A Multiple Classifier System for Classification of LIDAR Remote Sensing Data Using Multi-class SVM

Samadzadegan, Farhad; Bigdeli, Behnaz; Ramzi, Pouria

doi:10.1007/978-3-642-12127-2_26

Cited by 27 publications

(23 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Numerous studies have focused on extracting one object type in urban scenes, such as separation of ground from non-ground points in order to generate digital terrain models (DEMs) (Bartels et al 2006), building extraction (Huang et al, 2013), road extraction (Samadzadegan et al, 2009) and curbstones mapping (Zhou and Vosselman, 2012). The number of extracted classes has been extended to three main urban classes, including ground, vegetation and buildings (Samadzadegan et al, 2010). Nowadays, multi-classes extraction is a very important topic for building 3D city models, maps updating and emergency purposes.…”

Section: Introductionmentioning

confidence: 99%

Clustering of Multispectral Airborne Laser Scanning Data Using Gaussian Decomposition

Morsy

Shaker

El-Rabbany

2017

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:With the evolution of the LiDAR technology, multispectral airborne laser scanning systems are currently available. The first operational multispectral airborne LiDAR sensor, the Optech Titan, acquires LiDAR point clouds at three different wavelengths (1.550, 1.064, 0.532 m), allowing the acquisition of different spectral information of land surface. Consequently, the recent studies are devoted to use the radiometric information (i.e., intensity) of the LiDAR data along with the geometric information (e.g., height) for classification purposes. In this study, a data clustering method, based on Gaussian decomposition, is presented. First, a ground filtering mechanism is applied to separate non-ground from ground points. Then, three normalized difference vegetation indices (NDVIs) are computed for both non-ground and ground points, followed by histograms construction from each NDVI. The Gaussian function model is used to decompose the histograms into a number of Gaussian components. The maximum likelihood estimate of the Gaussian components is then optimized using Expectation -Maximization algorithm. The intersection points of the adjacent Gaussian components are subsequently used as threshold values, whereas different classes can be clustered. This method is used to classify the terrain of an urban area in Oshawa, Ontario, Canada, into four main classes, namely roofs, trees, asphalt and grass. It is shown that the proposed method has achieved an overall accuracy up to 95.1% using different NDVIs.

show abstract

Section: Introductionmentioning

confidence: 99%

Clustering of Multispectral Airborne Laser Scanning Data Using Gaussian Decomposition

Morsy

Shaker

El-Rabbany

2017

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…Ensemble classification is the training of multiple classifiers to solve a common labeling problem by different communities. This classification strategy is also called committee-based learning, multiple classifier systems, or a mixture of experts [16][17][18]. The criterion of a good ensemble system is that it should provide an increase in classification accuracy.…”

Section: Introductionmentioning

confidence: 99%

“…There are numerous ways of combining classifiers as suggested by previous studies [19] with a wide variety of applications contributed from text categorization [20] and hand-written word recognition [21]. In particular, a few examples of related studies in remote sensing include [17,[22][23][24].…”

Section: Introductionmentioning

confidence: 99%

Maximizing the Diversity of Ensemble Random Forests for Tree Genera Classification Using High Density LiDAR Data

Sohn

Remmel

et al. 2016

Remote Sensing

View full text Add to dashboard Cite

Recent research into improving the effectiveness of forest inventory management using airborne LiDAR data has focused on developing advanced theories in data analytics. Furthermore, supervised learning as a predictive model for classifying tree genera (and species, where possible) has been gaining popularity in order to minimize this labor-intensive task. However, bottlenecks remain that hinder the immediate adoption of supervised learning methods. With supervised classification, training samples are required for learning the parameters that govern the performance of a classifier, yet the selection of training data is often subjective and the quality of such samples is critically important. For LiDAR scanning in forest environments, the quantification of data quality is somewhat abstract, normally referring to some metric related to the completeness of individual tree crowns; however, this is not an issue that has received much attention in the literature. Intuitively the choice of training samples having varying quality will affect classification accuracy. In this paper a Diversity Index (DI) is proposed that characterizes the diversity of data quality (Q i ) among selected training samples required for constructing a classification model of tree genera. The training sample is diversified in terms of data quality as opposed to the number of samples per class. The diversified training sample allows the classifier to better learn the positive and negative instances and; therefore; has a higher classification accuracy in discriminating the "unknown" class samples from the "known" samples. Our algorithm is implemented within the Random Forests base classifiers with six derived geometric features from LiDAR data. The training sample contains three tree genera (pine; poplar; and maple) and the validation samples contains four labels (pine; poplar; maple; and "unknown"). Classification accuracy improved from 72.8%; when training samples were selected randomly (with stratified sample size); to 93.8%; when samples were selected with additional criteria; and from 88.4% to 93.8% when an ensemble method was used.

show abstract

“…Therefore, we identified these five land cover classes for training site selection, and implemented the classification with different combinations of features: 1) intensity data only, 2) intensity and digital surface model (DSM), 3) intensity and texture (TEX) features generated from the intensity, and 4) intensity, TEX and DSM. Previous studies reported that the use of entropy texture and homogeneity texture can significantly contribute to the enhancement of classification accuracy (Samadzadegan et al, 2010;Huang et al, 2011); therefore, these two texture features were generated for both OI and RCNI with a window size of 9 × 9 to support the experimental testing. A total of eight classification scenarios were implemented by using the traditional maximum likelihood classifier, and 1,000 random checkpoints were generated to assess the classification results.…”

Section: Land Cover Classificationmentioning

confidence: 99%

Reduction of Striping Noise in Overlapping Lidar Intensity Data by Radiometric Normalization

Yan

Shaker

2016

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:To serve seamless mapping, airborne LiDAR data are usually collected with multiple parallel strips with one or two cross strip(s). Nevertheless, the overlapping regions of LiDAR data strips are usually found with unbalanced intensity values, resulting in the appearance of stripping noise. Despite that physical intensity correction methods are recently proposed, some of the system and environmental parameters are assumed as constant or not disclosed, leading to such an intensity discrepancy. This paper presents a new normalization technique to adjust the radiometric misalignment found in the overlapping LiDAR data strips. The normalization technique is built upon a second-order polynomial function fitted on the joint histogram plot, which is generated with a set of pairwise closest data points identified within the overlapping region. The method was tested on Teledyne Optech's Gemini dataset (at 1064 nm wavelength), where the LiDAR intensity data were first radiometrically corrected based on the radar (range) equation. Five land cover features were selected to evaluate the coefficient of variation (cv) of the intensity values before and after implementing the proposed method. Reduction of cv was found by 19% to 59% in the Gemini dataset, where the striping noise was significantly reduced in the radiometrically corrected and normalized intensity data. The Gemini dataset was also used to conduct land cover classification, and the overall accuracy yielded a notable improvement of 9% to 18%. As a result, LiDAR intensity data should be pre-processed with radiometric correction and normalization prior to any data manipulation.

show abstract

A Multiple Classifier System for Classification of LIDAR Remote Sensing Data Using Multi-class SVM

Cited by 27 publications

References 8 publications

Clustering of Multispectral Airborne Laser Scanning Data Using Gaussian Decomposition

Clustering of Multispectral Airborne Laser Scanning Data Using Gaussian Decomposition

Maximizing the Diversity of Ensemble Random Forests for Tree Genera Classification Using High Density LiDAR Data

Reduction of Striping Noise in Overlapping Lidar Intensity Data by Radiometric Normalization

Contact Info

Product

Resources

About