No abstract

Artificial intelligence has smoothly penetrated several economic activities, especially monitoring and control applications, including the agriculture sector. However, research efforts toward low-power sensing devices with fully functional machine learning (ML) on-board are still fragmented and limited in smart farming. Biotic stress is one of the primary causes of crop yield reduction. With the development of deep learning in computer vision technology, autonomous detection of pest infestation through images has become an important research direction for timely crop disease diagnosis. This paper presents an embedded system enhanced with ML functionalities, ensuring continuous detection of pest infestation inside fruit orchards. The embedded solution is based on a low-power embedded sensing system along with a Neural Accelerator able to capture and process images inside common pheromone-based traps. Three different ML algorithms have been trained and deployed, highlighting the capabilities of the platform. Moreover, the proposed approach guarantees an extended battery life thanks to the integration of energy harvesting functionalities. Results show how it is possible to automate the task of pest infestation for unlimited time without the farmer's intervention.

Single layer mirrors have been prepared by evaporating gold and iridium on silicon substrates. The samples have been exposed to 4 keV He + ion flux at different total fluences, simulating the effect of solar wind ions on optical coatings. We show that the ion implantation significantly affects the optical characteristics of the metallic films. The phenomena are explained and modeled also considering the related material modifications observed with chemical and morphological analysis. (C) 2014 Optical Society of Americ

Future solar missions will investigate the Sun from very close distances and optical components are constantly exposed to low energy ions irradiation. In this work we present the results of a new experiment related to low energy alpha particles bombardments on Mo/Si multilayer optical coatings. Different multilayer samples, with and without a protecting capping layer, have been exposed to low energy alpha particles (4keV), fixing the ions fluency and varying the time of exposure in order to change the total dose accumulated. The experimental parameters have been selected considering the potential application of the coatings to future solar missions. Results show that the physical processes occurred at the uppermost interfaces can strongly damage the structure.

Apple is one of the most produced fruit crops in the world. Recent advances in Artificial Intelligence and the Internet of Things can reduce production costs and improve crop quality by providing prompt detection of dangerous parasites. This paper presents an effective solution to automate the detection of the Codling Moths. The system takes pictures of trapped insects in the orchard, analyzes them through a DNN algorithm, and sends alarms to the farmer in case of a positive detection. The system is fully autonomous and can operate unattended for the entire crop season. Detection reports are used for optimizing the treatment with chemicals only when threats are identified. The prototype is designed with an embedded platform powered by a small solar panel to achieve an energy-neutral balance.

Sol‐gel derived TiO2 thin films were prepared by spin‐ coating from an alcoholic solution of titanium isopropoxide. With the aim to develop titania layers suitable for memristive devices, the films were deposited onto test structures based on fused silica quartz substrates patterned with a Ti (5nm)/Pt (50nm) layer. The reported fabrication protocol is suitable for the development of a memristive device. Optical, structural, and morphological features of the samples were investigated with complementary techniques, such as scanning electron microscopy (SEM), prism coupling m‐line and micro‐Raman spectroscopy as well as transmittance and profilometry measurements. The quality of the surface of the obtained films was evaluated by SEM technique, and the morphology of samples deposited with different fabrication protocols was investigated. Additionally, a computer code for the refractive index and thickness estimation from the transmittance spectra was developed by unconstrained optimization procedure. The results of simulation were in good agreement with the experimental data obtained by m‐line measurements. Moreover, the porosity of a specific set of test unannealed films has been estimated. TiO2 films exhibit thickness of tens of nm, and micro‐Raman spectroscopy in conjunction with SEM indicate the presence of anatase phase after thermal annealing at 400 °C. (© 2015 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)

Batteryless image sensors present an opportunity for pervasive widespread remote sensor deployments that require little maintenance and have low cost. However, the reliance of these devices on energy harvesting presents tight constraints in the quantity of energy that can be stored and used, as well as limited, energydependent availability. In this work, we develop Camaroptera, the first batteryless, energy-harvesting image sensing platform to support active, long-range communication. Camaroptera reduces the high latency and energy cost of communication by using nearsensor processing pipelines to identify interesting images and transmit them to a faraway base station, while discarding uninteresting images. Camaroptera also dynamically adapts its processing pipeline to maximize system availability and responsiveness to interesting events in different harvesting conditions. We fully prototype the Camaroptera hardware platform in a compact, 2cm x 3cm x 5cm volume, composed of three adjoined circuit boards. We evaluate Camaroptera demonstrating the viability of a batteryless remote sensing platform in a small package. We show that compared to a system that transmits all image data, Camaroptera's processing pipelines and adaptive processing scheme captures and sends 2-5X more images of interest to an application. CCS CONCEPTS • Computer systems organization → Sensor networks; Embedded software; • Hardware → Wireless integrated network sensors; • Computing methodologies → Object detection; Neural networks.

No abstract