Análise de acurácia para o mapeamento de áreas queimadas utilizando uma cena VIIRS 1Km e classificação por Random Forest

A disponibilidade gratuita de dados de sensoriamento remoto em áreas atingidas por incêndios florestais em escala global oferece a oportunidade de geração sistemática de produtos terrestres de média resolução espacial, porém as conhecidas limitações de precisão é objeto de estudo em todo o mundo. Este artigo tem como objetivo analisar a acurácia da detecção de áreas queimadas utilizando o classificador Random Forest (RF) por meio de uma cena do sensor Radiômetro de Imagem Infravermelho Visível (VIIRS) (1Km) em quatro pontos da savana brasileira. Os resultados foram validados através dos produtos de referência espacial de áreas queimadas: Aq30m, Fire_cci e MCD64A1 por meio de uma abordagem estratificada possibilitando a amostragem dos dados no espaço e tempo. Os modelos de RF avaliados com seus parâmetros de entrada, em que, incluiu-se 400 árvores e um atributo, fornecendo uma taxa de erro abaixo de 4%. Os resultados mostraram que o mapeamento validado com o produto Aq30m apresentou importantes estimativas de Coeficiente de Sorensen-Dice enquanto a validação realizada entre os modelos globais, o MCD64A1 mostrou-se com maior exatidão (>50%) principalmente em feições de áreas queimadas de grandes proporções (> 200Km²). Em particular, a análise sugere que a validação de produtos de área queimada sempre deve estar ligada ao tempo mínimo da data dos dados de validação e o tamanho da área atingida pelo fogo. Os resultados mostram que esta abordagem é muito útil para ser usado para determinar áreas de floresta queimada. Accuracy analysis for mapping burnt areas using a 1Km VIIRS scene and Random Forest classification A B S T R A C TThe availability of remote sensing data with medium spatial resolution has offered several mapping possibilities for areas affected by forest fires on the Earth's surface. In this context, the analysis of sensor spatial accuracy limitations has been the subject of global research. The objective of this study was to analyze the mapping accuracy of the VIIRS sensor on board the NOAA satellite, using the Random Forest (RF) classifier for the detection of burned areas, in four points of the Chapada dos Veadeiros National Park - Goiás, inserted in the Brazilian savanna. The methodology consisted in validating the classification using the Sorensen-Dice coefficient (SD) in a stratified approach, using as reference the products: Aq30m, Fire_cci and MCD64A1. As a result, the RF models, included 400 trees and one attribute, with an error of less than 4%. Among the global models, the MCD64A1 presented a significant accuracy, greater than 50%, especially in features of burned areas greater than 200Km². Thus, the data suggest that the quality of accuracy of the validation process of mapping products for burned areas is associated with the minimum time interval of availability of validation data and the size of the area affected by fire. Based on this, the results show effectiveness in using the RF algorithm on medium spatial resolution images for fire detection in seasonally dry forests, such as the Cerrado.Keywords: Cerrado, fires, Random Forest.

Indonesian Online News Topics Classification using Word2Vec and K-Nearest Neighbor

News is information disseminated by newspapers, radio, television, the internet, and other media. According to the survey results, there are many news titles from various topics spread on the internet. This of course makes newsreaders have difficulty when they want to find the desired news topic to read. These problems can be solved by grouping or so-called classification. The classification process is carried out of course by using a computerized process. This study aims to classify several news topics in Indonesian language using the KNN classification model and word2vec to convert words into vectors which aim to facilitate the classification process. The use of KNN in this study also determines the optimal K value to be used. In addition to using the classification model, this study also uses a word embedding-based model, namely word2vec. The results obtained using the word2vec and KNN models have an accuracy of 89.2% with a value of K=7. The word2vec and KNN models are also superior to the support vector machine, logistic regression, and random forest classification models.

Matlab Based Potent Algorithm for WBc cancer Detection and classification

This paper aims to automate the detection of cancer using digital image processing techniques in MATLAB software. The analysis of white blood cells (WBC) is a powerful diagnostic tool for the prediction of Leukemia. The automatic detection of leukemia is a challenging task, which remains an unresolved problem in the medical imaging field. This Automation in Biological laboratories can be done by extracting the features of the blood film images taken from the digital microscopes and processed using MATLAB software. The aim of this approach is to discover the WBC cancer cells in an earlier stage and to reduce the discrepancies in diagnosis, by improving the system learning methodology. This paper presents the potent algorithm, which will eliminate the dubiety, in diagnosing the cancers with similar symptoms. This Algorithm concentrates on major WBC cancers, such as Acute Lymphocytic Leukemia, Acute Myeloid Leukemia, Chronic Lymphocytic Leukemia and Chronic Myeloid Leukemia. As they are life threatening diseases, rapid and precise differentiation is necessary in clinical settings. These cancers are categorized by segmentation and feature extraction, which will be further, classified using Random forest classification (RFC). RFC will classify the cancer using a decision tree learning method, which uses predictors at each node to make better decision.

