Segmentation-Based Classification of Laser Scanning Data
Zahra Lari2012
Sign up for access to the world's latest research
checkGet notified about relevant papers
checkSave papers to use in your research
checkJoin the discussion with peers
checkTrack your impact
Abstract
Over the past few years, laser scanning has been established as a leading technology for the acquisition of high density 3D spatial information. Digital Terrain Models (DTMs), which can be used for different engineering applications, are obtained by classification of laser data and removing the points that do not belong to terrain surface. The commonly used methods for the classification of laser scanning data are point-based. The major drawback of these methods is focusing on the discontinuities between neighbouring points regardless of the nature of the objects they belong to, which might lead to unreliable classification results. A segmentation-based approach for the classification of both airborne and terrestrial point clouds is presented in this paper. This approach is designed to overcome the drawbacks of point-based classification methods. As the first step, the laser point cloud is segmented by clustering the points with common attributes. To compute precise attributes, an a...
Related papers
Using Training Samples Retrieved from a Topographic Map and Unsupervised Segmentation for the Classification of Airborne Laser Scanning Data
Remote Sensing, 2020
The labeling of point clouds is the fundamental task in airborne laser scanning (ALS) point clouds processing. Many supervised methods have been proposed for the point clouds classification work. Training samples play an important role in the supervised classification. Most of the training samples are generated by manual labeling, which is time-consuming. To reduce the cost of manual annotating for ALS data, we propose a framework that automatically generates training samples using a two-dimensional (2D) topographic map and an unsupervised segmentation step. In this approach, input point clouds, at first, are separated into the ground part and the non-ground part by a DEM filter. Then, a point-in-polygon operation using polygon maps derived from a 2D topographic map is used to generate initial training samples. The unsupervised segmentation method is applied to reduce the noise and improve the accuracy of the point-in-polygon training samples. Finally, the super point graph is used ...
Object-Based Point Cloud Analysis of Full-Waveform Airborne Laser Scanning Data for Urban Vegetation Classification
Sensors, 2008
Airborne laser scanning (ALS) is a remote sensing technique well-suited for 3D vegetation mapping and structure characterization because the emitted laser pulses are able to penetrate small gaps in the vegetation canopy. The backscattered echoes from the foliage, woody vegetation, the terrain, and other objects are detected, leading to a cloud of points. Higher echo densities (>20 echoes/m 2) and additional classification variables from full-waveform (FWF) ALS data, namely echo amplitude, echo width and information on multiple echoes from one shot, offer new possibilities in classifying the ALS point cloud. Currently FWF sensor information is hardly used for classification purposes. This contribution presents an object-based point cloud analysis (OBPA) approach, combining segmentation and classification of the 3D FWF ALS points designed to detect tall vegetation in urban environments. The definition tall vegetation includes trees and shrubs, but excludes grassland and herbage. In the applied procedure FWF ALS echoes are segmented by a seeded region growing procedure. All echoes sorted descending by their surface roughness are used as seed points. Segments are grown based on echo width homogeneity. Next, segment statistics (mean, standard deviation, and coefficient of variation) are calculated by aggregating echo features such as amplitude and surface roughness. For classification a rule base is derived automatically from a training area using a statistical classification tree. To demonstrate our method we present data of three sites with around 500,000 echoes each. The accuracy of the classified vegetation segments is evaluated for two independent validation sites. In a point-wise error assessment, where the classification is compared with manually classified 3D points, completeness and correctness better than 90% are reached for the validation sites. In comparison to many other algorithms the proposed 3D point classification works on the original measurements directly, i.e. the acquired points. Gridding of the data is not necessary, a process which is inherently coupled to loss of data and precision. The 3D properties provide especially a good separability of buildings and terrain points respectively, if they are occluded by vegetation.
