TRAFFIC LIGHT ARROW SHAPE RECOGNITION USING HOG DESCRIPTORS AND SVM CLASSIFIER

Abstract – Autonomous intelligent vehicles are keep expanding under a real development in this era after many researches have been made to implement a better safety features in a transportation sector. In this paper, an algorithm of recognition of the traffic light arrow in the daytime specifically for the operation of the video system has been proposed using the technology of the image processing and machine learning. In the machine learning technique, the Support Vector Machine, SVM classifier has been used for the learning process and conduct a classification process as well. Firstly, the images are converted to Hue Saturation Value (HSV) color space while the color thresholding has been used to obtain the region of interest for the traffic light arrow in the image. In the learning process, each of the traffic light arrows in the dataset will go through the pre-processing phase, detection phase before going through the hog descriptor for feature extraction purpose. Prior to send the image to the SVM, the hog descriptor process is to extract the features for each of the shape of the traffic light arrow. As a result, this algorithm has achieved 98.52% accuracy, 1.48% error, 98.52% precision and 98.68% of recall through the testing process.