Tuesday 04 March 2025
A team of researchers has developed a new approach to counting vehicles in videos captured by static cameras. The method, which combines two existing technologies – YOLO (You Only Look Once) for object detection and Visual Rhythm for summarizing video content – is faster and more efficient than traditional methods.
The problem with current vehicle counting systems is that they typically involve detecting objects frame-by-frame, which can be computationally intensive. This approach also requires tracking individual vehicles across multiple frames, which can lead to errors and reduce accuracy. The new method, on the other hand, uses Visual Rhythm to create a single image that summarizes the most important information from a sequence of video frames.
This image, known as a time- spatial representation, shows the position of objects over time in a single frame. By analyzing this image, the system can detect vehicles and count them without needing to process each individual frame. The result is a faster and more efficient method that achieves similar accuracy levels to traditional approaches.
The researchers tested their approach on four publicly available videos captured from different locations and scenarios. They used a dataset of 960 frames randomly selected from these videos, annotated with six vehicle classes – bus, car, motorbike, pickup, truck, and van. The system was able to achieve counting accuracy of around 99% across the test set, with an efficiency improvement of approximately three times compared to traditional methods.
One of the key advantages of this approach is its ability to handle diverse environments and scenarios. The researchers used a combination of transfer learning and fine-tuning to train their model for task-specific detection, which allowed it to adapt to different conditions and improve accuracy.
However, there are still some limitations to the system. For example, the classification accuracy for certain vehicle classes – such as pickup and motorbike – was lower than for others. This suggests that the dataset may be biased towards more common vehicle types, and further work is needed to improve detection and classification of less frequent vehicles.
Overall, this new approach has the potential to revolutionize video-based traffic management systems by providing a faster, more efficient, and more accurate way to count vehicles. As cities continue to grow and urban populations increase, the need for effective traffic management solutions will only become more pressing. This innovative technology could play a key role in helping to address this challenge.
Cite this article: “Fast and Accurate Vehicle Counting in Videos Using Visual Rhythm and YOLO”, The Science Archive, 2025.
Vehicle Counting, Video Analysis, Object Detection, Yolo, Visual Rhythm, Time-Spatial Representation, Traffic Management, Computer Vision, Image Processing, Deep Learning.
