Convolutional Neural Networks to Estimate Dry Matter Yield in a Guineagrass Breeding Program Using UAV Remote Sensing

Oliveira, Gabriel Silva de; Marcato, José; Polidoro, Caio H. S.; Osco, Lucas Prado; Siqueira, Henrique; Rodrigues, Lucas de Souza; Jank, Liana; Barrios, S. C. L.; Valle, Cacilda Borges do; Simeão, Rosângela Maria; Carromeu, Camilo; Silveira, Eloise; Jorge, Lúcio André de Castro; Gonçalves, Wesley Nunes; Santos, M. F.; Matsubara, Edson Takashi

doi:10.3390/s21123971

Cited by 18 publications

(14 citation statements)

References 38 publications

(55 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, we recognize many other species, from herbaceous annual crops (Grenzdörffer, 2019) all the way up to perennial trees (López‐Granados et al., 2019), are using UAS tools to enhance their small‐plot research and breeding. Forage grasses and silage provide an especially interesting case to look at breeding for the aboveground biomass throughout the season (Alvarez‐Mendoza et al., 2022; de Oliveira et al., 2021; Nakasagga et al., 2022), while cassava ( Manihot esculenta Crantz), peanut ( Arachis hypogaea ), and potato ( Solanum tuberosum L.) UAS research (de Jesus Colwell et al., 2021; Sarkar et al., 2021; Selvaraj et al., 2020) has shown that below ground biomass yield can be estimated from UAS. Each of these species had to develop a unique methodology.…”

Section: Discussionmentioning

confidence: 99%

Unoccupied aerial systems imagery for phenotyping in cotton, maize, soybean, and wheat breeding

et al. 2023

View full text Add to dashboard Cite

High‐throughput phenotyping (HTP) with unoccupied aerial systems (UAS), consisting of unoccupied aerial vehicles (UAV; or drones) and sensor(s), is an increasingly promising tool for plant breeders and researchers. Enthusiasm and opportunities from this technology for plant breeding are similar to the emergence of genomic tools ∼30 years ago, and genomic selection more recently. Unlike genomic tools, HTP provides a variety of strategies in implementation and utilization that generate big data on the dynamic nature of plant growth formed by temporal interactions between growth and environment. This review lays out strategies deployed across four major staple crop species: cotton (Gossypium hirsutum L.), maize (Zea mays L.), soybean (Glycine max L.), and wheat (Triticum aestivum L.). Each crop highlighted in this review demonstrates how UAS‐collected data are employed to automate and improve estimation or prediction of objective phenotypic traits. Each crop section includes four major topics: (a) phenotyping of routine traits, (b) phenotyping of previously infeasible traits, (c) sample cases of UAS application in breeding, and (d) implementation of phenotypic and phenomic prediction and selection. While phenotyping of routine agronomic and productivity traits brings advantages in time and resource optimization, the most potentially beneficial application of UAS data is in collecting traits that were previously difficult or impossible to quantify, improving selection efficiency of important phenotypes. In brief, UAS sensor technology can be used for measuring abiotic stress, biotic stress, crop growth and development, as well as productivity. These applications and the potential implementation of machine learning strategies allow for improved prediction, selection, and efficiency within breeding programs, making UAS HTP a potentially indispensable asset.

show abstract

Section: Discussionmentioning

confidence: 99%

Unoccupied aerial systems imagery for phenotyping in cotton, maize, soybean, and wheat breeding

et al. 2023

View full text Add to dashboard Cite

show abstract

“…As for the difference in the average latency of both frame-based object detectors, we note that YOLOv3's 12 architecture utilizes the DarkNet-53 feature encoder as its backbone, which has 42 million parameters. 51 Meanwhile, SSD-300's architecture uses VGG-19 19 instead, which has more than 143 million parameters, thus justifying the 134-ms variance in their average latency, and the difference in their singlemodal object detection and tracking performance (at 24 Hz) as demonstrated in Tables 1 and 2. In contrast, event-based methods presented have shown at least an order of magnitude less latency than their frame-based counterpart.…”

Section: Computational Latency Analysismentioning

confidence: 99%

High-temporal-resolution event-based vehicle detection and tracking

Shair

Rawashdeh

2022

Opt. Eng.

