Development of a speed restriction road sign recognition system for motor vehicles.

Abstract

Road transport being a major means of transport used in Uganda at large and the world across, faces the problem of road accidents which are a major hazard on the road. Uganda still suffers these road accidents as it’s an aspirant of development and these claim lives. The major cause of these road accidents is the speeding problem, negligence and carelessness while driving on the road. To cub some of these road accidents, road signs are put in place on the road to enhance the safety of drivers and all road users. Speed limit road signs here limit the speed of the driver on certain road stretches, however more so at times are missed by drivers. According to the research made during this project, some drivers in Uganda are most times not aware of the existence of the speed restriction road signs. Some of them it is after being warned by other drivers to reduce speed because ahead of them lies traffic police officers arresting and charging speeding vehicles in offensive of the speed limit on the road stretch they are driving, however when they are asked of the speed limit they are driving at, they are not aware what the speed limit should be. This project aims to inform the drivers and increase their awareness of the speed limit road signs on the roads in Uganda. This is achieved in a way that a system is setup to read and inform drivers of the speed limit road signs and this consists of; - raspberry Pi 3B, Raspberry Pi camera and LCD screen as the hard ware components of the system installed in the car. However the system functional performance is based on the software developed to activate and make the hardware components work and also to perform some processes on the input so as to give an output. In the system that is setup, the hardware components like the raspberry Pi 3B is used as the minicomputer to work as the processing unit and the raspberry Pi camera is used for computer vision as the eye of the system to be able to capture images of the speed limit road signs. The LCD is used to display the output of the system for communication of feed back to the driver. The system software developed involved different stages and processes. It involved image processing techniques like gray scaling of images (converting images from colour BGR to black and white colour). It also involved image detection techniques like contour detection and Hough circle detection. The other process involved in the system is machine learning where the system is taught using examples to recognize the unseen before road sign images. The examples which are termed as a dataset was developed that contains over 1300 positive images and 1200 negative images. These were used in the Haar cascade classifier to train the system to recognise other speed limit road signs. The system performance stages include, the input as an image and if the image is a speed limit road sign image it is taken through the image processing techniques and detection stages. The classifier is then applied to the image to identify the speed limit and then it is displayed on the screen to inform the driver of the actual recommended speed limit on that specific road stretch. The system was tested of its performance with a sample speed limit road signs like 50, 70 and 80km/h at a distance range of 2 to 3meters. The system performance accuracy was noted on various speed limit road signs, 50km/h recognition was 100%, 70km/h recognition was 85%, and 80km/h recognition was 60%.