Structure from Motion (SfM) from Unmanned Aerial Vehicles (UAV) is increasingly being utilised for a wide range of applications including characterisation of forests. With continued developments in 3D Vision; namely the Graphics Processing Unit (GPU) and parallel computing, complex image matching algorithms used in SfM which used to be impractical are becoming quite computationally efficient (Leberl et al., 2010). Further to this development is the reality of obtaining a large number of high resolution (>5 megapixels) aerial photographs with a large overlap (>80%) at low cost using small UAVs and consumer-grade cameras. This implies that SfM from UAVs has an actual potential to be a low cost solution for monitoring forests, particularly for developing countries where funding for forest monitoring initiatives such as the Reducing Emissions from Deforestation and Degradation (REDD) may not always be readily available.

The aim of this study was to evaluate SfM from UAVs as a potential low cost method for forest monitoring for developing countries in the context of REDD. This was done by comparing individual tree heights obtained from SfM and from LiDAR and ground measurements.

Study Areas

Two study areas were chosen for this study, Meshaw and the University of Edinburgh Dryden Farm (Figure 1). The Meshaw site is located in the district of Devon, United Kingdom. The area has a forest with a sparse canopy structure which making it ideal for evaluating SfM in open canopy. The Dryden site is located near Roslin in Midlothian, United Kingdom and is a small forest plot (2.3 ha) comprising mainly of Sycamore (Acer pseudoplatanus) and Scots pine (Pinus sylvestris). Compared to the Meshaw site, the Dryden site had a dense canopy structure making it suitable for evaluating the performance of SfM in closed canopies.


Figure 1: Test sites (a) Dryden and (b) Meshaw in the United Kingdom (left). For Meshaw calibration GCPs were obtained from LiDAR DEM and at Dryden GPS was used.



UAV images were collected using a Quest QPod UAV and Sony NEX-7 24.3 megapixel camera at Meshaw, and a DJI Phantom 2 UAV and GoPro Hero 3+ 12.0 megapixel camera at Dryden. LiDAR data for Meshaw was obtained from the Centre for Ecology and Hydrology (CEH) Tellus South West Project website (Ferraccioli et al., 2014), while ground measurements were only done for the Dryden site.

Structure from Motion and Multi-view Stereo (MVS) reconstruction was done using open source software VisualSFM (Wu, 2011) and CMP-MVS (Jancosek & Pajdla, 2011) respectively (Figure 2left). Geo-referencing of the point clouds to the OSGB projection was done in the CloudCompare open source software while Post-processing was done using LAStools (Rapidlasso, 2014) (Figure 2right).


Figure 2: (left)SfM and MVS workflow in VisualSFM, CMP-MVS and CloudCompare. Four different algorithms are executed in VisualSFM to generate a dense point cloud. (right)Summary of point cloud post-processing using LAStools. The commands were executed in a batch script.



SfM point clouds for both sites had very high point densities on average but very few ground points were generated below the canopies (Figure 3). Results obtained at Meshaw showed a strong correlation between SfM and LiDAR digital surface models (R2=0.89) and canopy height models (R2=0.75) (Figure 4). However a poor correlation was observed between SfM tree heights and ground measured heights (R2=0.19) at Dryden.


Figure 3: Penetration of (left) SfM and (right) LiDAR in canopy covered areas at Meshaw. The brown points are all ground points and black lines delimit canopy areas. Point density in canopy areas was 0.27 points/m2 for LiDAR, 1.56 points/m2 for SfM.




This study demonstrated the utility of SfM from UAVs in generating high density point clouds and its potential of to be a low cost EO method for REDD monitoring for developing countries.

Although a number of challenges were observed with this solution, there are also strengths which can be useful for developing nations (Table 1). With continued improvements in the software and sensors, SfM from UAVs can become a real contender to airborne LiDAR for forestry applications in the near future.


Table 1: A summary of strengths and weakness of SfM from UAV against defined criteria.

  • Performs well over bare ground
  • Performs poorly with poor image coverage
  • Cost-effective for small areas
  • Cheap hobbyist UAVs available
  • Open source SfM/MVS software available
  • Open source might not be as accurate as commercial software
  • Cheap camera models (e.g. GoPro) introduce large distortions in SfM models
Ease of use/ Learning curve
  • Full autonomous missions
  • Automated data processing
  • Post-processing still requires experienced users
Amount of data
  • High density point clouds
  • Easy interpretation of point cloud because of true colour rendering
  • Classification of points based only on point height (no return number)



Ferraccioli, F.,Gerard, F.,Robinson, C.,Jordan, T.,Biszczuk, M.,Ireland, L.,Beasley, M.,Vidamour, A.,Barker, A.,Arnold, R.,Dinn, M.,Fox, A.,Howard, A. (2014). LiDAR based Digital Surface Model (DSM) data for South West England. NERC Environmental Information Data Centre. [Online] Available at: 4e31-8506-cabe899f989a (Accessed 10 June, 2015).

Jancosek, M., & Pajdla, T. (2011). Multi-view reconstruction preserving weakly-supported surfaces. In Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on (3121-3128). IEEE.

Leberl, F., Irschara, A., Pock, T., Meixner, P., Gruber, M., Scholz, S., & Wiechert, A. (2010). Point Clouds. Photogrammetric Engineering & Remote Sensing, 76(10), 1123-1134.

Rapidlasso (2014). LAStools: converting, filtering, viewing, processing and compressing LiDAR data in LAS format. [Online] Available at: (Last Accessed 4th August, 2015).

Wu, C. (2011). VisualSFM: A visual structure from motion system. [Online] Available at: (Last Accessed 15th July, 2015).