Accurate Yield Estimation Research Articles

Rice-Yield Prediction with Multi-Temporal Sentinel-2 Data and 3D CNN: A Case Study in Nepal

Crop yield estimation is a major issue of crop monitoring which remains particularly challenging in developing countries due to the problem of timely and adequate data availability. Whereas traditional agricultural systems mainly rely on scarce ground-survey data, freely available multi-temporal and multi-spectral remote sensing images are excellent tools to support these vulnerable systems by accurately monitoring and estimating crop yields before harvest. In this context, we introduce the use of Sentinel-2 (S2) imagery, with a medium spatial, spectral and temporal resolutions, to estimate rice crop yields in Nepal as a case study. Firstly, we build a new large-scale rice crop database (RicePAL) composed by multi-temporal S2 and climate/soil data from the Terai districts of Nepal. Secondly, we propose a novel 3D Convolutional Neural Network (CNN) adapted to these intrinsic data constraints for the accurate rice crop yield estimation. Thirdly, we study the effect of considering different temporal, climate and soil data configurations in terms of the performance achieved by the proposed approach and several state-of-the-art regression and CNN-based yield estimation methods. The extensive experiments conducted in this work demonstrate the suitability of the proposed CNN-based framework for rice crop yield estimation in the developing country of Nepal using S2 data.

Evaluation of Data Needs for Assessments of Aquifers Supporting Irrigated Agriculture

AbstractOne of the primary means of addressing the depletion of aquifers supporting irrigated agriculture is to reduce pumping. In this work, we address the key question of how much data is needed to reliably predict the impact of pumping reductions. Previous studies have demonstrated the effectiveness of a simple water balance‐based approach for predicting the near‐term impact of changes in total pumping on changes in average water levels for areas ranging in size from 256 to 21,600 km2 in the High Plains aquifer (HPA) in the state of Kansas in the central U.S. This method shows considerable promise as a management tool in the Kansas HPA and potentially other highly stressed aquifers. However, one characteristic that sets the Kansas HPA apart from many other aquifers is an abundance of data on both water levels, from a decades‐old annual water level measurement program, and annual pumping volumes, due to reporting requirements and regulatory oversight. Most regions will have considerably less information. This study investigates the impact of more limited data availability through random subsampling of the water level and water use data sets from Groundwater Management District 4 (GMD4) in northwest Kansas and GMD5 in south‐central Kansas. The results indicate that accurate estimates of net inflow and specific yield, the key parameters needed to predict near‐term aquifer responses to proposed pumping reductions, can be obtained with much sparser water level and water use data than are available in the Kansas HPA, as long as the data are not systematically biased.

Assimilation of LAI Derived from UAV Multispectral Data into the SAFY Model to Estimate Maize Yield

In this study, we develop a method to estimate corn yield based on remote sensing data and ground monitoring data under different water treatments. Spatially explicit information on crop yields is essential for farmers and agricultural agencies to make well-informed decisions. One approach to estimate crop yield with remote sensing is data assimilation, which integrates sequential observations of canopy development from remote sensing into model simulations of crop growth processes. We found that leaf area index (LAI) inversion based on unmanned aerial vehicle (UAV) vegetation index has a high accuracy, with R2 and root mean square error (RMSE) values of 0.877 and 0.609, respectively. Maize yield estimation based on UAV remote sensing data and simple algorithm for yield (SAFY) crop model data assimilation has different yield estimation accuracy under different water treatments. This method can be used to estimate corn yield, where R2 is 0.855 and RMSE is 692.8kg/ha. Generally, the higher the water stress, the lower the estimation accuracy. Furthermore, we perform the yield estimate mapping at 2 m spatial resolution, which has a higher spatial resolution and accuracy than satellite remote sensing. The great potential of incorporating UAV observations with crop data to monitor crop yield, and improve agricultural management is therefore indicated.

Rapid Yield Estimation of Microwave Passive Components Using Model-Order Reduction Based Neuro-Transfer Function Models

In this letter, we propose a novel technique for rapid and accurate yield estimation of microwave passive components using model-order reduction (MOR)-based neuro-transfer function (neuro-TF) models. In the proposed technique, the frequency responses of microwave components are represented by transfer functions in the pole-zero-gain format. The poles, zeros, and gain in the transfer functions are computed by the MOR technique. Neural networks are trained to capture the dynamic changes of the poles/zeros/gain as the statistical/geometrical variables change. A refinement training process is designed to further align the outputs of the neuro-TF model. Once developed, the MOR-based neuro-TF model can provide rapid and accurate prediction of electromagnetic (EM) behavior of microwave passive components, thereby accelerating EM-based yield estimation. To achieve similar yield estimation accuracy, the proposed technique requires a shorter CPU time than existing yield estimation methods. The advantages of the proposed technique are illustrated by two microwave examples.

