Estimation of soybean yield from machine learning techniques and multispectral RPAS imagery

Eugênio, Fernando Coelho; Grohs, Mara; Venâncio, Luan Peroni; Schuh, Mateus Sabadi; Bottega, Eduardo Leonel; Ruoso, Régis; Schons, Cristine Tagliapietra; Mallmann, Caroline Lorenci; Badin, Tiago Luis; Fernandes, Pablo

doi:10.1016/j.rsase.2020.100397

Cited by 19 publications

(13 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Because of the prediction differences of different traits, many studies have explored the best traits for modeling. This study selected 8 potential traits for modeling based on previous findings [ 32 , 37 ]. Although the performance of each trait in yield (GPC) prediction is compared, the main purpose, unlike previous studies, is to select the best structural and spectral traits for comparative analysis of data fusion and multitask learning.…”

Section: Discussionmentioning

confidence: 99%

Simultaneous Prediction of Wheat Yield and Grain Protein Content Using Multitask Deep Learning from Time-Series Proximal Sensing

Sun

Jin

et al. 2022

Plant Phenomics

View full text Add to dashboard Cite

Wheat yield and grain protein content (GPC) are two main optimization targets for breeding and cultivation. Remote sensing provides nondestructive and early predictions of yield and GPC, respectively. However, whether it is possible to simultaneously predict yield and GPC in one model and the accuracy and influencing factors are still unclear. In this study, we made a systematic comparison of different deep learning models in terms of data fusion, time-series feature extraction, and multitask learning. The results showed that time-series data fusion significantly improved yield and GPC prediction accuracy with R2 values of 0.817 and 0.809. Multitask learning achieved simultaneous prediction of yield and GPC with comparable accuracy to the single-task model. We further proposed a two-to-two model that combines data fusion (two kinds of data sources for input) and multitask learning (two outputs) and compared different feature extraction layers, including RNN (recurrent neural network), LSTM (long short-term memory), CNN (convolutional neural network), and attention module. The two-to-two model with the attention module achieved the best prediction accuracy for yield (R2=0.833) and GPC (R2=0.846). The temporal distribution of feature importance was visualized based on the attention feature values. Although the temporal patterns of structural traits and spectral traits were inconsistent, the overall importance of both structural traits and spectral traits at the postanthesis stage was more important than that at the preanthesis stage. This study provides new insights into the simultaneous prediction of yield and GPC using deep learning from time-series proximal sensing, which may contribute to the accurate and efficient predictions of agricultural production.

show abstract

Section: Discussionmentioning

confidence: 99%

Simultaneous Prediction of Wheat Yield and Grain Protein Content Using Multitask Deep Learning from Time-Series Proximal Sensing

Sun

Jin

et al. 2022

Plant Phenomics

View full text Add to dashboard Cite

show abstract

“…Some scholars use RGB images to assess the grain yield of durum wheat at a low cost [12] . Other scholars use machine learning technology and multi-spectral RPAS images to estimate soybean yield [13] . In addition, grain impurities are also measured, such as by installing a rice grain impurity sensor in the combine harvester's granary [14] .…”

Section: Introduction mentioning

confidence: 99%

Influence of vibration on the grain flow sensor during the harvest and the difference elimination method

Qian¹,

Lü²,

Shen³

et al. 2021

International Journal of Agricultural and Biological Engineering

View full text Add to dashboard Cite

The grain yield data collected by the intelligent yield measurement system of the combine harvester is generated into a field plot yield distribution map, which is of great significance for guiding agricultural production. However, in the process of drawing the yield map, the combine harvester is affected by vibration during operation and the generated error data in the process of collecting data which will cause the drawing results to be inaccurate. This study researched two factors that cause errors, then, the influence of vibration interference on the measurement signal was eliminated by filtering, vibration isolation, and designing a double-plates differential grain flow sensor. Three methods were taken to eliminate random errors, gross errors and systematic errors, including using the arithmetic average value to replace the true value, the 3σ criterion, and removing the filling time data and the delaying time data. Finally, the grain yield distribution map was obtained through Matlab and Excel. The results showed that the interference frequency above 50 Hz could be eliminated by filtering, but it was difficult to filter the low-frequency signal which was close to the grain impact frequency. The vibration amplitude was reduced to 14.29% by adding a vibration isolation plate, and the SNR was increased from −4.67 dB to 29.21 dB by combining low-pass filtering and damping vibration isolation. When the grain feeding rate was 2 kg/s, the natural vibration amplitude of the sensor after difference was about 0.02 V and evenly distributed around the zero voltage 0.2 V. The influence of positive and negative offset on the average value of grain impact signal could be ignored, and the signal-to-noise ratio was increased from 29.21 dB to 62.49 dB. The results of field experiments showed that the yield map drawn can clearly display the yield value of the harvest area, which is used to guide agricultural production.

show abstract

“…The model was trained by crop growth and environment variables, which include weather data, MODIS Land Surface Temperature data, and MODIS Surface Reflectance data. In Cachoeira do Sul, Brazil, a Multi-Layer Perceptron (MLP) was used to adjust a predictive model for estimating the yield of soybean crop based on 9 vegetation indices (Eugenio et al, 2020). A soybean yield model was created by deep learning framework using CNN and recurrent neural networks (Khaki et al, 2020).…”

mentioning

confidence: 99%

A surrogate model based on feature selection techniques and regression learners to improve soybean yield prediction in southern France

Corrales

Schoving

Raynal

et al. 2022

Computers and Electronics in Agriculture

View full text Add to dashboard Cite

Empirical and process-based models are currently used to predict crop yield at field and regional levels. A mechanistic model named STICS (Multidisciplinary Simulator for Standard Crops) has been used to simulate soybean grain yield in several environments, including southern France. STICS simulates at a daily step the effects of climate, soil and management practices on plant growth, development and production. In spite of good performances to predict total aboveground biomass, poor results were obtained for final grain yield. In order to improve yield prediction, a surrogate model was developed from STICS dynamic simulations, feature selection techniques and regression learners. STICS was used to simulate functional variables at given growth stages and over selected phenological phases. The most representative variables were selected through feature selection techniques (filter, wrapper and embedded), and a subset of variables were used to train the regression learners Linear regression (LR), Support vector regression (SVR), Back propagation neural network (BPNN), Random forest (RF), Least Absolute Shrinkage and Selection Operator (LASSO) and M5 decision tree. The subset of variables selected by wrapper method combined with regression models SVR (R 2 = 0. 7102; subset of variables = 6) and LR (R 2 = 0. 6912; subset of variables = 14) provided the best results. SVR and LR models improved significantly the soybean yield predictions in southern France in comparison to STICS simulations (R 2 = 0.040).

show abstract

Estimation of soybean yield from machine learning techniques and multispectral RPAS imagery

Cited by 19 publications

References 49 publications

Simultaneous Prediction of Wheat Yield and Grain Protein Content Using Multitask Deep Learning from Time-Series Proximal Sensing

Simultaneous Prediction of Wheat Yield and Grain Protein Content Using Multitask Deep Learning from Time-Series Proximal Sensing

Influence of vibration on the grain flow sensor during the harvest and the difference elimination method

A surrogate model based on feature selection techniques and regression learners to improve soybean yield prediction in southern France

Contact Info

Product

Resources

About