APLIKASI PENGENALAN RAMBU LALU LINTAS MENGGUNAKAN PRINCIPAL COMPONENT ANALYSIS (PCA) DAN LEARNING VECTOR QUANTIZATION (LVQ)

Abstract

Traffic signs are one of the instruments in traffic road which give instruction to the driver in order to create safety and discipline on the road. The driver knowledge about traffic signs that still low can cause accident on the road that cause casualties. The purpose of this research is to make an application to recognize traffic signs using extraction feature of Principal Component Analysis and artificial neural network Learning Vector Quantization. The traffic sign recognition application was developed using preprocessing on the initial image with thresholding method using HSV Color Space to determine color of the traffic image, then do segmentation process using Blob Counter to determine the Region of Interest. The feature of initial image that had been through the process of pre-processing would be extracted using Principal Component Analysis. Feature of sign image was processed using Learning Vector Quantization as an algorithm for training and testing. The result of implementation was an application that can identify five kinds of traffic signs, i.e. no parking, stop, winding road, caution, go left. The testing used 50 images as data with the distribution of data used 10-Fold Cross Validation. The traffic sign recognition application that has been built showed that optimum LVQ configuration is 1x10 -3 as learning rate and 1x10 -5 as epsilon, the result of this configuration produces an average accuracy of 86% and average error rate of 14%.