Development an edge-computing sensing unit for continuous measurement of canopy cover percentage of dry edible beans

Authors

Wei-zhen Liang, University of Nebraska-Lincoln
Joseph Oboamah, University of Nebraska-Lincoln
Xin Qiao, University of Nebraska - LincolnFollow
Yufeng Ge, University of Nebraska-LincolnFollow
Robert M. Harveson, University of Nebraska-LincolnFollow
Daran R. Rudnick, University of Nebraska-LincolnFollow
Jun Wang, University of IowaFollow
Haishun Yang, University of Nebraska - LincolnFollow
Angie Gradiz, University of Nebraska-LincolnFollow

Date of this Version

7-20-2022

Citation

An ASABE Meeting Presentation DOI: https://doi.org/10.13031/aim.202201238

Comments

Used by permission.

Abstract

Canopy cover (CC) is an important indicator for crop development. Currently, CC can be estimated indirectly by measuring leaf area index (LAI), using commercially available hand-held meters. However, it does not capture the dynamics of CC. Continuous CC monitoring is essential for dry edible beans production since it can affect crop water use, weed, and disease control. It also helps growers to closely monitor “yellowness”, or senescence of dry beans to decide proper irrigation cutoff to allow the crop to dry down for harvest. The goal of this study was to develop a device – CanopyCAM, containing software and hardware that can monitor dry bean CC continuously. CanopyCAM utilized an in-house developed image-based algorithm, edge-computing, and Internet of Things (IoT) telemetry to transmit and report CC in real-time. In the 2021 growing season, six CanopyCAMs were developed with three installed in fully irrigated dry edible beans research plots and three installed at commercial farms. CC measurements were recorded at 15 min interval from 7:00 am to 7:00 pm each day. Initially, the overall trend of CC development increased over time but there were many fluctuations in daily readings due to lighting conditions which caused some overexposed images. A simple filtering algorithm was developed to remove the “noisy images”. CanopyCAM measured CC (CC_CanopyCAM) were compared with CC obtained from a Li-COR Plant Canopy Analyzer (CC_LAI). The average error between CC_CanopyCAM and CC_LAI was 2.3%, and RMSE and R² were 2.95% and 0.99, respectively. In addition, maximum CC (CC_max) and duration of the maximum CC (t_{max_canopy}) were identified at each installation location using the generalized reduced gradient (CRG) algorithm with nonlinear optimization. An improvement of correlation was found between dry bean yield and combination of CC_max and t_{max_canopy} (R² = 0.77, Adjusted R² = 0.62) as compared to yield vs. CC_max (R² = 0.58) or yield vs. t_{max_canopy} (R² = 0.45). This edge-computing, IoT enabled capability of CanopyCAM, provided accurate CC readings which could be used by growers and researchers for different purpose.