information to classify the image frame into fire and non- fire regions. This method requires that the camera be stationary, thus it is not very effective to detect fire in the moving recorded videos. Yamagishi and Yamaguchi present a flame detection algorithm based on the spatial-temporal fluctuation data of the frame contour. A neural network is employed to determine fire from the Fourier transform of the fluctuation data. Their method only uses color information to extract flame pixels. However, we also observe that many non-fire objects have the same color distributions as fire. False extraction exists inevitably. Besides, the computational complexity of the algorithm is too high for practical appli- cation. Phillips et al. exploit both the color and temporal variation to detect fire. They use the Gaussian-smoothed color histogram to generate a color lookup table of fire pixel and then take advantage of temporal variation of pixel values to determine whether it is a fire pixel or not. The method is insensitive to the motion of camera. But it requires a close proximity to the fire. Liu and Ahuja present spectral, spatial and temporal models of fire regions in video sequences. They suggest that the shape of a fire region was represented in terms of spatial frequency content of the region contour using its Fourier coefficients and the temporal changes in these coefficients are used as the temporal signatures of the fire region. Their method can not detect flame in some situations, such as low burning power and relatively steady burning. Toreyin et al. integrate motion, flicker, edge blurring and color features for video flame detection. Temporal and spatial wavelet transform are performed to extract the characteristics of flicker and edge blurring. Although they show good results for several test data, they use many heuris- tic thresholds. Celik et al. propose a real-time fire detector that combines foreground object information with color pixel statistics for fire. The foreground information is extracted using an adaptive background subtraction algorithm, and then verified using a statistical fire color model. Although all the previous approaches achieve some level of success, they do not fully exploit the spatio-temporal information contained in video.
Many studies have found that topographic elements (elevation, slope and aspect) and fuel characteristics (type, moisture and inflammability) are prominent factors on shaping spatial pattern of natural caused fires (Kushla and Ripple 1997, Vasconcelos et al. 2001, Rayan 2002, Yang et al. 2007). Those variables determine the fire regime by controlling fire spread, intensity and extent (Guyette and Dey 2000). For example, in the (Silva et al. 2010) research elevation positively influenced ignition distribution. It is assumed that this effect may be due to some human activities typical for higher elevation such as renovation of pastures for livestock using traditional burning, which are also known as frequent cause of wildfires in the Iberian Peninsula. On the other hand, (Vasilakos et al. 2009) found that elevations have a small contribution to fire ignitions in Lesvos Island in Greece.
Abstract—The localization and recognition of moving objects from single monocular intensity images has been a popular issue in image analysis and computer vision over many years. It is also one of the fundamental crisis in model based vehicle localization and recognition. The recently used scheme is model based on simple object recognition and localization of road vehicles using the position and orientation of vehicle image data. But the drawback of the approach is that the shape of the vehicle and its pose varies in multiple junction coordination, the model based recognition is an inefficient one. To overcome the issues, our first work implemented a surveillance image object recognition and localization using improved local gradient model. The vehicle-object shape recognition and pose recovery in the traffic junction is carried out for varied traffic densities. But the drawback of the approach is that it considers only the vehicle shape and pose variations in the road network and does not discuss about the occurrences of event at the vehicle junction points. Now we have to focus on the process of occurrences of event like accident met at traffic junctions. For this, in this work, spatial time invariant model is introduced to measure the event occurrences of the vehicle traffic location points. The event which has been takes place is recorded as the reference context for standardization of the traffic modality. With the reference context, the detector can easily find out the reason of the event takes place. An experimental evaluation is carried out to estimate the performance of the proposed event detection at vehicle location points usingspatial time invariant model (EDSTIM) in terms of spatial events, multiple time scales, traffic controlling time and compared with an existing model based on simple object recognition and localization and the previous work Surveillance of Vehicle Object Recognition and Localization.
EOFs, representing 40% of the total variance. The relatively low value of retained variance from 9 EOFs indicates that fire anomalies are very scattered in time and space, being less condensable into a few dimensions than for example temper- ature, which classically has larger scale patterns and exhibits higher proportions of variance explained by the first EOFs. The complexity of those patterns is enhanced by our use of 3-monthly data, allowing a high temporal resolution of the observed patterns that would not be observed with annual anomalies. Standardisation also contributes to low values of explained variance, since it tends to give equal importance to all grid cells. However, as mentioned before this choice is justified so that fire sensitive ecosystems are not ignored. Performing the analysis with non-standardised data would result in focusing almost exclusively on regions of very high fire incidence (e.g. tropical savannahs and woodlands, pri- marily those located in Africa), which was considered unde- sirable. Figure 4 illustrates the PCs time series of the nine components retained, and Fig. 5 displays the spatial patterns extracted associated to EOF1 and EOF2.
