集成基于对象影像分析与贝叶斯软融合的土地覆被变化检测

Land cover change detection based on the object-based image analysis and Bayesian soft fusion method

应用遥感数据进行分类后变化检测时,土地覆被分类误差会造成严重的误差累积效应,同时会导致土地覆被类型转换结果中经常出现不尽合理的情况.采用基于对象的影像分析方法,利用两期国产高分一号影像数据分析了兰州市建成区及其东南部边缘地区的土地覆被变化.通过分析研究区的特征,得到了研究区的土地覆被转移逻辑.利用贝叶斯软融合方法降低了土地覆被变化分类误差的累积效应,运用土地覆被逻辑,消除了不合逻辑的结果,变化检测结果总精度达到81.61%.最终得出了兰州市总体上沿沟谷地向周边东南方向延伸的发展模式.

Using remote sensing data for post-classification change detection will result in land cover classification errors that will, in turn, cause serious errors in accumulation effects and, at the same time, will lead to unreasonable situations in land cover type conversion results. The object-based image analysis method was used to analyze the land cover changes in the built-up area of Lanzhou City and its southeastern marginal area based upon the two-stage domestic high score image data. With an analysis of the characteristics of the study area, the land cover transfer logic thereof was obtained. The Bayesian soft fusion method was used to reduce the cumulative effect of land cover change classification errors and the land cover logic was used to eliminate the illogical results. The total accuracy of the change test results was81.61%. It was concluded that a development model had been generally formed along the valley to the southeast in Lanzhou.