FOREST SPECIES CLASSIFICATION BASED ON STATISTICAL POINT PATTERN ANALYSIS USING AIRBORNE LIDAR DATA
- 1Dept. of Earth and Space Science and Engineering, York University, Toronto, ON, M3J1P3, Canada
- 2Ontario Ministry of Natural Resources, 1235 Queen St. East, Sault Ste. Marie, ON, P6A 2E5, Canada
Keywords: Remote Sensing, LiDAR, Point Pattern Analysis, K-function, PCA, Classification
Abstract. The paper investigated the effectiveness of point pattern methods in the application of forest species classification using airborne LiDAR data. The forest stands and individual trees in our study area were classified as either shade tolerant or intolerant species. The purpose of adopting the point pattern methods is to develop new features to effectively characterize the pattern of internal foliage distribution of forest stands or individual trees. Three methods including Quadrat Count, Ripley's K-function, and Delaunay Triangulation were applied, and six feature groups were derived for a stand or tree sample. Feature selection was performed based on the derived features in order to find the best ones for the following classification procedure, which was implemented by two supervised and two unsupervised methods. These newly derived features were proved effective for the classification. The highest classification accuracy 97% was achieved at stand level and 90% at individual tree level. The sensitivity of classification accuracy to the number of features used was also investigated in this paper.