A New Method for Estimating Wheat Powdery Mildew Severity Using Features Extracted from Wavelet-Transformed Spectrograms

Yang Liu; Ruomei Zhao; Zhihui Feng; Wei Zhou; Jibo Yue; Hong Sun; Haikuan Feng; Lulu An; Hui Zhang; Haiyang Zhang; Fangkui Zhao; Xiaojing Yan; Meiyan Shu; Wei Guo

doi:10.1016/j.eng.2026.04.022

Engineering ›› :202604022 DOI: 10.1016/j.eng.2026.04.022

research-article

A New Method for Estimating Wheat Powdery Mildew Severity Using Features Extracted from Wavelet-Transformed Spectrograms

Author information +

^a College of Information and Management Science, Henan Agricultural University , Zhengzhou 450046, China

^bState Key Laboratory of High—Efficiency Production of Wheat—Maize Double Cropping, College of Agronomy, Henan Agricultural University , Zhengzhou 450046, China

^c Zhengzhou Tobacco Research Institute of CNTC , Zhengzhou 450001, China

^d Key Lab of Smart Agriculture System (Ministry of Education), China Agricultural University , Beijing 100083, China

^e Information Technology Research Center, Beijing Academy of Agriculture and Forestry Sciences , Beijing 100097, China

^f School of Information Engineering, Inner Mongolia University of Technology , Hohhot 010000, China

^g State Key Laboratory of Efficient Utilization of Agricultural Water Resources , Beijing 100083, China

^h State Key Laboratory for Biology of Plant Diseases and Insect Pests, Institute of Plant Protection, Chinese Academy of Agricultural Sciences , Beijing 100193, China

^*E-mail addresses: hafzhh@163.com (Z. Feng),

zhouwei02@caas.cn (W. Zhou),

guowei@henau.edu.cn (W. Guo).

Show less

History +

PDF (4492KB)

Abstract

Wheat powdery mildew (WPM) spreads rapidly and can severely disrupt crop physiological structure, thereby suppress growth and cause yield loss. Spectroscopy-based analysis offers a rapid and non-destructive approach for WPM detection because disease development induces measurable changes in spectral signals. However, extracting informative features from complex spectral data remains a key challenge for accurate WPM monitoring. Accordingly, this study developed a new feature-extraction framework that combines continuous wavelet transform (CWT)-derived spectrograms with a convolutional neural network (CNN) for WPM severity estimation. Field experiments were conducted at the Xinxiang Plant Protection Institute in China from 2022 to 2024. Ground canopy spectra (325–1075 nm) and the corresponding WPM severity were collected on nine sampling dates over the three-year period. Three spectral preprocessing methods, namely Savitzky–Golay smoothing, standard normal variate (SNV), and first-order differentiation (FOD), were first compared in the spatial domain. CWT was then used to decompose the spectral signals into wavelet coefficients across multiple scales and wavelengths in the frequency domain. Because WPM altered both sensitive wavelengths and responsive scales, we hypothesized that integrating both sources of information would improve disease estimation. To test this, two-dimensional spectrograms were constructed from the wavelet coefficients and used as CNN inputs for integrated feature extraction. Three feature sets, including spectral features, wavelet coefficients at different scales, and CNN-derived spectrogram features, were then used to build random forest (RF) regression models for comparison. On the validation set, FOD-based features achieved better estimation accuracy (the coefficient of determination (R²) = 0.87, root mean square error (RMSE) = 10.03%, and normalized RMSE (NRMSE) = 25.75%) than the other spectral preprocessing methods. Wavelet-coefficient features further improved performance (R² = 0.91, RMSE = 8.25%, and NRMSE = 21.18%), indicating that spectral decomposition strengthened disease-related responses. Spectrogram features performed best, and the RF model based on these features achieved the highest accuracy on the validation set (R² = 0.93, RMSE = 7.53%, and NRMSE = 19.34%). The spectrogram-based model also showed good applicability across years and cultivars, with R² values above 0.95 across the three years and above 0.88 for the two cultivars. Overall, the proposed framework captured deeper WPM-induced spectral information and improved model generalizability, providing a promising tool for field-scale WPM monitoring and management.

Keywords

Wheat powdery mildew / Spectroscopy / Continuous wavelet transform / Random forest

Cite this article

Download citation ▾

Yang Liu, Ruomei Zhao, Zhihui Feng, Wei Zhou, Jibo Yue, Hong Sun, Haikuan Feng, Lulu An, Hui Zhang, Haiyang Zhang, Fangkui Zhao, Xiaojing Yan, Meiyan Shu, Wei Guo. A New Method for Estimating Wheat Powdery Mildew Severity Using Features Extracted from Wavelet-Transformed Spectrograms. Engineering 202604022 DOI:10.1016/j.eng.2026.04.022

登录浏览全文

4963

注册一个新账户忘记密码

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	Liu Y, An L, Wang N, Tang WJ, Li M, Liu G, et al. Leaf area index estimation under wheat powdery mildew stress by integrating UAV—based spectral, textural and structural features. Comput Electron Agric 2023; 213:108169.