Multi-Index Ecoacoustics Analysis for Terrestrial Soundscapes: A New Semi-Automated Approach Using Time-Series Motif Discovery and Random Forest Classification

High rates of biodiversity loss caused by human-induced changes in the environment require new methods for large scale fauna monitoring and data analysis. While ecoacoustic monitoring is increasingly being used and shows promise, analysis and interpretation of the big data produced remains a challenge. Computer-generated acoustic indices potentially provide a biologically meaningful summary of sound, however, temporal autocorrelation, difficulties in statistical analysis of multi-index data and lack of consistency or transferability in different terrestrial environments have hindered the application of those indices in different contexts. To address these issues we investigate the use of time-series motif discovery and random forest classification of multi-indices through two case studies. We use a semi-automated workflow combining time-series motif discovery and random forest classification of multi-index (acoustic complexity, temporal entropy, and events per second) data to categorize sounds in unfiltered recordings according to the main source of sound present (birds, insects, geophony). Our approach showed more than 70% accuracy in label assignment in both datasets. The categories assigned were broad, but we believe this is a great improvement on traditional single index analysis of environmental recordings as we can now give ecological meaning to recordings in a semi-automated way that does not require expert knowledge and manual validation is only necessary for a small subset of the data. Furthermore, temporal autocorrelation, which is largely ignored by researchers, has been effectively eliminated through the time-series motif discovery technique applied here for the first time to ecoacoustic data. We expect that our approach will greatly assist researchers in the future as it will allow large datasets to be rapidly processed and labeled, enabling the screening of recordings for undesired sounds, such as wind, or target biophony (insects and birds) for biodiversity monitoring or bioacoustics research.

Random Forest Classification of Inundation Following Hurricane Florence (2018) via L-Band Synthetic Aperture Radar and Ancillary Datasets

In response to Hurricane Florence of 2018, NASA JPL collected quad-pol L-band SAR data with the Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) instrument, observing record-setting river stages across North and South Carolina. Fully-polarized SAR images allow for mapping of inundation extent at a high spatial resolution with a unique advantage over optical imaging, stemming from the sensor’s ability to penetrate cloud cover and dense vegetation. This study used random forest classification to generate maps of inundation from L-band UAVSAR imagery processed using the Freeman–Durden decomposition method. An average overall classification accuracy of 87% is achieved with this methodology, with areas of both under- and overprediction for the focus classes of open water and inundated forest. Fuzzy logic operations using hydrologic variables are used to reduce the number of small noise-like features and false detections in areas unlikely to retain water. Following postclassification refinement, estimated flood extents were combined to an event maximum for societal impact assessments. Results from the Hurricane Florence case study are discussed in addition to the limitations of available validation data for accuracy assessments.

Characterization of soils conducive and non-conducive to Prunus replant disease

Successive orchard plantings of almond and other Prunus species exhibit reduced growth and yield in many California soils. This phenomenon, known as Prunus replant disease (PRD), can be prevented by preplant soil fumigation or anaerobic soil disinfestation, but its etiology is poorly understood and its incidence and severity are hard to predict. We report here on relationships among physicochemical variables, microbial community structure, and PRD induction in 25 diverse replant soils from California. In a greenhouse bioassay, soil was considered to be “PRD-inducing” when growth of peach seedlings in it was significantly increased by preplant fumigation and pasteurization, compared to an untreated control. PRD was induced in 18 of the 25 soils, and PRD severity correlated positively with soil exchangeable-K, pH, %clay, total %N, and electrical conductivity. The structure of bacterial, fungal, and oomycete communities differed significantly between the PRD-inducing and non-inducing soils, based on PERMANOVA of Bray Curtis dissimilarities. Bacterial class MB-A2-108 of phylum Actinobacteria had high relative abundances among PRD-inducing soils, while Bacteroidia were relatively abundant among non-inducing soils. Among fungi, many ASVs classified only to kingdom level were relatively abundant among PRD-inducing soils whereas ASVs of Trichoderma were relatively abundant among non-inducing soils. Random forest classification effectively discriminated between PRD-inducing and non-inducing soils, revealing many bacterial ASVs with high explanatory values. Random forest regression effectively accounted for PRD severity, with soil exchangeable-K and pH having high predictive value. Our work revealed several biotic and abiotic variables worthy of further examination in PRD etiology.

Estimating Actual Abundance of European Sousliks: Using UAV Imagery, Pixel Based Image Analysis and Random Forest Classification to Count Souslik Burrows

Burrowing mammals are widespread and contribute significantly to soil ecosystem services. However, how to conduct a non-invasive estimation of their actual population size has remained a challenge. Results support that the number of burrow entrances is positively correlated with population abundance and burrows’ location indicates their area of occupancy consequently it provides a benchmark for estimating population size. European souslik is an endangered burrowing species in decline across its range. We present an imagery-based method to identify and count animals’ burrows semi-automatically by combining remotely recorded RGB images, pixel-based imagery (PBI) and Random Forest (RF) classification. Field images recorded in four colonies were collected, combined and then processed by histogram matching and spectral band normalisation to improve the spectral distinction between the categories BURROW, SOIL, TREE, GRASS. Raw or processed images were analysed by RF classification to compare the change in accuracy metrics as a result of processing. From accuracy metrics kappa of precision (κBURROWP) and sensitivity (κBURROWS) for BURROW were 95 and 90% respectively. A 10-time bootstrapping of the final model resulted in coefficients of variation (CV%) of κBURROWS and κBURROWP lower than 5%, moreover CV% values were not significantly different between precision and sensitivity scores. The consistency of classification results and balanced precision and sensitivity confirmed the applicability of this approach. Our method provides an accurate and user-friendly tool to count the number of burrow openings and delineate the areas of occupancy as compared to traditional, more invasive approaches or other computer capacity and end-user expertise demanding methods.