Traffic Density Estimation Based on Block Occupancy Classification

High levels of traffic density cause negative impacts in society. Due to this
problem, the authors of this research proposed a method to estimate traffic density. The
proposed method was implemented on Junction 1 dataset. We utilized Local Binary Patterns
(LBP), Histogram of Oriented Gradients (HOG), and Gray-Level Cooccurrence
Matrix (GLCM) for extracting image features, and Support Vector Machine (SVM) for
classification. The results showed that CNN can be considered as the best model when
it comes to classification accuracy with a score of up to 99.6%. Meanwhile, the fastest
processing speed was obtained by LBP+SVM with an average of 30.2 FPS.