Estimating winter wheat yield by assimilation of remote sensing data with a four-dimensional variation algorithm considering anisotropic background error and time window

To further improve the yield estimation accuracy of regional winter wheat, based on En4DVar algorithm, an anisotropic background error and time window four-dimensional variation (ABT-4DVar) algorithm was constructed by introducing a variation time window. Based on a comparison with the simulated crop yield without data assimilation at a single point scale and in a typical test area, the effectiveness and feasibility of the constructed ABT-4DVar algorithm were effectively verified. On this basis, using the official statistics as verification data, the ABT-4DVar algorithm was applied to a winter wheat yield estimation system by assimilating remotely sensed LAI into the WOFOST (WOrld FOod STudies) crop growth model at a regional scale. The yield estimation results of winter wheat in Hengshui city showed that the RE between the average estimated yield and official statistics was 3.18%, and the REs between the average estimated yields and the official statistics of the 11 counties were all within 14.80%. The yield estimation results of winter wheat in Hengshui city indicated that the ABT-4DVar algorithm had good performance in regional yield estimation and that the assimilation algorithm could provide technology for crop growth evaluation and yield simulation.

Microwave Sensing for Estimating Cranberry Crop Yield: A Pilot Study Using Simulated Canopies and Field Measurement Testbeds

We present the results of an experimental and computational pilot study of cranberry crop yield prediction using low-power microwave sensing and machine learning. We simulated backscattered radiation from cranberry canopies using plane-wave illumination with frequency content from 300 to 2400 MHz. The computational canopy domains allowed for variable soil moisture and cranberry yield in terms of cranberry mass per 1 ft <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of canopy surface area. We collected experimental field data with a prototype open-ended waveguide sensor operating between 600 and 1300 MHz. We measured experimental microwave signals by placing our sensor directly on top of cranberry-crop bed canopies in central Wisconsin and recording reflection coefficients across the operating band. We implemented a machine learning approach to map the microwave reflection coefficients to yield. The mapping procedure involves dimensionality reduction with principal component analysis, supervised learning with linear discriminant analysis, and error-correcting output codes. The idealized computational results demonstrate the feasibility of discriminating three cranberry volume fractions that are representative of central Wisconsin field conditions, at frequencies below 2 GHz. Performance evaluations of the machine learning algorithm applied to the measured field data indicated that, in 81% of test cases, the predicted crop yield had less than 8% error. Most importantly, the average yield prediction error was less than 1.3%. These pilot study results provide strong evidence that machine learning enables accurate cranberry yield estimation when trained with <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in situ</i> (field) microwave backscattered signals.

Regional winter wheat yield estimation based on the WOFOST model and a novel VW-4DEnSRF assimilation algorithm

To further improve the accuracy of regional crop yield estimation based on data assimilation, a novel EnSRF assimilation algorithm based on a variable time window and four-dimensional extension (VW-4DEnSRF) was proposed. In this research, taking Hengshui City of Hebei Province as the study area and winter wheat as the research crop, based on the WOFOST crop model and the proposed VW-4DEnSRF algorithm, a crop yield assimilation system was successfully constructed after parameter sensitivity analysis and parameter calibration of the crop model. Supported by the field-measured crop yield data and based on the effective validation of the yield assimilation system at a single point scale and in a typical experimental area, the scale optimization of grid size for regional yield estimation was effectively selected. Finally, combining the WOFOST model and inverted remotely sensed LAI, the regional winter wheat yield simulation under the optimal grid size of 500 m was carried out effectively through comparison with the field-measured yield data and official statistical yield data at the county level. Among them, the R2, adjusted R2 and RMSE between the simulated yield and ground-measured yield were 0.481, 0.471 and 801.4 kg.ha−1, respectively. The mean value of the estimated yield of winter wheat in Hengshui City was 6787 kg.ha−1, and the RMSE and RE between the estimated yield and official yield were 416.7 kg.ha−1 and 4.56%, respectively. These above results showed that the crop yield assimilation system based on the WOFOST model and proposed VW-4DEnSRF algorithm had good performances at both the single-point level and regional level, which proved that the proposed algorithm was feasible and effective at simulating crop yield over a large area

