PlateSegFL: a privacy-preserving license plate detection system using federated segmentation learning

Abstract

ALPR is an integral component of an intelligent transport system with extensive applications in secure transportation, vehicle-to-vehicle communication, vehicular networks, passenger protection, and automatic vehicle authorization. In addition, it has numerous applications in the detection of stolen vehicles, traffic violations, and traffic flow management. Focusing on the identification of regions of interest from available recordings or image sources, learning-based APLR system development focuses on the detection of regions of interest. The existing license plate detection system focuses on one-shot learners or pre-trained models that operate with a geometric bounding box, limiting the model’s performance to selecting a geometric bounding box and thus providing more or less information than necessary. Furthermore, continuous video data streams uploaded to the central server as a result of an increase in data traffic which in return creates network and complexity issues. To combat this regulatory obfuscation, we proposed a method that implements federated learning via Unet-based segmentation with Federated learning. U-Net is renowned for its capacity to analyze high-resolution images and generate segmentation maps with a high degree of precision. UNet is well-suited for multiclass image segmentation tasks because it can analyze a large number of classes and generate a pixel-level segmentation map for each class. U-Net is capable of segmenting images at the pixel level. Federated learning is used to reduce the quantity of data required while safeguarding the user’s privacy. Different computing platforms, such as mobile phones, are able to collaborate on the development of a standard prediction model, it makes efficient use of one’s time, it incorporates more diverse data, it delivers projections in real-time, and it requires no physical effort from the user. The F1 dependability of our model hovers around 95%.