Transportation currently has very large and essential role in the needs of the community, namely vehicles. Some problems that often arise related to license plates, such as the length of parking management. Research on this number plate identification system can help with several issues in parking management. The machine learning method related to license plate images is Convolutional Neural Networks. Therefore, the use of the Convolutional Neural Network in this study so that the accuracy of the number plate image is better, then we need a system that in this research process uses the number plate image as input. Many pre-process such as grayscale, segmentation, and plate finder are performed, then this number plate image is identified using Convolutional Neural Networks. This study will build a computational model in which identification uses Convolutional Neural Networks. The results of this study, found the license plate number in the vehicle image using a plate finder. The results of introducing number plate characters using CNN resulted in an accuracy of 96.87% for training data and 93.71% test data using the SGD optimistic model.