Methods and uncertainty modeling for land cover change detection from multi-resolution remotely sensed data

时间序列遥感环境数据已被广泛应用于监测地球表面和生物物理变量的变化。随着遥感技术的快速发展和广泛应用，积累了大量具有广泛空间、光谱和时间尺度的航空和卫星数据。这些发展要求分析和整合这些数据的方法发生重大变化。过去发展了许多传统的时间序列遥感数据变化检测方法。理想情况下，变化检测是通过利用来自同一传感器的按时间分隔的图像来进行的。然而，许多条件将不可避免地带来理想的时间变化检测情况的终结。许多传统的方法在涉及多光谱和空间分辨率的图像的变化检测中不能很好地工作。与快速增长的数据可用性相比，在开发有效的变化检测模型和多尺度分析框架来处理这些日益复杂和数量越来越多的多时相遥感数据方面存在明显的滞后。 该项目的目标是：(A)审查误差和不确定性对土地覆盖分类和变化检测程序的尺度敏感性；(B)评估绘制不同地貌的土地覆盖及其变化的不同亚像素方法的准确性；(C)开发新的方法，改进多光谱、多空间分辨率遥感数据的土地覆盖变化检测。这项研究的成果将极大地促进遥感数据在许多应用中的最大限度地利用。在拟议的预算和研究范围内，每年将至少培训4名研究生和2名本科生。

Time series remotely sensed environmental data has been widely used in monitoring changes in earth surfaces and biophysical variables. With rapid development and widespread use of remote sensing technologies, a great deal of airborne and satellite data with a wide range of spatial, spectral and temporal scales has been accumulated. These developments demand significant changes in the approach of analyzing and integrating these data. Many traditional change detection methods in time series remotely sensed data have been developed in the past. Change detection is ideally conducted by utilizing images, separated by time, from the same sensor. However, many conditions will inevitably bring an end to the ideal temporal change detection situation. Many traditional approaches do not work well in change detection involving images with multiple spectral and spatial resolutions. Compared with rapidly growing data availability, there is a clear lag in the development of efficient change detection models and multi-scale analysis framework for handling these multi-temporal remotely sensed data of increasing complexity and quantity. The objectives of this project is to (a) to examine the scale sensitivity of error and uncertainty on land cover classification and change detection procedures from multi-sensor temporal remotely sensed data; (b) to evaluate the accuracy of different sub-pixel approaches for mapping land cover and its change for different landscapes; and (c) to develop novel approaches to improving land cover change detection from multi-spectral, multi-spatial resolution remotely sensed data. The outcomes from this research will significantly contribute to the maximum use of remote sensing data in many applications.Within the proposed budget and research scope at least four graduate students and two undergraduate students will be trained each year.