Crop canopy volume weighted by color parameters from UAV-based RGB imagery to estimate above-ground biomass of potatoes

• Published in: Computers and electronics in agriculture 2024-12, Vol.227, p.109678, Article 109678
• Main Authors: Liu, Yang, Yang, Fuqin, Yue, Jibo, Zhu, Wanxue, Fan, Yiguang, Fan, Jiejie, Ma, Yanpeng, Bian, Mingbo, Chen, Riqiang, Yang, Guijun, Feng, Haikuan
• Format: Article
• Language: English
• Subjects: Aboveground biomass; Canopy spectra; Potato; Structural indicators; Unmanned aerial vehicle
• ISSN: 0168-1699
• DOI: 10.1016/j.compag.2024.109678

•A color weighted crop canopy volume index was proposed.•AGB estimation model of potato in multiple growth stages was constructed.•The model performance built by single and complex metrics was compared.•The problem of underestimating AGB by using CP was solved. Current techniques to estimate crop aboveground biomass (AGB) across the multiple growth stages mainly used optical remote-sensing techniques. However, this technology was limited by saturation of the canopy spectrum. To meet this problem, this study used digital images obtained by an unmanned aerial vehicle to extract the spectral and structural indicators of the crop canopy in three key potato growth stages. We took the color parameters (CP) of assorted color space transformations as the canopy spectral information, and crop height (CH), crop coverage (CC), and crop canopy volume (CCV) as the canopy structural indicators. Based on the complementary advantages of CP and CCV, we proposed a new metric: the color parameter-weighted crop-canopy volume (CCVCP). Results showed that the CH, CCV, and CCVCP correlated more strongly with potato AGB during the multi-growth stages than do CP and CC. The hue-weighted crop-canopy volume (CCVH) correlated most strongly with the potato AGB among all structural indicators. Using CH was more accurate in estimating potato AGB compared to CP and CC. Combining indicators (CP + CC/CH, CP + CC + CH) improved the accuracy of potato AGB estimation over the multi-growth stages. Except for the CP + CC + CH model, other AGB estimation models produced inaccurate AGB estimation than the models based on CCV and CCVH. The AGB estimation accuracy produced by the univariate-based CCVH model (R2 = 0.65, RMSE = 281 kg/hm2, and NRMSE = 23.61 %) was comparable to that of the complex model [CP + CC + CH using random forest (RF) or multiple stepwise regression (MSR)]. Compared with CP + CC + CH using RF and MSR, the RMSE decreased and increased by 0.35 % and 4.24 %, respectively. Compared with CP, CP + CC, CP + CH, and CCV, the use of CCVH to estimate AGB decreased the RMSE by 10.24 %, 7.42 %, 6.36 %, and 6.33 %, respectively. Meanwhile, the performance of CCVH was verified at different stages and among varieties. Thus, this indicator can be used for monitoring potato growth to help guide field production management.