A new tool for supervised classification of satellite images available on web servers: Google Maps as a case study

Author(s):  
Agustín García-Flores ◽  
Abel Paz-Gallardo ◽  
Antonio Plaza ◽  
Jun Li
Author(s):  
Priscila Siqueira Aranha ◽  
Flavia Pessoa Monteiro ◽  
Paulo Andre Ignacio Pontes ◽  
Jorge Antonio Moraes de Souza ◽  
Nandamudi Lankalapalli Vijaykumar ◽  
...  

2021 ◽  
Vol 34 (1) ◽  
Author(s):  
Zhe Yang ◽  
Dejan Gjorgjevikj ◽  
Jianyu Long ◽  
Yanyang Zi ◽  
Shaohui Zhang ◽  
...  

AbstractSupervised fault diagnosis typically assumes that all the types of machinery failures are known. However, in practice unknown types of defect, i.e., novelties, may occur, whose detection is a challenging task. In this paper, a novel fault diagnostic method is developed for both diagnostics and detection of novelties. To this end, a sparse autoencoder-based multi-head Deep Neural Network (DNN) is presented to jointly learn a shared encoding representation for both unsupervised reconstruction and supervised classification of the monitoring data. The detection of novelties is based on the reconstruction error. Moreover, the computational burden is reduced by directly training the multi-head DNN with rectified linear unit activation function, instead of performing the pre-training and fine-tuning phases required for classical DNNs. The addressed method is applied to a benchmark bearing case study and to experimental data acquired from a delta 3D printer. The results show that its performance is satisfactory both in detection of novelties and fault diagnosis, outperforming other state-of-the-art methods. This research proposes a novel fault diagnostics method which can not only diagnose the known type of defect, but also detect unknown types of defects.


2012 ◽  
Author(s):  
Ángel Ferrán ◽  
Sergio Bernabé ◽  
Pablo García-Rodríguez ◽  
Antonio Plaza

2020 ◽  
Author(s):  
Zhe Yang ◽  
Dejan Gjorgjevikj ◽  
Jian-Yu Long ◽  
Yan-Yang Zi ◽  
Shao-Hui Zhang ◽  
...  

Abstract Novelty detection is a challenging task for the machinery fault diagnosis. A novel fault diagnostic method is developed for dealing with not only diagnosing the known type of defect, but also detecting novelties, i.e. the occurrence of new types of defects which have never been recorded. To this end, a sparse autoencoder-based multi-head Deep Neural Network (DNN) is presented to jointly learn a shared encoding representation for both unsupervised reconstruction and supervised classification of the monitoring data. The detection of novelties is based on the reconstruction error. Moreover, the computational burden is reduced by directly training the multi-head DNN with rectified linear unit activation function, instead of performing the pre-training and fine-tuning phases required for classical DNNs. The addressed method is applied to a benchmark bearing case study and to experimental data acquired from a delta 3D printer. The results show that it is able to accurately diagnose known types of defects, as well as to detect unknown defects, outperforming other state-of-the-art methods.


Author(s):  
Jannai Tokotoko ◽  
Frederic Flouvat ◽  
Claire Goiran ◽  
Laetitia Hedouin ◽  
Antoine Collin ◽  
...  

Geografie ◽  
1997 ◽  
Vol 102 (1) ◽  
pp. 17-30
Author(s):  
Jaromír Kolejka ◽  
Jásim K. Shallal

Surface soil data have been processed using the unsupervised classification (cluster analysis). Three soil categories with different erosional characteristics have been detected: heavily, moderately and slightly/no damaged soils. The supervised satellite image classification (MLC) was based on the data taken from case study areas in the proximity of classified soil sample sites on the vegetation free-fields.


Author(s):  
Jones Remo Barbosa Vale ◽  
Jamer Andrade da Costa ◽  
Jefferson Ferreira dos Santos ◽  
Elton Luis Silva da Silva ◽  
Artur Trindade Favacho

