Abstract: Object classification is an important technique for autonomous vehicles to identify surrounding dynamic objects and execute safe path planning. In this paper, a multi-class object classification method is proposed to classify objects around the vehicle into car, pedestrian bicyclist, and background using a LIDAR sensor. The various effective features of each object are computed using 3D point clouds, and Real AdaBoost algorithm is applied for multi-class classification. In addition, class probabilities are calculated and updated in a time series framework for tracking objects. Classification is evaluated using a dataset that includes long-range objects over 50m. The overall accuracy of the clusters for each frame is 92.7% and for tracking objects is 95.5%. Moreover, processing time for object classification is 0.07ms. Thus, this method can be used in real time for autonomous driving on urban roads.