The invention discloses a fast-growing forest
felling detection method and a fast-growing forest
felling detection device taking sentinel No.1
time sequence autocorrelation into account, aiming at the problem that the
false alarm rate of a traditional difference value accumulation and
felling detection method is relatively high due to the fact that sentinel No.1 back scattering data generates obvious short-term time autocorrelation due to the influence of rainfall, soil
moisture and
surface moisture content change, and discloses the fast-growing forest felling detection method and the fast-growing forest felling detection device taking sentinel No.1
time sequence autocorrelation into account. An improved difference cumulative sum detection method fusing a
sliding time window and a random block
resampling significance test is provided, the method uses the
sliding time window to calculate a local difference cumulative sum curve, and candidate
change points are extracted based on a maximum value; in order to overcome statistical test bias errors caused by
time sequence self-correlation, a random block
resampling strategy retaining a time sequence local dependence structure is innovatively introduced, and range zero
hypothesis distribution for significance test is constructed. According to the method, the precision and robustness of automatic identification of short-period fast-growing forest felling events in a large-range and long-time-sequence monitoring scene are effectively improved.