ENTITIES AND FEATURES FOR CLASSIFCATION OF AIRBORNE LASER SCANNING DATA IN URBAN AREA
ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences, 2012
A Hierarchical Procedure for Segmentation and Classification of Airborne LIDAR Images
… , 2005. IGARSS'05. Proceedings. 2005 IEEE …, 2005
<title>Classification of mobile terrestrial laser point clouds using semantic constraints</title>
Videometrics, Range Imaging, and Applications X, 2009
With mobile terrestrial laser scanning, laser point clouds of large urban areas can be acquainted rapidly during normal speed driving. Classification of the laser points is beneficial to the city reconstruction from laser point cloud, but a manual classification process can be rather time-consuming due to the huge amount of laser points. Although the pulse return is often used to automate classification, it is only possible to distinguish limited types such as vegetation and ground. In this paper we present a new method which classifies mobile terrestrial laser point clouds using only coordinate information. First, a point of a whole urban scene is segmented, and geometric properties of each segment are computed. Then semantic constraints for several object types are derived from observation and knowledge. These constraints concern not only geometric properties of the semantic objects, but also regulate the topological and hierarchical relations between objects. A search tree is formulated from the semantic constraints and applied to the laser segments for interpretation. 2D map can provide the approximate locations of the buildings and roads as well as the roads' dominant directions, so it is integrated to reduce the search space. The applicability of this method is demonstrated with a Lynx data of the city Enschede and a Streetmapper data of the city Esslingen. Four object types: ground, road, building façade, and traffic symbols, are classified in these data sets.
RECOGNISING STRUCTURE IN LASER SCANNER POINT CLOUDS1
2004
Both airborne and terrestrial laser scanners are used to capture large point clouds of the objects under study. Although for some applications, direct measurements in the point clouds may already suffice, most applications require an automatic processing of the point clouds to extract information on the shape of the recorded objects. This processing often involves the recognition of specific geometric
New Approaches for Estimating the Local Point Density and its Impact on Lidar Data Segmentation
Photogrammetric Engineering and Remote Sensing, 2013
Lidar systems have been proven as a cost-effective tool for the collection of high density and accurate point cloud over physical surfaces. The collected point cloud does not exhibit homogenous point distribution due to the characteristics of the scanning system and/or the physical properties of the scanned surfaces. In order to effectively process the lidar point clouds, local point density variations should be quantified and taken into account for the definition of processing parameters. In this paper, new approaches are presented for the estimation of local point density indices while considering the 3D relationship among lidar points, the physical properties of the reflecting surfaces, and the noise level in the datasets collected by different laser scanners. The impact of considering the estimated local point density variations on the quality of lidar data segmentation results is then investigated by performing a quality control procedure. Quantitative evaluation of segmentation results highlights the efficacy of utilizing the estimated local point density indices for the derivation of more accurate segmentation.
Automated DTM Extraction and Classification of Airborne LiDAR Data
Light Detection And Ranging (LiDAR) is an active remote sensing technology that has the ability to collect spatial information for objects in the form of 3D point cloud data. The classification of LiDAR data is an essential step for its processing techniques and the sub-sequent features extraction. Thus, the classification of 3D point cloud data into terrain and off-terrain points is very important to obtain useful spatial information for several applications, such as the generation of contour lines, 3D city model reconstruction, road network planning, and the determination of flood zone. This paper introduces a new approach for detecting buildings and vegetation classes from multiple returned LiDAR data. In the proposed approach, the height profile is processed in a local fashion over a sliding window. This sliding approach ensures successful DTM computation over hilly terrains. The proposed method uses eigenvalues for reliable surface detection. The conclusion and recommendation w...
Automatic Filtering and Classification of Low-Density Airborne Laser Scanner Clouds in Shrubland Environments
Remote Sensing
The monitoring of shrublands plays a fundamental role, from an ecological and climatic point of view, in biodiversity conservation, carbon stock estimates, and climate-change impact assessments. Laser scanning systems have proven to have a high capability in mapping non-herbaceous vegetation by classifying high-density point clouds. On the other hand, the classification of low-density airborne laser scanner (ALS) clouds is largely affected by confusion with rock spikes and boulders having similar heights and shapes. To identify rocks and improve the accuracy of vegetation classes, we implemented an effective and time-saving procedure based on the integration of geometric features with laser intensity segmented by K-means clustering (GIK procedure). The classification accuracy was evaluated, taking into account the data unevenness (small size of rock class vs. vegetation and terrain classes) by estimating the Balanced Accuracy (BA range 89.15–90.37); a comparison with a standard geom...
