ABSTRACT
This paper studies the machine learning sign recognition based on the adaptive algorithm. In allusion to the problem that the contrast degree of lane image obtained at night is low, a night-time lane mark identification algorithm based on adaptive threshold segmentation is put forward. First of all, the preprocessed image is divided into blocks to enhance the edge information. In combination with the Otsu threshold method and the neighbourhood value method, the adaptive threshold is obtained by weighted distribution, and the lane image is segmented. Finally, partition domain search and Hough transformation are used to extract lane marking line accurately. The field test shows that, as for lane line of structured road, the lane line extraction method is highly accurate and provides good in real-time performance.
