Image Classification Results Research Articles

The Application of RS and GIS Technologies on Landslide Information Extraction of ALOS Images in Yanbian Area, China

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.

Dimensionality Reduction by Integrating Sparse Representation and Fisher Criterion and its Applications.

Sparse representation shows impressive results for image classification, however, it cannot well characterize the discriminant structure of data, which is important for classification. This paper aims to seek a projection matrix such that the low-dimensional representations well characterize the discriminant structure embedded in high-dimensional data and simultaneously well fit sparse representation-based classifier (SRC). To be specific, Fisher discriminant criterion (FDC) is used to extract the discriminant structure, and sparse representation is simultaneously considered to guarantee that the projected data well satisfy the SRC. Thus, our method, called SRC-FDC, characterizes both the spatial Euclidean distribution and local reconstruction relationship, which enable SRC to achieve better performance. Extensive experiments are done on the AR, CMU-PIE, Extended Yale B face image databases, the USPS digit database, and COIL20 database, and results illustrate that the proposed method is more efficient than other feature extraction methods based on SRC.

Design and Implementation of Earth Image Classification Using Unmanned Aerial Vehicle

Research in the field of image classification has been widely applied and developed, especially in the field of satellite imagery. Image classification is the process of grouping the pixels in an image into a number of classes, so that each class can describe an entity with certain characteristics. The research aims to build software that can perform the classification of earth image results from UAV (Unmanned Aerial Vehicle) monitoring. The Image converted into YUV format then classified using Fuzzy Support Vector Machine (FSVM). This research designed elements that UAVs will be used for monitoring as follows: (1) the control station, which designed the software on a computer that is used to send or receive data, and display the data in graphical form, (2) payload, using the camera to capture images and send to the control station, (3) communication system using TCP/IP protocol, and (4) UAV, using X650 quadcopter products from xaircraft. All of data can be received if it is sent by several segmented package into smaller parts. The results of image classification, the image of the monitoring carried out on the UAV sized 256 x 256 pixels with a total number of 450 training data size. It is 16x16 pixel image data. Tests performed to classify the image into 3 classes, namely agricultural area, residential area, and water area. The highest accuracy value of 77.69% obtained by the number of training data as much as 375.

Картографирование сельскохозяйственных культур Вьетнама с использованием ГИС и данных многозональной космической съемки LANDSAT-7

Abstract: The article is devoted to presenting of a technique of mapping of crops based on GIS and remote sensing. The generalized method of automatic supervised classification of crops for multispectral satellite imagery LANDSAT 7 in Vietnam and the results of image classification by means of mathematical modeling, and provides vector processing technology results in a GIS environment. Images and statistics were chosen on the basis of the frequency of four months in one calendar year, starting in January. To show the accuracy of the classification used calculation tool matrix of contradictions, which were calculated errors and omissions attorney. When creating a map of crops made a generalization of image classification results. Separate out crops: rice, vegetables, corn and other cereals. In general, the map can be used in the planning and development of the rural economy of Vietnam and the countries of Southeast Asia.

Laplacian regularized locality-constrained coding for image classification

Feature coding, which encodes local features extracted from an image with a codebook and generates a set of codes for efficient image representation, has shown very promising results in image classification. Vector quantization is the most simple but widely used method for feature coding. However, it suffers from large quantization errors and leads to dissimilar codes for similar features. To alleviate these problems, we propose Laplacian Regularized Locality-constrained Coding (LapLLC), wherein a locality constraint is used to favor nearby bases for encoding, and Laplacian regularization is integrated to preserve the code consistency of similar features. By incorporating a set of template features, the objective function used by LapLLC can be decomposed, and each feature is encoded by solving a linear system. Additionally, k nearest neighbor technique is employed to construct a much smaller linear system, so that fast approximated coding can be achieved. Therefore, LapLLC provides a novel way for efficient feature coding. Our experiments on a variety of image classification tasks demonstrated the effectiveness of this proposed approach.

