Decision tree regression for soft classification of remote sensing data

Xu, Min; Watanachaturaporn, Pakorn; Varshney, Pramod K.; Arora, Manoj K.

doi:10.1016/j.rse.2005.05.008

Cited by 469 publications

(198 citation statements)

References 20 publications

Supporting

Mentioning

191

Contrasting

Unclassified

Order By: Relevance

“…Understanding and recognizing the uncertainty in image classification and the desire to fully exploit the information content of the produced land-cover maps were the driving forces in the development of soft classification of remotely sensed data [18], [43]. In this paper, we compare the proposed classification methods, described in Sections II and III, by looking at their soft answers on the test set, as well as on the remaining (not labeled) data set.…”

Section: Uncertainty Analysismentioning

confidence: 99%

Uncertainty Analysis for the Classification of Multispectral Satellite Images Using SVMs and SOMs

Giacco

Thiel

Pugliese

et al. 2010

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Abstract-Classification of multispectral remotely sensed data with textural features is investigated with a special focus on uncertainty analysis in the produced land-cover maps. Much effort has already been directed into the research of satisfactory accuracyassessment techniques in image classification, but a common approach is not yet universally adopted. We look at the relationship between hard accuracy and the uncertainty on the produced answers, introducing two measures based on maximum probability and α quadratic entropy. Their impact differs depending on the type of classifier. In this paper, we deal with two different classification strategies, based on support vector machines (SVMs) and Kohonen's self-organizing maps (SOMs), both suitably modified to give soft answers. Once the multiclass probability answer vector is available for each pixel in the image, we studied the behavior of the overall classification accuracy as a function of the uncertainty associated with each vector, given a hard-labeled test set. The experimental results show that the SVM with one-versus-one architecture and linear kernel clearly outperforms the other supervised approaches in terms of overall accuracy. On the other hand, our analysis reveals that the proposed SOM-based classifier, despite its unsupervised learning procedure, is able to provide soft answers which are the best candidates for a fusion with supervised results.

show abstract

Section: Uncertainty Analysismentioning

confidence: 99%

Uncertainty Analysis for the Classification of Multispectral Satellite Images Using SVMs and SOMs

Giacco

Thiel

Pugliese

et al. 2010

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Decision tree classification [13], neutral network classification [14], and support vector machine [8] are classifiers that are commonly applied to remote sensing images. In most cases, these methodologies are based on the spectral and textural features of the relevant classes [15,16].…”

Section: Introductionmentioning

confidence: 99%

Classification of Small-Scale Eucalyptus Plantations Based on NDVI Time Series Obtained from Multiple High-Resolution Datasets

Qiao

Shakir

et al. 2016

Remote Sensing

View full text Add to dashboard Cite

Eucalyptus, a short-rotation plantation, has been expanding rapidly in southeast China in recent years owing to its short growth cycle and high yield of wood. Effective identification of eucalyptus, therefore, is important for monitoring land use changes and investigating environmental quality. For this article, we used remote sensing images over 15 years (one per year) with a 30-m spatial resolution, including Landsat 5 thematic mapper images, Landsat 7-enhanced thematic mapper images, and HJ 1A/1B images. These data were used to construct a 15-year Normalized Difference Vegetation Index (NDVI) time series for several cities in Guangdong Province, China. Eucalyptus reference NDVI time series sub-sequences were acquired, including one-year-long and two-year-long growing periods, using invested eucalyptus samples in the study region. In order to compensate for the discontinuity of the NDVI time series that is a consequence of the relatively coarse temporal resolution, we developed an inverted triangle area methodology. Using this methodology, the images were classified on the basis of the matching degree of the NDVI time series and two reference NDVI time series sub-sequences during the growing period of the eucalyptus rotations. Three additional methodologies (Bounding Envelope, City Block, and Standardized Euclidian Distance) were also tested and used as a comparison group. Threshold coefficients for the algorithms were adjusted using commission-omission error criteria. The results show that the triangle area methodology out-performed the other methodologies in classifying eucalyptus plantations. Threshold coefficients and an optimal discriminant function were determined using a mosaic photograph that had been taken by an unmanned aerial vehicle platform. Good stability was found as we performed further validation using multiple-year data from the high-resolution Gaofen Satellite 1 (GF-1) observations of larger regions. Eucalyptus planting dates were also estimated using invested eucalyptus samples and the Root Mean Square Error (RMSE) of the estimation was 84 days. This novel and reliable method for classifying short-rotation plantations at small scales is the focus of this study.

show abstract

“…Towards this argument, a comparative experiment was conducted. In additional to our implemented SSGPR, we tried 5 famous non-Bayesian regression models -Kernal Ridge Regression (KRR) with RBF kernel [20,21] , Decision Tree (DT) [22], Ada Boosted Decision Tree (AdaDT) [23], Gradient Boosting Machine (GBM) [24] and Random Forests (RF) [25]. We used Scikit-learn [26] to implement these models, in which the model parameters are selected by 5-fold cross validation with grid search algorithm.…”

Section: Comparing To Other Regression Modelsmentioning

confidence: 99%

One-Match-Ahead Forecasting in Two-Team Sports with Stacked Bayesian Regressions

Lam

2018

Journal of Artificial Intelligence and Soft Computing Research

View full text Add to dashboard Cite

There is a growing interest in applying machine learning algorithms to real-world examples by explicitly deriving models based on probabilistic reasoning. Sports analytics, being favoured mostly by the statistics community and less discussed in the machine learning community, becomes our focus in this paper. Specifically, we model two-team sports for the sake of one-match-ahead forecasting. We present a pioneering modeling approach based on stacked Bayesian regressions, in a way that winning probability can be calculated analytically. Benefiting from regression flexibility and high standard of performance, Sparse Spectrum Gaussian Process Regression (SSGPR) -an improved algorithm for the standard Gaussian Process Regression (GPR), was used to solve Bayesian regression tasks, resulting in a novel predictive model called TLGProb. For evaluation, TLGProb was applied to a popular sports event -National Basketball Association (NBA). Finally, 85.28% of the matches in NBA 2014/2015 regular season were correctly predicted by TLGProb, surpassing the existing predictive models for NBA.

show abstract

Decision tree regression for soft classification of remote sensing data

Cited by 469 publications

References 20 publications

Uncertainty Analysis for the Classification of Multispectral Satellite Images Using SVMs and SOMs

Uncertainty Analysis for the Classification of Multispectral Satellite Images Using SVMs and SOMs

Classification of Small-Scale Eucalyptus Plantations Based on NDVI Time Series Obtained from Multiple High-Resolution Datasets

One-Match-Ahead Forecasting in Two-Team Sports with Stacked Bayesian Regressions

Contact Info

Product

Resources

About