Toward intelligent training of supervised image classifications: directing training data acquisition for SVM classification

By Giles M. Foody, Ajay Mathur. 2004

647 citations recorded by the source corpus.

Subfields: Analytical Chemistry, Ecology, Plant Science

View Paper: https://www.semanticscholar.org/paper/ae549c5b195ceb0f0f97c5ff3e9b7ed4126def73

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Analytical Chemistry

In Ecology

In Plant Science

Citations in the corpus

Support Vector Machine Versus Random Forest for Remote Sensing Image Classification: A Meta-Analysis and Systematic Review (2020, background): “In particular, land use and land cover (LULC) mapping is the most common application of RS data for a variety of environmental studies, given the increased availability of RS image archives [9]–[12].”
Classification accuracy comparison: hypothesis tests and the use of confidence intervals in evaluations of difference, equivalence and non-inferiority (2009, methodology): “…a fast approximation of the maximum likelihood classification (Settle and Briggs, 1987) or a knowledge-based procedure for training site selection (Foody and Mathur, 2004; Mathur and Foody, 2008) or perhaps in some feature reduction analyses (Demir and Ertürk, 2008; Wang and Li, 2008) the aim is…”
Vision-Language Models for Vision Tasks: A Survey (2023, background): “Before the deep learning era [7], visual recognition studies rely heavily on feature engineering with hand-crafted features [13], [45], [46], [47], [48], [49], [50], [51] and lightweight learning models [10], [11], [12] that classify the handcrafted features into pre-defined semantic categories.”
Multiclass and Binary SVM Classification: Implications for Training and Classification Users (2008, background): “Moreover, it may sometimes be possible to identify the most useful training sites for the provision of support vectors before the classification [3], [4].”
A Survey of Active Learning Algorithms for Supervised Remote Sensing Image Classification (2011, background): “However, the performances of supervised algorithms strongly depend on the representativeness of the data used to train the classifier [12].”
Toward an Optimal SVM Classification System for Hyperspectral Remote Sensing Images (2006, methodology): “Indeed, for training sample collection strategies based on an intelligent preselection of training samples by appropriately exploiting prior knowledge and/or ancillary information like in [42], the SV bound can appear unsuitable since it may be close to unity.”
Support vector machines in remote sensing: A review (2011, neutral): “Foody and Mathur (2004a) proposed a single multiclass SVM classification method while typical multi-class SVMs are based mainly on the use of multiple binary analyses.”
Recent Advances in Techniques for Hyperspectral Image Processing (2009, background): “For instance, urban area mapping requires sufficient spatial resolution to distinguish small spectral classes, such as trees in a park, or cars on a street (Gamba et al., 2004; Chanussot et al., 2006).”
Support vector machine to map oil palm in a heterogeneous environment (2014, methodology): “It then operates to find a wide separation boundary between class pairs by marking each pixel as belonging to a class, based on the inputs (Foody and Mathur 2004a; Kavzoglu and Colkesen 2009).”
Feature Selection for Classification of Hyperspectral Data by SVM (2010, neutral): “Given that training data acquisition may be difficult and costly [11]–[13], some means to accommodate”
On the Importance of Training Data Sample Selection in Random Forest Image Classification: A Case Study in Peatland Ecosystem Mapping (2015, methodology): “, not clustered) [14] and representative of the entire landscape [10,12], and there should be abundant training data in all classes [15].”
Monitoring land cover change in urban and peri-urban areas using dense time stacks of Landsat satellite data and a data mining approach (2012, methodology): “SVM are designed to use only the training data that that lie on the edge of the class distribution to define class boundaries; all other training data provide no contribution to the classification analysis (Foody & Mathur, 2004).”