Forest cover change detection based on multi-scale segmentation and tasseled cap transformation over plateau area

  Objective  Based on multi-scale segmentation and tasseled cap transformation, this paper aims to develop a detection method for forest cover change in complex plateau area.

  Method  Using Landsat 5 TM and Landsat 8 OLI multispectral images, taking the transitional mountainous area of the Yunnan Middle East Plateau of northwestern China as the research area, based on multi-scale segmentation, using the tasseled cap transformation (TCT) and principal component analysis (PCA), the object-oriented multi feature change vector was reconstructed, the optimal threshold was determined, forest cover change information was extracted, and the accuracy of the results was verified.

  Result  (1) The overall accuracy of the test results by PMKT-D was 92.32%, and the Kappa coefficient was 0.843 7, which was significantly better than the control method. (2) From 2010 to 2020, the area of forest cover change in the transition mountain zone of Middle East Yunnan Plateau was 356.88 km², accounting for 1.06% of the total image area, indicating significant changes of forest cover.

  Conclusion  Based on multi-scale segmentation, the method combining TCT and PCA can effectively reduce the adverse effects of complex topography and phenology, enhance spectral characteristics of forest cover changes and improve the accuracy of the change detection significantly while take advantages of object-oriented multi-feature change vector.