Search Results
Optimized Deep Learning Model as a Basis for Fast UAV Mapping of Weed Species in Winter Wheat Crops
Weed maps should be available quickly, reliably, and with high detail to be useful for site-specific management in crop protection and to promote more sustainable agriculture by reducing pesticide use. Here, the optimization of a deep residual convolutional neural network (ResNet-18) for the classification of weed and crop plants in UAV imagery is proposed. The target was to reach sufficient performance on an embedded system by maintaining the same features of the ResNet-18 model as a basis for fast UAV mapping. This would enable online recognition and subsequent mapping of weeds during UAV flying operation. Optimization was achieved mainly by avoiding redundant computations that arise when a classification model is applied on overlapping tiles in a larger input image. The model was trained and tested with imagery obtained from a UAV flight campaign at low altitude over a winter wheat field, and classification was performed on species level with the weed species Matricaria chamomilla L., Papaver rhoeas L., Veronica hederifolia L., and Viola arvensis ssp. arvensis observed in that field. The ResNet-18 model with the optimized image-level prediction pipeline reached a performance of 2.2 frames per second with an NVIDIA Jetson AGX Xavier on the full resolution UAV image, which would amount to about 1.78 ha h−1 area output for continuous field mapping. The overall accuracy for determining crop, soil, and weed species was 94%. There were some limitations in the detection of species unknown to the model. When shifting from 16-bit to 32-bit model precision, no improvement in classification accuracy was observed, but a strong decline in speed performance, especially when a higher number of filters was used in the ResNet-18 model. Future work should be directed towards the integration of the mapping process on UAV platforms, guiding UAVs autonomously for mapping purpose, and ensuring the transferability of the models to other crop fields.
Impact of the precursor chemistry and process conditions on the cell-to-cell variability in 1T-1R based HfO2 RRAM devices
The Resistive RAM (RRAM) technology is currently in a level of maturity that calls for its integration into CMOS compatible memory arrays. This CMOS integration requires a perfect understanding of the cells performance and reliability in relation to the deposition processes used for their manufacturing. In this paper, the impact of the precursor chemistries and process conditions on the performance of HfO2 based memristive cells is studied. An extensive characterization of HfO2 based 1T1R cells, a comparison of the cell-to-cell variability, and reliability study is performed. The cells’ behaviors during forming, set, and reset operations are monitored in order to relate their features to conductive filament properties and process-induced variability of the switching parameters. The modeling of the high resistance state (HRS) is performed by applying the Quantum-Point Contact model to assess the link between the deposition condition and the precursor chemistry with the resulting physical cells characteristics.
Fabrication of metastable crystalline nanocomposites by flash annealing of Cu47.5Zr47.5Al5 metallic glass using joule heating
Flash Joule-heating was applied to the Cu47.5Zr47.5Al5 metallic glass for designing fully crystalline metastable nanocomposites consisting of the metastable B2 CuZr and low-temperature equilibrium Cu10Zr7 phases. The onset of crystallization was in situ controlled by monitoring resistivity changes in the samples. The effect of heating rate and annealing time on the volume fraction of the crystalline phases and mechanical properties of the nanocomposites was studied in detail. Particularly, an increase of the heating rate and a decrease of the annealing time lead to a lower number of equilibrium Cu10Zr7 precipitates and an increase of tensile ductility. Tailoring of these non-equilibrium microstructures and mechanical properties may not be possible unless one starts with a fully glassy material that opens new perspectives for designing metastable nanomaterials with unique physical properties.
Ordered states in the Kitaev-Heisenberg model: From 1D chains to 2D honeycomb
We study the ground state of the 1D Kitaev-Heisenberg (KH) model using the density-matrix renormalization group and Lanczos exact diagonalization methods. We obtain a rich ground-state phase diagram as a function of the ratio between Heisenberg (J = cosϕ) and Kitaev (K = sinϕ) interactions. Depending on the ratio, the system exhibits four long-range ordered states: ferromagnetic-z, ferromagnetic-xy, staggered-xy, Néel-z, and two liquid states: Tomonaga-Luttinger liquid and spiral-xy. The two Kitaev points ϕ=π2 and φ=3π2 are singular. The ϕ-dependent phase diagram is similar to that for the 2D honeycomb-lattice KH model. Remarkably, all the ordered states of the honeycomb-lattice KH model can be interpreted in terms of the coupled KH chains. We also discuss the magnetic structure of the K-intercalated RuCl3, a potential Kitaev material, in the framework of the 1D KH model. Furthermore, we demonstrate that the low-lying excitations of the 1D KH Hamiltonian can be explained within the combination of the known six-vertex model and spin-wave theory.
