Porting Rulex Software to the Raspberry Pi for Machine Learning Applications on the Edge

Edge Computing enables to perform measurement and cognitive decisions outside a central server by performing data storage, manipulation, and processing on the Internet of Things (IoT) node. Also, Artificial Intelligence (AI) and Machine Learning applications have become a rudimentary procedure in virtually every industrial or preliminary system. Consequently, the Raspberry Pi is adopted, which is a low-cost computing platform that is profitably applied in the field of IoT. As for the software part, among the plethora of Machine Learning (ML) paradigms reported in the literature, we identified Rulex, as a good ML platform, suitable to be implemented on the Raspberry Pi. In this paper, we present the porting of the Rulex ML platform on the board to perform ML forecasts in an IoT setup. Specifically, we explain the porting Rulex’s libraries on Windows 32 Bits, Ubuntu 64 Bits, and Raspbian 32 Bits. Therefore, with the aim of carrying out an in-depth verification of the application possibilities, we propose to perform forecasts on five unrelated datasets from five different applications, having varying sizes in terms of the number of records, skewness, and dimensionality. These include a small Urban Classification dataset, three larger datasets concerning Human Activity detection, a Biomedical dataset related to mental state, and a Vehicle Activity Recognition dataset. The overall accuracies for the forecasts performed are: 84.13%, 99.29% (for SVM), 95.47% (for SVM), and 95.27% (For KNN) respectively. Finally, an image-based gender classification dataset is employed to perform image classification on the Edge. Moreover, a novel image pre-processing Algorithm was developed that converts images into Time-series by relying on statistical contour-based detection techniques. Even though the dataset contains inconsistent and random images, in terms of subjects and settings, Rulex achieves an overall accuracy of 96.47% while competing with the literature which is dominated by forward-facing and mugshot images. Additionally, power consumption for the Raspberry Pi in a Client/Server setup was compared with an HP laptop, where the board takes more time, but consumes less energy for the same ML task.

Identifying urban areas: A new approach and comparison of national urban metrics with gridded population data

The measurement of urbanization and other key urban indicators depends on how urban areas are defined. The Degree of Urbanization (DEGURBA) has been recently adopted to support international statistical comparability, but its rigid criteria for classify areas as urban/non-urban based upon fixed population size and density criteria is controversial. Here we present an alternative approach to urban classification, using a flexible range of population density \& count thresholds. We then compare how these thresholds affect estimation of urbanization and urban settlement counts across three of the most popular gridded population datasets (GPD). Instead of introducing further uncertainties by matching GPD to built-up area datasets, we classify urban areas in a purely spatial demographic way. By calculating national urban shares and urban area counts, we highlight the often overlooked uncertainties when using GPD. We find that the choice of GPD is generally the dominant factor in altering both of these urban indicators but the choice of urban criteria is also important. Overall, this alternative urban classification method offers a more flexible approach to human settlements classification that can be applied globally for comparative research.

Considering Urban Social Functions at Fine Spatial Resolution to Understand the Distribution of Invasive Plant Species in Cities

ContextCities are high sources of plant invasions. To understand mechanisms of introduction and dispersion of invasive alien species (IAS) in city, we need a thoroughly description of the social and structural components of urban landscapes. ObjectivesWe assessed the effects of neighborhood types and their associated human activities and structural linear elements on the distributions of IAS in a French medium city (Blois). We examined how the relative contributions of these variables varied between scales of analysis. MethodsWe recorded the presence of seven IAS species in the entire city (22 km²), at three spatial resolutions: 100×100m, 200×200m and 400×400m. We characterized neighborhoods through their main covers, human uses, and ages and structural elements through impervious soil, area of and distance to roads and railways.ResultsNeighborhood type was the most important variable in explaining IAS distributions. This variable was especially important at the finest scale which allowed a fine urban classification. B. davidii and B. aquifolium were found in individual residential neighborhoods, whereas R. pseudoacacia and A. altissima were most encountered in industrial areas. The effects of the structural elements differed between species and were lower. ConclusionsCharacterizing the high spatial and functional heterogeneity of urban landscapes at fine scale is critical to understand IAS distribution patterns. We show that considering human uses and planting practices is determinant to understand IAS introduction patterns. Then, linear transport corridors and ruderal conditions explain the dispersion and establishment of IAS across the city and potentially to the surrounding natural spaces.

