Machine vision–based herbicide applications relying on object detection or image classification deep convolutional neural networks (DCNNs) demand high memory and computational resources, resulting in lengthy inference times. To tackle these challenges, this study assessed the effectiveness of three teacher models, each trained on datasets of varying sizes, including D-20k (comprising 10,000 true-positive and true-negative images) and D-10k (comprising 5,000 true-positive and true-negative images). Additionally, knowledge distillation was performed on their corresponding student models across a range of temperature settings. After the process of student–teacher learning, the parameters of all student models were reduced. ResNet18 not only achieved higher accuracy (ACC ≥ 0.989) but also maintained higher frames per second (FPS ≥ 742.9) under its optimal temperature condition (T = 1). Overall, the results suggest that employing knowledge distillation in the machine vision models enabled accurate and reliable weed detection in turf while reducing the need for extensive computational resources, thereby facilitating real-time weed detection and contributing to the development of smart, machine vision–based sprayers.

Johnsongrass [Sorghum halepense (L.) Pers.] is one of the most problematic perennial grass weed species in row-crop production across the southern United States. Control of this species is especially challenging in organic systems due to a lack of effective options. A field experiment was conducted at the Texas A&M research farm near College Station, TX, from fall 2019 to spring 2021 to evaluate various nonchemical options for managing S. halepense in the fallow season, implemented over 2 yr in the same locations. The treatments included disking once, disking twice, disking + immediate flooding, disking + flush irrigation + flooding, disking twice + flooding after the first frost, periodic mowing, acetic acid treatment, and disking + tarping. Disking + immediate flooding, disking + flush irrigation + flooding, and disking + tarping were the most effective treatments. Compared with the nontreated control plots, these treatments reduced S. halepense aboveground density (<9 plants m−2 vs. 64 plants m−2), aboveground biomass (<80 g m−2 vs. 935 g m−2), rhizome biomass (<4 g m−2 vs. 55 g m−2), rhizome node number (<25 nodes m−2 vs. 316 nodes m−2), and rhizome length (<42 cm m−2 vs. 660 cm m−2). Disking twice + flooding after the first frost did not show a consistent impact. Periodic mowing also reduced S. halepense density (12 plants m−2 vs. 64 plants m−2) and other variables compared with the control plots at the end of the study in spring 2021. Disking alone once or twice each growing season or repeated application of acetic acid failed to control S. halepense. These results indicate that well-timed nonchemical management practices such as tarping and flooding implemented during the winter fallow can be very effective in reducing S. halepense densities.

In this research, the deep-learning optimizers Adagrad, AdaDelta, Adaptive Moment Estimation (Adam), and Stochastic Gradient Descent (SGD) were applied to the deep convolutional neural networks AlexNet, GoogLeNet, VGGNet, and ResNet that were trained to recognize weeds among alfalfa using photographic images taken at 200×200, 400×400, 600×600, and 800×800 pixels. An increase in the image sizes reduced the classification accuracy of all neural networks. The neural networks that were trained with images of 200×200 pixels resulted in better classification accuracy than the other image sizes investigated here. The optimizers AlexNet and GoogLeNet trained with AdaDelta and SGD outperformed the Adagrad and Adam optimizers; VGGNet trained with AdaDelta outperformed Adagrad, Adam, and SGD; and ResNet trained with AdaDelta and Adagrad outperformed the Adam and SGD optimizers. When the neural networks were trained with the best-performing input image size (200×200 pixels) and the best-performing deep learning optimizer, VGGNet was the most effective neural network, with high precision and recall values (≥0.99) when validation and testing datasets were used. Alternatively, ResNet was the least effective neural network in its ability to classify images containing weeds. However, there was no difference among the different neural networks in their ability to differentiate between broadleaf and grass weeds. The neural networks discussed herein may be used for scouting weed infestations in alfalfa and further integrated into the machine vision subsystem of smart sprayers for site-specific weed control.

Wood vinegar, a product of pyrolysis, can induce phytotoxicity on plants when applied at an adequate rate and concentration. The objective of this research was to investigate wood vinegar obtained from the pyrolysis of apple tree branches for weed control in dormant zoysiagrass. In environment-controlled growth chambers, white clover visual injury and shoot mass reduction were evaluated and compared to the nontreated control after wood vinegar application at 1,000, 2,000, or 4,000 L ha−1 under 10 C or 30 C temperature conditions. Averaged across rates, wood vinegar rapidly desiccated white clover and caused 83% and 71% visual injury at 10 C and 30 C, respectively, at 1 d after treatment (DAT). Averaged across temperatures, wood vinegar at 1,000, 2,000, and 4,000 L ha−1 reduced white clover shoot mass by 56%, 81%, and 98% from the nontreated control at 10 DAT, respectively. In field experiments, weed control increased as wood vinegar rates increased from 1,000 to 5,000 L ha−1 in dormant zoysiagrass. The effective application dose of wood vinegar required to provide 90% control (ED90) of annual fleabane, Persian speedwell, and white clover was determined to be 2,450, 2,300, and 4,020 L ha−1, respectively, at 2 wk after treatment. Turf quality did not differ among the wood vinegar treatments and the nontreated control when zoysiagrass completely recovered from dormancy. Overall, results illustrate that wood vinegar resulting from the pyrolysis of apple tree branches can be used as a nonselective herbicide in dormant turfgrass, offering a new nonsynthetic herbicide option for weed control in managed turf.

