• ETRI Journal
• /
• v.42 no.1
• /
• pp.67-77
• /
• 2020

Image captioning has received significant interest in recent years, and notable results have been achieved. Most previous approaches have focused on generating visual descriptions from images, whereas a few approaches have exploited visual descriptions for image classification. This study demonstrates that a good performance can be achieved for both description generation and image classification through an end-to-end joint learning approach with a loss function, which encourages each task to reach a consensus. When given images and visual descriptions, the proposed model learns a multimodal intermediate embedding, which can represent both the textual and visual characteristics of an object. The performance can be improved for both tasks by sharing the multimodal embedding. Through a novel loss function based on class activation mapping, which localizes the discriminative image region of a model, we achieve a higher score when the captioning and classification model reaches a consensus on the key parts of the object. Using the proposed model, we established a substantially improved performance for each task on the UCSD Birds and Oxford Flowers datasets.

• Journal of Korean Society for Geospatial Information Science
• /
• v.23 no.3
• /
• pp.85-93
• /
• 2015

This paper mainly introduces the methods of extracting landslide information using ALOS(Advanced Land Observing Satellite) images and GIS(Geographical Information System) technology. In this study, we classified images using three different methods which are the unsupervised the supervised and the PCA(Principal Components Analysis) for extracting landslide information based on characteristics of ALOS image. From the image classification results, we found out that the quality of classified image extracted with PCA supervised method was superior than the other images extracted with the other methods. But the accuracy of landslide information extracted from this image classification was still very low as the pixels were very similar between the landslide and safety regions. It means that it is really difficult to distinguish those areas with an image classification method alone because the values of pixels between the landslide and other areas were similar, particularly in a region where the landslide and other areas coexist. To solve this problem, we used the LSM(Landslide Susceptibility Map) created with ArcView software through weighted overlay GIS method in the areas. Finally, the developed LSM was applied to the image classification process using the ALOS images. The accuracy of the extracted landslide information was improved after adopting the PCA and LSM methods. Finally, we found that the landslide region in the study area can be calculated and the accuracy can also be improved with the LSM and PCA image classification methods using GIS tools.

• Journal of Digital Contents Society
• /
• v.18 no.8
• /
• pp.1635-1640
• /
• 2017

In this paper, a hierarchical image tree model for weather classification is defined in a single outdoor image, and a weather classification algorithm using image intensity and k-mean segmentation image is proposed. In the first level of the hierarchical image tree model, the indoor and outdoor images are distinguished. Whether the outdoor image is daytime, night, or sunrise/sunset image is judged using the intensity and the k-means segmentation image at the second level. In the last level, if it is classified as daytime image at the second level, it is finally estimated whether it is sunny or foggy image based on edge map and fog rate. Some experiments are conducted so as to verify the weather classification, and as a result, the proposed method shows that weather features are effectively detected in a given image.

• Journal of the Korean Geographical Society
• /
• v.43 no.5
• /
• pp.761-774
• /
• 2008

Integration of GIS data and human expert knowledge into digital image processing has long been acknowledged as a necessity to improve remote sensing image analysis. We propose inductive machine learning algorithm for GIS data integration and rule-based classification method for land cover classification. Proposed method is tested with a land cover classification of a Landsat ETM+ multispectral image and GIS data layers including elevation, aspect, slope, distance to water bodies, distance to road network, and population density. Decision trees and production rules for land cover classification are generated by C5.0 inductive machine learning algorithm with 350 stratified random point samples. Production rules are used for land cover classification integrated with unsupervised ISODATA classification. Result shows that GIS data layers such as elevation, distance to water bodies and population density can be effectively integrated for rule-based image classification. Intuitive production rules generated by inductive machine learning are easy to understand. Proposed method demonstrates how various GIS data layers can be integrated with remotely sensed imagery in a framework of knowledge base construction to improve land cover classification.

• The KIPS Transactions:PartD
• /
• v.9D no.5
• /
• pp.939-944
• /
• 2002

In this paper, we propose and apply new classification method to the remotely sensed image acquired from airborne multi-spectral scanner. This is a neuro-fuzzy image classifier derived from the generic model of a 3-layer fuzzy perceptron. We implement a classification software system with the proposed method for land cover image classification. Comparisons with the proposed and maximum-likelihood classifiers are also presented. The results show that the neuro-fuzzy classification method classifies more accurately than the maximum likelihood method. In comparing the maximum-likelihood classification map with the neuro-fuzzy classification map, it is apparent that there is more different as amount as 7.96% in the overall accuracy. Most of the differences are in the "Building" and "Pine tree", for which the neuro-fuzzy classifier was considerably more accurate. However, the "Bare soil" is classified more correctly with the maximum-likelihood classifier rather than the neuro-fuzzy classifier.

