Abstract:
Satellite video with high temporal resolution can provide a broad view for surface dynamic monitoring, especially traffic monitoring. However, the limited vehicle feature...Show MoreMetadata
Abstract:
Satellite video with high temporal resolution can provide a broad view for surface dynamic monitoring, especially traffic monitoring. However, the limited vehicle features and complex background information make it difficult to detect vehicles simply from video frame features such as spectrum, shape, and texture. In this letter, a new satellite video dynamic vehicle detection method for traffic monitoring is proposed, which considers the relation between vehicles and the background in a time series. Road buffers are obtained according to the various road levels in order to focus on dynamic vehicles, and then, a multiframe brightness gradient (MBG) is proposed to detect potential dynamic vehicles on the buffers, and the detection results are optimized by the cumulative trajectory of motion. This approach can effectively suppress false detection caused by parallax motion and illumination variation in satellite videos. The experimental results of three satellite videos show that the proposed method can effectively achieve dynamic vehicle detection with a high positive detection rate and low false alarm rate.
Published in: IEEE Geoscience and Remote Sensing Letters ( Volume: 20)