Publication Details
Abstract
This paper analyzes the hardware and software architecture of video data processing and computer vision algorithms. The emergence of video data processing, computer vision algorithms, including deep learning, has revolutionized the field of computer vision, enabling real-time analysis that is essential for traffic management and vehicle identification. This study presents a system that combines vehicle make and model recognition with Automatic Number Plate Recognition (ANPR) and achieves an impressive 97.5% accuracy. Unlike traditional methods, which focus either on make and model recognition or ANPR, this study combines both aspects into a single, coherent system and provides a more comprehensive and efficient solution for vehicle identification, ensuring reliable operation even in harsh weather conditions. The video data processing process consists of several main stages: data acquisition, compression, frame processing, and analysis of the results. The software and hardware used in these processes are of great importance. Hardware components such as GPU (Graphics Processing Units), FPGA (Field Programmable Logic Arrays) significantly accelerate the video data processing process. Software platforms such as TensorFlow, OpenCV, PyTorch also play an important role in the development and implementation of algorithms.[1] The article examines the application of deep learning techniques, including OpenCV and the Python programming language. Thousands of tests were conducted on the system using MobileNet-V2 and YOLOx (You Only Look Once) for vehicle detection, and YOLOv4-tiny, Paddle OCR (Optical Character Recognition), and SVTR-tiny for ANPR