Remote sensing and probabilistic sampling based forest inventory method
View Patent ↗A remote sensing and probabilistic sampling based forest inventory method can correlate aerial data, such as LiDAR, CIR, and/or Hyperspectral data with actual sampled and measured ground data to facilitate obtainment, e.g., prediction, of a more accurate forest inventory. The resulting inventory can represent an empirical description of the height, DBH and species of every tree within the sample area. The use of probabilistic sampling methods can greatly improve the accuracy and reliability of the forest inventory.
1. A computer-implemented forest inventory method comprising:
a. processing imagery data using a computer system, said imagery data indicative of tree attribute information for said forest;
b. using said computer system, classifying tree polygons within said imagery data to derive said tree attribute information;
c. correlating field data using said computer system, said field data comprising at least one of actual tree attribute information and plot center location;
d. using said computer system, generating a correlated model utilizing said tree attribute information derived from said imagery data and said actual tree attribute information; and
e. generating a probabilistic forest inventory by applying said correlated model to all said imagery data using said computer system;
f. wherein said classifying tree polygons comprises:
i. superimposing at least one of CIR photography, stand shapes, tree crown polygon shapes and LiDAR data;
ii. creating a classifier formula using discriminant analysis;
iii. classifying polygons for all stands and strata;
iv. manually reviewing species strata;
v. calculating an average CIR band for individual tree crowns;
vi. calculating a second order gray level texture feature;
vii. selecting a subset of stands for classification; and
viii. creating a training set for species at strata level.
2. A computer-implemented forest inventory method comprising:
a. processing imagery data using a computer system, said imagery data indicative of tree attribute information for said forest;
b. using said computer system, classifying tree polygons within said imagery data to derive said tree attribute information;
c. correlating field data using said computer system, said field data comprising at least one of actual tree attribute information and plot center location;
d. using said computer system, generating a correlated model utilizing said tree attribute information derived from said imagery data and said actual tree attribute information; and
e. generating a probabilistic forest inventory by applying said correlated model to all said imagery data using said computer system;
f. wherein correlating said field data further comprises:
i. defining a sampling frame within said imagery data;
ii. wherein said field data is collected from a field plot corresponding to said sampling frame;
iii. determining a geo-referenced plot center such that said field plot corresponds to said sampling frame; and
iv. correcting said plot center based upon field measurements; and
g. using said computer system, correcting said plot center location to a location that results in the best match between tree locations indicated by said remote sensing data and tree locations verified by field plot measurements.