Anomaly detection for deep-learning based license plate recognition in real time video
Bada Kim, Taeyeon Won, Sangmin Park, Junyoung Heo · 2019
License plate recognition has a recognition rate of more than 98% in limited situations; however, the recognition rate sometimes falls to about 50% to 70% in unlimited situations for real time. The inability of filtering out anomaly data which is similar to a license plate results in the low recognition rate. This paper aims to suppress anomaly and improve the recognition rate. CNN consist of as few layers as possible for application with IoT edge technology. As a result of the experiment, the detection rate on real-time roads where situations are unlimited was 77% in normal models with the little filter performance. However, when applying the techniques of this paper, the detection rate was 88%.