• Korean Journal of Remote Sensing
• /
• v.39 no.6_3
• /
• pp.1771-1777
• /
• 2023

With the advancements of deep learning, many semantic segmentation-based methods for land cover classification have been proposed. However, existing deep learning-based models only use image information and cannot guarantee spatiotemporal consistency. In this study, we propose a land cover classification model using geographical coordinates. First, the coordinate features are extracted through the Coordinate Hash Encoder, which is an extension of the Multi-resolution Hash Encoder, an implicit neural representation technique, to the longitude-latitude coordinate system. Next, we propose an architecture that combines the extracted coordinate features with different levels of U-net decoder. Experimental results show that the proposed method improves the mean intersection over union by about 32% and improves the spatiotemporal consistency.

• Proceedings of the KSRS Conference
• /
• 2008.10a
• /
• pp.76-79
• /
• 2008

This study presents the use of multi-temporal JERS-1 SAR images to the land cover classification. So far, land cover classified by high resolution aerial photo and field survey and so on. The study site was located in Non-san area. This study developed on multi-temporal land cover status monitoring and coherence information mapping can be processing by L band SAR image. From July, 1997 to October, 1998 JERS SAR images (9 scenes) coherence values are analyzed and then classified land cover. This technique which forms the basis of what is called SAR Interferometry or InSAR for short has also been employed in spaceborne systems. In such systems the separation of the antennas, called the baseline is obtained by utilizing a single antenna in a repeat pass

• Proceedings of the KSRS Conference
• /
• 2007.10a
• /
• pp.475-478
• /
• 2007

This study presents the use of multi-temporal JERS-1 SAR images to extract the land cover information and possibility. So far, land cover information extracted by high resolution aerial photo and field survey. The study site was located in Non-san area. This study developed on multi-temporal land cover status monitoring and coherence information mapping can be processing by L band SAR image. From July, 1997 to October, 1998 JERS SAR images (9 scenes) coherence values are analyzed and then extracted land cover information factors, so on. This technique which forms the basis of what is called SAR Interferometry or InSAR for short has also been employed in spaceborne systems. In such systems the separation of the antennas, called the baseline is obtained by utilizing a single antenna in a repeat pass

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.20 no.1
• /
• pp.21-31
• /
• 2002

This study applied each Neural Networks theory and Fuzzy Set theory to improve accuracy in remotely sensed images. Remotely sensed data have been used to map land cover. The accuracy is dependent on a range of factors related to the data set and methods used. Thus, the accuracy of maps derived from conventional supervised image classification techniques is a function of factors related to the training, allocation, and testing stages of the classification. Conventional image classification techniques assume that all the pixels within the image are pure. That is, that they represent an area of homogeneous cover of a single land-cover class. But, this assumption is often untenable with pixels of mixed land-cover composition abundant in an image. Mixed pixels are a major problem in land-cover mapping applications. For each pixel, the strengths of class membership derived in the classification may be related to its land-cover composition. Fuzzy classification techniques are the concept of a pixel having a degree of membership to all classes is fundamental to fuzzy-sets-based techniques. A major problem with the fuzzy-sets and probabilistic methods is that they are slow and computational demanding. For analyzing large data sets and rapid processing, alterative techniques are required. One particularly attractive approach is the use of artificial neural networks. These are non-parametric techniques which have been shown to generally be capable of classifying data as or more accurately than conventional classifiers. An artificial neural networks, once trained, may classify data extremely rapidly as the classification process may be reduced to the solution of a large number of extremely simple calculations which may be performed in parallel.

• Korean Journal of Remote Sensing
• /
• v.33 no.4
• /
• pp.401-410
• /
• 2017

Estimation of snow depth using optical image is conducted by using correlation with Snow Cover Fraction (SCF). Various algorithms have been proposed for the estimation of snow cover fraction based on Normalized Difference Snow Index (NDSI). In this study we tested linear, quadratic, and exponential equations for the generation of snow cover fraction maps using data from the Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua satellite in order to evaluate their applicability to the complex terrain of South Korea and to search for improvements to the estimation of snow depth on this landscape. The results were validated by comparison with in-situ snowfall data from weather stations, with Root Mean Square Error (RMSE) calculated as 3.43, 2.37, and 3.99 cm for the linear, quadratic, and exponential approaches, respectively. Although quadratic results showed the best RMSE, this was due to the limitations of the data used in the study; there are few number of in-situ data recorded on the station at the time of image acquisition and even the data is mostly recorded on low snowfall. So, we conclude that linear-based algorithms are better suited for use in South Korea. However, in the case of using the linear equation, the SCF with a negative value can be calculated, so it should be corrected. Since the coefficients of the equation are not optimized for this area, further regression analysis is needed. In addition, if more variables such as Normalized Difference Vegetation Index (NDVI), land cover, etc. are considered, it could be possible that estimation of national-scale snow depth with higher accuracy.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.16 no.2
• /
• pp.185-190
• /
• 2013

