A traffic light automatic recognition method based on machine learning

The invention provides a method for realizing automatic identification of traffic lights based on machine learning, which comprises the following steps: converting RGB value of an image into HSV value, and dividing the traffic light colors into yellow, red and green according to the color space of HSV; A region of interest is determined according to the initial lamp group information to binarize the image, obtaining a connected region from the binarized region of interest of the image, and screening out candidate regions from the connected regions, wherein the connected region is a region of the same color surrounded by a closed curve, determining a permutation and combination of traffic lights according to the candidate area, determining lamp group information of a current image frame, the lamp group information including position coordinates, color information and brightness information of a traffic signal lamp group, and providing the lamp group information to an image of a next frame for use. The method r