Abstract: Crabs of the genus Persephona are intensely captured in shrimp fisheries as bycatch in the Cananéia region off the coast of the state of São Paulo, Brazil. The analysis of the spatial and temporal distribution of Persephona punctata and Persephona mediterranea could provide information about variation in the abundance of these species, as well as the environmental variables affecting their distribution and the existence of a possible habitat partitioning. Using a shrimp fishery boat equipped with double-rig nets, crabs were monthly captured from July 2012 to June 2014 in seven sites: four in the coastal area adjacent to the Cananéia region and three in the Mar Pequeno estuarine area. The abundances of both species were compared according to spatial (among sites) and temporal (years and seasons) scale distribution. A total of 396 individuals of P. punctata and 64 of P. mediterranea were captured. The abundance of both species was higher in the second sampling year (July 2013–June 2014) and in coastal areas; only one individual of each species was captured in the estuarine area due to the low salinity at this location (approximately 27.7‰). The temperature was the environmental variable that most affected the distribution of both species, which was more abundant in warmer periods. The temporal variation in abundance was modulated by temperature, while salinity modulated the spatial distribution of P. punctata and P. mediterranea. The spatial-temporal distribution of both species differered in Cananéia, pointing to a similar use of the environment’s resources.
Technological advancement in the field of mobile devices, especially smart phones, availability of ubiquitous internet connectivity and positioning GPS sensors have revolutionize how people access to global information. Global social media research summary 2016 shows, out of 7.395 billion population of the world as of January 2016, 3.419 billion people are using internet and 2.307 billion people are active social media users (Chaffey 2016). Facebook received 1.04 billion daily active users on average for December 2015 (Facebook 2015). Statistics shows that Facebook is the most used social networking sites with 87.69% usage followed by Twitter with 6.43% of usage. Pininterest stands third with 4.55% and Tumblr stands at fourth position with 0.43% of usage (StatsMonkey 2015). In Nepal, social media like Facebook, Twitter, Flickr, and many more has replaced postal system as the swift communication medium. Record shows that 5,700,000 people have access to internet and are using Facebook as of November 15, 2015 (Internet World Stats 2015) and more than one million people are using Twitter in Nepal.
VGI has been implemented in several fields but as a source of information for Forest FireDetection there are only few applications. Nuñéz-Redó et al. 2011 recently presented a prototype of an information system that leverages scientific data and web 2.0 content and applies it to a forest fire monitoring situation. The Joint Research Center of the European Comission is also focusing on using VGI, in particular Location-Based Social Networks (LBSN) Information to gather spatio-temporal data on forest fires. A Case Study of a major forest fire event in the South of France during the month of July 2009 was used to demonstrate the reliability of LBSN Information as a source of valid information for the management, planing, risk and damage assessment of forest fires [Longueville et al. 2009]. Although not a specific tool for the detection of the fires, it could be adapted in order to serve this purpose.
