This article describes the detection of the characters of the license plate through of computer vision techniques: such as cascade of classifiers based in sobel algorithm, analysis of peaks and valleys, and support vector machines; the search for the region of the plate begins by detecting vehicles, then character segmentation and concludes with the recognition of these. The system was tested in different scenarios such as: different lighting provided by the day or night and weather changes. The average time to execute these phases it was 434 ms.