Development and application of a model for the automatic evaluation and classification of onions (Allium cepa L.) using a Deep Neural Network (DNN) DOI Creative Commons
P. Rybacki, Przemysław Przygodziński, Przemysław Łukasz Kowalczewski

et al.

Acta Scientiarum Polonorum Hortorum Cultus, Journal Year: 2024, Volume and Issue: 23(5), P. 39 - 57

Published: Nov. 30, 2024

Evaluating onions for size, shape, damage, colour and discolouration is the first most important step in classifying them raw material quality, processing horticultural agri-food sectors. Current methods of geometric evaluation grading involve mechanical extremely invasive sorting, which causes additional reduces quality also labour time-consuming. As a result, non-invasive classification that are both fast accurate being sought. One such method digital image analysis, which, when combined with instrumentation deep neural networks, can fully automate process. The main aim this study was development model automatic using convolutional network (CNN) model. A fixed-architecture built, computational algorithm developed Python 3.9 published at https://github.com/piotrrybacki/onion-CNN.git (accessed on 4 October 2024). Hyduro F1 onion variety, hybrid all-purpose variety Rijnsburger type, used to build, teach test classified images qualitatively an accuracy 91.85%. This based parameters onion, i.e. diameter, height, transversal longitudinal circumference, estimated area damage or skin. root mean square error (MSE) RGB space varied between 87.99 91.24, maximum time 28.98 ms/image. has very high utility, as it automates process, reducing its intensity.

Language: Английский

Enhanced Hyperspectral Forest Soil Organic Matter Prediction Using a Black-Winged Kite Algorithm-Optimized Convolutional Neural Network and Support Vector Machine DOI Creative Commons
Yun Deng, Linsong Xiao, Yuanyuan Shi

et al.

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(2), P. 503 - 503

Published: Jan. 7, 2025

Soil Organic Matter (SOM) is crucial for soil fertility, and effective detection methods are of great significance the development agriculture forestry. This study uses 206 hyperspectral samples from state-owned Yachang Huangmian Forest Farms in Guangxi, using SPXY algorithm to partition dataset a 4:1 ratio, provide an spectral data preprocessing method novel SOM content prediction model area similar regions. Three denoising (no denoising, Savitzky–Golay filter discrete wavelet transform denoising) were combined with nine mathematical transformations (original reflectance (R), first-order differential (1DR), second-order (2DR), MSC, SNV, logR, (logR)′, 1/R, ((1/R)′) form 27 combinations. Through Pearson heatmap analysis modeling accuracy comparison, SG-1DR combination was found effectively highlight features. A CNN-SVM based on Black Kite Algorithm (BKA) proposed. leverages powerful parameter tuning capabilities BKA, CNN feature extraction, SVM classification regression, further improving prediction. The results RMSE = 3.042, R2 0.93, MAE 4.601, MARE 0.1, MBE 0.89, PRIQ 1.436.

Language: Английский

Citations

0
A multi-component concentration spectral modeling method with parallel drift resistance based on disorderly difference DOI
Qilong Wan, Hongqiu Zhu, Chunhua Yang

et al.

Talanta, Journal Year: 2025, Volume and Issue: 292, P. 127943 - 127943

Published: March 13, 2025

Language: Английский

Citations

0
Federated learning applications in soil spectroscopy DOI Creative Commons

Giannis Gallios,

Nikolaos Tsakiridis, Nikolaos Tziolas

et al.

Geoderma, Journal Year: 2025, Volume and Issue: 456, P. 117259 - 117259

Published: March 25, 2025

Language: Английский

Citations

0
Characteristic wavelength selection based on multi-strategy fusion zebra optimization algorithm for PLSR DOI
Yü Liu, C.S. Tong,

Simin Wu

et al.

Measurement, Journal Year: 2025, Volume and Issue: unknown, P. 117486 - 117486

Published: April 1, 2025

Language: Английский

Citations

0
Spectral Data-Driven Prediction of Soil Properties Using LSTM-CNN-Attention Model DOI Creative Commons
Yiqiang Liu, Luming Shen, Xinghui Zhu

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(24), P. 11687 - 11687

Published: Dec. 14, 2024

Accurate prediction of soil properties is essential for sustainable land management and precision agriculture. This study presents an LSTM-CNN-Attention model that integrates temporal spatial feature extraction with attention mechanisms to improve predictive accuracy. Utilizing the LUCAS dataset, analyzes spectral data estimate key properties, including organic carbon (OC), nitrogen (N), calcium carbonate (CaCO3), pH (in H2O). The Long Short-Term Memory (LSTM) component captures dependencies, Convolutional Neural Network (CNN) extracts features, mechanism highlights critical information within data. Experimental results show proposed achieves excellent performance, coefficient determination (R2) values 0.949 0.916 0.943 0.926 (pH), along corresponding ratio percent deviation (RPD) 3.940, 3.737, 5.377, 3.352. Both R2 RPD exceed those traditional machine learning models, such as partial least squares regression (PLSR), support vector (SVR), random forest (RF), well deep models like CNN-LSTM Gated Recurrent Unit (GRU). Additionally, outperforms S-AlexNet in effectively capturing patterns. These findings emphasize potential significantly enhance accuracy reliability property predictions by both patterns effectively.

Language: Английский

Citations

2
Development and application of a model for the automatic evaluation and classification of onions (Allium cepa L.) using a Deep Neural Network (DNN) DOI Creative Commons
P. Rybacki, Przemysław Przygodziński, Przemysław Łukasz Kowalczewski

et al.

Acta Scientiarum Polonorum Hortorum Cultus, Journal Year: 2024, Volume and Issue: 23(5), P. 39 - 57

Published: Nov. 30, 2024

Evaluating onions for size, shape, damage, colour and discolouration is the first most important step in classifying them raw material quality, processing horticultural agri-food sectors. Current methods of geometric evaluation grading involve mechanical extremely invasive sorting, which causes additional reduces quality also labour time-consuming. As a result, non-invasive classification that are both fast accurate being sought. One such method digital image analysis, which, when combined with instrumentation deep neural networks, can fully automate process. The main aim this study was development model automatic using convolutional network (CNN) model. A fixed-architecture built, computational algorithm developed Python 3.9 published at https://github.com/piotrrybacki/onion-CNN.git (accessed on 4 October 2024). Hyduro F1 onion variety, hybrid all-purpose variety Rijnsburger type, used to build, teach test classified images qualitatively an accuracy 91.85%. This based parameters onion, i.e. diameter, height, transversal longitudinal circumference, estimated area damage or skin. root mean square error (MSE) RGB space varied between 87.99 91.24, maximum time 28.98 ms/image. has very high utility, as it automates process, reducing its intensity.

Language: Английский

Citations

0