A review of methods for automatic individual tree-crown detection and delineation from passive remote sensing

Efficient forest management demands detailed, timely information. As high spatial resolution remotely sensed imagery becomes more available, there is a great potential for conducting high accuracy forest inventory and analysis automatically and cost-efficiently. Recent research aimed at providing tree-based forest inventory measurements has generated numerous algorithms for automatic individual tree-crown detection and delineation. This article reviews this research with a focus on algorithms applied to passive remote-sensing imagery. The article categorizes and evaluates methods for automatic tree-crown detection and delineation. It considers the types of imagery and the characteristics of the study areas these algorithms are applied to and evaluates the influence of these factors on the methods. The article also reviews and evaluates quantitative accuracy assessment methods for tree-crown delineation and detection. Finally, the article summarizes the commonalities of current algorithms, and the new development that can be expected in the future.