Pölönen I., Sarkeala J., Viitala R. (2015). Unmanned aerial system imagery and photogrammetric canopy height data in area-based estimation of forest variables. Silva Fennica vol. 49 no. 5 article id 1348. 19 p. Highlights• Orthoimage mosaic and 3D canopy height model were derived from UAV-borne colourinfrared digital camera imagery and ALS-based terrain model. • Features extracted from orthomosaic and canopy height data were used for estimating forest variables.• The accuracy of forest estimates was similar to that of the combination of ALS and digital aerial imagery. AbstractIn this paper we examine the feasibility of data from unmanned aerial vehicle (UAV)-borne aerial imagery in stand-level forest inventory. As airborne sensor platforms, UAVs offer advantages cost and flexibility over traditional manned aircraft in forest remote sensing applications in small areas, but they lack range and endurance in larger areas. On the other hand, advances in the processing of digital stereo photography make it possible to produce three-dimensional (3D) forest canopy data on the basis of images acquired using simple lightweight digital camera sensors. In this study, an aerial image orthomosaic and 3D photogrammetric canopy height data were derived from the images acquired by a UAV-borne camera sensor. Laser-based digital terrain model was applied for estimating ground elevation. Features extracted from orthoimages and 3D canopy height data were used to estimate forest variables of sample plots. K-nearest neighbor method was used in the estimation, and a genetic algorithm was applied for selecting an appropriate set of features for the estimation task. Among the selected features, 3D canopy features were given the greatest weight in the estimation supplemented by textural image features. Spectral aerial photograph features were given very low weight in the selected feature set. The accuracy of the forest estimates based on a combination of photogrammetric 3D data and orthoimagery from UAV-borne aerial imaging was at a similar level to those based on airborne laser scanning data and aerial imagery acquired using purpose-built aerial camera from the same study area.

Abstract. Remotely sensing vegetation condition and health hazards requires modeling the connection of plants’ biophysical and biochemical parameters to their spectral response. Even though many models exist already, the field suffers from lack of access to program code. In this study, we will assess the feasibility of open-source 3D-modeling and rendering software Blender in simulating hyperspectral reflectance and transmittance of leaf tissue to serve as a base for a more advanced large-scale simulator. This is the first phase of a larger HyperBlend project, which will provide a fully open-source, canopy scale leaf optical properties model for simulating remotely sensed hyperspectral images. Test results of the current HyperBlend model show good agreement with real-world measurements with root mean squared error around 1‰. The program code is available at https://github.com/silmae/ hyperblend.

Abstract. The idea is to create a self-learning Minimal Learning Machine (MLM) model that is computationally efficient, easy to implement and performs with high accuracy. The study has two hypotheses. Experiment A examines the possibilities of introducing new classes with Recursive Least Squares (RLS) updates for the pre-trained self learning-MLM model. The idea of experiment B is to simulate the push broom spectral imagers working principles, update and test the model based on a stream of pixel spectrum lines on a continuous scanning process. Experiment B aims to train the model with a significantly small amount of labelled reference points and update it continuously with (RLS) to reach maximum classification accuracy quickly.The results show that the new self-learning MLM method can classify new classes with RLS update but with a cost of decreasing accuracy. With a larger amount of reference points, one class can be introduced with reasonable accuracy. The results of experiment B indicate that self-learning MLM can be trained with a few reference points, and the self-learning model quickly reaches accuracy results comparable with nearest-neighbour NN-MLM. It seems that the self-learning MLM could be a comparable machine learning method for the application of hyperspectral imaging and remote sensing.