Iraqi Plate Number Detection Based on Deep Learning and Recognize by SURF

License plate recognition (LPR) is an abstraction of the image or a series of images from the car license plate details. The camera which takes photos, the quality of the images obtained is an important factor in the LPR's achievement. LPR processes license plates below diverse environmental conditions with efficiency. Lighting, gravel, and towing devices inside the car will partially occlude the L.P. In this paper, a comprehensive evaluation of state-of-the-art LPR procedures was offered which detected a licensing plate using deep learning(DL) with the convolutional neural network (CNN) algorithm. Then used the speeded-up robust feature (SURF) algorithm to recognized this license plate. The goal behind this paper to predict the traffic flow by detecting and recognize the license plate cars. The accuracy was giving for detection of training 99.2% and loss of 0.8% and for testing 96% and loss of 4%. For recognition testing, given 96% by the SURF algorithm and using a database to save the personal information if this plate exists can be entered to private park-cars otherwise not enter. The waiting for cars in the gate was considered as The traffic flow.