Reliable automatic classification and segmentation of laser point clouds by statistical analysis of surface curvature values
Applied Geomatics, 2009
Dealing with a procedure for the automatic classification of laser point clouds based on surface curvature values, one of the most critical aspects is the correct interpretation of the estimated results. This care is required since the measurements are characterised by errors of different kind, and simplified analytical models are applied to estimate the differential terms used to locally compute the object surface curvature values. Following a non-parametric approach, the differential terms are the firstand second-order partial derivatives of a Taylor's expansion used to determine the Gaussian K and the mean H local curvatures. Therefore, a statistical analysis is proposed in this paper. It is based at first on a chi-square test applied to verify the fulfilment of the second-order Taylor's expansion. Successively, the variance-covariance propagation law is applied to the estimated differential terms, in order to calculate the covariance matrix of a two-row vector containing the Gaussian and the mean curvature estimates, and an F ratio test is then applied to verify their significance. By analysing the test acceptance or rejection for K and H and their sign, a reliable classification of the whole point cloud into its geometrical basic types is carried out. A robust parametric modelling is then applied to estimate the analytical function of each classified surface.
Related papers
An Adaptive Approach for Segmentation of 3D Laser Point Cloud
2011
Automatic processing and object extraction from 3D laser point cloud is one of the major research topics in the field of photogrammetry. Segmentation is an essential step in the processing of laser point cloud, and the quality of extracted objects from laser data is highly dependent on the validity of the segmentation results. This paper presents a new approach for reliable and efficient segmentation of planar patches from a 3D laser point cloud. In this method, the neighbourhood of each point is firstly established using an adaptive cylinder while considering the local point density and surface trend. This neighbourhood definition has a major effect on the computational accuracy of the segmentation attributes. In order to efficiently cluster planar surfaces and prevent introducing ambiguities, the coordinates of the origin's projection on each point's best fitted plane are used as the clustering attributes. Then, an octree space partitioning method is utilized to detect and extract peaks from the attribute space. Each detected peak represents a specific cluster of points which are located on a distinct planar surface in the object space. Experimental results show the potential and feasibility of applying this method for segmentation of both airborne and terrestrial laser data.
Water classification using 3D airborne laser scanning point clouds
2009
Airborne laser scanning (ALS), also referred to as airborne LiDAR (Light Detection And Ranging), provides highly accurate measurements of the Earth surface. In the last twenty years, ALS has been established as a standard technique for delineating objects (e.g. buildings, trees, roads) and mapping changes. Studies on hydrology or geomorphology such as monitoring of braided river structures, calculation of erosion and accumulation potential in watercourses, or floodplain mapping require all the precise location of the water surface. This paper shows a 3D point cloud based method, which allows an automatic water surface classification by using geometric and radiometric ALS information and the location of modeled lost reflections, which are called laser shot dropouts. The classification result can be used to map the watercourse, to improve DTM filtering routines or to replace water points with river bed heights for hydraulic modeling etc. Themethod relies on a threshold based classific...
Classification of ground objects using laser radar data
2003
Accurate 3D models of natural environments are important for many modelling and simulation applications, for both civilian and military purposes. When building 3D models from high resolution data acquired by an airborne laser scanner it is desirable to separate and classify the data to be able to process it further. For example, to build a polygon model of a building the samples belonging to the building must be found. In this thesis we have developed, implemented (in IDL and ENVI), and evaluated algorithms for classification of buildings, vegetation, power lines, posts, and roads. The data is gridded and interpolated and a ground surface is estimated before the classification. For the building classification an object based approach was used unlike most classification algorithms which are pixel based. The building classification has been tested and compared with two existing classification algorithms. The developed algorithm classified 99.6 % of the building pixels correctly, while the two other algorithms classified 92.2 % respective 80.5 % of the pixels correctly. The algorithms developed for the other classes were tested with the following result (correctly classified pixels)
Terrain modeling and airborne laser data classification using multiple pass filtering
2004
Airborne laser scanner technique provides a 3D perception of the terrestrial topography, including true ground and objects belonging either to vegetated areas or to human made features. The high intrinsic accuracy and regularity of airborne laser sensors makes highly conceivable the extraction of semantic information related to the recorded 3D-points. In this respect, a new algorithm has been developed in order to classify the initial cloud of points into ground/non ground earth points and generate accurate Digital Terrain Models (DTMs) on a regular grid. Our approach is based on a multiple pass classification process. An estimation of the ground is performed within overlapping neighborhood and laser points are classified with regard to this ground estimation. The algorithm moves toward the neighbor where the average altitude is the lowest. We then compare the vicinity of the terrain with the estimated ground and apply a linear correction. As it goes along, points are filtered many times until we vote for the final label. The estimated ground surface is then the input of an energy minimization algorithm (ICM) which consider laser points as a set of attractors. The final DTM will be a trade off between internal properties and its closeness to ground laser points. The resolution may be fine enough to proceed relevant micro relief analysis especially in a rural environment.
