Supervised Classification of Satellite Images to Analyze Multi-Temporal Land Use and Coverage : A Case Study for the Town of MARABA, State of PARA, Brazil

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.

Download Full-text

Pemetaan Klasifikasi Dan Analisa Perubahan Ekosistem Mangrove Menggunakan Citra Satelit Multi Temporal Di Karimunjawa, Jepara, Indonesia

Jurnal Kelautan Tropis ◽

10.14710/jkt.v21i2.2977 ◽

2018 ◽

Vol 21 (2) ◽

pp. 97

Author(s):

Nurul Latifah ◽

Sigit Febrianto ◽

Hadi Endrawati ◽

Muhammad Zainuri

Keyword(s):

Supervised Classification ◽

Satellite Images ◽

Mangrove Ecosystem ◽

Nursery Ground ◽

Fish Ponds ◽

Natural Phenomena ◽

Key Factors ◽

Marine Life ◽

Fisheries Resources ◽

Multi Temporal

Mapping of Classification and Analysis of Changes in Mangrove Ecosystem Using Multi-Temporal Satellite Images in Karimunjawa, Jepara, Indonesia Mangrove ecosystem is one of the three ecosystem in the coastal area which has important ecological role in supporting marine life and fisheries resources. These important roles include spawning ground and nursery ground for various marine organisms. However, in the last decades, mangrove ecosystem has been undergoing significant degradation. The aim of this research is to quantify the changes of mangrove coverage and density in Karimunjawa as well as key-factors leading to the changes. Supervised classification method (83% accuracy and Kappa coefficient 0.73%) was applied to satellite images to identify the temporal changes in mangrove coverage. Mangrove density was quantified using NDVI algorithm and NIR-RED wavelength. The result shows that during three periods of observed data, changes in mangrove coverage were significant: 126.81 ha increase (1992 – 2003); 82.37 ha decrease (1992 – 2017); and 209.18 ha decrease (2003 – 2017). Mangrove density in most part of Karimunjawa belongs to the category of ‘low’ (NDVI value: -1 – 0.33). Key factors contributing to the decrease mangrove coverage are deforestation, natural phenomena, land conversion into fish ponds and hotels. The only increase in the year 1992 – 2003 was caused by high sedimentation level that allows more mangroves to grow. Overall, the methods in this research could be used as an approach to describe to effectively monitor the changes of mangrove coverage in an area as basic data for sustainable environmental management. Ekosistem mangrove merupakan salah satu dari tiga ekosistem pesisir yang memiliki peranan ekologis penting dalam mendukung kehidupan dan keberlangsungan dari sumberdaya perikanan. Hal tersebut dikarenakan fungsi mangrove sebagai tempat memijah dan asuhan bagi banyak biota air. Beberapa dekade terakhir keberadaan dari ekosisitem mangrove mengalami degradasi yang sangat signifikan. Tujuan dari penelitian ini adalah untuk mengetahui perubahan luasan dan kerapatan mangrove dan mengidentifikasi faktor penyebabnya. Metode analisa perubahan luasan mangrove menggunakan citra satelit multi temporal dengan dilakukan pembuatan klasifikasi menggunakan metode supervised classification dengan nilai akurasi total 83% dengan Kappa koefisien 0,73%. Setelah terseleksi antara mangrove dan non mangrove maka dilakukan perhitungan kerapatan tajuk menggunakan algoritma NDVI dengan memanfaatkan panjang gelombang NIR dan RED. Hasil analisa spasial dengan rentang 3 tahun berbeda didapatkan perubahan penurunan dan penambahan luasan mangrove terjadi secara signifikan: tahun 1992 – 2003 telah terjadi penambahan luasan sebesar 126,81 ha; tahun 1992–2017 terjadi perubahan luasan sebesar 82,37 ha; tahun 2003–2017 terjadi perubahan luasan sebesar 209,18 ha. Kerapatan mangrove di Karimunjawa sebagian besar tergolong kategori kerapatan jarang dengan nilai NDVI antara -1 – 0,33. Faktor utama penyebab penurunan luasan mangrove antara lain penebangan liar, faktor alam, perubahan fungsi lahan menjadi pertambakan dan perhotelan. Penambahan luasan mangrove terjadi pada antara tahun1992 sampai 2003 hal tersebut disebabkan sedimentasi yang menumpuk di pantai dan sudah ditumbuhi oleh mangrove. Secara keseluruhan metode ini dapat menggambarkan perubahan secara efektif serta hasilnya dapat dipergunakan untuk pemantauan dan perencanaan ekosistem mangrove di suatu wilayah.

Download Full-text

Sparse Autoencoder-Based Multi-Head Deep Neural Networks for Machinery Fault Diagnostics With Novelty Detection

10.21203/rs.3.rs-122416/v1 ◽

2020 ◽

Author(s):

Zhe Yang ◽

Dejan Gjorgjevikj ◽

Jian-Yu Long ◽

Yan-Yang Zi ◽

Shao-Hui Zhang ◽

...

Keyword(s):

Supervised Classification ◽

Deep Neural Networks ◽

State Of The Art ◽

Novelty Detection ◽

Activation Function ◽

Reconstruction Error ◽

Fine Tuning ◽

Sparse Autoencoder

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.

Download Full-text

Advanced radar-interpretation of InSAR time series for mapping and characterization of geological processes

Natural Hazards and Earth System Science ◽

10.5194/nhess-11-865-2011 ◽

2011 ◽

Vol 11 (3) ◽

pp. 865-881 ◽

Cited By ~ 69

Author(s):

F. Cigna ◽

C. Del Ventisette ◽

V. Liguori ◽

N. Casagli

Keyword(s):

Time Series ◽

Field Surveys ◽

Synthetic Aperture Radar Interferometry ◽

Geological Processes ◽

Multi Temporal ◽

Under Sampling ◽

Temporal Measures

Abstract. We present a new post-processing methodology for the analysis of InSAR (Synthetic Aperture Radar Interferometry) multi-temporal measures, based on the temporal under-sampling of displacement time series, the identification of potential changes occurring during the monitoring period and, eventually, the classification of different deformation behaviours. The potentials of this approach for the analysis of geological processes were tested on the case study of Naro (Italy), specifically selected due to its geological setting and related ground instability of unknown causes that occurred in February 2005. The time series analysis of past (ERS1/2 descending data; 1992–2000) and current (RADARSAT-1 ascending data; 2003–2007) ground movements highlighted significant displacement rates (up to 6 mm yr−1) in 2003–2007, followed by a post-event stabilization. The deformational behaviours of instable areas involved in the 2005 event were also detected, clarifying typology and kinematics of ground instability. The urban sectors affected and unaffected by the event were finally mapped, consequently re-defining and enlarging the influenced area previously detected by field observations. Through the integration of InSAR data and conventional field surveys (i.e. geological, geomorphologic and geostructural campaigns), the causes of instability were finally attributed to tectonics.

Download Full-text