Histogram of dense subgraphs for image representation

Modelling spatial information of local features is known to improve performance in image categorisation. Compared with simple pairwise features and visual phrases, graphs can capture the structural organisation of local features more adequately. Besides, a dense regular grid can guarantee a more reliable representation than the interest points and give better results for image classification. In this study, the authors introduced a bag of dense local graphs approach that combines the performance of bag of visual words expressing the image classification process with the representational power of graphs. The images were represented with dense local graphs built upon dense scale-invariant feature transform descriptors. The graph-based substructure pattern mining algorithm was applied on the local graphs to discover the frequent local subgraphs, producing a bag of subgraphs representation. The results were reported from experiments conducted on four challenging benchmarks. The findings show that the proposed subgraph histogram improves the categorisation accuracy.

Crop Area Estimation from UAV Transect and MSR Image Data Using Spatial Sampling Method

Using remote sensing data to estimate crop area is efficient to a wide range of end-users, including government agencies, farmers and researchers. Moderate spatial resolution (MSR) image data are widely used to estimate crop area. But its accuracy can’t meet the demands of precision. Spatial sampling techniques integrated the strengths of remote sensing and sampling survey are being widely used. This method need large sample size which is cannot be guaranteed by remote sensing due to weather. The Unmanned Aerial Vehicle (UAV) can be used as an effective way to guarantee enough sample size. This paper proposed a spatial sampling method using MSR image classification results and UAV transects, a stratified random sampling method was proposed, area-scale (from MSR image classification) was used as auxiliary variable to guide the distribution of UAV transects, which had proved that 2% sampling ratio can make the crop area estimation accuracy more than 95% with a 95% confidence interval.

Interfacing robotic tactile sensation with human vibrotactile perception for digit recognition

To achieve human like functionality, artificial arms are required to be incorporated with the sense of touch that can be achieved by affixing tactile sensors to them. The user operating these artificial arms, for any Human–Computer Interactive task, should be able to perceive the same tactile sensation as is acquired by the tactile sensor fitted artificial arms. This work is an initial step to interface artificial tactile sensations of robots with human perception. Pressure distribution images are acquired while a robot arm fitted with a tactile sensor palpates embossed digits (0–9) and are classified. The image classification results are transmitted via control signals as vibrotactile stimuli to humans using vibration motors attached on their backs. Vibrotactile stimulus is chosen as a feedback sensation in an attempt to not engage visual/auditory channels such that the work can be extended in a multitasking environment in future. Vibrotactile stimuli are also essential for a person using prosthetic arm, suffering from sensory-motor disabilities that inhibit the use of the audio and/or visual sensory channels as well as for applications such as tele-operation. Electroencephalogram (EEG) signal analysis during vibrotactile stimulus presentation as well as subjects’ response evaluation show that vibrotactile stimuli for particular digits as recognized from tactile data is capable of imparting information from artificial somato-sensation to humans. Fuzzy k-Nearest Neighbor classifier classifies both tactile images and vibrotactually stimulated EEG signals online with average accuracies of 85.25% and 72.60%, and runtimes of the order of 0.11 and 35 s respectively.

이미지 단어집과 관심영역 자동추출을 사용한 이미지 분류

플리커, 페이스북과 같은 대용량 소셜 미디어 공유 사이트의 발전으로 이미지 정보가 매우 빠르게 증가하고 있다. 이에 따라 소셜 이미지를 정확하게 검색하기 위한 다양한 연구가 활발히 진행되고 있다. 이미지 태그들의 의미적 연관성을 이용하여 태그기반의 이미지 검색의 정확도를 높이고자 하는 연구를 비롯하여 이미지 단어집(Bag of Visual Words)을 기반으로 웹 이미지를 분류하는 연구도 다양하게 진행되고 있다. 본 논문에서는 이미지에서 배경과 같은 중요도가 떨어지는 정보를 제거하여 중요부분을 찾는 GBVS(Graph Based Visual Saliency)모델을 기존 연구에 사용할 것을 제안한다. 제안하는 방법은 첫 번째, 이미지 태그들의 의미적 연관성을 이용해 1차 분류된 데이터베이스에 SIFT알고리즘을 사용하여 이미지 단어집(BoVW)을 만든다. 두 번째, 테스트할 이미지에 GBVS를 통해서 이미지의 관심영역을 선택하여 테스트한다. 의미연관성 태그와 SIFT기반의 이미지 단어집을 사용한 기존의 방법에 GBVS를 적용한 결과 더 높은 정확도를 보임을 확인하였다. As social multimedia sites are getting popular such as Flickr and Facebook, the amount of image information has been increasing very fast. So there have been many studies for accurate social image retrieval. Some of them were web image classification using semantic relations of image tags and BoVW(Bag of Visual Words). In this paper, we propose a method to detect salient region in images using GBVS(Graph Based Visual Saliency) model which can eliminate less important region like a background. First, We construct BoVW based on SIFT algorithm from the database of the preliminary retrieved images with semantically related tags. Second, detect salient region in test images using GBVS model. The result of image classification showed higher accuracy than the previous research. Therefore we expect that our method can classify a variety of images more accurately.