Automatic Classification of Trees from Laser Scanning Point Clouds
ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2015
Development of laser scanning technologies has promoted tree monitoring studies to a new level, as the laser scanning point clouds enable accurate 3D measurements in a fast and environmental friendly manner. In this paper, we introduce a probability matrix computation based algorithm for automatically classifying laser scanning point clouds into 'tree' and 'non-tree' classes. Our method uses the 3D coordinates of the laser scanning points as input and generates a new point cloud which holds a label for each point indicating if it belongs to the 'tree' or 'non-tree' class. To do so, a grid surface is assigned to the lowest height level of the point cloud. The grids are filled with probability values which are calculated by checking the point density above the grid. Since the tree trunk locations appear with very high values in the probability matrix, selecting the local maxima of the grid surface help to detect the tree trunks. Further points are assigned to tree trunks if they appear in the close proximity of trunks. Since heavy mathematical computations (such as point cloud organization, detailed shape 3D detection methods, graph network generation) are not required, the proposed algorithm works very fast compared to the existing methods. The tree classification results are found reliable even on point clouds of cities containing many different objects. As the most significant weakness, false detection of light poles, traffic signs and other objects close to trees cannot be prevented. Nevertheless, the experimental results on mobile and airborne laser scanning point clouds indicate the possible usage of the algorithm as an important step for tree growth observation, tree counting and similar applications. While the laser scanning point cloud is giving opportunity to classify even very small trees, accuracy of the results is reduced in the low point density areas further away than the scanning location. These advantages and disadvantages of two laser scanning point cloud sources are discussed in detail.
Forest delineation and structure assessment based on Airborne Laser Scanning data
Airborne laser scanning (ALS) data has been established as the standard method for the acquisition of high precision topographic data. In addition to the derivation of topographic models, such as digital terrain models (DTM) or digital surface models (DSM), ALS data is the main input data source for a variety of applications, e.g. building modelling, power line modelling or forestry applications. Until now a severe limitation is the availability of tools allowing computations directly on the 3D point cloud for district wide calculations. In complex 3D scenarios such as forests, the point cloud content is commonly converted to raster data (e.g. DTM and DSM) with a notable loss of information. As a result, the information on the vertical structure of vegetation is irretrievably lost. Therefore, a methodology for the delineation of forest areas and subsequent derivation of vertical vegetation strata is proposed. The presented approach combines processing steps directly in the 3D point ...
Classification and filtering of laser data
International Archives of …, 2003
A strategy for the classification of raw LIDAR data as terrain, buildings and vegetation is presented. Its main features are a preliminary classification of grid data based on a geometric and topological description and a final filtering of raw data, guided by the previous classification. After raw data have been interpolated to a grid and segmented in connected regions bordered by a step edge, the topology of these regions is built up. Noise, vegetation and data gaps are classified first, mainly based on size and region fragmentation. Then, regions enclosing terrain and building points are labelled analysing their relationships with adjacent regions. Since regions may enclose more than one instance of different classes, a first check is made on grid data looking for consistency of gradient orientation with class characteristics. Finally, a local analysis is performed on each grid cell to label raw data point, based on the information on the surroundings inferred by the classification. Results obtained with Toposys and Optech systems on datasets with different ground point density gathered over the town of Pavia are shown to illustrate the effectiveness of the procedure.
