中文摘要：

针对传统非结构化道路检测算法在光照变化、阴影及水迹等干扰因素下不能同时满足鲁棒性与实时性的问题，提出一种结合图像灰度特征和道路模型的非结构化道路检测方法。通过二维最大熵分类对图像进行初步分割，采用模糊熵对错分点进行重新分类，优化分割效果。利用改进分块分类方法对分割图像进行分块，得到包含道路边界的混合区域，进而快速准确提取道路边界点。采用实时性较好的二次曲线模型，在最小二乘法拟合曲线的基础上提出改进的拟合方法，最大程度降低了干扰点对曲线拟合精度的影响。仿真结果表明，改进的方法不仅实时性好，检测精度高，且鲁棒性较强。

英文摘要：

Traditional unstructured road detection algorithm cannot meet the require of robustness and real-time under the factors such as illumination change, shadow and water stain. To solve this problem, a detection algorithm based on gray feature and road model was proposed in the paper. We used 2-D maximum entropy segment road image and fuzzy entropy to reclassify the wrong equinox and optimize segmentation results. Then, the segmented image was divided into blocks with the improved block classification method and the mixed area including road boundary was obtained, which makes the extraction of road boundary points quickly and accurately. A real-time quadratic curve model was choosen, and an improved method of fitting was put forward on the basis of least squares curve fitting, which can reduce the effect of interference on the precision of curve fitting. Simulation results show that the improved method has good real-time performance, high detection precision and strong robustness.