The main goal of this study is to extract traffic statistics throughout the day, week and season based on video surveillance. For our propose, we focus on traffic analysis that is related to vehicle counting. We provide the following process chain: objects detection and tracking, trajectories extraction, detection of entry/exit zones and vehicle counting. Experiments show that the information obtained is reliable and gives effective vehicles counting that are compared to previously published results obtained on a public data set. Additional comparisons are done with ground truth for a private data set.