Sakari Tuominen

Airborne laser scanning and digital stereo imagery measures of forest structure: comparative results and implications to forest mapping and inventory update

Airborne laser scanning (ALS) has demonstrated utility for forestry applications and has renewed interest in other forms of remotely sensed data, especially those that capture three-dimensional (3-D) forest characteristics. One such data source results from the advanced processing of high spatial resolution digital stereo imagery (DSI) to generate 3-D point clouds. From the derived point cloud, a digital surface model and forest vertical information with similarities to ALS can be generated. A key consideration is that when developing forestry related products such as a canopy height model (CHM), a high spatial resolution digital terrain model (DTM), typically from ALS, is required to normalize DSI elevations to heights above ground. In this paper we report on our investigations into the use of DSI-derived vertical information for capturing variations in forest structure and compare these results to those acquired using ALS. An ALS-derived DTM was used to provide the spatially detailed ground surface elevations to normalize DSI-derived heights. Similar metrics were calculated from the vertical information provided by both DSI and ALS. Comparisons revealed that ALS metrics provided a more detailed characterization of the canopy surface including canopy openings. Both DSI and ALS metrics had similar levels of correlation with forest structural attributes (e.g., height, volume, and biomass). DSI-based models predicted height, diameter, basal area, stem volume, and biomass with root mean square (RMS) accuracies of 11.2%, 21.7%, 23.6%, 24.5%, and 23.7%, respectively. The respective accuracies for the ALS-based predictions were 7.8%, 19.1%, 17.8%, 17.9%, and 17.5%. Change detection between ALS-derived CHM (time 1) and DSI-derived CHM (time 2) provided change estimates that demonstrated good agreement (r = 0.71) with two-date, ALS only, change outputs. For the single-layered, even-aged stands under investigation in this study, the DSI-derived vertical information is an appropriate and cost-effective data source for estimating and updating forest information. The accuracy of DSI information is based on a capability to measure the height of the upper canopy envelope with performance analogous to ALS. Forest attributes that are well captured and subsequently modeled from height metrics are best suited to estimation from DSI metrics, whereas ALS is more suitable for capturing stand density. Further investigation is required to better understand the performance of DSI-derived height products in more complex forest environments. Furthermore, the difference in variance captured between ALS and DSI-derived CHM also needs to be better understood in the context of change detection and inventory update considerations.

Unmanned Aerial Vehicle (UAV) operated spectral camera system for forest and agriculture applications

VTT Technical Research Centre of Finland has developed a Fabry-Perot Interferometer (FPI) based hyperspectral imager compatible with the light weight UAV platforms. The concept of the hyperspectral imager has been published in the SPIE Proc. 7474 and 7668. In forest and agriculture applications the recording of multispectral images at a few wavelength bands is in most cases adequate. The possibility to calculate a digital elevation model of the forest area and crop fields provides means to estimate the biomass and perform forest inventory. The full UAS multispectral imaging system will consist of a high resolution false color imager and a FPI based hyperspectral imager which can be used at resolutions from VGA (480 x 640 pixels) up to 5 Mpix at wavelength range 500 - 900 nm at user selectable spectral resolutions in the range 10...40 nm @ FWHM. The resolution is determined by the order at which the Fabry- Perot interferometer is used. The overlap between successive images of the false color camera is 70...80% which makes it possible to calculate the digital elevation model of the target area. The field of view of the false color camera is typically 80 degrees and the ground pixel size at 150 m flying altitude is around 5 cm. The field of view of the hyperspectral imager is presently is 26 x 36 degrees and ground pixel size at 150 m flying altitude is around 3.5 cm. The UAS system has been tried in summer 2011 in Southern Finland for the forest and agricultural areas. During the first test campaigns the false color camera and hyperspectral imager were flown over the target areas at separate flights. The design and calibration of the hyperspectral imager will be shortly explained. The test flight campaigns on forest and crop fields and their preliminary results are also presented in this paper.