Object-based classification of terrestrial laser scanning point clouds for landslide monitoring
The Photogrammetric Record
Terrestrial laser scanning (TLS) is often used to monitor landslides and other gravitational mass movements with high levels of geometric detail and accuracy. However, unstructured TLS point clouds lack semantic information, which is required to geomorphologically interpret the measured changes. Extracting meaningful objects in a complex and dynamic environment is challenging due to the objects' fuzziness in reality, as well as the variability and ambiguity of their patterns in a morphometric feature space. This work presents a point-cloud-based approach for classifying multitemporal scenes of a hillslope affected by shallow landslides. The 3D point clouds are segmented into morphologically homogeneous and spatially connected parts. These segments are classified into seven target classes (scarp, eroded area, deposit, rock outcrop and different classes of vegetation) in a two-step procedure: a supervised classification step with a machine-learning classifier using morphometric features, followed by a correction step based on topological rules. This improves the final object extraction considerably.
Planar segmentation of indoor terrestrial laser scanning point clouds via distance function from a point to a plane
Optics and Lasers in Engineering, 2015
Terrestrial laser scanners are frequently used in most of measurement application, particularly in documentation and restoration studies of indoor historical structures, and in acquiring facade reliefs. When compared to a photogrammetric method, terrestrial laser scanners have the ability to give three dimensional point cloud data directly in a fast and detailed way. High data density of point cloud data is a challenging factor in texture-map operations during documentation and restoration of historical artifacts with more indoor spaces. When coordinate information for terrestrial laser scanner point cloud data is documented, it is seen that there is no regular order and classification for the data. The aim of this study is to suggest the mathematical filtering algorithm for segmentation work towards separation of planar surfaces which have different depths and parallel to each other and which can be frequently encountered in the indoor spaces from the data of terrestrial laser scanner. Filtering function for segmentation used, is based on the distance of a point to the plane. This algorithm has been chosen for the advantage of the rapid and easy results for extracting 3D coordinate data in texture mapping process. The MatLAB interface has been developed for using this method and analyzing the results for application which is detected how many different surfaces exist according to the statistical deviation amount. In the application, test data with 21932 points was segmented by separating it into 16 points in total with four different planes and four corner points per plane. Surfaces with four different depths were obtained as the result of the research. Each of them included four points. These segmented surfaces consisting of four points will facilitate integrated data production by integrating vectorial terrestrial laser scanner data into raster camera data, without the need to conventional measurements that accelerate particularly documentation and modeling in the fields of historical indoor areas.
Area-wide roof plane segmentation in airborne LiDAR point clouds
Computers, Environment and Urban Systems, 2012
Most algorithms performing segmentation of 3D point cloud data acquired by, e.g. Airborne Laser Scanning (ALS) systems are not suitable for large study areas because the huge amount of point cloud data cannot be processed in the computer's main memory. In this study a new workflow for seamless automated roof plane detection from ALS data is presented and applied to a large study area. The design of the workflow allows area-wide segmentation of roof planes on common computer hardware but leaves the option open to be combined with distributed computing (e.g. cluster and grid environments). The workflow that is fully implemented in a Geographical Information System (GIS) uses the geometrical information of the 3D point cloud and involves four major steps: (i) The whole dataset is divided into several overlapping subareas, i.e. tiles. (ii) A raster based candidate region detection algorithm is performed for each tile that identifies potential areas containing buildings. (iii) The resulting building candidate regions of all tiles are merged and those areas overlapping one another from adjacent tiles are united to a single building area. (iv) Finally, three dimensional roof planes are extracted from the building candidate regions and each region is treated separately. The presented workflow reduces the data volume of the point cloud that has to be analyzed significantly and leads to the main advantage that seamless area-wide point cloud based segmentation can be performed without requiring a computationally intensive algorithm detecting and combining segments being part of several subareas (i.e. processing tiles). A reduction of 85% of the input data volume for point cloud segmentation in the presented study area could be achieved, which directly decreases computation time.
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.