Studies on Stress Corrosion Cracking of Vit 105 Bulk Metallic Glass
The project “Stress Corrosion Cracking of Zr-based Bulk Metallic Glasses” (SCC of Zr-BMGs) within PP1594 mainly dealt with mechanical–corrosive interactions and failure of this class of metastable materials. It focused on one of the most application-relevant zirconium (Zr)-BMG, Vit(reloy) 105, with composition Zr52.5Cu17.9Ni14.6Al10Ti5 (at.%). Even though this BMG is known as an extraordinary glass former, the metallurgical processing is still a critical issue. In contrast to conventional processing, i.e., arc melting of master alloy ingots from single constituents, a different route using binary pre-alloys for the master alloys production was applied and led to superior mechanical properties upon mechanical testing under tensile and three-point-bending (3PB) conditions in air. As a reference and for a detailed understanding of failure, fracture, and cracking of Zr-based BMG in air, notched specimen 3PB experiments with in situ microscopic observation were done and the still controversial interpretation of the mechanical behavior of BMG in the framework of fracture mechanics was addressed. The specimen from the in situ 3PB tests served for transmission electron microscopy (TEM) investigations on the structural nature of shear bands in BMG on the atomistic scale. Altogether, complete crack paths could be observed and analyzed, and based on this, details of the shear band-driven crack growth are described. While in first SCC studies using a newly developed setup full cross section (3PB) bars were investigated, in recent in situ experiments, notched specimens were tested in 0.01 M NaCl, yielding strong evidence for a catastrophic failure due to hydrogen embrittlement (HE). The known susceptibility to pitting corrosion in halide-containing environments is only the initial stage for failure under SCC conditions. Once pitting is initiated, the local electrode potential is severely reduced. Further, the hydrolysis reaction of oxidized Zr4+ to zirconyl ions ZrO2+ during local BMG dissolution produces H+ and, thus, a local acidic environment that enables proton reduction and hydrogen absorption in the stressed BMG region. The peculiar failure and fracture surface characteristics as well as the proven local reduction of the pH value in the vicinity of the notch during in situ experiments clearly account for the proposed HE-SCC failure mechanism.
Combining magnetic forces for contactless manipulation of fluids in microelectrode-microfluidic systems
A novel method to drive and manipulate fluid in a contactless way in a microelectrode-microfluidic system is demonstrated by combining the Lorentz and magnetic field gradient forces. The method is based on the redox-reaction [Fe(CN) 6 ] 3− /[Fe(CN) 6 ] 4− performed in a magnetic field oriented perpendicular to the ionic current that crosses the gap between two arrays of oppositely polarized microelectrodes, generating a magnetohydrodynamic flow. Additionally, a movable magnetized CoFe micro-strip is placed at different positions beneath the gap. In this region, the magnetic flux density is changed locally and a strong magnetic field gradient is formed. The redox-reaction changes the magnetic susceptibility of the electrolyte near the electrodes, and the resulting magnetic field gradient exerts a force on the fluid, which leads to a deflection of the Lorentz force-driven main flow. Particle Image Velocity measurements and numerical simulations demonstrate that by combining the two magnetic forces, the flow is not only redirected, but also a local change of concentration of paramagnetic species is realized.
Using machine-learning to optimize phase contrast in a low-cost cellphone microscope
Cellphones equipped with high-quality cameras and powerful CPUs as well as GPUs are widespread. This opens new prospects to use such existing computational and imaging resources to perform medical diagnosis in developing countries at a very low cost. Many relevant samples, like biological cells or waterborn parasites, are almost fully transparent. As they do not exhibit absorption, but alter the light’s phase only, they are almost invisible in brightfield microscopy. Expensive equipment and procedures for microscopic contrasting or sample staining often are not available. Dedicated illumination approaches, tailored to the sample under investigation help to boost the contrast. This is achieved by a programmable illumination source, which also allows to measure the phase gradient using the differential phase contrast (DPC) [1, 2] or even the quantitative phase using the derived qDPC approach [3]. By applying machine-learning techniques, such as a convolutional neural network (CNN), it is possible to learn a relationship between samples to be examined and its optimal light source shapes, in order to increase e.g. phase contrast, from a given dataset to enable real-time applications. For the experimental setup, we developed a 3D-printed smartphone microscope for less than 100 $ using off-the-shelf components only such as a low-cost video projector. The fully automated system assures true Koehler illumination with an LCD as the condenser aperture and a reversed smartphone lens as the microscope objective. We show that the effect of a varied light source shape, using the pre-trained CNN, does not only improve the phase contrast, but also the impression of an improvement in optical resolution without adding any special optics, as demonstrated by measurements.
