Automatic number plate recognition recent the recent Trend in current environment, various systems has already developed by existing authors to detect the license plate as well as number plate using machine learning and deep learning algorithms. Internet of things (IoT) and some machine learning techniques for image processing is also contributed many classical systems. In this paper we proposed automatic number plate recognition for smart toll system, in traditional approach it is very tedious to perform the entire process with manual human interfaces. Sometime it also generate long traffic queue on toll plaza, to eliminate such a problem with this propose systems. This system initially deals with some IoT devices which contains camera sensor and arduino based microcontroller. The deep learning libraries have used to process the capture image by IoT environment with OCR library. This system also deals with user applications to auto debit toll charges. The state of art provides the conceptual model of entire system execution with real time as well as synthetic dataset, and theoretical experiment analysis shows the effectiveness of system.