View full text Add to dashboard Cite

Event-based vision has been rapidly growing in recent years justified by the unique characteristics, such as its high temporal resolutions (∼1 μs), high dynamic range (>120 dB), and output latency of only a few microseconds. Our work further explores a hybrid, multimodal approach for object detection and tracking that leverages state-of-the-art frame-based detectors complemented by hand-crafted event-based methods to improve the overall tracking performance with minimal computational overhead. The methods presented include event-based bounding box (BB) refinement that improves the precision of the resulting BBs, as well as a continuous event-based object detection method, to recover missed detections and generate interframe detections that enable a high-temporal-resolution tracking output. The advantages of these methods are quantitatively verified by an ablation study using the higher order tracking accuracy (HOTA) metric. Results show significant performance gains resembled by an improvement in the HOTA from 56.6%, using only frames, to 64.1% and 64.9%, for the event and edgebased mask configurations combined with the two methods proposed, at the baseline frame rate of 24 Hz. Likewise, incorporating these methods with the same configurations has improved HOTA from 52.5% to 63.1% and from 51.3% to 60.2% at the high-temporal-resolution tracking rate of 384 Hz. Finally, a validation experiment is conducted to analyze the real-world singleobject tracking performance using high-speed LiDAR. Empirical evidence shows that our approaches provide significant advantages compared to using frame-based object detectors at the baseline frame rate of 24 Hz and higher tracking rates of up to 500 Hz.

show abstract

“…To address the need for data, several methods have been proposed to train robust models with a limited amount of labeled data. One approach is to use pre-trained models with transfer learning, which has been successful in estimating forage biomass ( Castro et al., 2020 ; de Oliveira et al., 2021 ). However, when dealing with multispectral images, pre-trained models that are generally trained on RGB images may not perform well.…”

Section: Introductionmentioning

confidence: 99%

Comparing CNNs and PLSr for estimating wheat organs biophysical variables using proximal sensing

Carlier,

Dandrifosse,

Dumont

et al. 2023

Front. Plant Sci.

View full text Add to dashboard Cite

Estimation of biophysical vegetation variables is of interest for diverse applications, such as monitoring of crop growth and health or yield prediction. However, remote estimation of these variables remains challenging due to the inherent complexity of plant architecture, biology and surrounding environment, and the need for features engineering. Recent advancements in deep learning, particularly convolutional neural networks (CNN), offer promising solutions to address this challenge. Unfortunately, the limited availability of labeled data has hindered the exploration of CNNs for regression tasks, especially in the frame of crop phenotyping. In this study, the effectiveness of various CNN models in predicting wheat dry matter, nitrogen uptake, and nitrogen concentration from RGB and multispectral images taken from tillering to maturity was examined. To overcome the scarcity of labeled data, a training pipeline was devised. This pipeline involves transfer learning, pseudo-labeling of unlabeled data and temporal relationship correction. The results demonstrated that CNN models significantly benefit from the pseudolabeling method, while the machine learning approach employing a PLSr did not show comparable performance. Among the models evaluated, EfficientNetB4 achieved the highest accuracy for predicting above-ground biomass, with an R² value of 0.92. In contrast, Resnet50 demonstrated superior performance in predicting LAI, nitrogen uptake, and nitrogen concentration, with R² values of 0.82, 0.73, and 0.80, respectively. Moreover, the study explored multi-output models to predict the distribution of dry matter and nitrogen uptake between stem, inferior leaves, flag leaf, and ear. The findings indicate that CNNs hold promise as accessible and promising tools for phenotyping quantitative biophysical variables of crops. However, further research is required to harness their full potential.

show abstract

Convolutional Neural Networks to Estimate Dry Matter Yield in a Guineagrass Breeding Program Using UAV Remote Sensing

Cited by 18 publications

References 38 publications

Unoccupied aerial systems imagery for phenotyping in cotton, maize, soybean, and wheat breeding

Unoccupied aerial systems imagery for phenotyping in cotton, maize, soybean, and wheat breeding

High-temporal-resolution event-based vehicle detection and tracking

Comparing CNNs and PLSr for estimating wheat organs biophysical variables using proximal sensing

Contact Info

Product

Resources

About