• Korean Journal of Remote Sensing
• /
• v.34 no.4
• /
• pp.671-680
• /
• 2018

Compared to the imagery produced by traditional satellites, PlanetScope satellite imagery has made it possible to easily capture remotely-sensed imagery every day through dozens or even hundreds of satellites on a relatively small budget. This study aimed to detect changed areas and update a land cover map using a PlanetScope image. To generate a classification map, pixel-based Random Forest (RF) classification was performed by using additional features, such as the Normalized Difference Water Index (NDWI) and the Normalized Difference Vegetation Index (NDVI). The classification result was converted to vector data and compared with the existing land cover map to estimate the changed area. To estimate the accuracy and trends of the changed area, the quantitative quality of the supervised classification result using the PlanetScope image was evaluated first. In addition, the patterns of the changed area that corresponded to the classification result were analyzed using the PlanetScope satellite image. Experimental results found that the PlanetScope image can be used to effectively to detect changed areas on large-scale land cover maps, and supervised classification results can update the changed areas.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.33 no.3
• /
• pp.203-209
• /
• 2015

This paper discusses the effectiveness of using Landsat 8 TIR (Thermal Infrared) band images to improve the accuracy of landuse/landcover classification of urban areas. According to classification results for the study area using diverse band combinations, the classification accuracy using an image fusion process in which the TIR band is added to the visible and near infrared band was improved by 4.0%, compared to that using a band combination that does not consider the TIR band. For urban area landuse/landcover classification in particular, the producer’s accuracy and user’s accuracy values were improved by 10.2% and 3.8%, respectively. When MLC (Maximum Likelihood Classification), which is commonly applied to remote sensing images, was used, the TIR band images helped obtain a higher discriminant analysis in landuse/landcover classification.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.33 no.3
• /
• pp.173-179
• /
• 2015

In this paper, object-based classification of urban areas based on a combination of information from lidar and aerial images is introduced. High resolution images are frequently used in automatic classification, making use of the spectral characteristics of the features under study. However, in urban areas, pixel-based classification can be difficult since building colors differ and the shadows of buildings can obscure building segmentation. Therefore, if the boundaries of buildings can be extracted from lidar, this information could improve the accuracy of urban area classifications. In the data processing stage, lidar data and the aerial image are co-registered into the same coordinate system, and a local maxima filter is used for the building segmentation of lidar data, which are then converted into an image containing only building information. Then, multiresolution segmentation is achieved using a scale parameter, and a color and shape factor; a compactness factor and a layer weight are implemented for the classification using a class hierarchy. Results indicate that lidar can provide useful additional data when combined with high resolution images in the object-oriented hierarchical classification of urban areas.

• Proceedings of the Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography Conference
• /
• 2003.04a
• /
• pp.277-281
• /
• 2003

We fused RADARSAT image and SPOT panchromatic image by wavelet transform in order to improve the accuracy of classification on the water area. Fused image in water not only maintained the characteristic of SAR image (low pixel value)but also had boundary information improved. This leads to accurate method to classify water areas.

• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.703-712
• /
• 2003

With the explosive growth n the numbers and sizes of imaging technologies, Content-Based Image Retrieval (CBIR) has been attacked the interests of researchers in the fields of digital libraries, image processing, and database systems. In general, in the case of query-by-image, in user has to select an image from database to query, even though it is not his completely desired one. However, since query-by-sketch approach draws a query shape according to the user´s desire it can provide more high-level searching interface to the user compared to the query-b-image. As a result, query-by-sketch has been widely used. In this paper, we propose a Java-based image retrieval system that consists of sketch query and image classification. We use two features such as color histogram and Haar wavelets coefficients to search similar images. Then the Leave-One-Out method is used to classify database images. The categories of classification are photo & painting, city & nature, and sub-classification of nature image. By using the sketch query and image classification, w can offer convenient image retrieval interface to user and we can also reduce the searching time.