High-Density SRAM Read Access Yield Estimation Methodology

As high-density SRAMs must be designed to ensure a substantially small failure rate, the accurate yield estimation with practically acceptable runtime of circuit simulations is highly challenging. Here, a read access yield estimation method for high-density static random access memory (SRAM) is proposed. Instead of performing SPICE runs for the entire SRAM circuit, the proposed method partitions the SRAM into three parts—the control signal generation circuit, bitcell array, and sense amplifier (SA)—that determine three key parameters: word-line to SA enable delay, bit-line voltage difference, and SA offset voltage. Subsequently, the proposed method derives the probability density of these key parameters from each of the three partitioned circuits. Here, different methods are applied to derive the probability of the key parameters, considering the respective characteristics of each circuit part and parameter. According to our experimental results, the proposed method can accelerate the yield estimation by 500– $3000\times $ , compared with the brute-force Monte Carlo simulation method, and 10– $100\times $ compared with the other state-of-art methods. In addition, the proposed method can accelerate the circuit optimization procedure accompanied by multiple circuit revisions, that is, the circuit revisions can be reflected with SPICE runs only for the revised circuit part, unlike the previous methods that require SPICE runs for the entire SRAM.

Kalman filter based method for tracking dynamic transmission line parameters

Transmission line parameters are required by various power system applications. Thus, it is important to develop accurate transmission line parameter estimation methods. This paper puts forward Kalman filter based methods to track the dynamically changing line parameters. The proposed methods can overcome the estimation inaccuracy problem caused by the effect of the noise in the measurements. The proposed methods are based upon the equivalent- circuit of long line and utilize voltage and current data recorded by the Phasor Measurement Units at both ends of the line. Non-compensated and differently configured series compensated lines have been considered in this paper. The case studies demonstrate that the proposed methods have the capability of dynamically tracking changing line parameters and yield accurate estimates with the presence of measurement noises.

Estimation of Nitrogen Content and Marketable Yield Using Chlorophyll Readings

Accurate estimation of nitrogen content and yield is crucial and these can be predicted through chlorophyll readings as indicators. To explore the diagnosis model, a field experiment was conducted in a greenhouse at Yang ling, northwest China. The SPAD values results estimated the nitrogen status, total marketable yield, and rates of reduction in market yield. There was a positive correlation between SPAD value at the middle leaf (MD-SPAD) and nitrogen accumulation, and between SPAD value under the leaf (UD-SPAD) and nitrogen accumulation (R²=0.37-0.78, P<0.01, n=27). The MD-SPAD and UD-SPAD were positively correlated with total marketable yield during the growth period (R²=0.43-0.59, P<0.01, n=27). However, a poor correlation was observed at the upper leaf (UP-SPAD) in the expanding fruit period but was better in the mature period. Moreover, the nitrogen nutrition index (NNI) and MD-SPAD were positively correlated (R²=0.42-0.55, P<0.01, n=54). There SPAD values and reduction rates of ranks were positively correlated in marketable yield (R²=0.48-0.52, P<0.01, n=54), including middle fruit yield (MFY) and big fruit yield (BFY). This was similar between SPAD values and reduction rates of total marketable yield (R²=50, P<0.01, n=54). This indicated that nitrogen accumulation estimates including nitrogen accumulation in stem, leaf, fruit, and the overall nitrogen accumulation, as well as total marketable yield, is not accurate via UP-SPAD during fruit expanding period but it is opposite in mature period. When simultaneously considering estimation of nitrogen accumulation, NNI and marketable yield using chlorophyll readings, which were adopted by mean of SPAD values the different leaf positions or SPAD values at the middle leaf. On the other hand, UP-SPAD should be cautiously used to estimate nitrogen status, yield, and yield reduction rates. The overall marketable yield was optimal under W2 of 2488 m³hm^-2 and N2 of 300 kghm^-2. The reduction rates during middle fruit yield, big fruit yield, and total marketable yield ranged from 31%-38%. The big fruit yield was the highest, indicating that the big fruit yield is highly affected by severe water stress in the blossom and fruit period.