[2]	Xuan GT, Li QK, Shao Y, Shi YK . Early diagnosis and pathogenesis monitoring of wheat powdery mildew caused by blumeria graminis using hyperspectral imaging. Comput Electron Agric 2022; 197:106921.

[3]	Liu W, Sun CF, Zhao YN, Xu F, Song YL, Fan JR, et al. Monitoring of wheat powdery mildew under different nitrogen input levels using hyperspectral remote sensing. Remote Sens 2021; 13(18):3753.

[4]	Liu LY, Dong Y, Huang WJ, Du XP, Luo JH, Shi Y, et al. Enhanced regional monitoring of wheat powdery mildew based on an instance—based transfer learning method. Remote Sens 2019; 11(3):298.

[5]	He L, Qi SL, Duan JZ, Guo TC, Feng W, He DX . Monitoring of wheat powdery mildew disease severity using multiangle hyperspectral remote sensing. IEEE Trans Geosci Remote Sens 2021; 59(2):979—90.

[6]	Song L, Cai JX, Wu K, Li YH, Hou G, Du SL, et al. Early diagnosis of wheat powdery mildew using solar—induced chlorophyll fluorescence and hyperspectral reflectance. Eur J Agron 2025; 162:127427.

[7]	Abdulridha J, Ampatzidis Y, Roberts P, Kakarla SC . Detecting powdery mildew disease in squash at different stages using UAV—based hyperspectral imaging and artificial intelligence. Biosyst Eng 2020; 197:135-48.

[8]	Berdugo CA, Zito R, Paulus S, Mahlein AK . Fusion of sensor data for the detection and differentiation of plant diseases in cucumber. Plant Pathol 2014; 63(6):1344-56.

[9]	Chandel AK, Khot LR, Sallato CB . Apple powdery mildew infestation detection and mapping using high—resolution visible and multispectral aerial imaging technique. Sci Hortic 2021; 287:110228.

[10]	Poblete T, Camino C, Beck PA, Hornero A, Kattenborn T, Saponari M, et al. Detection of Xylella fastidiosa infection symptoms with airborne multispectral and thermal imagery: assessing bandset reduction performance from hyperspectral analysis. ISPRS J Photogramm Remote Sens 2020; 162:27-40.

[11]	Mustafa G, Zheng HB, Khan IH, Tian L, Jia HY, Li GQ, et al. Hyperspectral reflectance proxies to diagnose in—field fusarium head blight in wheat with machine learning. Remote Sens 2022; 14(12):2784.

[12]	Song L, Wang LY, Yang ZQ, He L, Feng ZH, Duan JZ, et al. Comparison of algorithms for monitoring wheat powdery mildew using multi—angular remote sensing data. Crop J 2022; 10(5):1312—22.

[13]	Feng W, Shen WY, He L, Duan JZ, Guo B, Li YX, et al. Improved remote sensing detection of wheat powdery mildew using dual—green vegetation indices. Precis Agric 2016; 17(5):608-27.

[14]	Cao XR, Luo Y, Zhou YL, Duan XY, Cheng DF . Detection of powdery mildew in two winter wheat cultivars using canopy hyperspectral reflectance. Crop Prot 2013; 45:124—31.

[15]	An L, Liu Y, Liu GH, Zhao RM, Tang W, Li MZ, et al. Estimation on powdery mildew of wheat canopy based on in—situ hyperspectral responses and characteristic wavelengths optimization. Crop Prot 2024; 184:106804.

[16]	Watt MS, Poblete T, de Silva D, Estarija HJC, Hartley RJL, Leonardo EMC, et al. Prediction of the severity of Dothistroma needle blight in radiata pine using plant based traits and narrow band indices derived from UAV hyperspectral imagery. Agric For Meteorol 2023; 330:109294.

[17]	Liu Y, Ma XY, An L, Sun H, Zhao FK, Yan XJ, et al. Exploring UAV narrow—band hyperspectral indices and crop functional traits derived from radiative transfer models to detect wheat powdery mildew. Int J Appl Earth Obs Geoinf 2025; 141:104627.

[18]	Rumpf T, Mahlein AK, Steiner U, Oerke EC, Dehne HW, Plümer L . Early detection and classification of plant diseases with support vector machines based on hyperspectral reflectance. Comput Electron Agric 2010; 74(1):91-9.

