Vehicle Detection Based on Convolutional Neural Networks
Accurate vehicle detection or classification plays an important role in Intelligent Transportations Systems. Ability to detect vehicles in traffic scenes allows analyzing drivers' behavior as well as detect traffic offenses and accidents. Detection and classification of vehicles is a challenging task due to weather and light conditions and vehicle type diversity. Several solutions use feature extraction algorithms along with support vector machine classifier. However, convolutional neural networks have proved to be potentially more effective. In this thesis, we present a convolutional neural network trained to classify and detect vehicles from multiple angles. Moreover, Fast Fourier Transform is used during data preprocessing. The effect of such preprocessing is examined on the developed vehicle classifier and detector.
Graduation Thesis language
Graduation Thesis type
Master - Computer Science