High temperature behavior of rual thin films on piezoelectric CTGS and LGS substrates
This paper reports on a significant further improvement of the high temperature stability of RuAl thin films (110 nm) on the piezoelectric Ca3TaGa3Si2O14 (CTGS) and La3Ga5SiO14 (LGS) substrates. RuAl thin films with AlN or SiO2 cover layers and barriers to the substrate (each 20 nm), as well as a combination of both were prepared on thermally oxidized Si substrates, which serve as a reference for fundamental studies, and the piezoelectric CTGS, as well as LGS substrates. In somefilms, additional Al layers were added. To study their high temperature stability, the samples were annealed in air and in high vacuum up to 900 °C, and subsequently their cross-sections, phase formation, film chemistry, and electrical resistivity were analyzed. It was shown that on thermally oxidized Si substrates, all films were stable after annealing in air up to 800 °C and in high vacuum up to 900 °C. The high temperature stability of RuAl thin films on CTGS substrates was improved up to 900 °C in high vacuum by the application of a combined AlN/SiO2 barrier layer and up to 800 °C in air using a SiO2 barrier. On LGS, the films were only stable up to 600 °C in air; however, a single SiO2 barrier layer was sufficient to prevent oxidation during annealing at 900 °C in high vacuum.
cellSTORM - Cost-effective Super-Resolution on a Cellphone using dSTORM
High optical resolution in microscopy usually goes along with costly hardware components, such as lenses, mechanical setups and cameras. Several studies proved that Single Molecular Localization Microscopy can be made affordable, relying on off-the-shelf optical components and industry grade CMOS cameras. Recent technological advantages have yielded consumer-grade camera devices with surprisingly good performance. The camera sensors of smartphones have benefited of this development. Combined with computing power smartphones provide a fantastic opportunity for “imaging on a budget”. Here we show that a consumer cellphone is capable of optical super-resolution imaging by (direct) Stochastic Optical Reconstruction Microscopy (dSTORM), achieving optical resolution better than 80 nm. In addition to the use of standard reconstruction algorithms, we used a trained image-to-image generative adversarial network (GAN) to reconstruct video sequences under conditions where traditional algorithms provide sub-optimal localization performance directly on the smartphone. We believe that “cellSTORM” paves the way to make super-resolution microscopy not only affordable but available due to the ubiquity of cellphone cameras.High optical resolution in microscopy usually goes along with costly hardware components, such as lenses, mechanical setups and cameras. Several studies proved that Single Molecular Localization Microscopy can be made affordable, relying on off-the-shelf optical components and industry grade CMOS cameras. Recent technological advantages have yielded consumer-grade camera devices with surprisingly good performance. The camera sensors of smartphones have benefited of this development. Combined with computing power smartphones provide a fantastic opportunity for “imaging on a budget”. Here we show that a consumer cellphone is capable of optical super-resolution imaging by (direct) Stochastic Optical Reconstruction Microscopy (dSTORM), achieving optical resolution better than 80 nm. In addition to the use of standard reconstruction algorithms, we used a trained image-to-image generative adversarial network (GAN) to reconstruct video sequences under conditions where traditional algorithms provide sub-optimal localization performance directly on the smartphone. We believe that “cellSTORM” paves the way to make super-resolution microscopy not only affordable but available due to the ubiquity of cellphone cameras.
Advection of Biomass Burning Aerosols towards the Southern Hemispheric Mid-Latitude Station of Punta Arenas as Observed with Multiwavelength Polarization Raman Lidar
In this paper, we present long-term observations of the multiwavelength Raman lidar PollyXT conducted in the framework of the DACAPO-PESO campaign. Regardless of the relatively clean atmosphere in the southern mid-latitude oceans region, we regularly observed events of long-range transported smoke, originating either from regional sources in South America or from Australia. Two case studies will be discussed, both identified as smoke events that occurred on 5 February 2019 and 11 March 2019. For the first case considered, the lofted smoke layer was located at an altitude between 1.0 and 4.2 km, and apart from the predominance of smoke particles, particle linear depolarization values indicated the presence of dust particles. Mean lidar ratio values at 355 and 532 nm were 49 ± 12 and 24 ± 18 sr respectively, while the mean particle linear depolarization was 7.6 ± 3.6% at 532 nm. The advection of smoke and dust particles above Punta Arenas affected significantly the available cloud condensation nuclei (CCN) and ice nucleating particles (INP) in the lower troposphere, and effectively triggered the ice crystal formation processes. Regarding the second case, the thin smoke layers were observed at altitudes 5.5–7.0, 9.0 and 11.0 km. The particle linear depolarization ratio at 532 nm increased rapidly with height, starting from 2% for the lowest two layers and increasing up to 9.5% for the highest layer, indicating the possible presence of non-spherical coated soot aggregates. INP activation was effectively facilitated. The long-term analysis of the one year of observations showed that tropospheric smoke advection over Punta Arenas occurred 16 times (lasting from 1 to 17 h), regularly distributed over the period and with high potential to influence cloud formation in the otherwise pristine environment of the region.