[19]	Liu Y, Feng HK, Yue JB, Jin XL, Fan YG, Chen RQ, et al. Improved potato AGB estimates based on UAV RGB and hyperspectral images. Comput Electron Agric 2023; 214:108260.

[20]	Zhao RM, An L, Tang WJ, Gao DH, Qiao L, Li MZ, et al. Deep learning assisted continuous wavelet transform—based spectrogram for the detection of chlorophyll content in potato leaves. Comput Electron Agric 2022; 195:106802.

[21]	Liu Y, Yang FQ, Yue JB, Zhu WX, Fan YG, Fan J, et al. Crop canopy volume weighted by color parameters from UAV—based RGB imagery to estimate above—ground biomass of potatoes. Comput Electron Agric 2024; 227:109678.

[22]	Mohammadi V, Minaei S, Gouton P, Mahdavian AR, Khoshtaghaza MH . Spectral discrimination of crops and weeds using deep learning assisted by wavelet transform and statistical preprocessing. Weed Sci 2024; 72(5):536—45.

[23]	Lv ZA, Meng R, Chen GS, Zhao F, Xu BY, Zhao YT, et al. Combining multiple spectral enhancement features for improving spectroscopic asymptomatic detection and symptomatic severity classification of southern corn leaf blight. Precis Agric 2023; 24(4):1593-618.

[24]	Guo W, Sun HG, Qiao HB, Zhang H, Zhou L, Dong P, et al. Spectral detection of peanut southern blight severity based on continuous wavelet transform and machine learning. Agriculture 2023; 13(8):1504.

[25]	Ren KH, Dong Y, Huang WJ, Guo AT, Jing X . Monitoring of winter wheat stripe rust by collaborating canopy SIF with wavelet energy coefficients. Comput Electron Agric 2023; 215:108366.

[26]	Duan Z, Li H, Li C, Zhang J, Zhang D, Fan X, et al. A CNN model for early detection of pepper Phytophthora blight using multispectral imaging, integrating spectral and textural information. Plant Methods 2024; 20(1):115.

[27]	Ahad MT, Li Y, Song B, Bhuiyan T . Comparison of CNN—based deep learning architectures for rice diseases classification. Artif Intell Agric 2023; 9:22-35.

[28]	Kendler S, Aharoni R, Young S, Sela H, Kis—Papo T, Fahima T, et al. Detection of crop diseases using enhanced variability imagery data and convolutional neural networks. Comput Electron Agric 2022; 193:106732.

[29]	Sun HG, Zhou L, Shu MY, Zhang J, Feng ZH, Feng HK, et al. Estimation of peanut southern blight severity in hyperspectral data using the synthetic minority oversampling technique and fractional—order differentiation. Agriculture 2024; 14(3):476.

[30]	Ong P, Jian J, Li X, Zou C, Yin J, Ma G . Sugarcane disease recognition through visible and near—infrared spectroscopy using deep learning assisted continuous wavelet transform—based spectrogram. Spectrochim Acta A Mol Biomol Spectrosc 2025; 324:125001.

[31]	Haghbin N, Bakhshipour A, Zareiforoush H, Mousanejad S . Non—destructive pre—symptomatic detection of gray mold infection in kiwifruit using hyperspectral data and chemometrics. Plant Methods 2023; 19(1):53.

[32]	Azadbakht M, Ashourloo D, Aghighi H, Radiom S, Alimohammadi A . Wheat leaf rust detection at canopy scale under different LAI levels using machine learning techniques. Comput Electron Agric 2019; 156:119—28.

[33]	An L, Liu Y, Wang N, Liu GH, Liu MJ, Tang WJ, et al. Classification of wheat powdery mildew based on hyperspectral: from leaves to canopy. Crop Prot 2023; 177:106599.

[34]	Xue BW, Tian L, Wang ZY, Wang X, Yao X, Zhu Y, et al. Quantification of rice spikelet rot disease severity at organ scale with proximal imaging spectroscopy. Precis Agric 2023; 24(3):1049-71.

[35]	Dhau I, Adam E, Mutanga O, Ayisi K, Abdel—Rahman EM, Odindi J, et al. Testing the capability of spectral resolution of the new multispectral sensors on detecting the severity of grey leaf spot disease in maize crop. Geocarto Int 2018; 33(11):1223—36.

[36]	Zhang J, Jing X, Song XY, Zhang T, Duan WA, Su J . Hyperspectral estimation of wheat stripe rust using fractional order differential equations and Gaussian process methods. Comput Electron Agric 2023; 206:107671.

[37]	Almoujahed MB, Apolo—Apolo OE, Whetton RL, Kazlauskas M, Kriauciuniene Z, Sarauskis E, et al. Field—based hyperspectral imaging for detection and spatial mapping of fusarium head blight in wheat. Eur J Agron 2025; 164:127485.