Manifold-Based Sparse Representation for Hyperspectral Image Classification

A sparsity-based model has led to interesting results in hyperspectral image (HSI) classification. Sparse representation from a test sample is used to identify the class label. However, an $\ell_{1}$ -based sparse algorithm sometimes yields unstable sparse representation. Inspired by recent progress in manifold learning, two manifold-based sparse representation algorithms are proposed to exploit the local structure of the test samples in corresponding sparse representations for enforcing smoothness across neighboring samples' sparse representations. Using techniques from regularization and local invariance, two manifold-based regularization terms are incorporated into the $\ell_{1}$ -based objective function. Extensive experiments show that our proposed algorithms obtain excellent classification performance on three classic HSIs.

Elastic net regularized dictionary learning for image classification

Dictionary learning plays a key role in image representation for classification. A multi-modal dictionary is usually learned from feature samples across different classes and shared in the feature encoding process. Ideally each atom in dictionary corresponds to a single class of images, while each class of images corresponds to a certain group of atoms. Image features are encoded as linear combinations of selected atoms in a given dictionary. We propose to use elastic net as regularizer to select atoms in feature coding and related dictionary learning process, which not only benefits from the sparsity similar as l1 penalty but also encourages a grouping effect that helps improve image representation. Experimental results of image classification on benchmark datasets show that with dictionary learned in the proposed way outperforms state-of-the-art dictionary learning algorithms.

Performance evaluation of new colour histogram-based interest point detectors

Interest point detection is an active area in computer vision due to its importance in many applications. Measuring the pixel-wise difference between image pixel intensities is the mechanism of most detectors that have been proposed in literature. Recently, interest point detectors were proposed that incorporated the histogram representation instead of image pixel intensity. In this paper, research that extends histogram-based interest point detectors is introduced. Four colour-space representations were used to construct new detectors: HSV, Opponent, Transformed and Ohta colour spaces. Several experiments were performed to evaluate the new colour histogram-based detectors and compare them with previous detectors. First, the proposed detectors were evaluated in an image-matching task. Then, we studied and evaluated the performance of some of the local image descriptors that were extracted from the interest points and regions detected by the proposed detectors. Finally, the four top-ranked descriptors in the descriptor evaluation experiments were used to evaluate the new colour histogram-based detectors in an image-classification task using different object and scene image datasets. The experimental results demonstrate that our new detectors possess an increased ability to distinguish and more robust in regards to image matching, particularly with respect to textured scene images that involve transformations, such as illumination, viewpoint and blur changes. Furthermore, the descriptor performance may change depending on the detector and data set type. The image-classification results demonstrate that the proposed detectors exhibit higher classification accuracy for certain descriptors and data sets than the other detectors.

