Discrimination of Deciduous Tree Species from Time Series of Unmanned Aerial System Imagery

Abstract

Technology advances can revolutionize Precision Forestry by providing accurate and fine forest information at tree level. This paper addresses the question of how and particularly when Unmanned Aerial System (UAS) should be used in order to efficiently discriminate deciduous tree species. The goal of this research is to determine when is the best time window to achieve an optimal species discrimination. A time series of high resolution UAS imagery was collected to cover the growing season from leaf flush to leaf fall. Full benefit was taken of the temporal resolution of UAS acquisition, one of the most promising features of small drones. The disparity in forest tree phenology is at the maximum during early spring and late autumn. But the phenology state that optimized the classification result is the one that minimizes the spectral variation within tree species groups and, at the same time, maximizes the phenologic differences between species. Sunlit tree crowns (5 deciduous species groups) were classified using a Random Forest approach for monotemporal, two-date and three-date combinations. The end of leaf flushing was the most efficient single-date time window. Multitemporal datasets definitely improve the overall classification accuracy. But single-date high resolution orthophotomosaics, acquired on optimal time-windows, result in a very good classification accuracy (overall out of bag error of 16%).

Individual Tree Detection and Classification with UAV-Based Photogrammetric Point Clouds and Hyperspectral Imaging

“…Previous UAV-based studies have been based on only multispectral data captured using NIR-modified cameras. Lisein et al [41] performed classification using RandomForest and achieved an overall classification error of 16% (out of bag error) with five different tree species. Michez et al [39] achieved a classification accuracy of 84% with five different tree species but the results were closer to pixel-based classification where a single tree was not represented as a whole.…”

“…Multispectral data Remote Sens. 2017, 9, 185 3 of 34 acquired from UAVs has been used in tree species classification, but the data has been limited to RGB images and one near-infrared (NIR) channel using NIR-modified cameras [39][40][41]. The development of low-weight hyperspectral imaging sensors is likely to increase the use of spectral data collected from UAVs in forestry applications [34].…”

Mapping forest attributes using data from stereophotogrammetry of aerial images and field data from the national forest inventory

“…3). Some of the high volume stands had large deciduous stem volume and Q2 images were acquired during leaf-off season, this affect the image matching to generate points mostly on the ground and with more noise and thus the canopy height metric values negatively Lisein et al 2015;Bohlin et al 2015), clearly underestimating raster cells with high deciduous proportion. Another reason for underestimation is that the mature stands, which are under estimated (Fig.…”

Estimating tree heights with images from an unmanned aerial vehicle

“…They claimed that UAV flights with short time intervals can refresh LiDAR CHMs in order to create canopy height series. Lisein et al (2015) reported classification of deciduous trees with UAV imagery with high classification accuracy. Zhang et al (2016) showed that long-term ecological monitoring can be effectively done by UAV systems with high resolution data.…”