A Practical Solution to the Problem of High Quality Detecting People and Vehicles from Video Frames

The research focuses on addressing the challenge of localizing people and vehicles in video frames. Test data consists of video frames from forested and road areas. The DeepSORT algorithm from the extended version of “deep_sort_realtime” package is utilized for object tracking. Additionally, YOLO v8 has been integrated for object detection, and MobileNet v3 is used for extracting informative features. Input images undergo letterbox preprocessing, and various optimizations have been implemented to enhance the quality and speed of the results.For license plate recognition, the “tflite_avto_num_recognition” software package is employed, which uses Canny and Hough transformations along with a CNN-LSTM-CTC neural network. A study was performed on the identification and recognition of license plates on automobiles, buses, and trucks using machine learning algorithms. These solutions operate in real time and are based on open-source libraries. © 2025 Elsevier B.V., All rights reserved.