Expansion of a glacial lake, Tsho Chubda, Chamkhar Chu Basin, Hindukush Himalaya, Bhutan

A wide range of methods for detection of glacial lakes and their expansions using multi-temporal remote sensing images have been employed in the past. This paper presents a framework for semiautomatic detection of glacial lakes and estimation of its expansion in Chamkhar Chu Basin, Hindukush Himalaya, Bhutan, with the help of ASTER multispectral image classifications. Lakes in the glacierized area tend to have a varying spectral response ranging from light blue or green to almost black which makes them difficult to be differentiated from shadows in the region. Detection of glacial lakes has been performed using NDWI technique and support vector machine image classification approach and results were compared. An integrated thermal–optical dataset was generated for applying SVM technique, and results showed that the lakes under cast shadow were accurately detected. Water spread extent of the lake has been estimated and it was observed that the present expansion rate of glacial lake has increased by twofolds from its previous rate in the last decade. The formations, deformations around glacial lake and related geomorphological variations occurring around lake were employed in defining the expansion mechanism of the lake. This study also demonstrates that ASTER data provide the possibility of accurate detections and estimation of water spread of glacial lake without simultaneous field observations.

A Nonlocal Weighted Joint Sparse Representation Classification Method for Hyperspectral Imagery

As a powerful and promising statistical signal modeling technique, sparse representation has been widely used in various image processing and analysis fields. For hyperspectral image classification, previous studies have shown the effectiveness of the sparsity-based classification methods. In this paper, we propose a nonlocal weighted joint sparse representation classification (NLW-JSRC) method to improve the hyperspectral image classification result. In the joint sparsity model (JSM), different weights are utilized for different neighboring pixels around the central test pixel. The weight of one specific neighboring pixel is determined by the structural similarity between the neighboring pixel and the central test pixel, which is referred to as a nonlocal weighting scheme. In this paper, the simultaneous orthogonal matching pursuit technique is used to solve the nonlocal weighted joint sparsity model (NLW-JSM). The proposed classification algorithm was tested on three hyperspectral images. The experimental results suggest that the proposed algorithm performs better than the other sparsity-based algorithms and the classical support vector machine hyperspectral classifier.

Assessing the relationship between water quality parameters and changes in landuse patterns in the Upper Manyame River, Zimbabwe

For the past 30years, the increases in population pressure and external influences, such as economic growth, have accelerated the demand for land within the Upper Manyame River catchment in Zimbabwe which has caused substantial changes in landuse. The general objective of this research was to assess the impacts of landuse activities on the water quality of the Upper Manyame River which drains the rural and urbanised part of the catchment up to flow gauging station C21. Landcover data for the month of April in years of 1984, 1995, 2003 and 2011 were acquired from available Landsat TM and ETM images and were classified through the maximum likelihood digital image classification using the supervised classification approach. The status of water quality of the Upper Manyame River was also assessed through analyses of historical concentrations and pollution loads for TP, DO, COD, NH3-N, SS, Pb, NO3, BOD5, EC, PO4-P and TN at the Environmental Management Agency (EMA) gauging station CR21 sampling point for 1996, 2000/1 and 2008/9. Water quality of 15 monitoring sites comprising 25 water quality parameters were monitored monthly from January to June 2012. These locations were selected to reflect a wide array of landuse for both the dry and wet seasons. The results indicated that there was an increase in pollution load from 1995 to 2012; for TP from 130kg/day to 376kg/d, and for TN from 290kg/day to 494kg/d. This indicates high pollution levels which have severe impacts on downstream users and also severe sewage contamination. Significant deviations occurred in DO (0.1–6.8) mg/L, COD (11–569) mg/L, BOD5 (5–341) mg/L, PO4-P (0.01–4.45) mg/L, NH3-N (0.001–6.800) mg/L and EC (38–642) μS/cm. Hydrologic Response Unit and buffer analysis were used to determine the dominant landuse which contributes to a certain water quality. Results of digital image classification indicate that woodland/forest, grassland and bareland decreased between years 1984 to 2011 by 24.0%, 22.6% and 31.7% respectively. This was mainly due to expansion or increase of agriculture and urban areas by 24.4% and 41.6% respectively over the same time period. It was concluded that settlements and agricultural areas are the ones mainly affecting the water quality in Upper Manyame River with a Pearson’s Correlation Coefficient of r=0.97 for COD and r=0.78 for TSS respectively. It was also observed that the water quality status of the Upper Manyame River and its tributaries is very poor and the level of pollution generally increases from upstream to downstream. The results obtained provided baseline information which may be used in the development of appropriate water quality management systems in the catchment. Thus the study recommended a combined programme of point source control and landuse modification.