Integration of RGB-based vegetation index, crop surface model and object-based image analysis approach for sugarcane yield estimation using unmanned aerial vehicle

Estimation of yield is a major challenge in the production of many agricultural crops, including sugarcane. Mapping the spatial variability of plant height (PH) and the stalk density is important for accurate sugarcane yield estimation, and this estimation can aid in the planning of upcoming labor- and cost-intensive actions like harvesting, milling, and forward selling decisions. The objective of this research is to assess the potential of a consumer-grade red-green-blue (RGB) camera mounted on an unmanned aerial vehicle (UAV) for sugarcane yield estimation with minimal field dataset. The study mapped the spatial variability of PH and stalk density at the grid level (4 m × 4 m) on a farm. The average PH was estimated at the grid level by masking the sugarcane area. An object-based image analysis (OBIA) approach was used to extract the sugarcane area by integrating the plant height model (PHM), extracted by subtracting the digital elevation model (DEM) from the crop surface model (CSM). Both CSM and DEM were generated from UAV images, where CSM was produced approximately one month before the harvest and the DEM after the sugarcane was harvested. The PHM improved the overall accuracy of classification from 61.98% to 87.45%. The UAV estimated PH showed a high correlation (r = 0.95) with ground observed PH, with an average overestimation of 0.10 m. An ordinary least square (OLS) linear regression model was developed to estimate millable stalk height (MSH) from PH, weight from estimated MSH, and stalk density from vegetation indices (VIs) at the grid-level. Excess green (ExG) derived from RGB showed R2 of 0.754 with the stalk density. Likewise, R2 of 0.798 and 0.775 were obtained between MSH and PH, and weight and MSH. Eventually, the yield was estimated by integrating the variability of PH and stalk density and weight information. The estimated yield from ExG (200.66 tons) was close to the actual harvest yield (192.1 tons). The very high-resolution RGB-based images from the UAV and OBIA approach demonstrate significant potential for mapping the spatial variability of PH and stalk density and for estimating sugarcane yield. This can aid growers and millers in decision making.

Wheat yield predictions at a county and field scale with deep learning, machine learning, and google earth engine

To meet the challenges of climate change, increasing population and food demand, a timely, accurate and reliable estimation of crop yield at a large scale is more imperative than ever for crop management, food security evaluation, food trade and policy-making. In this study, taking the major winter wheat production regions of China as an example, we compared a traditional machine learning method (random forest, RF) and three deep learning (DL) models, including DNN (deep neural networks), 1D-CNN (1D convolutional neural networks), and LSTM (long short-term memory networks) to predict crop yields by integrating publicly available data within the GEE (Google Earth Engine) platform, including climate, satellite, soil properties, and spatial information data. The results showed that all four models could capture winter wheat yield variations in all the county-years, with R2 of recorded and simulated yields ranging from 0.83 to 0.90 and RMSE ranging from 561.18 to 959.62 kg/ha. They all performed well for winter wheat yield prediction at a county level from 2011 to 2015, with mean R2≥0.85 and RMSE ≤ 768 kg/ha. At a field level, the spatial pattern of estimated winter wheat yield could capture the spatial heterogeneity and yield differences between individual fields across a county fairly well. However, only the DNN and RF models had relatively good performance at the field level, with mean R2 values of 0.71, 0.66 and RMSE values of 1127 kg/ha and 956 kg/ha, respectively. The model comparisons showed that the performance of RF was not always worse than DL at both the county and field levels. Our findings demonstrated a scalable, simple and inexpensive framework for estimating crop yields at various scales in a timely manner and with reliable accuracy, which has important implications for crop yield forecasting, agricultural disaster monitoring, food trade policy, and food security warning.

A novel green apple segmentation algorithm based on ensemble U-Net under complex orchard environment

Fruit segmentation is a critical step for fruit recognition, and key to the efficiency of a robotic vision system for harvesting and accurate orchard yield estimation. Due to the unstructured characteristics of orchard environment, the segmentation performance of traditional segmentation algorithms is insufficient for handling green target fruits in a complex environment. Deep learning algorithms bring a fresh perspective to target segmentation. This study proposes an ensemble U-Net segmentation model suitable for small sample datasets. Edge structures are designed by integrating residual blocks and gated convolutions to obtain the boundary semantic information of the target image; atrous convolutions are applied to resolve the contradiction between the resolution of the feature map and the receiving field, retain more multiscale context information and achieve target fruit segmentation. An atrous spatial pyramid pooling (ASPP) structure is applied to merge the edge features and the high-level features of U-Net. The experimental results show that the proposed method effectively improves the segmentation accuracy of the target fruit and the generalization ability of the model. The proposed method extends the application scope of the harvesting robot and orchard yield measurements, thereby providing a theoretical reference for other fruit and vegetable target fruit segmentation efforts.

