【摘要】 应用图像低级特征提取技术中的边缘检测技术能够确定和提取车道线边界信息，为后续完成道路识别、实现辅助驾驶系统等提供了重要依据和基础。传统sobel算子进行边缘检测虽然算法简单、速度快，能满足道路图像边缘检测实时性的要求，但由于只获取了水平和垂直两个方向的边缘梯度信息，且阈值选取全靠人为设置，因此边缘信息检测效果不理想。文中提出改进sobel算子的边缘检测算法，算法通过扩大构造5×5的梯度权值模板计算出图像边缘信息，应用K-means聚类算法获取的最佳自适应阈值完成图像分割，最后通过改进的边缘细化算法，保留了更完整有效的边缘信息。通过实验表明，该算法能够较快、较好的提取道路图像边缘信息，而且具有一定的抗干扰能力。更多还原

【Abstract】 The application of edge detection technology in image low-level feature extraction technology can determine and extract lane line boundary information, which provides an essential basis and foundation for the subsequent completion of road recognition and realization of an auxiliary driving system. The traditional Sobel operator for edge detection is simple and fast, and it can meet the real-time requirements of road image edge detection. However, the edge information detection effect could be better because only the horizontal and vertical edge gradient information is obtained, and the threshold selection is all manually set. This paper proposes an edge detection algorithm based on an improved Sobel operator. The algorithm calculates the image edge information constructing the gradient weight template of 5×5, uses the best adaptive threshold obtained by K-means clustering algorithm to complete image segmentation, and finally retains more complete and adequate edge information through the improved edge thinning algorithm. The experiment shows that the algorithm can quickly and well extract the edge information of road images and has a specific anti-interference ability.更多还原