Support Vector Machines for Object Based Building Extraction in Suburban Area using Very High Resolution Satellite Images, a Case Study: Tetuan, Morocco

Many fields of artificial intelligence have been developed such as computational intelligence and machine learning involving neural networks, fuzzy systems, genetic algorithms, intelligent agents and Support Vector Machines (SVM). SVM is a machine learning methodology with great results in image classification. In this paper, we present the potential of SVMs to automatically extract buildings in suburban area using Very High Resolution Satellite (VHRS) images. To achieve this goal, we use object based approach: Segmentation before classification in order to create meaningful image objects using color features. In the first step, we form objects with the aid of mean shift clustering algorithm. Then, SVM classifier was used to extract buildings. The proposed method has been applied on a suburban area in Tetuan city (Morocco) and 83.76% of existing buildings have been extracted by only using color features. This result can be improved by adding other features (e.g., spectral, texture, morphology and context). DOI: http://dx.doi.org/10.11591/ij-ai.v2i1.1781

Knowledge Extraction from HEp2 Dataset

Pattern recognition is the process of taking the raw data, extract the features and categorize the patterns based on the features. Pattern recognition has been widely used in the area of diseases diagnosis. Autoimmune diseases are proven to be connected with the occurrence of autoantibodies in patient serum. Antinuclear autoantibodies (ANAs) identification can be accomplished in a laboratory using indirect immunofluorescence (IIF) imaging. In this paper we present the results of image analysis image analysis, feature extraction and classification of HEp-2 cell images. The experimental studies are performed on images of HEp-2 cells (1). 

A Method of Spatial Mapping and Reclassification for High‐Spatial‐Resolution Remote Sensing Image Classification

This paper presents a new classification method for high-spatial-resolution remote sensing images based on a strategic mechanism of spatial mapping and reclassification. The proposed method includes four steps. First, the multispectral image is classified by a traditional pixel-based classification method (support vector machine). Second, the panchromatic image is subdivided by watershed segmentation. Third, the pixel-based multispectral image classification result is mapped to the panchromatic segmentation result based on a spatial mapping mechanism and the area dominant principle. During the mapping process, an area proportion threshold is set, and the regional property is defined as unclassified if the maximum area proportion does not surpass the threshold. Finally, unclassified regions are reclassified based on spectral information using the minimum distance to mean algorithm. Experimental results show that the classification method for high-spatial-resolution remote sensing images based on the spatial mapping mechanism and reclassification strategy can make use of both panchromatic and multispectral information, integrate the pixel- and object-based classification methods, and improve classification accuracy.

On Confidence-Constrained Rank Recovery in Topic Models

Topic models have been proposed to model a collection of data such as text documents and images in which each object (e.g., a document) contains a set of instances (e.g., words). In many topic models, the dimension of the latent topic space (the number of topics) is assumed to be a deterministic unknown. The number of topics significantly affects the prediction performance and interpretability of the estimated topics. In this paper, we propose a confidence-constrained rank minimization (CRM) to recover the exact number of topics in topic models with theoretical guarantees on recovery probability and mean squared error of the estimation. We provide a computationally-efficient optimization algorithm for the problem to further the applicability of the proposed framework to large real world datasets. Numerical evaluations are used to verify our theoretical results. Additionally, to illustrate the applicability of the proposed framework to practical problems, we provide results in image classification for two real world datasets and text classification for three real world datasets.

Adaptive Hybrid Feature Extraction for Leaf Image Classification by Support Vector Machine

In leaf image classification or retrieval fields, hybrid features are widely used to represent the information in various aspects by combining a number of sub-features linearly. However, the importance degrees of sub-features are often ignored by assigning the weights in an ad-hoc fashion without a solid theoretical basis. In this paper, a new type of adaptive hybrid features is proposed by using kernel trick of support vector machine (SVM), in which the weights can be adaptively selected. All weights are obtained by solving an optimization problem to maximize the discriminability of features. Experimental results of leaf image classification show that SVMs with new features significantly outperform those with traditional ones in terms of test accuracy.