UP - logo
Knjižnica tehniških fakultet, Maribor (KTFMB)
  • Avtomatiziran sistem za spremljanje vlage z uporabo zemeljsko prodirajočega radarja v vrtinah in konvolucijske nevronske mreže [Elektronski vir]
    Pongrac, Blaž, 1993- ; Gleich, Dušan
    This paper shows a cross-hole ground penetrating radar system and a deep-learning method for soil moisture estimation in lined water canal embankment. A custom-made nano-second pulse-based radar ... system that can detect variations in soil moisture content through a cross-hole approach was developed. The proposed system consists of a pulse generator, three transmitting antennas positioned in a 12 m deep borehole, and three receiving antennas placed in a separate borehole 100 m away from the transmitter. The receiver utilizes a high-frequency data acquisition card to capture signals at 3 Giga Bytes per second. The borehole antennas have been designed to operate in a broad frequency band to ensure optimal signal propagation within the soil. This paper proposes a deep regression convolutional network that model changes in wave propagation between the transmitted and received signals and estimates volumetric soil moisture using timesampled signals. The training dataset includes soil moisture measurements taken at three points between the transmitter and receiver, spaced 25 meters apart, to provide reliable ground truth data. To further enhance convolutional neural networks’ training and evaluation process, radar data and soil moisture measurements were collected over 73 days between the two boreholes. Additional data was acquired through an experiment involving pouring water into specially prepared boreholes between the transmitter and receiver antennas. This additional data was then used for training, validation, and testing. The experimental results indicate that the proposed system successfully detects changes in volumetric soil moisture using the provided Tx and Rx antennas.
    Vir: Zbornik dvaintridesete mednarodne Elektrotehniške in računalniške konference ERK 2023 [Elektronski vir] = Proceedings of the 32nd International Electrotechnical and Computer Science Conference ERK 2023 : Portorož, Slovenija, 28. - 29. september 2022 (Str. 101-105)
    Vrsta gradiva - prispevek na konferenci ; neleposlovje za odrasle
    Leto - 2023
    Jezik - slovenski
    COBISS.SI-ID - 166883587

    Povezava(-e):

    https://erk.fe.uni-lj.si/2023/papers/pongrac(avtomatiziran_sistem).pdf

Vnos na polico