License Plate Recognition (LPR) is an advent research in Intelligent Transportation System (ITS) since it poses serious challenges in relation with image recognition. The robustness of ALPR gets reduced in real-world complex due to low quality images occurs from poor illumination, weather condition, complex background, perspective distortions and night light. In this paper, a deep neural network is used for classification and recognition of number plates in real-time moving vehicle under complex scenarios that mitigates the challenges associated. We develop a classification model using a deep learning algorithm namely Generative Adversarial Network(GAN) for the detection of number plates. The GAN network structure is optimized with 20-classes of CNN for the detection of number plates with higher accuracy. Such GAN forms an efficient network that performs classification with reduced errors. The simulation is conducted using python simulator to obtain the classification rates of the model. The extensive simulation shows that the GAN model obtains improved robustness against the real-time vehicle detection and improved scalability in terms of high classification accuracy.