STT-MRAM Sensing: A Review

This brief presents a review of developments in spin-transfer-torque magnetoresistive random access memory (STT-MRAM) sensing over the past 20 years from a circuit design perspective. Various sensing schemes are categorized and described according to the data-cell variation-tolerant characteristics, pre-amplifiers, and offset tolerance. Key breakthroughs for achieving the optimal reference scheme, read disturbance prevention, read energy reduction, accurate yield estimation, and overcoming other non-idealities are discussed. This review is intended to facilitate further enhancement of STT-MRAM sensing in advanced technology nodes, thereby fulfilling STT-MRAM’s potential as a universal memory.

UAV Hyperspectral Remote Sensing Estimation of Soybean Yield Based on Physiological and Ecological Parameter and Meteorological Factor in China

Accurate soybean yield estimates are important for establishing effective agricultural and soybean trade policies aimed at ensuring food security in China. Unmanned aerial vehicles (UAV) have been used to collect hyperspectral remote sensing data for soybean yield inversion owing to the flexibility, efficiency, and low cost of the technology. Traditional yield inversion algorithms are based on physiological and ecological parameters. However, meteorological factors also influence crop growth and yield and can even eliminate differences caused by physiological and ecological parameters at certain growth stages. The paper proposed a new yield estimation parameter index of effective accumulated temperature, incorporating the influence of meteorological factors into yield inversion models, which can improve the accuracy and universality of crop yield estimation. In the current study, soybean yield inversion models were developed for leaf area index (LAI), fresh biomass, and in-season estimated yield (INSEY) at different growth stages. The accuracies of the various models were evaluated by comparing yield estimates with ground-based sampling data. The analyses revealed that plant parameters, including plant height, LAI, dry biomass, and fresh biomass, correlated significantly with soybean yields, and that LAI and fresh biomass demonstrated the highest correlations with soybean yields. The results of soybean yield estimation at the single growth stage showed that the R2 value of the model and validation increased by 50.00, 26.60, and 34.40%, and 47.20, 12.50, and 31.90%, respectively, at the flowering stage, pod stage, and seed filling stage, indicating the highest accuracy of yield estimation at the beginning seed stage. The analysis of yield estimation results of soybean at multiple growth stages showed that R2 values of model and validation of soybean yield estimation models at three different growth stages increased by 4.50, 7.20, and 15.80%, and 4.00, 5.30, and 8.90%, respectively, compared with the beginning seed stage with the highest yield estimation accuracy. Compared with the yield estimation model of the single growth stage, the yield estimation accuracy of multiple growth stages is higher.

Estimating soybean yield using time series of anomalies in vegetation indices from MODIS

ABSTRACT An accurate estimation of soybean yield while the plants are still in the field is highly necessary for industry applications and decision-making policies related to planning. Remote sensing is a powerful tool, due to its spatio-temporal coverage, for developing empirical models to predict and evaluate crop yields at regional and national scales. In Argentina, soybean (Glycine max (L.) Merr.) is the most important crop, particularly in Córdoba province the 89% of the sown area and 88% of the production is concentrated in eleven departments. The objectives of this work were to evaluate the performance of three vegetation indices: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI) and Normalized Differential Water Index (NDWI), from Moderate Resolution Imaging Spectroradiometer (MODIS), to explain the anomalies in the soybean yield at department-level in Córdoba, and to develop regressions models for estimate this variable using anomalies of these indices and average crop yield, considering time series of historical records and different sources of data. The results showed that the anomalies of the three vegetation indices fit, with very good precision, the anomalies of soybean yield (Pearson correlation coefficient values from 0.71 to 0.85). The evolution of the NDVI anomalies of mid-season crop development stage, for all periods considered, showed a similar pattern to yield anomalies, particularly differentiating years where droughts or highest soybean yields occurred, independently of data sources used. The regression models estimated soybean yield with NDVI anomalies, obtained prior to harvest, with % RMSE values between 8% and 17%. These simple and versatile models show that using free MODIS data, we can produce reasonable real-time estimates of soybean yield at department-level without previous crop classification.

Current rice models underestimate yield losses from short-term heat stresses.

