Research on traffic signal image processing technology for autonomous vehicles

The progression of scientific innovation has driven evolution of the automotive sector, and autonomous driving has gradually gained favor with the public. Visual recognition is a key issue for the safe operation of autonomous vehicles, which involves identifying vehicles, pedestrians, traffic lights, traffic signs, and other road information to control the vehicle’s safe movement in accordance with traffic signage requirements. This paper investigates an image processing algorithm for traffic signal lights, importing RGB (Red and Green and Blue) images recognized by the in-vehicle camera into image processing software in real-time. The algorithm converts RGB area photographs to HSV house images, obtaining threshold values of the HSV image components. Based on the standard ranges for red, green, and yellow thresholds, the Length feature is utilized to remember the quantity of red, green, and yellow pixels, thereby figuring out the shade of the site visitors sign light. The color parts of the visitors sign light images are subjected to binarization, dilation, erosion, and closing operations to process the image. The main subject of the processed image is extracted, and constraints are applied to the size of the connected pixels to segment and filter out the signal light portion. The numbers and directions in the photos are identified, and the final results are outputted.