Figure 1 - uploaded by Raghavendra Ramachandra
Content may be subject to copyright.
Source publication
Cloud analysis provides information which is vital to the detection, understanding and prediction of meteorological trends and environmental changes. Cloud classification is needed for the purpose of achieving an automatic extraction of information on cloud occurrence and cloud types. In this work, we propose a method of cloud classification using...
Context in source publication
Similar publications
In this research we propose a new scheme for generating binary vectors, which can be used as keys for cryptographic purposes. These vectors are obtained from the speech signal and from the spoken user passphrase. The key bits are built using the Automatic Speech Recognition Technology to detect the phoneme limits in the speech utterance and the Sup...
In this notebook paper we describe the technical details of the submissions to TRECVID 2006 from CMU Informedia team. We participated in the high-level feature extraction and the search (automatic and interactive) tasks. Our emphasis is on various techniques used for the search task, where our interactive runs won the first place in the interactive...
High computational costs and large memory requirements which are particularly relevant in embedded systems often result from a large support vector (SV) set of a non linear kernel support vector machine (SVM). We propose effective and straight forward methods, an improved RANSAC-SVM algorithm, called RANSAC-SV, and a novel algorithm SortedVotes, fo...
In this paper we propose two ways of incomplete data representation. The first one is a generalization of a flag representation, where a vector with missing attributes is filled with some values and joined with flag vectors indicating missing components. Our generalization uses pointed affine subspaces, which in addition to flag representation allo...
An important approach for efficient support vector machine (SVM) model selection is to use differentiable bounds of the leave-one-out (loo) error. Past efforts focused on finding tight bounds of loo (e.g., radius margin bounds, span bounds). However, their practical viability is still not very satisfactory. Duan, Keerthi, and Poo (2003) showed that...
Citations
... Deng et al. [32] used the Gabor transform for extracting the texture features in cloud detection to improve the ability of the algorithm to distinguish between cloud regions and highlighted snow regions. Chethan et al. [43] also used Gabor transform to extract texture features for cloud detection primarily. Because Gabor transform can extract various texture features that are important for cloud detection, we have introduced the Gabor feature extraction module into the network in order to enhance the learning ability of the network for texture features. ...
... The value of λ is usually greater than 2. Therefore, the range of values is [2,5], and the interval is 1. The size of the Gabor filters is 3 × 3. Deng et al. [32] and Chethan et al. [43] both used Gabor transform to extract texture features for cloud detection primarily, and improve the ability of the algorithm in order to distinguish between cloud regions and highlighted snow regions. They both use eight different orientations. ...
Cloud detection, as a crucial step, has always been a hot topic in the field of optical remote sensing image processing. In this paper, we propose a deep learning cloud detection Network that is based on the Gabor transform and Attention modules with Dark channel subnet (NGAD). This network is based on the encoder-decoder framework. The information on texture is an important feature that is often used in traditional cloud detection methods. The NGAD enhances the attention of the network towards important texture features in the remote sensing images through the proposed Gabor feature extraction module. The channel attention module that is based on the larger scale features and spatial attention module that is based on the dark channel subnet have been introduced in NGAD. The channel attention module highlights the important information in a feature map from the channel dimensions, weakens the useless information, and helps the network to filter this information. A dark channel subnet with spatial attention module has been designed in order to further reduce the influence of the redundant information in the extracted features. By introducing a “dark channel”, the information in the feature map is reconstructed from the spatial dimension. The NGAD is validated while using the Gaofen-1 WFV imagery in four spectral bands. The experimental results show that the overall accuracy of NGAD reaches 97.42% and the false alarm rate reaches 2.22%. The efficiency of cloud detection using NGAD exceeds the state-of-art image segmentation network model and remote sensing image cloud detection model.