Sulfuryl fluoride (SF) is currently used as a fumigant for control of drywood termites and insects in building structures, vehicles, wood products, postharvest commodities, and food processing facilities. This research investigated the feasibility of using SF as a preplant soil fumigant for purple nutsedge control in plastic-mulched tomato production. SF treatments included SF injected through drip tapes or SF injected through drip tapes a few hours following shank injection of chloropicrin (Pic). Results revealed that SF alone at 224, 336, or 448 kg ha−1 was generally less effective compared with when it was applied in conjunction with Pic at 168 kg ha−1. SF alone provided inconsistent control of purple nutsedge. In contrast, SF + Pic was as efficacious or more efficacious on purple nutsedge than the industry standards, including 1,3-dichloropropene (1,3-D) plus Pic and metam potassium. None of the fumigant treatments visually injured tomato plants, stunted growth, or adversely affected tomato yield. In one of the four tomato seasons, tomato plants growing in plots fumigated with SF + Pic resulted in taller tomato plants and higher markable yields. Results indicate that soil fumigation with SF + Pic is safe on plastic-mulched tomato and effectively controls purple nutsedge.

Resistance to penoxsulam among barnyardgrass populations is prevalent in rice fields in China. Seeds of penoxsulam-resistant (AXXZ-2) and penoxsulam-susceptible (JLGY-3) barnyardgrass populations, as well as the seeds of two rice varieties, including Wuyungeng32 (WY) and Liangyou669 (LY), were planted in plastic pots and then treated with a rate titration of acetochlor, anilofos, butachlor, clomazone, oxadiazon, pendimethalin, pretilachlor, pyraclonil, or thiobencarb. The two barnyardgrass populations exhibited similar susceptibility to acetochlor, anilofos, butachlor, oxadiazon, pretilachlor, or pyraclonil. However, the susceptibility differed between the barnyardgrass populations in response to clomazone, pendimethalin, and thiobencarb. For AXXZ-2, herbicide rates that caused 50% reduction in shoot biomass from the nontreated control (GR50) were 179, >800, and 1,798 g ha−1 for clomazone, pendimethalin, and thiobencarb, respectively; whereas JLGY-3 GR50 values were 61, 166, and 552 g ha−1, respectively. Both rice varieties demonstrated excellent tolerance to acetochlor, butachlor, oxadiazon, pretilachlor, and thiobencarb. However, substantial rice damage was observed when anilofos and clomazone were used. Anilofos at 352 g ha−1 and clomazone at 448 g ha−1 reduced rice shoot biomass by 41% and 50% from the nontreated, respectively. Averaged across herbicide rates, clomazone use resulted in a reduction in rice shoot biomass from that of the nontreated control by 52% and 34% for WY and LY, respectively; and pendimethalin use resulted in a reduction in rice shoot biomass from the nontreated control by 25% and 9% for WY and LY, respectively.

Multi-device radio frequency power amplifiers (PAs) often exhibit strongly non-linear behavior in combination with long-term memory effects, leading to an extremely challenging model development cycle. This paper presents a new, dynamic, behavioral modeling technique, based on a combination of the real-valued decomposed piecewise method and concepts from the field of machine learning. The underlying theory of the proposed modeling technique is provided, along with a detailed modeling procedure. Experimental results show that the proposed decomposed piecewise support vector regression (SVR) model leads to significant performance improvements when compared with standard SVR models for both single transistor and multi-transistor PAs. Different model thresholds are used to test the proposed model performance for both PA types. For the single-transistor PA, modeled using only one partition, an approximately 10 dB normalized mean square error (NMSE) reduction is seen when compared with the standard SVR model. For the same PA, when utilizing two partitions, the reduction improves to 14 dB. When applied to a multi-device Doherty PA, the NMSE between model and measurement data is −50 dB, representing more than 10 dB improvement compared with the standard SVR model.

