MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641071106 A) filed by Dr. L. V. Raja; Mr. R. K. Sivaratheesh; Mr. R. Daryll Paul; Dr. Sinthiya Jayapal; Dr. M. Jaithoon Bibi; Dr. D. Kavitha; Mr. T. Senthil Ganesh; Rohit Singh; Ms. K. Iswarya; and Ms. Karpagam Kulandaivel on June 08, 2026, for Real-Time Vehicle Detection Using Deep Learning And Object Tracking.
Inventors include Dr. L. V. Raja; Mr. R. K. Sivaratheesh; Mr. R. Daryll Paul; Dr. Sinthiya Jayapal; Dr. M. Jaithoon Bibi; Dr. D. Kavitha; Mr. T. Senthil Ganesh; Rohit Singh; Ms. K. Iswarya; and Ms. Karpagam Kulandaivel.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: Real-Time Vehicle Detection Using Deep Learning and Object Tracking Abstract: Real-time vehicle detection has become a critical component in modern intelligent transportation systems, smart city infrastructure, and traffic surveillance applications. Accu-rate detection and tracking of vehicles enable efficient traffic monitoring, congestion analysis, automated law enforcement, and accident prevention. Traditional computer vision techniques relied on handcrafted feature extraction methods that often struggle in complex traffic environments. With the advancement of deep learning and convolutional neural networks (CNNs), object detection systems have significantly improved in both accuracy and real-time performance. This research presents a real-time vehicle detection and tracking system based on the YOLO (You Only Look Once) object detection model integrated with a tracking mechanism to maintain vehicle identities across video frames. The system processes video input, detects vehicles using the YOLO model, tracks them through consecutive frames, and performs additional classification tasks such as vehicle color and model recognition using MobileNet-based neural networks. The implementation is carried out using Python with libraries such as OpenCV, Ultralytics YOLO, and TensorFlow. Experimental evaluation demonstrates that the proposed sys-tem achieves efficient real-time performance while maintaining high detection accuracy in dynamic traffic environments. The de-veloped framework can be applied in various real-world scenarios including traffic monitoring systems, intelligent transportation infrastructure, automated surveillance, and smart city traffic management. Today's world is witnessing the rapid urbanization of the global population and the rising number of vehicles. This has led to the formidable challenge of managing vehicle traffic in cities. Effective monjtoring of traffic is one of the most important factors for optimizing the transportation system, minimizing traffic jams, and protecting the safety of the public. An important feature of intelligent transportation systems is detecting vehicles and analyzing them in real time. With this feature, numerous applications are able to be imple-mented, such as traffic monitoring, vehicle counting, accident detection, parking management, and automatic enforcement of traffic laws. Historically, these systems depended on human supervision or rudimentary computer vision, wruch were weak in richly structured real-world situations. Hand-crafted feature engineering such as edge detection, Haar-like features, or Histogram of Oriented Gradients (HOG) features combined with a lighting classifier defined the field of early computer vision vehicle detection. While these vehicles detectors are valid in ideal scenarios, their detection ability drops as a result of varying lighting, traffic congestion and occlusions. Currently, the field of computer vision has been transformed as a result of convolutional neural networks (CNNs) and the field of deep learning. With deep learning and the use of advanced neural networks, developers are able to create soprusticated models that automatically learn the necessary features to achjeve higher detection accuracy and more robust-ness to changes in the environment. Of the many advanced neural network models available, YOLO (You Only Look Once) is the most successful on the real time detection of objects. YOLO achieves object detection using a single pass of a neural network by splitting an image into grid cells, and predicting bounding boxes and class probabilities for each grid cell. This technique enables YOLO to maintain both impressive detection accuracy and rapid processing speeds, even for real time applications. Besides detecting vehicles, tracking them throughout video frames is critical for analyzing vehicle movement. In addition to recognizing vehicles, knowing how to follow them through individual video frames is key to understanding how vehicles behave, as well as how to eliminate repeated detections. Ve-hicle detection and tracking systems integrate object detection and tracking with certain classifications, such as categorizing vehicles by color and model. The system is meant to analyze video streams and offer insights for traffic monitoring systems. YOLO performs object detection by dividing an image into grid cells and predicting bounding boxes along with class probabilities in a single forward pass of the neural network. This approach allows YOLO to achieve both high detection accuracy and fast processing speed, making it suitable for real-time applications. In addition to detecting vehicles, tracking them across video frames is essential for understanding vehicle movement patterns and preventing duplicate detections. Object tracking algorithms assign unique identifiers to detected vehicles and maintain their identities across consecutive frames. This project proposes a real-time vehicle detection and tracking system that integrates YOLO-based object detection with a tracking system and additional classification modules for vehicle attributes such as color and model. The system is designed to process video streams efficiently and provide meaningful insights for traffic monitoring applications. The main contributions of this work include: • Development of a real-time vehicle detection system using YOLO. • Integration of an object tracking mechanism to maintain vehicle identities. • Implementation of vehicle color and model classification using MobileNet-based neural networks. • Evaluation of system performance in real-time traffic scenarios.
Disclaimer: Curated by HT Syndication.