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

Small unmanned aerial vehicle (UAV) based remote sensing is a rapidly evolving technology. Novel sensors and methods are entering the market, offering completely new possibilities to carry out remote sensing tasks. Three-dimensional (3D) hyperspectral remote sensing is a novel and powerful technology that has recently become available to small UAVs. This study investigated the performance of UAV-based photogrammetry and hyperspectral imaging in individual tree detection and tree species classification in boreal forests. Eleven test sites with 4151 reference trees representing various tree species and developmental stages were collected in June 2014 using a UAV remote sensing system equipped with a frame format hyperspectral camera and an RGB camera in highly variable weather conditions. Dense point clouds were measured photogrammetrically by automatic image matching using high resolution RGB images with a 5 cm point interval. Spectral features were obtained from the hyperspectral image blocks, the large radiometric variation of which was compensated for by using a novel approach based on radiometric block adjustment with the support of in-flight irradiance observations. Spectral and 3D point cloud features were used in the classification experiment with various classifiers. The best results were obtained with Random Forest and Multilayer Perceptron (MLP) which both gave 95% overall accuracies and an F-score of 0.93. Accuracy of individual tree identification from the photogrammetric point clouds varied between 40% and 95%, depending on the characteristics of the area. Challenges in reference measurements might also have reduced these numbers. Results were promising, indicating that hyperspectral 3D remote sensing was operational from a UAV platform even in very difficult conditions. These novel methods are expected to provide a powerful tool for automating various environmental close-range remote sensing tasks in the very near future.

Comparing Accuracy of Airborne Laser Scanning and TerraSAR-X Radar Images in the Estimation of Plot-Level Forest Variables

Abstract: In this study we compared the accuracy of low-pulse airborne laser scanning (ALS) data, multi-temporal high-resolution noninterferometric TerraSAR-X radar data and a combined feature set derived from these data in the estimation of forest variables at plot level. The TerraSAR-X data set consisted of seven dual-polarized (HH/HV or VH/VV) Stripmap mode images from all seasons of the year. We were especially interested in distinguishing between the tree species. The dependent variables estimated included mean volume, basal area, mean height, mean diameter and tree species-specific mean volumes. Selection of best possible feature set was based on a genetic algorithm (GA). The nonparametric k -nearest neighbour ( k -NN) algorithm was applied to the estimation. The research material consisted of 124 circular plots measured at tree level and located in the vicinity of Espoo, Finland. There are large variations in the elevation and forest structure in the study area, making it demanding for image interpretation. The best feature set contained

Segment-level stand inventory for forest management

The aim of this study was to develop a method for segment-based forest inventory and determine whether segment-level inventories can be used in forest management planning. The study area covered 76 ha located in two different aerial photographs in eastern Finland. The study area was segmented into 220 segments with the aid of aerial photographs and the segment-level forest characteristics were assessed in the field using relascope sample plots and a field computer which displayed the aerial photographs, segment borders and surveyors location on the screen. The segment estimates were calculated as weighted averages of k nearest neighbours (kNN) for the segments and the sample plots. The estimates were tested with a cross-validation technique. The averages and the standard deviations of the spectral values of aerial images extracted for the segments and the sample plots were used in the kNN estimation. The relative root mean square error of the mean volume was 58.1% (bias –6.4%) at the segment level and 57.9% (bias –0.9%) at the sample plot level. The segment-based approach studied here needs further research and improvement before it can be applied to forest management planning.

Data Combination and Feature Selection for Multi-source Forest Inventory

Both satellite images and aerial photographs are now used operationally in Finland’s forestry for different tasks; satellite images are used for national forest inventory purposes and aerial images for forest management planning. Due to the double coverage, it could be advantageous to utilize the strengths of both image types. The aim of this study was to evaluate the potential of

Comparison of Grid-Based and Segment-Based Estimation of Forest Attributes Using Airborne Laser Scanning and Digital Aerial Imagery

Forest management planning in Finland is currently adopting a new-generation forest inventory method, which is based on interpretation of airborne laser scanning data and digital aerial images. The inventory method is based on a systematic grid, where the grid elements serve as inventory units, for which the laser and aerial image data are extracted and the forest variables estimated. As an alternative or a complement to the grid elements, image segments can be used as inventory units. The image segments are particularly useful as the basis for generation of the silvicultural treatment and cutting units since their boundaries should follow the actual stand borders, whereas when using grid elements it is typical that some of them cover parts of several forest stands. The proportion of the so-called mixed cells depends on the size of the grid elements and the average size and shape of the stands. In this study, we carried out automatic segmentation of two study areas on the basis of laser and aerial image data with a view to delineating micro-stands that are homogeneous in relation to their forest attributes. Further, we extracted laser and aerial image features for both systematic grid elements and segments. For both units, the feature set used for estimating the forest attributes was selected by means of a genetic algorithm. Of the features selected, the majority (61–79%) were based on the airborne laser scanning data. Despite the theoretical advantages of the image segments, the laser and aerial features extracted from grid elements seem to work better than features extracted from image segments in estimation of forest attributes. We conclude that estimation should be carried out at grid level with an area-specific combination of features and estimates for image segments to be derived on the basis of the grid-level estimates.

