Abstract: Existing lane identification algorithm is difficult to match the image adaptively. Due to low recognition rate when the vehicle cornering,poor robustness and poor real-time,an algorithm of fast lane identification is proposed and implemented based on overall optimal threshold. The algorithm first processes image by adaptive binarization. Then,the region of interest is extracted. The method of progressive retrieval for screening feature points inside lane is proposed and real parameters of lane are obtained to reconstruct the road model.Resultsshow that being different from the previous common Hough transform,this method has better real-time performance and accuracy and can provide more useful information when the vehicle is cornering.