Earth Observation Image Semantics: Latent Dirichlet Allocation Based Information Discovery

Asiyabi, Reza Mohammadi; Datcu, Mihai

doi:10.1109/igarss47720.2021.9553122

Cited by 1 publication

(5 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Usually, BOVW model is applied in the patch level for patchbased categorization [24], [28], [31]. However, kernel-based BOVW can provide pixel-based representation of the image [32]. In the kernel-based BOVW, an arbitrary weighted kernel is used and the BOVW histogram of the area covered by the kernel is assigned to the central pixel in the kernel.…”

Section: A Bag Of Visual Words (Bovw)modelmentioning

confidence: 99%

“…In the kernel-based BOVW, an arbitrary weighted kernel is used and the BOVW histogram of the area covered by the kernel is assigned to the central pixel in the kernel. After sliding the kernel window over the image and repeating the procedure for all the pixels in the image, each pixel will be represented by a separate BOVW histogram [32]. These histograms can be utilized for categorization and classification of the image.…”

Section: A Bag Of Visual Words (Bovw)modelmentioning

confidence: 99%

“…In the first step of this experiment, five semantic classes including buildings, roads, vegetation, water, and shadow are visually identified in the subset and the user-defined GT map is created. Kernel-based BOVW histograms of the scene are computed, according to the procedure explained in the section II-A [32]. In the BOVW model, K-means clustering algorithm is used to construct the visual words and the number of the visual words set to 20, considering the size of the image and histograms.…”

Section: B Scenario 1: Submeter Resolution Land Cover Mappingmentioning

confidence: 99%

“…In the first experiment, only three spectral bands in the visible portion of the electromagnetic spectrum (RGB bands) are used. In the first step, the kernel-based BOVW histograms of the subsets are created [32]. The same parameters as the previous scenario are used in the BOVW model.…”

Section: Scenario 2: Sentinel-2 Based Forest Fire Monitoringmentioning

confidence: 99%

“…Similar to the previous scenarios, 20 visual words are constructed by k-means algorithm. However, due to the patch-based processing in this scenario, patch-based BOVW is used, instead of the kernelbased model (See section II-A and [24], [28], [31], [32]). The BOVW histogram of the training patches, which are used in the initial training stage of the SVM classifier (1274 patches), are used to train the LDA model with 7 topics.…”

Section: Scenario 3: Sentinel-1 Patch-based Annotation Analysismentioning

confidence: 99%

See 4 more Smart Citations

Earth Observation Semantic Data Mining: Latent Dirichlet Allocation-Based Approach

Asiyabi

Datcu

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Self Cite

View full text Add to dashboard Cite

Recent advances in remote sensing technology have provided (very) high spatial resolution Earth Observation (EO) data with abundant latent semantic information. Conventional data processing algorithms are not capable of extracting the latent semantic information form these data and harness their full potential. As a result, semantic information discovery methods, based on data mining techniques, such as Latent Dirichlet Allocation (LDA) and Bag of Visual Words (BOVW) models, can discover the latent information. Despite their crucial rule, there are only a few studies in the field of semantic data mining for remote sensing applications. This study is focused on this shortage. Three different scenarios are used to evaluate the semantic information discovery in various remote sensing applications, including both optical and Synthetic Aperture Radar (SAR) data with different spatial resolutions. In the first scenario, semantic discovery method correlated the semantic perception of the user and machine to correct and enhance the user defined Ground Truth (GT) map in very high-resolution RGB data. The potential of the semantic discovery is evaluated for wildfire affected area detection in Sentinel-2 data in the second scenario. Finally, in the third scenario, the semantic discovery method is utilized to detect the misclassifications as well as the patches with ambiguous or multiple semantic labels in a Sentinel-1 SAR patch-based benchmark dataset, to enhance the robustness and accuracy of the annotation in the dataset. Our results in these three scenarios demonstrated the capability of the data mining-based semantic information discovery methods for various remote sensing applications.

show abstract