Land cover change detection using autocorrelation analysis on MODIS time-series data : detection of new human settlements in the Gauteng province of South Africa

摘要

Human settlement expansion is one of the most pervasiveforms of land cover change in the Gauteng province of SouthAfrica. A method for detecting new settlement developments inareas that are typically covered by natural vegetation using 500 mMODIS time-series satellite data is proposed. The method is a perpixel change alarm that uses the temporal autocorrelation to infera change index which yields a change or no-change decision afterthresholding. Simulated change data was generated and used todetermine a threshold during an off-line optimization phase. Afteroptimization the method was evaluated on examples of known landcover change in the study area and experimental results indicate a92% change detection accuracy with a 15% false alarm rate. Themethod shows good performance when compared to a traditionalNDVI differencing method that achieved a 75% change detectionaccuracy with a 24% false alarm rate for the same study area.