Estimation of Shade Tree Density in Tea Garden using Remote Sensing Images and Deep Convolutional Neural Network
Arati Paul, Sayari Bhattacharyya, Debasish Chakraborty
Abstract
A specific amount of shade tree density is essential for quality tea production. Here, deep convolutional neural network (DCNN) based architectures are used for detecting and measuring the canopy area of shade trees in high-resolution remote sensing (RS) images covering tea gardens with precision, recall, F1 score and Intersection-over-Union value of 98.9%, 85.1%, 91.36 and 0.96 respectively. Subsequently, shade tree density is estimated with average error of 0.03. In the present paper a fully automated DCNN-based process is established which not only detects shade trees in RS imagery, but also estimates their canopy density for assisting tea garden management.
Topics & Concepts
Convolutional neural networkCanopyTree (set theory)Tree canopyRemote sensingIntersection (aeronautics)Artificial intelligenceGeographyComputer scienceMathematicsForestryCartographyArchaeologyMathematical analysisRemote Sensing in AgricultureRemote Sensing and LiDAR ApplicationsRemote Sensing and Land Use