|
Locating and recognition of vehicle license plate has always been a research direction of image processing. Compared to traditional methods using color and edge features for license plate detection, the proposed block scanning of edge image could immediately identify candidate plate regions of different distances. By using edge informant, the proposed is also effective and robust under variety of weather conditions.
Firstly, the proportion of license plate is used to setup possible block sizes for edge image scanning. While doing block scanning, we could quickly filter out false positives by characteristics of license plates such as upper/lower blank area, as well as fixed font of license plate characters. Furthermore, license plate character could be adjusted and recognized by vertical projection.
The used filtering criteria not only can reduce the complexity of image operations but also robust under varying weather conditions for license plate locating and recognition. In order to verify the feasibility of the system, we tested 155 images in daytime/evening/night and sunny/rainy day and the average license plate detection rate was 95%. In the future, the effectiveness of the overall system performance could be improved by developing more criteria for traffic control or parking access control.
|