Identifying tree crown areas in undulating eucalyptus plantations using JSEG multi-scale segmentation and unmanned aerial vehicle near-infrared imagery

In the present study, we propose an identification method for tree crown areas from imagery captured by a near-infrared (NIR) camera mounted on an unmanned aerial vehicle (UAV) platform over an undulating area planted with eucalyptus in Guangdong Province, China. The method extracts the eucalyptus crown areas by applying mathematical morphology, unsupervised segmentation based on J-value segmentation, local spatial statistics, and Iterative Self-Organising Data Analysis Technique Algorithm. Two morphology filters and four segmentation scales were compared in densely sunlit, sparsely sunlit, densely shaded, and sparsely shaded plantation plots. Analyses of variance (ANOVAs) revealed that the identification of crowns is more accurate in sparse plantation areas than in dense plantation areas, and more accurate in sunlit plantation areas than in shaded plantation areas. Although there is not an optimal morphological filtering operation for identifying crown areas, the combination of opening operation by the kernel size of 9 × 9 pixels and segmentation by the seed area sized 65 × 65 pixels achieved the best performance with overall accuracies (OAs) of 91%, 93%, 89%, and 91% in densely sunlit, sparsely sunlit, densely shaded, and sparsely shaded plots, respectively. Based on these results, our method could produce an acceptable tree crown area identification technique for eucalyptus trees.