... After obtaining the feature description of the ground-based cloud image, selecting a suitable classifier according to the extracted ground-based cloud image features also affects the final effect of ground-based cloud recognition to a certain extent. Chethan et al. [17] proposed a texture feature cloud classification method based on Gabor transform, and used the improved classifier Support Vector Machine (SVM) for classification. Alireza et al. [18] use multilayer perceptron (MLP) neural networks and support vector machine (SVM)] capabilities for automatic cloud detection in whole-sky images. ...
The recognition of ground-based cloud images has rich application prospects in many aspects such as weather prediction, astronomical site selection, and meteorological observation. Affected by factors such as rotation and illumination, the traditional feature extraction method is difficult to accurately describe the features of cloud images, resulting in low accuracy of ground-based cloud image recognition, and cannot meet the requirements of practical applications. With the popularity of convolutional neural networks in image processing, ground-based cloud image recognition algorithms based on convolutional neural network have become a research focus. However, the features of the ground-based cloud image are relatively shallow, and the cloud texture and other features are seriously lost in the convolution process, and it is difficult to achieve a good recognition effect. This paper proposes a ground-based cloud image recognition system based on multi-scale convolutional neural network (Multi-CNN) and multilayer perceptron neural networks (MLP). The multi-level and multi-scale convolution feature extraction is performed through convolution layers of Multi-CNN, and the local features with strong resolving power are selected through the feature screening algorithm based on DP clustering. Finally, the local features are encoded and fused for cloud image classification based on MLP. Filed test results show that our method was superior to other tested network models in terms of the recognition accuracy of 94.8% under 9 classification. In addition, ablation experiments show that the multi-scale feature extraction, screening and local feature coding in this paper have a significant effect on improving the algorithm's ability to distinguish different cloud images.
... A great number of cloud detection approaches for optical RS image have been proposed. In [6], Chethan et al. proposed an algorithm of cloud detection by exploiting Gabor transform to extract the texture information. Then the support vector machine (SVM) is adopted to classify the cloud region and non-cloud region. ...
Cloud detection plays a significant role in remote sensing (RS) image processing. Numbers of cloud detection algorithms have been developed in the literature. However, they suffer the weakness of omitting thin and small cloud, and poor ability of differentiating the cloud from confusing ground region (e.g. artificial building). In this study, a robust ragged cloud detection algorithm for RS image is proposed. First, the simple linear iterative clustering method is applied to segment ragged cloud. Then, the improved Qtsu's method is introduced to remove the redundant superpixel. Finally, the Natural Scene Statistic is designed to classify the cloud region. Finally, original image will be classified into thick cloud, thin cloud and non-cloud. Experimental results indicate that the proposed model outperforms the state-of-the-art methods for cloud detection.
... Nevertheless, the detection results of these algorithms are highly affected by image registration and light intensity, making it inconvenient and inaccurate for cloud detection. Later algorithms [7]- [11] extract more sophisticated characteristics, including color, texture, structure, and so on. These strategies also suffer from various drawbacks. ...
... These strategies also suffer from various drawbacks. For example, Chethan et al. [7] only extracted texture information using Gabor filtering without taking pixel intensities and regional structure into account. Rossi et al. [8] calculated the singular value decomposition of images, then detected clouds using a support vector machine (SVM), while neglecting small clouds. ...
... Again, our model outperformed the other cloud detection methods. The models compared are the SVM method in [7], the progressive refinement scheme (PRS) model in [15], and a CNN-based approach [11]. The quantitative performance measures we used are the precision and recall ...
Cloud detection is an important task in remote sensing (RS) image processing. Numerous cloud detection algorithms have been developed. However, most existing methods suffer from the weakness of omitting small and thin clouds, and from an inability to discriminate clouds from photometrically similar regions, such as buildings and snow. Here, we derive a novel cloud detection algorithm for optical RS images, whereby test images are separated into three classes: thick clouds, thin clouds, and noncloudy. First, a simple linear iterative clustering algorithm is adopted that is able to segment potential clouds, including small clouds. Then, a natural scene statistics model is applied to the superpixels to distinguish between clouds and surface buildings. Finally, Gabor features are computed within each superpixel and a support vector machine is used to distinguish clouds from snow regions. The experimental results indicate that the proposed model outperforms state-of-the-art methods for cloud detection.
