Real-Time Video For Road Lane Marker (Urban Type) Using Image Processing

Liang, Jin Chuan (2018) Real-Time Video For Road Lane Marker (Urban Type) Using Image Processing. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Real-Time Video For Road Lane Marker (Urban Type) Using Image Processing.pdf - Submitted Version Download (712kB)

Abstract

Road markers provide road information to driver to ensure road and their safety. Different type of markers indicate different kind of information. For example, double lane marker indicates that drivers are not allowed to overtake due to dangerous road condition. It can lead to road accident if the drivers do not follow the rule of the road marker or the road markers are not seen clearly by the drivers. Nowadays, road accidents become a concerning issue all over the world. In order to avoid the tragedy continues to happen, vision based system is developed to detect and classify those markers. Besides, the existing features are insufficient to allow a robust recognition. In this report, real-time video of three types of marker is developed using image processing and artificial neural network (ANN). The video is recorded in urban area at afternoon. The image processing and classification are done using MATLAB software. The image processing techniques include grayscale conversion, image sharpening, noise removal, cropping, resizing and feature extraction. The features selected are the length of the feature vector on HOG and LBP. These feature vectors are later to be trained using neural network pattern recognition tool, to classify the data getting from the feature. The validation for the sample is 15% while the testing for the sample is 15% and the rest is used for training. The result shows an accuracy of 99.4% with HOG and LBP features as input vectors.

Actions (login required)

View Item