Assessment of soybean injury from glyphosate using airborne multispectral remote sensing

Differentiate Soybean Response to Off-Target Dicamba Damage Based on UAV Imagery and Machine Learning

The wide adoption of dicamba-tolerant (DT) soybean has led to numerous cases of off-target dicamba damage to non-DT soybean and dicot crops. This study aimed to develop a method to differentiate soybean response to dicamba using unmanned-aerial-vehicle-based imagery and machine learning models. Soybean lines were visually classified into three classes of injury, i.e., tolerant, moderate, and susceptible to off-target dicamba. A quadcopter with a built-in RGB camera was used to collect images of field plots at a height of 20 m above ground level. Seven image features were extracted for each plot, including canopy coverage, contrast, entropy, green leaf index, hue, saturation, and triangular greenness index. Classification models based on artificial neural network (ANN) and random forest (RF) algorithms were developed to differentiate the three classes of response to dicamba. Significant differences for each feature were observed among classes and no significant differences across fields were observed. The ANN and RF models were able to precisely distinguish tolerant and susceptible lines with an overall accuracy of 0.74 and 0.75, respectively. The imagery-based classification model can be implemented in a breeding program to effectively differentiate phenotypic dicamba response and identify soybean lines with tolerance to off-target dicamba damage.

Multi-Modal Deep Learning for Weeds Detection in Wheat Field Based on RGB-D Images

Single-modal images carry limited information for features representation, and RGB images fail to detect grass weeds in wheat fields because of their similarity to wheat in shape. We propose a framework based on multi-modal information fusion for accurate detection of weeds in wheat fields in a natural environment, overcoming the limitation of single modality in weeds detection. Firstly, we recode the single-channel depth image into a new three-channel image like the structure of RGB image, which is suitable for feature extraction of convolutional neural network (CNN). Secondly, the multi-scale object detection is realized by fusing the feature maps output by different convolutional layers. The three-channel network structure is designed to take into account the independence of RGB and depth information, respectively, and the complementarity of multi-modal information, and the integrated learning is carried out by weight allocation at the decision level to realize the effective fusion of multi-modal information. The experimental results show that compared with the weed detection method based on RGB image, the accuracy of our method is significantly improved. Experiments with integrated learning shows that mean average precision (mAP) of 36.1% for grass weeds and 42.9% for broad-leaf weeds, and the overall detection precision, as indicated by intersection over ground truth (IoG), is 89.3%, with weights of RGB and depth images at α = 0.4 and β = 0.3. The results suggest that our methods can accurately detect the dominant species of weeds in wheat fields, and that multi-modal fusion can effectively improve object detection performance.

Dicamba Injury on Soybean Assessed Visually and with Spectral Vegetation Index

The recent availability of soybean cultivars with resistance to dicamba herbicide has increased the risk of injury in susceptible cultivars, mainly as a result of particle drift. To predict and identify the damage caused by this herbicide requires great accuracy. The objective of this work was to evaluate the injury caused by the simulated drift of dicamba on soybean (nonresistant to dicamba) plants assessed visually and using the Triangular Greenness Index (TGI) from images obtained from Remotely Piloted Aircraft (RPA). The study was conducted in a randomized complete block design with four replications during the 2019/2020 growing season, and the treatments consisted of the application of six doses of dicamba (0, 0.28, 0.56, 5.6, 28, and 112 g acid equivalent dicamba ha−1) on soybean plants at the third node growth stage. For the evaluation of treatments using the TGI technique, spectral data acquired through a Red Green Blue (RGB) sensor attached to an RPA was used. The variables studied were the visual estimation of injury, TGI response at 7 and 21 days after application, plant height, and crop yield. The exposure to the herbicide caused a reduction in plant height and crop yield. Vegetation indices, such as TGI, have the potential to be used in the evaluation of injury caused by dicamba, and may be used to cover large areas in a less subjective way than visual assessments.

Sensor-based evaluation of maize (Zea mays) and weed response to post-emergence herbicide applications of Isoxaflutole and Cyprosulfamide applied as crop seed treatment or herbicide mixing partner

BACKGROUND

Some maize post-emergence herbicides obtain their crop/weed selectivity only through the use of chemical crop safeners. Safeners improve the tolerance of maize to herbicidal active ingredients. In order to investigate the crop response to safener (cyprosulfamide) spray application and seed treatment, greenhouse and field trials were conducted on three maize development stages (2-, 4-, and 6-leaf stage). Visual estimations on crop vitality were compared to ground-based and airborne hyperspectral and multispectral sensors.

RESULTS

The reduction of cyprosulfamide by 88% when applied as seed treatment did not significantly reduce maize biomass yields at the field. The crop deterioration in both trials was stronger in the cyprosulfamide seed treatments compared to the spray applications but was found to be transient in the field trial. The hyperspectral sensor and multispectral camera data correlated with R² = 0.84 (CropSpec Vegetation Index) and R² = 0.64 (Green Normalized Difference Vegetation Index).

