Accurately retrieving tree stem location distributions is a basic requirement for biomass estimation of forest inventory. Combining Inertial Measurement Units (IMU) with Global Navigation Satellite Systems (GNSS) is a commonly used positioning strategy in most Mobile Laser Scanning (MLS) systems for accurate forest mapping. Coupled with a tactical or consumer grade IMU, GNSS offers a satisfactory solution in open forest environments, for which positioning accuracy better than one decimeter can be achieved. However, for such MLS systems, positioning in a mature and dense forest is still a challenging task because of the loss of GNSS signals attenuated by thick canopy. Most often laser scanning sensors in MLS systems are used for mapping and modelling rather than positioning. In this paper, we investigate a Simultaneous Localization and Mapping (SLAM)-aided positioning solution with point clouds collected by a small-footprint LiDAR. Based on the field test data, we evaluate the potential of SLAM positioning and mapping in forest inventories. The results show that the positioning accuracy in the selected test field is improved by 38% compared to that of a traditional tactical grade IMU + GNSS positioning system in a mature forest environment and, as a result, we are able to produce a unambiguous tree distribution map.

Unmanned airborne vehicles (UAV) equipped with novel, miniaturized, 2D frame format hyper-and multispectral cameras make it possible to conduct remote sensing measurements cost-efficiently, with greater accuracy and detail. In the mapping process, the area of interest is covered by multiple, overlapping, small-format 2D images, which provide redundant information about the object. Radiometric correction of spectral image data is important for eliminating any external disturbance from the captured data. Corrections should include sensor, atmosphere and view/illumination geometry (bidirectional reflectance distribution function-BRDF) related disturbances. An additional complication is that UAV remote sensing campaigns are often carried out under difficult conditions, with varying illumination conditions and cloudiness. We have developed a global optimization approach for the radiometric correction of UAV image blocks, a radiometric block adjustment. The objective of this study was to implement and assess a combined adjustment approach, including comprehensive consideration of weighting of various observations. An empirical study was carried out using imagery captured using a hyperspectral 2D frame format camera of winter wheat crops. The dataset included four separate flights captured during a 2.5 h time period under sunny weather conditions. As outputs, we calculated orthophoto mosaics using the most nadir images and sampled multiple-view hyperspectral spectra for vegetation sample points utilizing multiple images in the dataset. The method provided an automated tool for radiometric correction, compensating for efficiently radiometric disturbances in the images. The global homogeneity factor improved from 12-16% to 4-6% with the corrections, and a reduction in disturbances could be observed in the spectra of the object points sampled from multiple overlapping images. Residuals in the grey and white reflectance panels were less than 5% of the reflectance for most of the spectral bands.

Interest in drone solutions in forestry applications is growing. Using drones, datasets can be captured flexibly and at high spatial and temporal resolutions when needed. In forestry applications, fundamental tasks include the detection of individual trees, tree species classification, biomass estimation, etc. Deep neural networks (DNN) have shown superior results when comparing with conventional machine learning methods such as multi-layer perceptron (MLP) in cases of huge input data. The objective of this research is to investigate 3D convolutional neural networks (3D-CNN) to classify three major tree species in a boreal forest: pine, spruce, and birch. The proposed 3D-CNN models were employed to classify tree species in a test site in Finland. The classifiers were trained with a dataset of 3039 manually labelled trees. Then the accuracies were assessed by employing independent datasets of 803 records. To find the most efficient set of feature combination, we compare the performances of 3D-CNN models trained with hyperspectral (HS) channels, Red-Green-Blue (RGB) channels, and canopy height model (CHM), separately and combined. It is demonstrated that the proposed 3D-CNN model with RGB and HS layers produces the highest classification accuracy. The producer accuracy of the best 3D-CNN classifier on the test dataset were 99.6%, 94.8%, and 97.4% for pines, spruces, and birches, respectively. The best 3D-CNN classifier produced ~5% better classification accuracy than the MLP with all layers. Our results suggest that the proposed method provides excellent classification results with acceptable performance metrics for HS datasets. Our results show that pine class was detectable in most layers. Spruce was most detectable in RGB data, while birch was most detectable in the HS layers. Furthermore, the RGB datasets provide acceptable results for many low-accuracy applications.