Estimation of potato above-ground biomass based on unmanned aerial vehicle red-green-blue images with different texture features and crop height

• DOI: 10.3389/fpls.2022.938216
• Publication Date: 2022-08-25T05:00:06Z
• Authors: Yang Liu; Haikuan Feng; Jibo Yue; Xiuliang Jin; Zhenhai Li; Guijun Yang
• Keywords: RGB color model; Texture (cosmology)

Obtaining crop above-ground biomass (AGB) information quickly and accurately is beneficial to farmland production management the optimization of planting patterns. Many studies have confirmed that, due canopy spectral saturation, AGB underestimated in multi-growth period crops when using only optical vegetation indices. To solve this problem, study obtains textures height directly from ultrahigh-ground-resolution (GDS) red-green-blue (RGB) images estimate potato three key growth periods. Textures include a grayscale co-occurrence matrix texture (GLCM) Gabor wavelet texture. GLCM-based were extracted seven-GDS (1, 5, 10, 30, 40, 50, 60 cm) RGB images. Gabor-based obtained magnitude on five scales (scales 1-5, labeled S1-S5, respectively). Potato was based generated model. Finally, AGB, we used (i) different GDS their combinations, (ii) (iii) all combined with height, (iv) (v) two types by least-squares support vector machine (LSSVM), extreme learning machine, partial least squares regression techniques. The results show that first increase then decrease over period; mainly affect correlation between GLCM- AGB; GDS1 GDS30 work best 1 5 cm 10 cm, respectively (however, estimating gradually deteriorates as convolution kernel scale increases); estimation single-type not good estimates multi-resolution multiscale (with latter being best); forms LSSVM technique improved 22.97, 14.63, 9.74, 8.18% (normalized root mean square error) compared textures, former respectively. Therefore, features acquired unmanned aerial vehicles improve accuracy under high coverage.