Deep learning-based early weed segmentation using motion blurred UAV images of sorghum fields

Genze, Nikita; Ajekwe, Raymond; Güreli, Zeynep; Haselbeck, Florian; Grieb, Michael; Grimm, Dominik G.

doi:10.1016/j.compag.2022.107388

Cited by 44 publications

(24 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, U-Net [5], another similar segmentation approach, introduces skip connections that allow it to capture both high-level and low-level features. In [6] and [7], U-Net outperforms DeepLabV3+ in crop and weed segmentation.…”

Section: Related Workmentioning

confidence: 99%

Localising Weeds Using a Prototype Weed Sprayer

Darbyshire¹,

Salazar-Gomez²,

Lennox³

et al. 2022

UK-RAS Conference for PhD and Early Career Researchers Proceedings

View full text Add to dashboard Cite

The application of convolutional neural networks (CNNs) to challenging visual recognition tasks has been shown to be highly effective and robust compared to traditional machine vision techniques. The recent development of small, powerful GPUs has enabled embedded systems to incorporate real-time, CNN-based, visual inference. Agriculture is a domain where this technology could be hugely advantageous. One such application within agriculture is precision spraying where only weeds are targeted with herbicide. This approach promises weed control with significant economic and environmental benefits from reduced herbicide usage. While existing research has validated that CNN-based vision methods can accurately discern between weeds and crops, this paper explores how such detections can be used to actuate a prototype precision sprayer that incorporates a CNNbased weed detection system and validates spraying performance in a simplified scenario.

show abstract

Section: Related Workmentioning

confidence: 99%

Localising Weeds Using a Prototype Weed Sprayer

Darbyshire¹,

Salazar-Gomez²,

Lennox³

et al. 2022

UK-RAS Conference for PhD and Early Career Researchers Proceedings

View full text Add to dashboard Cite

show abstract

“…Weeds compete with crops for resources such as nutrients, water, space, and sunlight, severely affecting crop production. , Chemical herbicides have become essential agricultural production materials because of their excellent performance in managing weeds and reducing labor costs . However, with increasing awareness of the importance of environmental protection and sustainable development, as well as the constant emergence of resistant weeds, innovative varieties of herbicides are needed to meet production and market demands. , …”

Section: Introductionmentioning

confidence: 99%

Discovery of Novel Pyrazole Acyl Thiourea Skeleton Analogue as Potential Herbicide Candidates

Ma,

Tian,

Wang

et al. 2024

J. Agric. Food Chem.

View full text Add to dashboard Cite

To discover novel transketolase (TKL, EC 2.2.1.1) inhibitors with potential herbicidal applications, a series of pyrazole acyl thiourea derivatives were designed based on a previously obtained pyrazolamide acyl lead compound, employing a scaffold hopping strategy. The compounds were synthesized, their structures were characterized, and they were evaluated for herbicidal activities. The results indicate that 7a exhibited exceptional herbicidal activity against Digitaria sanguinalis and Amaranthus retroflexus at a dosage of 90 g ai/ha, using the foliar spray method in a greenhouse. This performance is comparable to that of commercial products, such as nicosulfuron and mesotrione. Moreover, 7a showed moderate growth inhibitory activity against the young root and stem of A. retroflexus at 200 mg/L in the small cup method, similar to that of nicosulfuron and mesotrione. Subsequent mode-of-action verification experiments revealed that 7a and 7e inhibited Setaria viridis TKL (SvTKL) enzyme activity, with IC 50 values of 0.740 and 0.474 mg/L, respectively. Furthermore, they exhibited inhibitory effects on the Brassica napus acetohydroxyacid synthase enzyme activity. Molecular docking predicted potential interactions between these (7a and 7e) and SvTKL. A greenhouse experiment demonstrated that 7a exhibited favorable crop safety at 150 g ai/ha. Therefore, 7a is a promising herbicidal candidate that is worthy of further development.

show abstract

“…Object detection and instance segmentation models have also recently outperformed the ML algorithm Support Vector Machine [ 24 ]. Further research has been conducted to evaluate the performance of various DL semantic segmentation networks such as a U-Net, SegNet, FCN, and DeepLabv3+, employing pixel-based image classification to differentiate between the background, crops, and weeds [ 3 , 25 , 26 , 27 , 28 , 29 ]. In a study using multispectral UAV imagery, a U-Net proved most efficient in terms of computational resources as well as model performance [ 27 ].…”

Section: Introductionmentioning

confidence: 99%

“…In a study using multispectral UAV imagery, a U-Net proved most efficient in terms of computational resources as well as model performance [ 27 ]. This is further demonstrated by the U-Net’s proficiency in real-time segmentation during on-board processing [ 30 ] and its successful application to a blurry UAV-based dataset of sorghum [ 28 ]. However, the precise detection of weeds relies not only on the choice of a high-performance algorithm but also largely on the applied sensor.…”

Section: Introductionmentioning

confidence: 99%

Weed Detection from Unmanned Aerial Vehicle Imagery Using Deep Learning—A Comparison between High-End and Low-Cost Multispectral Sensors

Seiche,

Wittstruck,

Jarmer

2024

Sensors

View full text Add to dashboard Cite

In order to meet the increasing demand for crops under challenging climate conditions, efficient and sustainable cultivation strategies are becoming essential in agriculture. Targeted herbicide use reduces environmental pollution and effectively controls weeds as a major cause of yield reduction. The key requirement is a reliable weed detection system that is accessible to a wide range of end users. This research paper introduces a self-built, low-cost, multispectral camera system and evaluates it against the high-end MicaSense Altum system. Pixel-based weed and crop classification was performed on UAV datasets collected with both sensors in maize using a U-Net. The training and testing data were generated via an index-based thresholding approach followed by annotation. As a result, the F1-score for the weed class reached 82% on the Altum system and 76% on the low-cost system, with recall values of 75% and 68%, respectively. Misclassifications occurred on the low-cost system images for small weeds and overlaps, with minor oversegmentation. However, with a precision of 90%, the results show great potential for application in automated weed control. The proposed system thereby enables sustainable precision farming for the general public. In future research, its spectral properties, as well as its use on different crops with real-time on-board processing, should be further investigated.

show abstract

Deep learning-based early weed segmentation using motion blurred UAV images of sorghum fields

Cited by 44 publications

References 39 publications

Localising Weeds Using a Prototype Weed Sprayer

Localising Weeds Using a Prototype Weed Sprayer

Discovery of Novel Pyrazole Acyl Thiourea Skeleton Analogue as Potential Herbicide Candidates

Weed Detection from Unmanned Aerial Vehicle Imagery Using Deep Learning—A Comparison between High-End and Low-Cost Multispectral Sensors

Contact Info

Product

Resources

About