The good, the bad, and the environment: developing an area-based measure of access to health-promoting and health-constraining environments in New Zealand

Background Accounting for the co-occurrence of multiple environmental influences is a more accurate reflection of population exposure than considering isolated influences, aiding in understanding the complex interactions between environments, behaviour and health. This study examines how environmental ‘goods’ such as green spaces and environmental ‘bads’ such as alcohol outlets co-occur to develop a nationwide area-level healthy location index (HLI) for New Zealand. Methods Nationwide data were collected, processed, and geocoded on a comprehensive range of environmental exposures. Health-constraining ‘bads’ were represented by: (i) fast-food outlets, (ii) takeaway outlets, (iii) dairy outlets and convenience stores, (iv) alcohol outlets, (v) and gaming venues. Health-promoting ‘goods’ were represented by: (i) green spaces, (ii) blue spaces, (iii) physical activity facilities, (iv) fruit and vegetable outlets, and (v) supermarkets. The HLI was developed based on ranked access to environmental domains. The HLI was then used to investigate socio-spatial patterning by area-level deprivation and rural/urban classification. Results Results showed environmental ‘goods’ and ‘bads’ co-occurred together and were patterned by area-level deprivation. The novel HLI shows that the most deprived areas of New Zealand often have the most environmental ‘bads’ and less access to environmental ‘goods’. Conclusions The index, that is now publicly available, is able to capture both inter-regional and local variations in accessibility to health-promoting and health-constraining environments and their combination. Results in this study further reinforce the need to embrace the multidimensional nature of neighbourhood and place not only when designing health-promoting places, but also when studying the effect of existing built environments on population health.

Urban Classification from Aerial and Satellite Images

When talking about land cover, we need to find a proper way to extract information from aerial or satellite images. In the field of photogrammetry, aerial images are generally acquired by optical sensors that deliver images in four bands (red, green, blue and near-infrared). Recent researches in this field demonstrated that for the image classification process is still place for improvement. From satellites are obtained multispectral images with more bands (e.g. Landsat 7/8 has 36 spectral bands). This paper will present the differences between these two types of images and the classification results using support-vector machine and maximum likelihood classifier. For the aerial and the satellite images we used different sets of classification classes and the two methods mentioned above to highlight the importance of choosing the classes and the classification method.

County Reclassifications and Rural–Urban Mortality Disparities in the United States (1970–2018)

Objectives. To demonstrate how inferences about rural–urban disparities in age-adjusted mortality are affected by the reclassification of rural and urban counties in the United States from 1970 to 2018. Methods. We compared estimates of rural–urban mortality disparities over time, produced through a time-varying classification of rural and urban counties, with counterfactual estimates of rural–urban disparities, assuming no changes in rural–urban classification since 1970. We evaluated mortality rates by decade of reclassification to assess selectivity in reclassification. Results. We found that reclassification amplified rural–urban mortality disparities and accounted for more than 25% of the rural disadvantage observed from 1970 to 2018. Mortality rates were lower in counties that reclassified from rural to urban than in counties that remained rural. Conclusions. Estimates of changing rural–urban mortality differentials are significantly influenced by rural–urban reclassification. On average, counties that have remained classified as rural over time have elevated mortality. Longitudinal research on rural–urban health disparities must consider the methodological and substantive implications of reclassification. Public Health Implications. Attention to rural–urban reclassification is necessary when evaluating or justifying policy interventions focusing on geographic health disparities.

Characterising populations living close to intensive farming and composting facilities in England

Bioaerosol exposure has been linked to adverse respiratory conditions. Intensive farming and composting facilities are important anthropogenic sources of bioaerosols. We aimed to characterise populations living close to intensive farming and composting facilities. We also infer whether the public are becoming more concerned about anthropogenic bioaerosol emissions, using reports of air pollution related incidents attributed to facilities. We mapped the location of 1,257 intensive farming and 310 composting facilities in England in relation to the resident population and its characteristics (sex and age), area characteristics (deprivation proxy and rural/urban classification) and school locations stratified by pre-defined distance bands from these bioaerosol sources. We also calculated the average number of air pollution related incidents per year per facility. We found that more than 16% of the population and 15% of schools are located within 4,828 m of an intensive farming facility or 4,000 m of a composting facility; few people (0.01 %) live very close to these sites and tend to be older people. Close to composting facilities, populations are more likely to be urban and more deprived. The number of incidents were attributed to a small proportion of facilities; population characteristics around these facilities were similar. Results indicate that populations living near composting facilities (particularly>250 to ⩽ 4,000 m) are mostly located in urban areas (80%–88% of the population), which supports the need for more community health studies to be conducted. Results could also be used to inform risk management strategies at facilities with higher numbers of incidents.