Clearview: Real-time traffic signal and license plate recognition

This paper introduces an exploratory innovative Android application called ‘Clearview’ to directly address a particular prerequisite needed in ITS, namely, real-time traffic light and license plate identification in situations that include adverse weather that include fogs and hazes. The Traffic Management System of Clearview involves impressive technologies used in traffic surveillance and managing incidences of traffic congestions. The core of the developed application entails the use of Generative Adversarial Networks GAN in dehazing aspect that makes it rather simple to improve clarity and image brightness particularly under poor visibility conditions. Thus, dehazed images were fed into YOLOv4, the real-time object detection model to get a better prediction of traffic signals on the road. To find the outline of the objects like a vehicle, Clearview utilizes single shot multiBox detector (SSD) while the characters in the license plate are identified with the help of Tesseract OCR. The developed application also schedules on-device inference through optimizing the deep- learning model with the TensorFlow Lite for boosting real-time responses. Using TensorFlow/Keras for deep learning, OpenCV for computer vision, and Android Studio for app development, Clearview is created to be a plug-and-play system on this basis where all of these technologies are integrated into one functioning system. The real life detection rates were presented after the evaluative experimentation of Clearview and supported by evidences that is possible to achieve over 90% of the detection rates while working on real time. The following paper attempts at outlining the structure and architecture of the system under consideration, the technologies utilized in the formulation of Clearview and the various performance indices that will qualify ITS technology and traffic observation under difficult conditions hence the overall role played by Clearview in advancement of ITS technology.