Name: Jiraporn Kulsoontornrat
Title: PLANT IDENTIFICATION THROUGH OBJECT-BASED IMAGE ANALYSIS APPROACH
The rule-set have been develop to classify tree species andcompared with the classification between object based image analysis (OBIA) andpixel-based image analysis (PBIA) by using very high spatial resolution;GEOEYE-1 imagery data (0.5 meter). OBIA was employed by Ecognition softwarewhile ENVI 4.7 was applied for PBIA. The study area is Dhammakaya temple, KlongLuang, Thailand. Most of trees were planted for decoration and scenery based onhuman decisions. Study area was divided into three parts such as part1 (a densetree species and two sparse tree species), part2 (diverse of tree species), andpart3 (diverse of dense tree species and a sparse tree species). Additionally,imagery data were produced by 3×3 of sobel operator, NIR (Near-Infrared) and Panchromaticlayers. Also, median 3×3 filter with red, green, blue, NIR, and Panchromaticlayers. The segmentation by multi-scale levels was crated form size, smoothnessand compactness ratio, color and shape ratio. Three levels of segmented toclassify tree and non-tree, density and homogeneity, tree species respectively.The accuracy assessment between OBIA and PBIA are from confusion matrix and TTA(Test area and training mask). Higher accuracy assessment was obtained whenperformed by OBIA: part1, part2, and part3 as following; 100%, 79.34%, and61.26% of OBIA respectively while PBIA offered 86.31 %, 70.74 %, and 25.62 % respectively.