COMPARATIVE ANALYSIS THE METHODS OF SUPERVISED CLASSIFICATION APPLIED TO THE MAPPING OF SOIL COVER IN THE MUNICIPALITY OF MEDICILÂNDIA, PARÁANÁLISIS COMPARATIVO DE MÉTODOS DE CLASIFICACIÓN SUPERVISADA APLICADA AL MAPEO DE LA COBERTURA DEL SUELO EN EL MUNICIPIO DE MEDICILÂNDIA, PARÁAs imagens de satélite são produtos gerados por sensoriamento remoto e, estão associadas aos fenômenos e eventos que ocorrem na superfície a partir do registro e da análise das interações entre a radiação eletromagnética e os alvos. O objetivo do trabalho é comparar métodos de classificação supervisionada de imagens de satélite para o mapeamento da cobertura do solo. A área de estudo compreende o município de Medicilândia, localizado no sudoeste paraense. Para a realização do trabalho foram utilizados imagens do satélite Landsat 8, sensor OLI-TIRS, cenas 226/062 e 227/062. Foram realizados os testes de classificação, utilizando três classificadores: Distância Mínima, Distância Mahalanobis e Máxima Verossimilhança. Na etapa de classificação foram identificadas as seguintes classes: água, nuvem, sombra de nuvem, solo exposto, vegetação primária e vegetação secundária. Para fins de avaliação da fidedignidade da classificação de cada método foram calculados, o Índice Kappa e a Exatidão Global. A classificação pelo método Máxima Verossimilhança obteve maior exatidão apresentando Índice Kappa de 0,920 e Exatidão Global 96% quando comparada à classificação pelos métodos Distância Mínima e Distância Mahalanobis, que apresentaram Índice Kappa de 0,842 e 0,845 e Exatidão Global 92% respectivamente. As técnicas de classificação supervisionada são ferramentas essenciais no processo de mapeamento da cobertura do solo de grandes áreas, visto que dispondo-se de recursos limitados, imagens de baixo custo e de sistemas livres para processamento e integração das informações, é possível obter parâmetros com altos níveis de precisão, sendo fundamentais para subsidiar o planejamento territorial e ambiental.Palavras-chave: Sensoriamento Remoto; Classificação de Imagens de Satélite; Cobertura do Solo.ABSTRACTThe satellite images are products generated by remote sensing and are associated with phenomena and events that occur on the surface from the recording and analysis of interactions between electromagnetic radiation and targets. The objective of this work is to compare methods of supervised classification of satellite images for the mapping of the soil cover. The study area comprises the municipality of Medicilândia, located in southwest of Para. In order to perform the work, were used images from the Landsat 8 satellite, OLI-TIRS sensor, scenes 226/062 and 227/062. The classification tests were performed using three classifiers: Minimum Distance, Mahalanobis Distance and Maximum Likelihood. In the classification processe were identified the following classes: water, cloud, cloud shadow, exposed soil, primary vegetation and secondary vegetation. For the purposes of evaluating the reliability of the classification of each method were calculated, Kappa Index and Global Accuracy. The classification by the Maximum Likelihood method obtained a greater accuracy presenting Kappa Index of 0,920 and Global Accuracy 96% when compared to the classification by the Minimum Distance and Mahalanobis Distance, which presented Kappa Index of 0,842 and 0,845 and Global Accuracy 92% respectively. The supervised classification techniques are essential tools in the mapping process of large-area soil cover, since with limited resources, low-cost images and free systems for processing and integrating information, it is possible to obtain parameters with high levels of precision, being fundamental to subsidize territorial and environmental planning.Keywords: Remote Sensing; Classification of Satellite Images; Soil Cover.RESUMENLas imágenes de satélite son productos generados por la detección remota y están asociados a los fenómenos y eventos que ocurren en la superficie a partir del registro y del análisis de las interacciones entre la radiación electromagnética y los blancos. El objetivo del trabajo es comparar métodos de clasificación supervisada de imágenes de satélite para el mapeo de la cobertura del suelo. El área de estudio comprende el municipio de Medicilândia, ubicado en el suroeste paraense. Para la realización del trabajo se utilizaron imágenes del satélite Landsat 8, sensor OLI-TIRS, escenas 226/062 y 227/062. Se utilizaron tres clasificadores: Distancia Mínima, Distancia Mahalanobis y Máxima Verosimilitud. En la etapa de clasificación se identificaron las siguientes clases: agua, nube, sombra de nube, suelo expuesto, vegetación primaria y vegetación secundaria. Para evaluar la confianza de la clasificación de cada método se ha calculado, el Índice Kappa y la Exactitud Global. La clasificación por Máxima Verosimilitud obtuvo mayor exactitud presentando Índice Kappa de 0,920 y Exactitud Global 96% cuando comparada a la clasificación por Distancia Mínima y Distancia Mahalanobis, que presentaron Índice Kappa de 0,842 y 0,845 y Exactitud Global 92% respectivamente. Las técnicas de clasificación supervisada son herramientas esenciales en el proceso de mapeo de la cobertura del suelo de grandes áreas, ya que disponiendo de recursos limitados, imágenes de bajo costo y de sistemas libres para procesamiento e integración de la información, es posible obtener parámetros con altos niveles de precisión, siendo fundamentales para subsidiar la planificación territorial y ambiental.Palabras clave: Sensoramiento Remoto; Clasificación de Imágenes de Satélite; Cobertura del Suelo.


Sign in / Sign up

Export Citation Format

Share Document