Image Understanding for Alpine Environmental Monitoring

在一个人为全球变化的时代，测量和监测自然环境的需求越来越大。延时摄影可以提供令人信服的自然系统随时间变化的可视化，例如，全球变暖导致的冰川衰退。然而，除了可视化之外，相机还可以用作传感器，进行测量，帮助科学地理解自然系统。该项目的目的是开发新的计算机视觉技术来监测室外环境的长期变化，将单反相机变成一个有用的环境监测传感器。特别是，我们将使用延时图像来提取场景中的详细表面信息和细微运动，这是使用现有的基于卫星的技术无法实现的。实现这一目标需要改进的图像理解技术，能够提取有用的信息，如非常小的场景运动，在很长的时间尺度和广泛变化的外观和照明条件下。

In an era of anthropogenic global change, there is an increasing need to measure and monitor our natural environment. Time-lapse photography can provide compelling visualisations of how natural systems behave over time, for example, glacial recession due to global warming. However, in addition to visualisation, cameras can also be used as sensors, making measurements that aid the scientific understanding of natural systems.The aim of this project is to develop novel computer vision techniques to monitor long-term change in outdoor environments, turning an SLR into a useful sensor for environmental monitoring. In particular we will use time-lapse imagery to extract detailed surface information and subtle motion in scenes, something that is not possible using existing satellite-based techniques. Reaching this goal will require improved techniques for robust image understanding that are capable of extracting useful information, such as very small scene motion, over long timescales and with wildly varying appearance and illumination conditions.