An Approach to Instantly Detecting Fake Plates Based on Large-Scale ANPR Data

Traditional methods of detecting fake plates are mostly inefficient. They usually require lots of investments in advance. These methods cannot fully play potentials of ANPR (Automatic Number Plate Recognition) data and utilize them to detect fake plates quickly. In this paper, we propose a method, called as FP-Detector, to instantly detect fake plates through parallel analyzing the historical large-scale ANPR data with MapReduce. The main contributions include: we design a partition strategy, which can fully use the features of ANPR and maintain balances among different nodes. In addition, we also give a criterion of judging fake plates through analyzing spatio-temporal contradiction of plate information. Finally, we apply our method on a real large-scale data set and compare the performance of our method with default blocking strategy of MapReduce. The experiment results show the effectiveness of our method.