本文提出了一种从航空照片中提取主要道路交叉口的方法。本方法是基于道路交叉的几何—统计模型来实现自动检测的。首先检测可能存在道路交叉口的候选区域,然后再针对候选区做道路交叉口参数的提取,最后根据道路交叉口的模型进行证实。实验证明该方法具有检测准确性高、速度快、抗噪能力及鲁棒性强的优点。 This paper presents a method of extracting major road intersections from aerial photographs. This method is based on the geometry of cross-roads - statistical models to achieve automatic detection. Firstly, the candidate areas with possible road intersections are detected, and then the parameters of road intersections are extracted for the candidate areas. Finally, the model is verified according to the road intersections. Experiments show that the method has the advantages of high accuracy, fast speed, anti-noise ability and robustness.