In forest management, knowledge about a forest's distribution of tree species is key. Being able to automate tree species classification for large forest areas is of great interest, since it is tedious and costly labour doing it manually. In this project, the aim was to investigate the efficiency of classifying individual tree species (pine, spruce and deciduous forest) from 3D point clouds acquired by airborne laser scanning (ALS), using convolutional neural networks. Raw data consisted of 3D point clouds and photographic images of forests in northern Sweden, collected from a helicopter flying at low altitudes. The point cloud of each individual tree was connected to its representation in the photos, which allowed for manual labeling of training data to be used for training of convolutional neural networks. The training data consisted of labels and 2D projections created from the point clouds, represented as images. Two different convolutional neural networks were trained and tested; an adaptation of the LeNet architecture and the ResNet architecture. Both networks reached an accuracy close to 98 %, the LeNet adaptation having a slightly lower loss score for both validation and test data compared to that of ResNet. Confusion matrices for both networks showed similar F1 scores for all tree species, between 97 % and 98 %. The accuracies computed for both networks were found higher than those achieved in similar studies using ALS data to classify individual tree species. However, the results in this project were never tested against a true population sample to confirm the accuracy. To conclude, the use of convolutional neural networks is indeed an efficient method for classification of tree species, but further studies on unbiased data is needed to validate these results.

In this paper the brief overview of methods available for individual tree stems quality evaluation and tree species classification has been performed. The use of Near Infrared photometry based on conifer’s canopy reflectance measurement in near infrared range of spectrum has been evaluated for the use in autonomous forest harvesting. Photometric method based on the image processing of the bark pattern has been proposed to perform classification between main construction timber tree species in Scandinavia: Norway spruce (Picea abies) and Scots Pine (Pinus sylvestris). Several feature extraction algorithms have been evaluated, resulting two methods selected: Statistical Analysis using gray level co-occurrence matrix and maximally stable extremal regions feature detector. Feedforward Neural Network with Backpropagation training algorithm and Support Vector Machine classifiers have been implemented and compared. The verification of the proposed algorithm has been performed by real-time testing.

Vorliegende Arbeit zeigt die jüngsten Möglichkeiten hochauflösender Fernerkundung am Beispiel von Stadtbäumen in Berlin, Deutschland. Es wurden neuste methodische Ansätze eingesetzt, wie beispielsweise maschinelles Lernens und individuelle Baumdetektion. Sie erwiesen sich von großem Vorteil für die detaillierte Analyse urbaner Ökosystemdienstleistungen in einer heterogenen Umwelt. Neueste Fernerkundung von hoher zeitlicher Auflösung hat Möglichkeiten gezeigt, Veränderungen des Stadtwaldes präziser zu untersuchen. Diesbezüglich konnten Baumspezies klassifiziert werden auf Grundlage saisonaler Veränderungen, die mittels Fernerkundungsdaten aufgenommen wurden. Dies ist für den urbanen Bereich einmalig und über große Flächen noch nicht durchgeführt worden. Darüber hinaus haben diese Baumarten einzelnen Bäumen zugeordnet werden können, deren Abmessung fernerkundlich erfasst worden ist. Diese neu erzeugten Umweltinformationen einzelner Bäume können damit verbundene urbane Ökosystemdienstleistungen präzise aktualisieren. Zum Beispiel haben so Unsicherheiten in der Schätzung zur Kohlenstoffspeicherung städtischer Wälder reduziert werden können. Es ist zudem von Vorteil gewesen, den gegenwärtigen Mangel an räumlich expliziten dreidimensionalen Informationen über Stadtwälder anzusprechen. Allerdings ist die Rolle städtischen Wälder, das Treibhausgas CO2 langfristig auszugleichen, immer noch wenig untersucht. Gerade der Mangel an präzisen, konsistenten und aktuellen Details führt zu großen Unsicherheiten im Rahmen von Lebenszyklus-Analysen. Auf Grund des aktuellen Fortschritts in hochauflösender Fernerkundung könnten diese Unsicherheiten reduziert werden. Dazu werden Möglichkeiten ausgiebig kritisch bewertet und anhand einer Lebenszyklus-Analyse am Beispiel Berlin andiskutiert, inwieweit sie präzisere langfristige Prognosen zum Stadtwald als Kohlenstoffspeicher liefern. / This work shows recent options for implementing high resolution remote sensing in assessing urban trees in Berlin, Germany. State-of-the-art methodological approaches like machine learning and individual tree detection proved to be highly advantageous for analyzing details of urban ecosystem services within a heterogeneous urban environment. Recent remote sensing of high temporal resolution offers new options for more precisely addressing urban forest dynamics. This successfully shows that tree species could be identified from seasonal changes of remotely sensed imagery, though this has not yet been applied across cities. Furthermore, these tree species results could be combined with remotely sensed individual tree dimensions. This newly generated data can be suggested to update spatially explicit information on related urban ecosystem services. For example, this could reduce the uncertainties of such estimates as urban forest carbon storage, and also address the present lack of spatially explicit three-dimensional information on urban forests. However, few studies have considered the local scale of urban forests to effectively evaluate their potential long-term carbon offset. The lack of precise, consistent and up-to-date forest details is challenging within the scope of life cycle assessments. This can cause high uncertainties in urban forest carbon offset. Although, recent progress in high resolution remote sensing is promising to reduce these uncertainties. For this purpose, remote sensing options are extensively reviewed and briefly discussed using an example of life cycle assessment for Berlin, which allow more precise long-term prognoses of urban forest carbon offset.