Organizations: Department of Electrical and Electronic Engineering, Islamic University of Technology, Gazipur, Bangladesh · Department of Mechanical Engineering, Bangladesh University Of Engineering And Technology, Dhaka, Bangladesh
An Automatic License Plate Recognition (ALPR) system constitutes a crucial element in an intelligent traffic management system. However, the detection of Bangla license plates remains challenging because of the complicated character scheme and uneven layouts. This paper presents a robust Bangla License Plate Recognition system that integrates a deep learning-based object detection model for license plate localization with Optical Character Recognition for text extraction. Multiple object detection architectures, including U-Net and several YOLO (You Only Look Once) variants, are compared for license plate localization. This study proposes a novel two-stage adaptive training strategy built upon the YOLOv8 architecture to improve localization performance. The proposed approach outperforms the established models, achieving an accuracy of 97.83% and an Intersection over Union (IoU) of 91.3%. The text recognition problem is phrased as a sequence generation problem with a VisionEncoderDecoder architecture, with a combination of encoder-decoders evaluated. It was demonstrated that the ViT + BanglaBERT model gives better results at the character level, with a Character Error Rate of 0.1323 and Word Error Rate of 0.1068. The proposed system also shows a consistent performance when tested on an external dataset that has been curated for this study purpose. The dataset offers completely different environment and lighting conditions compared to the training sample, indicating the robustness of the proposed framework. Overall, our proposed system provides a robust and reliable solution for Bangla license plate recognition and performs effectively across diverse real-world scenarios, including variations in lighting, noise, and plate styles. These strengths make it well suited for deployment in intelligent transportation applications such as automated law enforcement and access control.
Figures & tables
Fig. 1 : BRTA Standard Vehicle Registration Plate
Fig. 2 : Images for License Plate Detection
Fig. 3 : Images for Text Extraction
Fig. 4 : Images for External Validation
Fig. 5 : Proposed Algorithm for Localization Model
Hyperparameter
Value
Decoder Mode ( is_decoder )
True
Add Cross Attention
True
Encoder Dropout Rate
0.3
Decoder Dropout Rate
0.3
Attention Probabilities Dropout
0.3
Max Sequence Length
20
TABLE I : Model Configuration and Hyperparameters
Model
Accuracy(%)
Precision(%)
Recall(%)
F1 Score(%)
IoU(%)
U-net
91.43
93.19
89.71
91.41
82.1
YOLOv5m
90.97
99.89
87.03
93.06
83.9
YOLOv7m
67.15
95.28
54.59
69.42
63.5
YOLOv8m
97.11
99.93
95.68
97.79
89.7
YOLOv8m + Multi Stage Learning
97.83
99.44
96.31
98.37
91.3
YOLOv9m
95.67
99.44
95.20
97.24
88.8
TABLE II : Evaluation Metrics of Localization Models
Fig. 6 : License plate detection
Model
Encoder
Decoder
CER ↓
WER ↓
ViT + B-BERT
ViT-Base
BanglaBERT
0.1323
0.1068
TrOCR
TrOCR-Base
TrOCR-Base
0.1363
0.1583
ViT + mBART
ViT-Base
mBART-50
0.4438
0.0935
TABLE III : Performance Comparison of OCR Models on external dataset
Metric
Best Value
Validation Loss
0.4101
Character Error Rate (CER)
0.1323
Word Error Rate (WER)
0.1068
Levenshtein Distance
3.02
TABLE IV : Summary of Best OCR Model Performance
Fig. 7 : Comparison of Character Error Rate (CER) over training steps.
Fig. 8 : Comparison of Levenshtein Distance convergence.
Fig. 9 : Comparison of Word Error Rate (WER) across models.
Fig. 10 : OCR outputs of the model using the external dataset.
Model
Accuracy(%)
Precision(%)
Recall(%)
F1 Score(%)
IoU(%)
U-net
80.43
92.10
82.59
80.79
60.4
YOLOv5m
52.53
94.26
63.20
72.81
45.20
YOLOv7m
55.12
97.46
62.74
75.43
43.04
YOLOv8m
61.11
99.87
69.13
79.30
52.51
YOLOv8m + Multi Stage Learning
82.9
99.92
89.14
92.10
67.5
YOLOv9m
53.62
99.86
63.68
75.83
47.42
TABLE V : Performance with External Dataset
Fig. 11 : Inference on External Dataset (Green = Annotation; Red = Prediction
Department of Mechatronics Engineering, Universidad Catolica Boliviana “San Pablo”, La Paz, Bolivia · School of Electronic Engineering and Computer Science, Queen Mary University of London, England, United Kingdom