Real-time 2D-3D door detection and state classification on a low-power device

In this paper, we propose three methods for door state classification with the goal to improve robot navigation in indoor spaces. These methods were also developed to be used in other areas and applications since they are not limited to door detection as other related works are. Our methods work offline, in low-powered computers as the Jetson Nano, in real-time with the ability to differentiate between open, closed and semi-open doors. We use the 3D object classification, PointNet, real-time semantic segmentation algorithms such as, FastFCN, FC-HarDNet, SegNet and BiSeNet, the object detection algorithm, DetectNet and 2D object classification networks, AlexNet and GoogleNet. We built a 3D and RGB door dataset with images from several indoor environments using a 3D Realsense camera D435. This dataset is freely available online. All methods are analysed taking into account their accuracy and the speed of the algorithm in a low powered computer. We conclude that it is possible to have a door classification algorithm running in real-time on a low-power device.

Comparison between the optical properties of aerosols in the fine and coarse fractions over Valladolid, Spain

Continuous measurements of the optical properties of aerosol particles have been made at Valladolid, Spain, covering the period from June 2011 to July 2012. The measurements were made at two size cuts: sub-10 μm and sub-1 μm (PM10 and PM1). The data measured were the scattering and backscattering coefficients, σ_s and σ_bs, obtained from an integrating nephelometer, and the absorption coefficient, σ_a, obtained from a particle soot absorption photometer. Spectrally resolved data were obtained from both instruments at 3 wavelengths (blue/green/red) at low relative humidity (RH < 40%). The statistical data for the instruments were calculated based on the hourly averages. For the PM10 fraction, the hourly mean values of σ_s and σ_a at 550 nm were 33 Mm^-1 (StD = 30 Mm^-1) and 4 Mm^-1 (StD = 3 Mm^-1), respectively. For the PM1 fraction, σ_s and σ_a mean values were 16 Mm^-1 (StD = 14 Mm^-1) and 4 Mm^-1 (StD = 3 Mm^-1), also at 550 nm. The derived parameters analyzed were the single scattering albedo, ω₀, the backscatter fraction, σ_bs/σ_s, and the Ångström exponents of scattering, absorption and single scattering albedo, α_s, α_a and α_ω0. The contribution of the PM10 and the PM1 fractions for all these parameters plays a central role throughout the paper, allowing an improved classification of aerosol types. Our data are dominated by elemental carbon (EC) and elemental carbon/organic carbon mixed (EC/OC). For the PM10 data, dust dominated aerosol is also observed. Although we found that fine particles contribute more than coarse particles for decreasing the ω₀ values, results suggest that it is also necessary to quantify the effect of coarse particles. Fine particles were found to produce ω₀ spectra that decrease with the wavelength, α_ω0 > 0, while PM10 fractions were found to produce spectra that can decrease or increase with the wavelength, 0 < α_ω0 < 0. Both daily cycle and monthly variations are analyzed and related to local features as well as the transport of particles from elsewhere. A diurnal pattern characteristic of urban areas is observed, but it is less evident on weekends. The main long range transport influences are Atlantic advection, anthropogenic events from Central Europe and dust events.

In situ UV-VIS-NIR absorbing properties of atmospheric aerosol particles: estimates of the imaginary refractive index and comparison with columnar values

In this work, a novel technique is used to estimate the aerosol complex index of refraction of in situ collected samples. Samples of atmospheric particulate matter were collected in El Arenosillo, southern Spain, on polycarbonate filters during summer 2004 as part of an aerosol characterization campaign. These samples were analyzed for the volumetric absorption coefficient in the 320-800 nm spectral region and an estimation of the effective imaginary refractive index was made. The values of the imaginary part of the complex refractive index ranged between 0.0009-0.0215 at 800 nm and 0.0015-0.0114 at 320 nm. Little dependence on the wavelength was observed. Several intense and long lasting desert outbreaks were registered during the campaign and the complex refractive index almost doubles its value during these dust events. Finally, we present a comparison of data obtained in situ with columnar data obtained from the AERONET network. A correlation factor of 0.64 was obtained between both data, which gives an idea of how accurately the in situ ground data represent the total column.