Typically, many features are utilized to detect pedestrian by using LIDAR. However, to implement a pedestrian detector based on SVM algorithm, it is necessary to manually determine the features and their parameters. In addition, there are a large number of combinations of parameters, and it is also necessary to change the parameters according to the variety of the performance of LIDAR. To overcome above problems, we proposed automatic selection of features by using Real AdaBoost which is a one of statistical learning techniques. In some pedestrian detection experiments, our proposal method was compare to previous method based on SVM.