A vision-based method for firedetection was presented. Experimental results showed that the method is able to segment fire regions in 93.1 % of the tested dataset. A novelty in the presented method is the use of an object detection and tracking pipeline in order to reduce false fire alarms caused by fire-coloured moving objects. It was shown that the object detection and tracking pipeline by itself is able to produce a success rate of roughly 92.8 % in the tested dataset. Overall, and without special code optimizations, the proposed method runs at 10 Hz. Background subtraction was implemented in a windowed manner so as to increase robustness to the presence of artefacts. To focus the application of a computationally expensive frequency analysis component and - therefore - reduce the computational cost, an attentive mechanism based on a rough temporalanalysis was proposed. An object detection and tracking algorithm was proposed and its output was integrated with the firedetection algorithm to further reduce the false alarm rate in firedetection. A new colour-based model of fire’s appearance and a new wavelet-based model of fire’s spatio-temporal frequency signature were proposed for improved accuracy. Finally, to avoid hard assumptions regarding the environment’s configuration when determining the camera-world mapping, a GPS-based learning procedure was proposed. In future work, we expect to improve the method in order to attain a full frame-rate on low-end computational units equipping affordable smart cameras. We also intend to include the ability to recognize multiple categories in the object detection and tracking pipeline and introduce the
Abstract. Eleven years of global total electron content (TEC) data derived from the assimilated thermosphere–ionosphere electrodynamics general circulation model are analyzed us- ing empirical orthogonal function (EOF) decomposition and the corresponding principal component analysis (PCA) tech- nique. For the daily averaged TEC field, the first EOF ex- plains more than 89 % and the first four EOFs explain more than 98 % of the total variance of the TEC field, indicating an effective data compression and clear separation of different physical processes. The effectiveness of the PCA technique for TEC is nearly insensitive to the horizontal resolution and the length of the data records. When the PCA is applied to global TEC including local-time variations, the rich spatial and temporal variations of field can be represented by the first three EOFs that explain 88 % of the total variance. The spectral analysis of the time series of the EOF coefficients re- veals how different mechanisms such as solar flux variation, change in the orbital declination, nonlinear mode coupling and geomagnetic activity are separated and expressed in dif- ferent EOFs. This work demonstrates the usefulness of using the PCA technique to assimilate and monitor the global TEC field.
ABSTRACT: Introduction: In 2010, there were 305 (37.8%) municipalities with malaria epidemics in the Brazilian Amazon. The epidemics spread can be explained by the spatial distribution pattern. Objective: To analyze the spatial dependence, autocorrelation, of the malaria epidemics in the municipalities of this region. Methods: An automated algorithm was used for the detection of epidemic municipalities in 2003, 2007 and 2010. Spatial dependence was analyzed by applying the global and local Moran index on the epidemic months proportion variable. The epidemic municipalities clusters were identiied using the TerraView software. Results: The global Moran index values were 0.4 in 2003; 0.6 in 2007; and 0.5 in 2010 (p = 0.01), conirming the spatial dependence among the epidemic municipalities. Box Map and Moran Map identiied inter-municipal, interstate and borders clusters with spatial autocorrelation (p < 0.05). There were 10 epidemic municipalities clusters in 2003; 9 in 2007 and 8 in 2010. Discussion: The epidemic municipalities clusters may be linked to the health facilities diiculties on acting together. The structural limitations of the health services can be overcome by territorial integration to support planning and control activities, strengthening the interventions. Conclusion: The routine analysis of the epidemic municipalities clusters with spatial and temporal persistence may provide a new indicator of planning and integrated control prioritization, contributing to malaria epidemics reducing in inter-municipal, interstate and borders areas.
A more fine-grained analysis, using Twitter data to track and characterize people flows in Greater London was conducted by . The authors modelled individual characteristics, as well as conventional measures of land use morphology and night-time residence. Their purpose consisted in characterizing the geography of information generation across the city to create geo-temporal demographical classification of users in London and using tweets to characterize the links between locations across the city. The researchers created a geodemographic classification based on variables such as residence, total number of tweets, number of countries visited, age, ethnicity, London land use category, tweets outside the United Kingdom, and temporal scales. Their study was conducted on three stages: first they evaluated the spatial evenness of Twitter usage at different times of day, with the purpose of get insights about the daily rhythm of activities in the city. The second stage focused on using onomatology and data linkage to infer age, ethnicity, residence, and socioeconomic characteristics of Twitter users, as well as the dominant economic land use at the tweet location. Finally, they measured the connectivity of different areas within
In this paper we presented a novel approach for hotspot detection exploiting tensor decom- position and eigenvector elements matching techniques. The experimental results reveal the effectiveness of these techniques. Our approach is not a replacement for ST-Scan and its variants rather can be considered as a helpful method to reduce the search space in ST-Scan process, even though one could adapt this method independently for monitor- ing spatiotemporal variance in data. We also showed that how multi-way data analysis improves the quality of detection as it already was expected. One major of our approach drawback, is its inability about automatically connecting of the separated spatial and temporal hotspots with aim of identification of spatiotemporal hotspots. Another issue is that the current version is retrospective. Adapting algorithm for online detection would be another feature research direction.
