Estimation of Mango Tree Count and Crown Cover Delineation using Template Matching Algorithm

Abstract

Light detection and Ranging (LiDAR) nowadays is becoming a widely used tool in forest inventory management. LiDAR data has three dimensional structure information that can be used to estimate tree height and crown diameter, the use of this available LiDAR information, availability of image processing software and well developed algorithm allow us to estimate and detect individual tree counts. Several algorithms were developed for different types of remote sensing images, and were tested in different forest areas. However, there were very few studies testing an algorithm in tree crop plantations due to the problems in detecting individual trees affected by varying conditions such as different size of canopy of young and old trees, clumped or clustered trees which sometimes lead to underestimation. This paper aims to estimate the number of trees using LiDAR data and high resolution imagery using a single detection method. The study was conducted in Mangifera indica plantation. The study area is composed of young and matured mango trees having different tree crown size, the planting orientation of mango in the area also vary from isolated and clustered mango tree crown which is a challenge in detecting individual trees. Template matching method in eCognition Developer 9 was applied to different layers derived from point cloud data and high resolution image. Canopy Height Model (CHM) was found to be the optimal layer in creating template sample based on the correlation value of 0.94. The results demonstrated that template matching algorithm can detect the mango tree crowns with 88 % accuracy based on the reference data which is manually delineated. This study suggests that template matching algorithm can be used to estimate mango tree counts. Further study is needed to further improve the algorithm to increase the percent detection. The study also suggested that the method be tested in other types of plantation.