Smart Traffic Monitoring through Drone Images via Yolov5 and Kalman Filter

Monitoring traffic is of vital relevance in our modern scenario. Over the years, standard ways of capturing data, such as induction loops and camcorders, were applied for this aim. Nevertheless, the introduction of unmanned aerial vehicles (UAVs) has opened new possibilities in this subject, leading to extensive study in computer vision. However, there are still issues in object detection and tracking because of the complexities that come from a high number of objects, shifting heights of unmanned aerial vehicles (UAVs), and variable lighting conditions. Our study provides a new approach that combines centroid tracking with the YOLOv5 algorithm to successfully monitor and identify vehicles. To acquire an accurate analysis, we began our method by aligning and referencing aerial images. These basic measurements assisted the advancement to next phases covering feature extraction, segmentation of the region of interest, and tasks connected to detection and tracking. The effectiveness of our technique was tested by applying it to the VAID dataset, resulting in amazing levels of accuracy demonstrated by our recommended solution. Our system exhibited an unparalleled accuracy of 98.5% for object detection and a reasonable 90.9% for tracking tests. The validation indicates the robustness and efficacy of our designed system in solving the obstacles given by sophisticated aerial surveillance conditions.