Algorithm: TD-ADO

1: Input: TJi = {P1, P2, …, Pm}, angle offset threshold θth, distance offset threshold dth; 2: Output: Feature Points FP; 3: P1→FP; 4: i = 1; 5: repeat 6: initial movement behavior = PiPi+1; 7: for j = 1 to m−2 do 8:  if θj > θth || dj > dth then 9:   Pj+2→FP; 10:   i = j+2; 11:   break; 12:  else 13:   if j = m−2 then 14:    Pj+2→FP; 15:   end if 16:  end if 17: end for 18: until Pm→FP.

19: return FP.