CONCLUSION

The sensor-based collection of crop responses to treatments enables early, quantifiable and auditor-independent assessments. In particular, the airborne multispectral imagery assessment of field experiments provides more detailed and comprehensive information than visually collected data. © 2019 The Authors. Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

Assessing crop damage from dicamba on non-dicamba-tolerant soybean by hyperspectral imaging through machine learning

BACKGROUND

Dicamba effectively controls several broadleaf weeds. The off-target drift of dicamba spray or vapor drift can cause severe injury to susceptible crops, including non-dicamba-tolerant crops. In a field experiment, advanced hyperspectral imaging (HSI) was used to study the spectral response of soybean plants to different dicamba rates, and appropriate spectral features and models for assessing the crop damage from dicamba were developed.

RESULTS

In an experiment with six different dicamba rates, an ordinal spectral variation pattern was observed at both 1 week after treatment (WAT) and 3 WAT. The soybean receiving a dicamba rate ≥0.2X exhibited unrecoverable damage. Two recoverability spectral indices (HDRI and HDNI) were developed based on three optimal wavebands. Based on the Jeffries-Matusita distance metric, Spearman correlation analysis and independent t-test for sensitivity to dicamba spray rates, a number of wavebands and classic spectral features were extracted. The models for quantifying dicamba spray levels were established using the machine learning algorithms of naive Bayes, random forest and support vector machine.

CONCLUSIONS

The spectral response of soybean injury caused by dicamba sprays can be clearly captured by HSI. The recoverability spectral indices developed were able to accurately differentiate the recoverable and unrecoverable damage, with an overall accuracy (OA) higher than 90%. The optimal spectral feature sets were identified for characterizing dicamba spray rates under recoverable and unrecoverable situations. The spectral features plus plant height can yield relatively high accuracy under the recoverable situation (OA = 94%). These results can be of practical importance in weed management. © 2019 Society of Chemical Industry.

Monitoring Glyphosate-Based Herbicide Treatment Using Sentinel-2 Time Series—A Proof-of-Principle

In this paper we aim to show a proof-of-principle approach to detect and monitor weed management using glyphosate-based herbicides in agricultural practices. In a case study in Germany, we demonstrate the application of Sentinel-2 multispectral time-series data. Spectral broadband vegetation indices were analysed to observe vegetation traits and weed damage arising from herbicide-based management. The approach has been validated with stakeholder information about herbicide treatment using commercial products. As a result, broadband NDVI calculated from Sentinel-2 data showed explicit feedback after the glyphosate-based herbicide treatment. Vegetation damage could be detected after just two days following of glyphosate-based herbicide treatment. This trend was observed in three different application scenarios, i.e., during growing stage, before harvest and after harvest. The findings of the study demonstrate the feasibility of satellite based broadband NDVI data for the detection of glyphosate-based herbicide treatment and, e.g., the monitoring of latency to harvesting. The presented results can be used to implement monitoring concepts to provide the necessary transparency about weed treatment in agricultural practices and to support environmental monitoring.

Optimizing band selection for spectral detection of Aphis glycines Matsumura in soybean

BACKGROUND

Soybean aphid, Aphis glycines Matsumura (Hemiptera: Aphididae), is a significant insect pest of soybean in North America. Accurate estimation of A. glycines densities requires costly, time-intensive weekly counts of adults and nymphs on plants. Field studies were conducted in 2013 and 2014 to assess the potential for spectral-based remote sensing to more efficiently quantify cumulative aphid-days (CADs) using soybean canopy reflectance.

RESULTS

Narrow-band wavelengths in the near-infrared spectral range were associated with CAD, but those in the visible spectral range were not associated with CAD. Simple linear regression models of CAD on reflectance were generally better than quadratic and cubic regression models. Simulated wide-band sensors centered at 740-1100 nm yielded better regression models than ones centered at 600-740 nm, regardless of bandwidth. Among the simulated wide-band sensors, increasing sensor bandwidth worsened CAD estimation or required more simulated sensors to optimize CAD estimation. Optimal combinations of spectral bands explained 83-96% of the experimentally manipulated variation in CAD.

CONCLUSION

Near-infrared wavelengths at 780 ± 50 nm can effectively estimate A. glycines abundance on soybean. Our approach of simulating wide-band multispectral sensors from ground-based hyperspectral data helped to refine spectral sensors and holds potential to reduce the cost and complexity of treat/no-treat classification tasks. This study will contribute to future research aiming to quantify insect injury using customized commercial-grade sensors for detection, quantification, and differentiation of A. glycines from other stressors. © 2018 Society of Chemical Industry.