The Fundamental Base of Geographic Data of the Czech Republic (hereinafter FBGD) is a national 2D geodatabase at a 1:10,000 scale with more than 100 geographic objects. This paper describes the design of the permanent updating mechanism of buildings in FBGD. The proposed procedure belongs to the category of hybrid change detection (HCD) techniques which combine pixel-based and object-based evaluation. The main sources of information for HCD are cadastral information and bi-temporal vertical digital aerial photographs. These photographs have great information potential because they contain multispectral, position and also elevation information. Elevation information represents a digital surface model (DSM) which can be obtained using the image matching technique. Pixel-based evaluation of bi-temporal DSMs enables fast localization of places with potential building changes. These coarse results are subsequently classified through the object-based image analysis (OBIA) using spectral, textural and contextual features and GIS tools. The advantage of the two-stage evaluation is the pre-selection of locations where image segmentation (a computationally demanding part of OBIA) is performed. It is not necessary to apply image segmentation to the entire scene, but only to the surroundings of detected changes, which contributes to significantly faster processing and lower hardware requirements. The created technology is based on open-source software solutions that allow easy portability on multiple computers and parallelization of processing. This leads to significant savings of financial resources which can be expended on the further development of FBGD.
Methods: We performed an ecological time-series study with data from the Brazilian Mortality Information System (Sistema de Informações sobre Mortalidade – SIM) about deaths by suicide occurring between 2000 and 2015. We considered as suicide deaths cases recorded as voluntary self-inflicted injuries. Suicide rates were estimated and age-adjusted in the population above 9 years. We analyzed temporal trends by sex and age groups using the simple linear regression model. For the spatialanalysis, we performed Kernel density estimation with the software TerraView version 4.2.2.
Portugal is the smallest country in southern Europe but has the highest fire incidence in the region. The total burnt area from 1980 to 2005 was 2 714 547 ha, with a mean annual value of 108 582 ha, or 1.18% of the total area of the country. For Spain, France, Italy and Greece, the mean annual percentages of burnt area during the same period were 0.37, 0.05, 0.39 and 0.35% respectively (European Commission 2006). The Landsat-based national fire atlas developed at the Forest Research Centre (Technical University of Lisbon, Lisbon, Portugal), which provides the data for our analysis, revealed large interannual variability in burnt area, ranging from a maximum of 440 000 ha in 2003 to a minimum of 15 462 ha in 1977. The years 1985 and 2005 were also severe fire years, whereas 1988 and 1997 experienced low fire incidence. Pereira et al. (2005) and Trigo et al. (2006) showed climate to be a major driver behind these annual fluctuations in the extent of area burnt. Therefore, it is appropriate for the fire frequency analysis to encompass a time span in the range required to define a climatological normal (i.e. ,30 years). Spatial variability in fire incidence is also high (Pereira and Santos 2003; Pereira et al. 2006). With the exception of the flat, coastal region, the northern half of the country is much more fire prone than the southern half, owing to differences in land use and cover, topography, climate and
Figure 1. Near simultaneous IR3.9 channel (3.9 µm) imagery of fires in southern Africa from (a) SEVIRI and (b) MODIS, along with zooms of a sunglint area in two versions of the SEVIRI data, both (c) non-geometrically corrected level 1.0 and (d) fully pre-processed level 1.5. These image subsets show pixels with elevated MWIR brightness temperatures as bright, and apart from the glint regions almost all of these are caused by actively burning fires. (a) and (b) show the area surrounding the Okavango delta wetland (around 250 km long). The SEVIRI data were collected at 11:57 UTC on 4 September 2003, soon after the end of the MSG-1 (Meteosat 8) commissioning phase and just over one year after the satellite launch, whilst the matching near-nadir Aqua MODIS image was collected on the same day shortly after (12:05 UTC). The polar orbiting MODIS and geostationary SEVIRI data are not exactly co-registered, but cover approximately the same area. Whilst the increased spatial resolution of the MODIS data is clear and allows more fires to be visually identified via their elevated MWIR signals, many of the fires can also clearly be seen in the SEVIRI imagery. SEVIRI provides 96 images per day (one every 15 min) at a consistent view zenith angle. At this latitude up MODIS provides up to four images per day, though some of these will be at extreme view zenith angles up to 65 ◦ under which conditions the MODIS spatial fidelity is far reduced, with each pixel covering approximately the same ground area as does a SEVIRI pixel (Freeborn et al., 2011). The local afternoon imaging time of MODIS Aqua, as used here, is also relatively close to the typical peak of the fire diurnal cycle (Roberts et al., 2009a), but the times of the other MODIS overpasses are significantly distant from this. The SEVIRI images in (c) and (d) indicate the similarity between sunglint and active fire signals in the MWIR channel, which requires consideration when designing active firedetection algorithms. The effect of the spatial resampling undertaken between the SEVIRI Level 1.0 and Level 1.5 data is also apparent.