The previous steganographic methods by using the interpolation were difficult to estimate the distortion because the size of cover image is extended by interpolation algorithms. In this paper, to solve the problems of previous methods proposed the improved steganographic method based on the pixel replacement algorithms. In our method, we cannot extend a cover image, but also can estimate exactly the distortion of the stego-images. In the experimental results, the estimated distortion and embedding capacity of stego-image are shown on three pixel replacement methods.

• Convergence Security Journal
• /
• v.10 no.2
• /
• pp.19-26
• /
• 2010

In this study, we try to make up for the vulnerability in steganography that it is easily revealed the hidden logo image in cover image by bit-plane extraction. For this, we apply some methods, the permutation which shift the scattered pieces of logo image to one side, bit-plane dispersion insertion method and pack-type compressor.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.30 no.6_1
• /
• pp.583-590
• /
• 2012

This study proposes fusion image matching method according to land cover property to generate a detailed DEM using the high resolution IKONOS-2 stereo pair. A classified image, consists of building, crop-land, forest, road and shadow-water, is produced by color image with four bands. Edges and points are also extracted from panchromatic image. Matching is performed by the cross-correlation computing after five classes are automatically selected in a reference image. In each of building class, crop-land class, forest class and road class, matching was performed by the grid and edge, only grid, only grid, grid and point, respectively. Shadow-water class was excepted in the matching because this area causes excessive error of the DEM. As the results, edge line, building and residential area could be expressed more dense than DEM by the conventional method.

• International Journal of Computer Science & Network Security
• /
• v.22 no.2
• /
• pp.214-222
• /
• 2022

Convolutional Neural networks (CNNs) are a category of deep learning networks that have proven very effective in computer vision tasks such as image classification. Notwithstanding, not much has been seen in its use for remote sensing image classification in developing countries. This is majorly due to the scarcity of training data. Recently, transfer learning technique has successfully been used to develop state-of-the art models for remote sensing (RS) image classification tasks using training and testing data from well-known RS data repositories. However, the ability of such model to classify RS test data from a different dataset has not been sufficiently investigated. In this paper, we propose a deep CNN model that can classify RS test data from a dataset different from the training dataset. To achieve our objective, we first, re-trained a ResNet-50 model using EuroSAT, a large-scale RS dataset to develop a base model then we integrated Augmentation and Ensemble learning to improve its generalization ability. We further experimented on the ability of this model to classify a novel dataset (Nig_Images). The final classification results shows that our model achieves a 96% and 80% accuracy on EuroSAT and Nig_Images test data respectively. Adequate knowledge and usage of this framework is expected to encourage research and the usage of deep CNNs for land cover mapping in cases of lack of training data as obtainable in developing countries.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.19 no.2
• /
• pp.98-106
• /
• 2016

Remote sensing is an efficient technology for observing and monitoring the land surfaces inaccessible to humans. This research proposes a methodology for improving the accuracy of the land cover classification using the Landsat-8 operational land imager(OLI) image. The proposed methodology consists of the following steps. First, the normalized difference vegetation index(NDVI) and normalized difference water index(NDWI) images are generated from the given Landsat-8 OLI image. Then, a new image is generated by adding both NDVI and NDWI images to the original Landsat-8 OLI image using the layer-stacking method. Finally, the maximum likelihood classification(MLC), and support vector machine(SVM) methods are separately applied to the original Landsat-8 OLI image and new image to identify the five classes namely water, forest, cropland, bare soil, and artificial structure. The comparison of the results shows that the utilization of the layer-stacking method improves the accuracy of the land cover classification by 8% for the MLC method and by 1.6% for the SVM method. This research proposes a methodology for improving the accuracy of the land cover classification by using the layer-stacking method.