Plant species classification using deep convolutional neural network

Dyrmann, Mads; Karstoft, Henrik; Midtiby, Henrik Skov

doi:10.1016/j.biosystemseng.2016.08.024

Cited by 528 publications

(279 citation statements)

References 8 publications

Supporting

Mentioning

268

Contrasting

Unclassified

Order By: Relevance

“…Deep learning may also cope with shadows using the shadows as additional species-specific structure information. Nevertheless, deep neural networks usually need a large amount of training data which could hamper their use for practical applications with limited field data (Dyrmann et al 2016). Other alternatives to address shadow effects are shadow correction methods, which consist in the radiometric enhancement of shaded pixels usually based on information extracted from neighboring non-shadowed regions (empirical methods; e.g.…”

Section: Introductionmentioning

confidence: 99%

How canopy shadow affects invasive plant species classification in high spatial resolution remote sensing

Lopatin

Dološ

Kattenborn

2019

Remote Sens Ecol Conserv

View full text Add to dashboard Cite

Plant invasions can result in serious threats for biodiversity and ecosystem functioning. Reliable maps at very‐high spatial resolution are needed to assess invasions dynamics. Field sampling approaches could be replaced by unmanned aerial vehicles (UAVs) to derive such maps. However, pixel‐based species classification at high spatial resolution is highly affected by within‐canopy variation caused by shadows. Here, we studied the effect of shadows on mapping the occurrence of invasive species using UAV‐based data. MaxEnt one‐class classifications were applied to map Acacia dealbata, Ulex europaeus and Pinus radiata in central‐south Chile using combinations of UAV‐based spectral (RGB and hyperspectral), 2D textural and 3D structural variables including and excluding shaded canopy pixels during model calibration. The model accuracies in terms of area under the curve (AUC), Cohen's Kappa, sensitivity (true positive rate) and specificity (true negative rate) were examined in sunlit and shaded canopies separately. Bootstrapping was used for validation and to assess statistical differences between models. Our results show that shadows significantly affect the accuracies obtained with all types of variables. The predictions in shaded areas were generally inaccurate, leading to misclassification rates between 65% and 100% even when shadows were included during model calibration. The exclusion of shaded areas from model calibrations increased the predictive accuracies (especially in terms of sensitivity), decreasing false positives. Spectral and 2D textural information showed generally higher performances and improvements when excluding shadows from the analysis. Shadows significantly affected the model results obtained with any of the variables used, hence the exclusion of shadows is recommended prior to model calibration. This relatively easy pre‐processing step enhances models for classifying species occurrences using high‐resolution spectral imagery and derived products. Finally, a shadow simulation showed differences in the ideal acquisition window for each species, which is important to plan revisit campaigns.

show abstract

Section: Introductionmentioning

confidence: 99%

How canopy shadow affects invasive plant species classification in high spatial resolution remote sensing

Lopatin

Dološ

Kattenborn

2019

Remote Sens Ecol Conserv

View full text Add to dashboard Cite

show abstract

“…The CNN methodology was also enhanced for recognizing plant species. A CNN model was built over 10 413 color images . It was noted that the performance of CNN is capable of distinguishing 22 weed and crop species with an accuracy of 86.2%.…”

Section: Introductionmentioning

confidence: 99%

Identification of wheat kernels by fusion of RGB, SWIR, and VNIR samples

2019

View full text Add to dashboard Cite

Background The sustainable management of agricultural resources requires the integration of cutting‐edge science with the observation and identification of crops. This assists experts to make correct decisions. The aim of this study is to assess the robustness of a commonly used deep learning tool, VGG16, in improving the categorization of wheat kernels. Two fusion methodologies were considered simultaneously. We performed experiments on visible light (RGB), short wave infrared (SWIR), and visible‐near infrared (VNIR) datasets, including 40 classes, with 200 samples in each class, giving 8000 samples in total. Results After making simulations with 6400 training and 1600 testing samples, we achieved excellent performance scores, with 98.19% and 100% accuracy rates, respectively. Conclusion The wheat identification system developed here serves as an effective identification framework and supports the view that deep learning tools can adequately discriminate between different types of wheat kernels. The proposed automated system would be useful for improving economic growth and in reducing the labor force, leading to greater efficiency and higher productivity in the wheat industry. © 2019 Society of Chemical Industry

