic algorithm is
used to search the global optimal value of the likelihood function to get the optimal parameter of the deforable
road model. Experimental results indicate that the algorithm has strong robustness for shadows, noise and
discontinuity of road edge or marked line.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第2期156-160,共5页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金优秀创新群体资助项目(No.60024301)
关键词
道路检测
变形模板
似然函数
遗传算法
Road Detection
Deformable Template
Likelihood Function
Genetic Algorithm
