Analysis and classification of the lane changing of vehicles by wavelet transform

Traffic management is getting more difficult because of increasing vehicle demand and urbanization. Many models are developed using intelligent transportation systems to improve traffic conditions, and solutions are offered using traffic datasets with these developed models. Lane changing is an important part of traffic and, it's one of the basic driving behaviors that has a major impact on traffic efficiency, safety and flow. In this thesis, a new model has been developed for the lane changing detection. In order to detect lane changing, the azimuth angles of the vehicles were calculated using the WGS-84 coordinates provided with the pNEUMA data set. In addition, the lateral deviations of the vehicles were calculated using the traveled distance data of the vehicles. The maximum amplitude of each vehicle was obtained by applying multilevel discrete wavelet transform to the azimuth series. The lane changing behavior of vehicles in urban roads has been determined by classifying the lateral deviation and maximum amplitude features with K-NN. The time interval at which the lane changing takes place was determined by the applied wavelet type and the maximum amplitude wavelet coefficient. In addition to the time interval detection, the target lane that the vehicle passes through was determined according to the wavelet coefficient sign of the maximum amplitude. Besides of the lane changing detection, lane changing was classified as smooth and sudden using the features including level information of lane changing were obtained by applying wavelet transform base functions of haar, symlet and daubechies. It is observed that the wavelet transform approach proposed in the thesis is successful in detecting the lane changing. Compared to other approaches in the literature, it was determined that the proposed method provides a high success rate and has less processing complexity.