Recognition and Possible Remediation of Automated Tree Delineations with Multiple Isolations per Tree (Split Cases) on High-Resolution Imagery

Abstract

Individual tree crown analysis from high-resolution imagery is gaining greater use in forest applications. Automated crown delineations (isols) that are poor can cause errors in species classification and inventory estimates. This study explores the issue of recognizing split cases (e.g., tree crowns oversegmented into several isols) and demonstrates 3 remediation procedures to improve delineations. Several methods for identifying split cases are proposed, but a conceptual framework for a template-matching approach is developed further. Candidate split cases are identified where there is a good match of a template model representing the appearance of trees with the imagery, and several isols are within the template. Candidates are further analyzed through evidence such as isol shape, species class, and match of templates centered on each isol. Procedures were demonstrated with a typical individual crown isolation on 40 cm multispectral imagery of a mixed species forest in northeastern Ontario. The process showed useful effectiveness in improving the isolation, with expected omission rates of 15%–20% and 25%–30% false alarms. Overall, almost all true split cases recognized had improved crown delineations. The work shows that approaches for recognizing and remediating split cases are possible, but will have to be complex and consider multiple evidence.

Résumé

Les analyses individuelles des couronnes d’arbres provenant de l'imagerie à haute résolution est de plus en plus utilisée dans les applications forestières. Les délimitations automatiques des couronnes (isols) qui sont imprécises peuvent provoquer des erreurs dans la classification des espèces et sur les estimations de l’inventaire. Cet article explore la question de la reconnaissance des cas fractionnés (par exemple, les couronnes d'arbres qui ont été sursegmentées en plusieurs isols) et montre 3 procédures de correction pour améliorer les délimitations. Plusieurs méthodes pour l'identification des cas fractionnés sont proposées, mais un cadre conceptuel pour une approche d'adaptation de gabarit est davantage développé. Des cas fractionnés candidats sont identifiés là où il y a une bonne correspondance entre le gabarit représentant l'apparence des arbres dans l'imagerie, et plusieurs isols sont dans le gabarit. Les candidats sont analysés plus en profondeur par des preuves telles que la forme de l’isol, la classe des espèces et les concordances des modèles centrés sur chaque isol. Les procédures ont été démontrées avec un isolement de la couronne individuelle typique sur l'imagerie multispectrale de 40 cm pour une forêt mixte dans le nord-est de l'Ontario. Le processus a montré une efficacité pour améliorer l'isolement, avec des taux d'omission attendus de 15%–20% et 25%–30% de fausses alarmes. Dans l'ensemble, la quasi-totalité des cas fractionnés réels observés a montré une amélioration des délimitations des couronnes. Ce travail a montré que des approches pour la reconnaissance et la remédiation des cas de partage sont possibles, mais devront être complexes et devront examiner de multiples preuves.

ACKNOWLEDGMENTS

Kathleen Oddleifson, Victoria Rogers, Derrick Plotsky, Laura Johnson, and Dave Johnson, all formerly with the Canadian Forest Service (CFS), Natural Resources Canada, made significant contributions to the analysis of the data. Ashlin Richardson and Dr. Rajeev Sharma assisted in exploring the isol cluster pattern and isol rerun techniques for split case identification. Stephen Gray of the CFS helped with the manual tree delineations. Field work logistics were supported by the Petawawa Research Forest (PRF) of the Canadian Wood Fibre Centre, Canadian Forest Service. Dr. Murray Woods of Ontario Ministry Natural Resources and Jason Bernard of PRF assisted with some of the field work. Dr. Darren Pouliot of Canada Centre for Mapping and Earth Observation, Natural Resources Canada, and Hao Chen of the CFS are thanked for their review of a previous version of this article. Joanne White and Sally Tinis of the CFS also helped review aspects of the manuscript. The work is part of a larger effort to develop enhanced forest inventories, led and supported by the Canadian Wood Fibre Centre.

Notes

Nine had no planarized template associated with them, 38 had planarized templates of very good quality for matching the tree and for split case identification, 31 of good quality, 16 moderate, 11 poor, and 5 very poor quality.

Of the 101 split test cases with templates, 51 had the planarized template as the best; 20 had similar qualities of template match among the planarized and best alternate templates. Of the 30 remaining cases in which the alternate template was better, 7 had planarized templates that were, nevertheless, of good quality.

Fifty would be of very good quality, 37 good, 10 moderate, only 4 poor, and none very poor.