Crop production will likely face enormous challenges against the occurrences of extreme climatic events projected under future climate change. Heat waves that occur at critical stages of the reproductive phase have detrimental impacts on the grain yield formation of rice (Oryza sativa). Accurate estimates of these impacts are essential to evaluate the effects of climate change on rice. However, the accuracy of these predictions by crop models has not been extensively tested. In this study, we evaluated 14 rice growth models against four year phytotron experiments with four levels of heat treatments imposed at different times after flowering. We found that all models greatly underestimated the negative effects of heat on grain yield, suggesting that yield projections with these models do not reflect food shocks that may occur under short-term extreme heat stress (SEHS). As a result, crop model ensembles do not help to provide accurate estimates of grain yield under heat stress. We examined the functions of grain-setting rate response to temperature (TRF_GS) used in eight models and showed that adjusting the effective periods of TRF_GS improved the model performance, especially for models simulating accumulative daily temperature effects. For TRF_GS which uses daily maximum temperature averaged for the effective period, the models provided better grain yield estimates by using maximum temperatures averaged only when daily maximum temperatures exceeded the base temperature (Tbase ). An alternative method based on heating-degree days and stage-dependent heat sensitivity parameters further decreased the prediction uncertainty of grain yield under heat stress, where stage-dependent heat sensitivity was more important than heat dose for model improvement under SEHS. These results suggest the limitation of the applicability of existing rice models to variable climatic conditions and the urgent need for an alternative grain-setting function accounting for the stage-dependent heat sensitivity.

Nondestructive estimation of potato yield using relative variables derived from multi-period LAI and hyperspectral data based on weighted growth stage

BackgroundThe accurate estimation of potato yield at regional scales is crucial for food security, precision agriculture, and agricultural sustainable development.MethodsIn this study, we developed a new method using multi-period relative vegetation indices (rVIs) and relative leaf area index (rLAI) data to improve the accuracy of potato yield estimation based on the weighted growth stage. Two experiments of field and greenhouse (water and nitrogen fertilizer experiments) in 2018 were performed to obtain the spectra and LAI data of the whole growth stage of potato. Then the weighted growth stage was determined by three weighting methods (improved analytic hierarchy process method, IAHP; entropy weight method, EW; and optimal combination weighting method, OCW) and the Slogistic model. A comparison of the estimation performance of rVI-based and rLAI-based models with a single and weighted stage was completed.ResultsThe results showed that among the six test rVIs, the relative red edge chlorophyll index (rCIred edge) was the optimal index of the single-stage estimation models with the correlation with potato yield. The most suitable single stage for potato yield estimation was the tuber expansion stage. For weighted growth stage models, the OCW-LAI model was determined as the best one to accurately predict the potato yield with an adjusted R2 value of 0.8333, and the estimation error about 8%.ConclusionThis study emphasizes the importance of inconsistent contributions of multi-period or different types of data to the results when they are used together, and the weights need to be considered.

Adaptively Weighted Yield-Driven EM Optimization Incorporating Neurotransfer Function Surrogate With Applications to Microwave Filters

In this article, we propose an adaptively weighted yield-driven EM optimization technique incorporating neurotransfer function (neuro-TF) surrogate. In the proposed technique, an adaptive weighting factor is set up for each frequency point of interest based on the degree to which the EM response may violate the design specification. These weighting factors are first incorporated into the error function for training the neuro-TF model and then involved in the objective function of yield optimization using the trained model. We identify the key frequency points where the EM response is likely to violate the design specification over the whole frequency range of interest. Using the adaptive weighting factor-incorporated error function to train the model enhances the model accuracy at the key frequency points while preserving the model accuracy at other ordinary frequency points. This improves the yield estimation accuracy using the trained surrogate model at each iteration of optimization and, consequently, facilitates the yield optimization process. By involving the weighting factors into the formulation of the objective function of neuro-TF-assisted yield optimization, higher priorities are given to the key frequency points than the ordinary frequency points. This allows the proposed technique to find a more effective update direction at each iteration of optimization and, consequently, achieves a similar yield increase with a fewer number of iterations compared with the conventional neuro-TF approach. Two microwave examples demonstrate the advantages of the proposed technique against other existing approaches, including the Monte Carlo (MC)-based approach, the polynomial chaos (PC)-based approach, and the conventional neuro-TF approach.

ESTIMATING CROP YIELDS WITH REMOTE SENSING AND DEEP LEARNING

Abstract. Increasing the accuracy of crop yield estimates may allow improvements in the whole crop production chain, allowing farmers to better plan for harvest, and for insurers to better understand risks of production, to name a few advantages. To perform their predictions, most current machine learning models use NDVI data, which can be hard to use, due to the presence of clouds and their shadows in acquired images, and due to the absence of reliable crop masks for large areas, especially in developing countries. In this paper, we present a deep learning model able to perform pre-season and in-season predictions for five different crops. Our model uses crop calendars, easy-to-obtain remote sensing data and weather forecast information to provide accurate yield estimates.