Accuracy of High-Resolution Radar Images in the Estimation of Plot-Level Forest Variables

In the present study, we used the airborne E-SAR radar to simulate the satellite-borne high-resolution TerraSAR radar data and determined the accuracy of the plot-level forest variable estimates produced. Estimation was carried out using the nonparametric k-nearest neighbour (k-nn) method. Variables studied included mean volume, tree species-specific volumes and their proportions of total volume, basal area, mean height and mean diameter. E-SAR-based estimates were compared with those obtained using aerial photographs and medium-resolution satellite image (Landsat ETM+) recording optical wavelength energy. The study area was located in Kirkkonummi, southern Finland. The relative RMSEs for ESAR were 45%, 29%, 28% and 38% for mean volume, mean diameter, mean height and basal area, respectively. For aerial photographs these were 51%, 26%, 27% and 42%, and for Landsat ETM+ images 58%, 40%, 35% and 49%. Combined datasets outperformed all single-source datasets, with relative RMSEs of 26%, 23%, 33% and 39%. Of the single-source datasets, the E-SAR images were well suited for estimating mean volume, while for mean diameter, mean height and basal area the E-SAR and aerial photographs performed similarly and far better than Landsat ETM+. The aerial photographs succeeded well in the estimation of species-specific volumes and their proportions, but the combined dataset was still significantly better in volume proportions. Due to its good temporal resolution, satellite-borne radar imaging is a promising data source for forest inventories, both in large-area forest inventories and operative forest management planning. Future high-resolution synthetic aperture radar (SAR) images could be combined with airborne laser scanner data when estimating forest or even tree characteristics.

Assessment of Classifiers and Remote Sensing Features of Hyperspectral Imagery and Stereo-Photogrammetric Point Clouds for Recognition of Tree Species in a Forest Area of High Species Diversity

Recognition of tree species and geospatial information on tree species composition is essential for forest management. In this study, tree species recognition was examined using hyperspectral imagery from visible to near-infrared (VNIR) and short-wave infrared (SWIR) camera sensors in combination with a 3D photogrammetric canopy surface model based on RGB camera stereo-imagery. An arboretum with a diverse selection of 26 tree species from 14 genera was used as a test area. Aerial hyperspectral imagery and high spatial resolution photogrammetric color imagery were acquired from the test area using unmanned aerial vehicle (UAV) borne sensors. Hyperspectral imagery was processed to calibrated reflectance mosaics and was tested along with the mosaics based on original image digital number values (DN). Two alternative classifiers, a k nearest neighbor method (k-nn), combined with a genetic algorithm and a random forest method, were tested for predicting the tree species and genus, as well as for selecting an optimal set of remote sensing features for this task. The combination of VNIR, SWIR, and 3D features performed better than any of the data sets individually. Furthermore, the calibrated reflectance values performed better compared to uncorrected DN values. These trends were similar with both tested classifiers. Of the classifiers, the k-nn combined with the genetic algorithm provided consistently better results than the random forest algorithm. The best result was thus achieved using calibrated reflectance features from VNIR and SWIR imagery together with 3D point cloud features; the proportion of correctly-classified trees was 0.823 for tree species and 0.869 for tree genus.

Methods for estimating forest stem volumes by tree species using digital surface model and CIR images taken from light UAS

In this paper we consider methods for estimating forest tree stem volumes by species using images taken from light unmanned aircraft systems (UAS). Instead of using LiDAR and additional multiband imagery a color infrared camera mounted to a light UAS is used to acquire both imagery and the DSM of target area. The goal of this study is to accurately estimate tree stem volumes in three classes. The status of the ongoing work is described and an initial method for delineating and classifying treetops is presented.