show abstract

“…The main advantage of CNN is the high performance in object detection and automated feature engineering. CNN models such as Inception‐v3 (Szegedy, Vanhoucke, Ioffe, Shlens, & Wojna, ), GoogleNet (Szegedy et al, ), DenseNet (Huang, Liu, Van Der Maaten, & Weinberger, ), and customized models were proven effective in crop/weed detection and classification even with uncontrolled illumination (Dyrmann, Jørgensen, & Midtiby, ; Dyrmann, Karstoft, & Midtiby, ; McCool, Perez, & Upcroft, ; Milioto, Lottes, & Stachniss, ; Potena, Nardi, & Pretto, ).…”

Section: Introductionmentioning

confidence: 99%

“…The main advantage of CNN is the high performance in object detection and automated feature engineering. CNN models such as Inception-v3 (Szegedy, Vanhoucke, Ioffe, Shlens, & Wojna, 2015), GoogleNet F I G U R E 1 The actuators (left) were designed as an implement of a tractor (right), employing rotating vertical tines as the weeding tool for effectively cutting, uprooting, and burying weeds [Color figure can be viewed at wileyonlinelibrary.com] (Szegedy et al, 2014), DenseNet (Huang, Liu, Van Der Maaten, & Weinberger, 2017), and customized models were proven effective in crop/weed detection and classification even with uncontrolled illumination (Dyrmann, Jørgensen, & Midtiby, 2017;Dyrmann, Karstoft, & Midtiby, 2016;Milioto, Lottes, & Stachniss, 2017;Potena, Nardi, & Pretto, 2017). CNN approaches, however, also face several challenges.…”

mentioning

confidence: 99%

Automated crop plant detection based on the fusion of color and depth images for robotic weed control

Gai

Tang

Steward

2019

Journal of Field Robotics

View full text Add to dashboard Cite

Robotic weeding enables weed control near or within crop rows automatically, precisely and effectively. A computer‐vision system was developed for detecting crop plants at different growth stages for robotic weed control. Fusion of color images and depth images was investigated as a means of enhancing the detection accuracy of crop plants under conditions of high weed population. In‐field images of broccoli and lettuce were acquired 3–27 days after transplanting with a Kinect v2 sensor. The image processing pipeline included data preprocessing, vegetation pixel segmentation, plant extraction, feature extraction, feature‐based localization refinement, and crop plant classification. For the detection of broccoli and lettuce, the color‐depth fusion algorithm produced high true‐positive detection rates (91.7% and 90.8%, respectively) and low average false discovery rates (1.1% and 4.0%, respectively). Mean absolute localization errors of the crop plant stems were 26.8 and 7.4 mm for broccoli and lettuce, respectively. The fusion of color and depth was proved beneficial to the segmentation of crop plants from background, which improved the average segmentation success rates from 87.2% (depth‐based) and 76.4% (color‐based) to 96.6% for broccoli, and from 74.2% (depth‐based) and 81.2% (color‐based) to 92.4% for lettuce, respectively. The fusion‐based algorithm had reduced performance in detecting crop plants at early growth stages.

show abstract

Plant species classification using deep convolutional neural network

Cited by 528 publications

References 8 publications

How canopy shadow affects invasive plant species classification in high spatial resolution remote sensing

How canopy shadow affects invasive plant species classification in high spatial resolution remote sensing

Identification of wheat kernels by fusion of RGB, SWIR, and VNIR samples

Automated crop plant detection based on the fusion of color and depth images for robotic weed control

Contact Info

Product

Resources

About