[2411.02136] Advanced computer vision for extracting georeferenced vehicle trajectories from drone imagery

View the paper PDF file titled Advanced computer vision to extract geological vehicle tracks from drones, written by Robert Fonoud, Haishan Chu, Hwasoo Yeo and Nikolas Geroliminis

PDF HTML (experimental) view

a summary:This paper provides a framework for extracting geological vehicle paths from drones highly altitude, and deals with the main challenges in monitoring traffic in urban areas and traditional ground systems restrictions. Our approach merges many new contributions, including the designer -designed object detector to the face of high bird display perspectives, a unique way to stabilize the track that uses vehicle oceans discovered as excluded masks while recording images, and a strategy for geographic parts based on the main frame that enhances consistent compatibility through multiple Drone views. In addition, our framework is characterized by a strong grade of vehicles dimensions and detailed methods of ways, allowing comprehensive traffic analysis. It was conducted in the SongDo International Business District, South Korea, the study used a multi -program experience covering 20 intersections, obtaining approximately 12 terabytes of 4K video data for four days. The frame produced high -quality data sets: SongDo traffic collection, which includes about 700,000 unique vehicle tracks, and the SongDo Vision Data set, which contains more than 5,000 pictures made of humans with about 300,000 complex cases in four categories. Comparisons with high -resolution sensor data from the investigation vehicle with a resolution and the consistency of the extraction pipeline in thick urban environments. The general version of Songdo Traffic and Songdo VISION, the full source of the extraction pipeline code, defines new standards in data quality, cloning, and expansion of traffic research. The results show the potential for incorporating drone technology with an advanced computer vision to monitor accurate and cost -effective urban traffic, and provide valuable resources to develop smart transportation systems and enhance traffic management strategies.