Annual sedge (Cyperus compressus L.) populations with resistance to halosulfuron were identified in turfgrass at two new locations in Georgia. Research was conducted to evaluate (1) resistance levels to two acetolactate synthase (ALS) inhibitors, (2) ALS enzyme susceptibility, (3) genetic differences associated with resistance, and (4) differential levels of ALS gene expression in these biotypes. In dose–response experiments, the biotypes were >160 times resistant to halosulfuron but only 12 times more resistant to imazaquin compared with a susceptible biotype. In vitro enzyme assays indicated that resistant (R) biotypes required 6.1-fold greater concentrations of imazaquin to reduce ALS activity 50% compared with the susceptible (S) biotype. Both R biotypes had a similar Pro-197-Ser amino acid substitution in the ALS gene that confers resistance to sulfonylureas. Compared with the S biotype, R biotypes had 4.4 times higher ALS gene expression than the S biotype. No differences in gene copy number were found between any biotypes for the ALS gene. Overall, ALS-resistant C. compressus selected by halosulfuron use in turfgrass may be exhibiting partial susceptibility to imazaquin but complete resistance to sulfonylureas. Differential levels of susceptibility to ALS inhibitors for these biotypes are associated with the Pro-197-Ser substitution and enhanced expression of the ALS gene.

In many cases, the efficiency and safety of a drilling project depend on the reliability of the electrical and electronic control system, as the process progresses without visual access of the operator. The electrical and electronic system provides and regulates the power supply for the drill, collects and monitors the drill data during the whole operating process, and sends and receives the control instructions and feedback signals. The entire system is composed of the surface, borehole and software subsystems. The surface subsystem serves for operating the drilling process, transmitting the drilling and environmental data, and supplying power for the drill motor and downhole control system. The borehole subsystem is generally intended for borehole data acquisition, drill motor control, power regulation and communication. The software subsystem is designed for human–computer interaction, data processing and storage, and programming of signal acquisition and transmission of data. The control system of Antarctic subglacial drilling rig was tested during the 2018–2019 summer season near Zhongshan Station, East Antarctica, in the course of drilling to the bedrock at a depth of 198 m. It exhibited a steady and efficient performance without significant system failures.

Spot spraying POST herbicides is an effective approach to reduce herbicide input and weed control cost. Machine vision detection of grass or grass-like weeds in turfgrass systems is a challenging task due to the similarity in plant morphology. In this work, we explored the feasibility of using image classification with deep convolutional neural networks (DCNN), including AlexNet, GoogLeNet, and VGGNet, for detection of crabgrass species (Digitaria spp.), doveweed [Murdannia nudiflora (L.) Brenan], dallisgrass (Paspalum dilatatum Poir.), and tropical signalgrass [Urochloa distachya (L.) T.Q. Nguyen] in bermudagrass [Cynodon dactylon (L.) Pers.]. VGGNet generally outperformed AlexNet and GoogLeNet in detecting selected grassy weeds. For detection of P. dilatatum, VGGNet achieved high F1 scores (≥0.97) and recall values (≥0.99). A single VGGNet model exhibited high F1 scores (≥0.93) and recall values (1.00) that reliably detected Digitaria spp., M. nudiflora, P. dilatatum, and U. distachya. Low weed density reduced the recall values of AlexNet at detecting all weed species and GoogLeNet at detecting Digitaria spp. In comparison, VGGNet achieved excellent performances (overall accuracy = 1.00) at detecting all weed species in both high and low weed-density scenarios. These results demonstrate the feasibility of using DCNN for detection of grass or grass-like weeds in turfgrass systems.

Experiments were conducted to determine the effect of various environmental factors and burial depth on germination and seedling emergence of common beggar’s-tick [Bidens alba (L.) DC.] seeds at two different stages of afterripening. Mature B. alba seeds were stored at 4 C for 3 to 5 mo (new seed lot) and 13 to 15 mo (old seed lot) until experiment initiation. Germination exponentially decreased with increasing moisture stress. Germination rate decreased from 87 ± 2.9% to 13 ± 6.1% as osmotic potential decreased from 0 to −0.5 MPa and was completely inhibited at osmotic potentials below −0.83 MPa. A large portion of the new seeds tested positively photoblastic, but seeds that had afterripened for 1 additional year were partially desensitized to the light requirement. New and old seeds still germinated to a greater percentage in the presence of light than under continuous dark at temperatures ranging from 15 to 35 C. Both new and old seeds germinated over a range of temperatures from 5 to 35 C, but the optimum temperatures for germination was 15 to 30 C in the presence of light. Regardless of seed lot, seedling emergence was the greatest when seeds were sown at the soil surface. Seedling emergence was abruptly reduced when burial depth was 1 cm or greater. Based on these results, we conclude that shallow cultivation could effectively suppress this population of B. alba from emerging when incorporated into an integrated control strategy. The information obtained in this research identifies some important factors that facilitate the widespread presence of B. alba in Florida and may contribute to weed management programs.