... However, the considerable number of factors to be considered leads to the necessity of performing the analysis by areas, because clouds are one of the main sources of uncertainty in the development of models for climate prediction and its conformation depends on geographic conditions. Other examples of image-based detection are [1], [6], [7] where the algorithms used can detect volcanic ash, atmospheric gas concentration, reluctance and temperature atmospheric changes, dust, fires, smoke and clouds according to Mahani, Rillo, and Chethan [8]- [10]. ...
... The interpretation and visual classification of cloud types is done manually by experts. This process is significantly dependent on the experience of the person who performs the analysis, requiring time, training and personal For automatic detection of convective system, based on Chetan, Peak, Desbois and Azimi-Sadjadi [10], [27]- [29], artificial intelligence (AI) and digital image processing (DIP) techniques are used typically to automate the meteorological pattern recognition. Other authors, like Yin and Qian, [7], have worked with unsupervised methods using propose intelligent segmentation techniques as a recognition method. ...
En la aviación, los fenómenos meteorológicos son uno de los aspectos más importantes para tener en cuenta en todas las etapas de vuelo, desde la planificación hasta el aterrizaje. El desarrollo de sistemas de predicción meteorológica aplicados a la aviación puede apoyar el proceso de toma de decisiones de los controladores de tráfico aéreo y los pilotos, facilitando el análisis de las variables meteorológicas y proporcionando una primera interpretación a disposición de todos los usuarios del sistema aéreo.
Por esta razón el Centro de desarrollo Tecnológico para la Defensa (CETAD) tiene como principal objetivo en este documento describir los resultados del desarrollo de una metodología sistematizada que utiliza técnicas inteligentes para la detección, identificación y seguimiento de cualquier tipo de formación que por sus características pueda representar un riesgo para la aviación, generando a su vez información de soporte al controlador aéreo.
Para esto es necesario primero identificar las formaciones convectivas, clasificarlas, filtrar el ruido e individualizarlas. Este tipo de procesos pueden ser automatizados a través del análisis inteligente de productos disponibles en cualquier sistema aéreo como las imágenes satelitales multiespectrales.
Posterior a una identificación, se deben determinar un grupo de características que permitan desarrollar algoritmos eficientes capaces de realizar un seguimiento del comportamiento de la formación convectiva, que permita generar pronósticos de las características de los sistemas convectivos en el corto plazo, para lo que se requiere conocer otras variables como el viento en las áreas de análisis.
Este tipo de aplicaciones integradas a los sistemas de control de tráfico aérea disminuirían los riesgos debidos factores meteorológicos.
... The SVM algorithm makes it possible to get an accurate classification from small training samples (Lizarazo 2008b;Addesso et al. 2012). SVM creates a hyper-plane between each pair of classes, such that it maximizes the distance between the support vectors of each class (Chethan, Raghavendra, and Hemantha 2009;Tso and Mather 2009;Buddhiraju and Rizvi 2010). If SVM cannot build the hyper-plane in the original spectral space, the separation is performed in a higher dimensional spectral space through kernel functions (Addesso et al. 2012). ...
... The most important feature of RF is the output of the variable importance (VI), a measure of the degree of association between a given attribute and the result of the classification, being highlighted the use of Mean Decrease Accuracy and the Mean Decrease Gini (Khalilia, Chakraborty, and Popescu 2011). For the SVM algorithm, the radial basis function kernel (Chethan, Raghavendra, and Hemantha 2009) was used in this study. As this study is an SVM multiclass classification, the model was fitted for every class, one at a time, using the one-against-all strategy (Anthony, Gregg, and Tshilidzi 2007;Lizarazo 2008b). ...
