EILPR: Toward End-to-End Irregular License Plate Recognition Based on Automatic Perspective Alignment

Xu, Hui; Zhou, Xiangdong; Li, Zhenghao; Liu, Liangchen; Li, Chaojie; Shi, Yu

doi:10.1109/tits.2021.3130898

Cited by 15 publications

(9 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Xu et al. [25] improved the ALPR performance when the LPs had multi‐line texts or arbitrary shooting angles, using only plate‐level annotations for training. Zhang et al.…”

Section: Related Workmentioning

confidence: 99%

“…A typical ALPR system usually includes two subtasks, namely license plate detection (LPD) which locates the license plate (LP) in the form of a bounding box from an input image and license plate recognition (LPR) which predicts the LP number displayed in the LP area. Based on the strategies adopted to complete two subtasks, the existing ALPR methods can be roughly divided into two categories as follows: 1) two-stage schemes [9][10][11][12][13][14][15][16][17][18][19][20], which first complete the LPD and then perform the LPR base on the LPD result; and 2) one-stage schemes [21][22][23][24][25], which utilize a unified framework to complete both LPD and LPR tasks simultaneously.…”

Section: Introductionmentioning

confidence: 99%

“…Recently, the one-stage schemes for ALPR [21][22][23][24][25] are becoming increasingly popular due to their advantages over the two-stage schemes. Specifically, a one-stage scheme trains one deep neural network for the end-to-end detection, localization, and recognition of the license plate in a single forward pass.…”

Section: Introductionmentioning

confidence: 99%

“…This allows the models to share parameters between two stages and the total number of parameters in onestage models is often much fewer than that in two-stage models. As a result, they are more efficient than two-stage methods while achieving comparable or even higher accuracy [21][22][23][24][25].…”

Section: Introductionmentioning

confidence: 99%

“…Although the existing one-stage methods [21][22][23][24][25] have achieved excellent results and demonstrate the power of a single CNN network to complete the two tasks of license plate detection and recognition, they still have some limitations in striking a good balance between accuracy and efficiency. In order to pursue high accuracy, they often excessively increase the model size and complexity, leading to much higher computational cost and thus longer detection and recognition time.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

FAFEnet: A fast and accurate model for automatic license plate detection and recognition

Zhou

Cheng

Jiang

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

Automatic License Plate detection and Recognition (ALPR) is a key problem in intelligent transportation systems with wide applications in traffic monitoring, electronic toll collection (ETC), intelligent parking lots (IPLs), and elsewhere. Although numerous methods have been proposed in the literature, it is still challenging to strike a good balance between the accuracy and efficiency of ALPR. In this paper, a novel end-to-end CNN-based model is proposed, called Fast and Accurate Network with Feature Enhancement (FAFEnet), to jointly detect the license plates and recognize the characters with high accuracy and efficiency. Specifically, the FAFEnet model seamlessly integrates two CNN-based models, namely the detection and recognition modules, into a unified framework to reduce accumulated errors and computational overheads in two separate steps. The detection module is a lightweight model with only seven convolutional layers yet achieves over 99.8% accuracy rates for license plate detection across all datasets. The recognition module utilizes two feature enhancement blocks to compensate and enhance the shallow character features extracted by the detection module. Furthermore, the joint optimization of detection and recognition modules exploits the feature association in two modules, and thus improves the prediction accuracy while reducing the execution time. Finally, extensive experimental results on several real-world datasets demonstrate that FAFEnet outperforms all the competitors in terms of both accuracy and efficiency.

show abstract

“…Xu et al. [25] improved the ALPR performance when the LPs had multi‐line texts or arbitrary shooting angles, using only plate‐level annotations for training. Zhang et al.…”

Section: Related Workmentioning

confidence: 99%