Object-based change detection is a powerful analysis tool for remote sensing data, but few studies consider the potential of temporal semivariogram indices for mapping land-cover changes using object-based approaches. In this study, we explored and evaluated the performance of semivariogram indices calculated from remote sensing imagery, using the Normalized Differential Vegetation Index (NDVI) to detect changes in spatial features related to land cover caused by a disastrous 2015 dam failure in Brazil’s Mariana district. We calculated the NDVI from Landsat 8 images acquired before and after the disaster, then created objects by multiresolution segmentation analysis based on post-disaster images. Experimental semivariograms were computed within the image objects and semivariogram indices were calculated and selected by principal component analysis. We used the selected indices as input data to a support vector machine algorithm for classifying change and no-change classes. The selected semivariogram indices showed their effectiveness as input data for object-based change detectionanalysis, producing highly accurate maps of areas affected by post-dam-failure flooding in the region. This approach can be used in many other contexts for rapid and accurate assessment of such land-cover changes.
Resistant Subclonal Populations Are Present in Pre-Treatment Disease The survival analyses suggested that the degree of CE could have effects on PFS and OS consis- tent with the hypothesis that cancers with high CE may have increased genetic diversity favour- ing the emergence of drug-resistant clones. We were able to explore this hypothesis in patients 5 and 8, who had additional samples collected at progression. Patient 8 had symptomatic pro- gressive disease, with the development of ascites at 12 mo after completing chemotherapy. The phylogenetic reconstruction of her cancer showed early divergence of the ascites sample from the root (Fig. 4). Examination of the relapsed copy number profile revealed a new focal deletion at NF1 that was not present in the pre-chemotherapy and interval debulking surgery samples (Fig. 4). NF1 is recurrently mutated in HGSOC [22,23], which suggests this was unlikely to be a passenger event. As copy number profiles detect the dominant clone in a sample, we investigat- ed the population structure of earlier samples using WGS to map the new NF1 deletion, and digital PCR to accurately estimate the number of cells containing the NF1 deletion in each sam- ple. To prevent confounding effects from differences in tumour cellularity, the counts for the NF1 deletion were expressed as a proportion of all mutant TP53 counts. The NF1 deletion was detected at 5% and 26% in the pre-treatment samples (absolute cellularity 80% and 41%) and in 25%–100% of the interval debulking samples (median cellularity 49%). Histological analysis of the left fallopian tube specimen removed at interval debulking confirmed a tubal primary site (S1 Fig). We therefore extended the digital PCR analysis to DNA from microdissected tis- sues from formalin-fixed tissue blocks including the left fallopian tube fimbria. The NF1
Before considering N-dimensional hyper planes, let’s look at a simple 2-dimensional example. Assume we wish to perform a classification, and our data has a categorical target variable with two categories. Also assume that there are two predictor variables with continuous values. If we plot the data points using the value of one predictor on the X axis and the other on the Y axis we might end up with an image such as shown below. One category of the target variable is represented by rectangles while the other category is represented by ovals as shown in Fig.2.In this example, the cases with one category are in the lower left corner and the cases with the other category are in the upper right corner; the cases are completely separated. The SVM analysis attempts to find a 1-dimensional hyper plane (i.e. a line) that separates the cases based on their target categories.
Moderate Resolution Imaging Spectoradiometer (MODIS) was launched onboard the National Aeronautics and Space Administration (NASA) Earth Observing System (EOS) Terra and Aqua platforms on 1999 and 2002, respectively. The instrument is a scanning spectroradiometer with 36 visible (VIS), near-infrared (NIR), and infrared (IR) spectral bands between 0.645 and 14.235 mm (King et al. 1992). After the launch, it brings many opportunities to study atmospheric temperature, soil moisture, land use change, vegetation change and other environmental aspects. These wide spectral range, high spatial resolution, and near-daily global coverage of MODIS enable it to observe the earth‟s atmosphere and continuously monitor changes. MODIS retrievals of atmospheric water vapor and temperature distributions are intended to advance understanding of the role-played by energy and water cycle processes in determining the earth‟s weather and climate. These MODIS temperature and moisture data can be used in numerical weather prediction models in regions where conventional meteorological observation are sparse.