Entity

Time filter

Source Type


Jiang M.,Bureau of Mineral Resources | Lin Y.,Finnish Geodetic Institute
IEEE Geoscience and Remote Sensing Letters | Year: 2013

This study proposed and tested a multistep method for the recognition of individual deciduous trees in leaf-off aerial ultrahigh spatial resolution remotely sensed (UHSRRS) imagery. This topic has received limited coverage in previous endeavors, which focused mainly on the detection and delineation of coniferous trees in remotely sensed images with relatively lower spatial resolutions. Thus, the traditional algorithms tend to fail in case of the referred scenario. In order to fill this technical gap, an algorithm that joins mathematical morphological operations and marker-controlled watershed segmentation was first assumed for the extraction of single trees in UHSRRS images. Next, a distribution-free support vector machine (SVM) classifier was applied to distinguish the extracted segments as deciduous or coniferous trees, merely in terms of two newly-derived morphological features. Experimental evaluations indicated that the integral solution plan can extract and classify the deciduous and coniferous trees in the leaf-off aerial UHSRRS images of local dense forest for test with correctness over 92% and 70%, respectively. Overall, the recognition results with >66% correctness have primarily validated the proposed technique. © 2004-2012 IEEE. Source


Hellesen T.,Copenhagen University | Matikainen L.,Finnish Geodetic Institute
Remote Sensing | Year: 2013

Due to the abandonment of former agricultural management practices such as mowing and grazing, an increasing amount of grassland is no longer being managed. This has resulted in increasing shrub encroachment, which poses a threat to a number of species. Monitoring is an important means of acquiring information about the condition of the grasslands. Though the use of traditional remote sensing is an effective means of mapping and monitoring land cover, the mapping of small shrubs and trees based only on spectral information is challenged by the fact that shrubs and trees often spectrally resemble grassland and thus cannot be safely distinguished and classified. With the aid of LiDAR-derived information, such as elevation, the classification of spectrally similar objects can be improved. In this study, we applied high point density LiDAR data and colour-infrared orthoimages for the classification of shrubs and trees in a study area in Denmark. The classification result was compared to a classification based only on colour-infrared orthoimages. The overall accuracy increased significantly with the use of LiDAR and, for shrubs and trees specifically, producer's accuracy increased from 81.2% to 93.7%, and user's accuracy from 52.9% to 89.7%. Object-based image analysis was applied in combination with a CART classifier. The potential of using the applied approach for mapping and monitoring of large areas is discussed. © 2013 by the authors. Source


Zhu L.,Finnish Geodetic Institute | Hyyppa J.,Finnish Geodetic Institute
Remote Sensing | Year: 2014

This paper presents methods for 3D modeling of railway environments from airborne laser scanning (ALS) and mobile laser scanning (MLS). Conventionally, aerial data such as ALS and aerial images were utilized for 3D model reconstruction. However, 3D model reconstruction only from aerial-view datasets can not meet the requirement of advanced visualization (e.g., walk-through visualization). In this paper, objects in a railway environment such as the ground, railroads, buildings, high voltage powerlines, pylons and so on were reconstructed and visualized in real-life experiments in Kokemaki, Finland. Because of the complex terrain and scenes in railway environments, 3D modeling is challenging, especially for high resolution walk-through visualizations. However, MLS has flexible platforms and provides the possibility of acquiring data in a complex environment in high detail by combining with ALS data to produce complete 3D scene modeling. A procedure from point cloud classification to 3D reconstruction and 3D visualization is introduced, and new solutions are proposed for object extraction, 3D reconstruction, model simplification and final model 3D visualization. Image processing technology is used for the classification, 3D randomized Hough transformations (RHT) are used for the planar detection, and a quadtree approach is used for the ground model simplification. The results are visually analyzed by a comparison with an orthophoto at a 20 cm ground resolution. © 2014 by the authors; licensee MDPI, Basel, Switzerland. Source


Lin Y.,Finnish Geodetic Institute | Hyyppa J.,Finnish Geodetic Institute
IEEE Transactions on Geoscience and Remote Sensing | Year: 2012

This paper presents a novel attempt at combining the mobile mapping mode and a multiecho-recording laser scanner, as well as a new methodology based on the resulting single-scan point clouds, for enhancing the integrity of individual tree crown reconstruction. The motive stemmed from the widespread but hard-to-reach demand in precision forestry, i.e., efficiently acquiring the integral 3-D structures of single crowns via single-scan light detection and ranging (LiDAR) surveys. For this task, aerospace and aerial LiDAR is generally subject to low sampling density, and static terrestrial LiDAR is restricted to high relocation cost. As a state-of-the-art mapping technology, mobile laser scanning (MLS) can somehow overcome these limitations owing to its strengths of high sampling density and moving efficiency. However, its laser emissions, even from the incorporated peculiar scanner, still suffer from leaf/branch occlusions. To address this challenge, mirroring the half crowns facing the MLS system to the other sides can be assumed as a solution strategy, in a point of view different from typically strengthening laser transmission penetrability. In the case of no multiscans available, this plan turns out to be unstable due to the shortage of reference data. For this issue, this study further expands the roles of multiechoes beyond penetration, namely, also as the self-indicators for correcting the mirrored half crowns. Quantitative evaluation about the integrity of reconstruction in terms of crown outer surface was conducted based on the sample trees, which were measured by the multiecho-recording MLS and a static terrestrial LiDAR from two opposite sides. The promising results basically validated the proposed technique. © 1980-2012 IEEE. Source


Lin Y.,Finnish Geodetic Institute | Hyyppa J.,Finnish Geodetic Institute
IEEE Geoscience and Remote Sensing Letters | Year: 2011

A novel κ-segments-based 2-D geometric modeling schematic is proposed for characterizing the scan lines of vehicle-based laser scanning (VLS) with just a few geometric primitives. VLS has been developing quickly as a new research focus recently, but the relevant data processing techniques lag behind the system establishment due to the associated huge standwise point clouds collected in a real 3-D sense. To solve this issue, the often-assumed sampling mode based on scan lines suggests an alternative frame for exploring new efficient methodologies in diverse applications. As we know, principal segments can reflect the morphological signatures of the scatter-point-represented objects, and besides, profiles comprised by various open and close outlines can be geometrically modeled by line segments and ellipses, respectively. By combining these two merits, the $k$-segments-based geometric modeling algorithm can be constructed, which segments and fits the center-clustered points, e.g., in crowns, into ellipses and the line-arranged points, e.g., on walls, into line segments. Eventually, the experiments based on real VLS data primarily validate this new algorithm. © 2010 IEEE. Source

Discover hidden collaborations