Accurate identification of precipitating clouds is a challenging task. In the present work, Support Vector Machines (SVMs), Decision Trees (DT), and Random Forests (RD) algorithms were applied to extract and track mesoscale convective precipitating clouds from a series of 22 Geostationary Operational Environmental Satellite-13 meteorological image sub-scenes over the continental territory of Colombia. This study’s aims are twofold: (i) to establish whether the use of five meteorological spectral channels, rather than a single infrared (IR) channel, improves rainfall objects detection and (ii) to evaluate the potential of machine learning algorithms to locate precipitation clouds. Results show that while the SVM algorithm provides more accurate classification of rainfall cloud objects than the traditional IR brightness temperature threshold method, such improvement is not statistically significant. Accuracy assessment was performed using STEP (shape (S), theme (T), edge (E), and position (P)) object-based similarity matrix method, taking as reference precipitation satellite images from the Tropical Rainfall Measuring Mission. Best thematic and geometric accuracies were obtained applying the SVM algorithm.
... Hughes et al. [29] developed a neural network approach to detect clouds in Landsat scenes using spectral and spatial information. Chenthan [30] has compared Linear, Polynomial, and Gaussian-RBF kernel in SVMs and concluded the combination of texture and Linear SVM with correct classification rate of 92.30% in cloud detection. Başeski [31] used color and line based elimination as well as texture based SVM classification to detect cloud on RGB images. ...
The accurate location of clouds in images is prerequisite for many high-resolution satellite imagery applications such as atmospheric correction, land cover classifications, and target recognition. Thus, we propose a novel approach for cloud detection using machine learning and multi-feature fusion based on a comparative analysis of typical spectral, textural, and other feature differences between clouds and backgrounds. To validate this method, we tested it on 102 Gao Fen-1(GF-1) and Gao Fen-2(GF-2) satellite images. The overall accuracy of our multi-feature fusion method for cloud detection was more than 91.45%, and the Kappa coefficient for all the tested images was greater than 80%. The producer and user accuracy were also higher at 93.67% and 95.67%, respectively; both of these values were higher than the values for the other tested feature fusion methods. Our results show that this novel multi-feature approach yields better accuracy than other feature fusion methods. In post-processing, we applied an object-oriented method to remove the influence of highly reflective ground objects and further improved the accuracy. Compared to traditional methods, our new method for cloud detection is accurate, exhibits good scalability, and produces consistent results when mapping clouds of different types and sizes over various land surfaces that contain natural vegetation, agriculture land, built-up areas, and water bodies.
... For single images, feature based and color based approaches have shown to be effective. Chethan et al. [7] uses Gabor based image features on down-scaled images and feeds these features to an SVM based classifier. Similarly, Changhui et al. [15] uses image based features and Laban et al. [16]uses texture based features, in order to identify cloud and noncloud regions. ...
With the increasing number of aerial and satellite image sources, automated interpretation algorithms are becoming more and more crucial. Automatically determining the cloud coverage reduces image preprocessing time and aids automatic image exploitation algorithms about where to look. The proposed method makes use of both color and texture characteristics of cloud regions. The image is divided into subimages in order to perform initial color and edge analysis. Further analysis is done by classifying patches as cloud and non-cloud with respect to texture based features.
... ----------Keywords: cloud mass classification, machine learning algorithms, weather images, decision trees, support vector machines, random forests Introducción El análisis de nubes proporciona información que es vital para la detección, comprensión y pronóstico de las tendencias meteorológicas y de los cambios ambientales [1], siendo su identificación y clasificación a partir de imágenes meteorológicas necesaria para extraer información acerca de su aparición y tipos [2,3]. Dichos aspectos son fundamentales en muchas aplicaciones meteorológicas, como el pronóstico del tiempo desde el espacio, el cual es realizado a partir de técnicas basadas en la evaluación y el seguimiento de masas de nubes [4][5][6]. ...
... Existen trabajos previos de utilización de imágenes meteorológicas para clasificar digitalmente los diferentes tipos de nubes, mediante la aplicación de distintos métodos y técnicas de acuerdo al propósito de la investigación [6,9]. Algunos de ellos reportan el uso de algoritmos de aprendizaje de máquina tales como Redes Neuronales Artificiales y Máquinas de Soporte Vectorial [1,2], entre otros. ...
