Speed sign recognition using artificial neural network and threshold rule for safety precautions / Mohd Azuddin Zakaria

The objective of this thesis is to build a system that able to read and extract the speed limit signs and remind the driver for safety application while driving vehicles. This research is focus on speed limits in Malaysian's highway which 90km/h and llOkm/h were selected as the sample for the input of the systems. The system consists of three processes which are image detection, image recognition and evaluation. In this thesis, the samples of speed limit signs are taken from the real scene on basis of circle shape, captured in 90° align between camera and the signs with specified distance. The image detection process is base on image processing technique including spatial image transformation, image segmentation and morphological operation. The recognition task is performed by using artificial neural network (ANN) and threshold rule to classify the speed limits type based on total white pixels of the digits. Next, the speed limit's sign is compared to the real speeds of vehicles for evaluation of allowable speed limit for driving on the highway condition. This system was developed using MATLAB 7.0. The experiment result proved the feasibility of this system.