... El principal atractivo de SVM es su capacidad de minimizar los errores de clasificación, creando un hiperplano entre cada par de clases, de tal manera que maximiza la distancia entre los vectores de soporte de cada clase [1,2,16]. Si no es posible construir ese hiperplano en el espacio espectral original, la separación se realiza en un espacio espectral de dimensión más alta [5]. ...
Resumen La identificación exacta de nubes precipitantes es una tarea difícil. En el presente trabajo se aplicaron los algoritmos Máquinas de Soporte Vectorial, Árboles de Decisión y Bosques Aleatorios para discriminar entre nubes precipitantes y nubes no precipitantes, a partir de una imagen meteorológica del satélite GOES-13 que cubre el territorio colombiano. El objetivo del trabajo fue evaluar el desempeño de los algoritmos de aprendizaje de máquina (ML), para la clasificación digital de masas nubosas, en términos de la exactitud temática de la clasificación usando como referencia el algoritmo convencional distancia de Mahalanobis. Los resultados muestran que los algoritmos ML proporcionan una clasificación de masas de nubes más exacta que la obtenida por algoritmos convencionales. La mejor exactitud fue obtenida usando Bosques Aleatorios (RF), con una exactitud temática global de 97%. Adicionalmente, la clasificación obtenida con RF fue comparada pixel a pixel con estimaciones de precipitación de la NASA Tropical Rainfall Measurement Mission (TRMM) obteniendo una exactitud global del 94%. De acuerdo con este estudio, los algoritmos ML pueden ser usados para mejorar los actuales métodos de identificación de nubes precipitantes.
... ----------Keywords: cloud mass classification, machine learning algorithms, weather images, decision trees, support vector machines, random forests Introducción El análisis de nubes proporciona información que es vital para la detección, comprensión y pronóstico de las tendencias meteorológicas y de los cambios ambientales [1], siendo su identificación y clasificación a partir de imágenes meteorológicas necesaria para extraer información acerca de su aparición y tipos [2,3]. Dichos aspectos son fundamentales en muchas aplicaciones meteorológicas, como el pronóstico del tiempo desde el espacio, el cual es realizado a partir de técnicas basadas en la evaluación y el seguimiento de masas de nubes [4][5][6]. ...
... Existen trabajos previos de utilización de imágenes meteorológicas para clasificar digitalmente los diferentes tipos de nubes, mediante la aplicación de distintos métodos y técnicas de acuerdo al propósito de la investigación [6,9]. Algunos de ellos reportan el uso de algoritmos de aprendizaje de máquina tales como Redes Neuronales Artificiales y Máquinas de Soporte Vectorial [1,2], entre otros. ...
... El principal atractivo de SVM es su capacidad de minimizar los errores de clasificación, creando un hiperplano entre cada par de clases, de tal manera que maximiza la distancia entre los vectores de soporte de cada clase [1,2,16]. Si no es posible construir ese hiperplano en el espacio espectral original, la separación se realiza en un espacio espectral de dimensión más alta [5]. ...
Accurate identification of precipitating clouds is a challenging task. In the present work, Support Vector Machines, Decision Trees and Random Forests algorithms were applied to discriminate between precipitating clouds and non-precipitating clouds from a satellite weather image GOES- 13 covering the Colombian territory. The objective of this study was to evaluate the performance of machine learning (ML) algorithms for digital classification of cloud masses in terms of thematic accuracy classification using the conventional Mahalanobis algorithm as benchmark. Results show that ML algorithms provide more accurate classification of cloud masses than conventional algorithms. The best accuracy was obtained using Random Forests (RF), with an overall thematic accuracy of 97%. Furthermore, the classification obtained with the RF algorithm was compared pixel-to-pixel with NASA Tropical Rainfall Measurement Mission (TRMM) rainfall estimates, obtaining an overall accuracy of 94%. ML algorithms can therefore be used to improve current precipitating clouds identification methods.
