Dataset Information

Instance segmentation for the fine detection of crop and weed plants by precision agricultural robots.

ABSTRACT: Premise:Weed removal in agriculture is typically achieved using herbicides. The use of autonomous robots to reduce weeds is a promising alternative solution, although their implementation requires the precise detection and identification of crops and weeds to allow an efficient action. Methods:We trained and evaluated an instance segmentation convolutional neural network aimed at segmenting and identifying each plant specimen visible in images produced by agricultural robots. The resulting data set comprised field images on which the outlines of 2489 specimens from two crop species and four weed species were manually drawn. We adjusted the hyperparameters of a mask region-based convolutional neural network (R-CNN) to this specific task and evaluated the resulting trained model. Results:The probability of detection using the model was quite good but varied significantly depending on the species and size of the plants. In practice, between 10% and 60% of weeds could be removed without too high of a risk of confusion with crop plants. Furthermore, we show that the segmentation of each plant enabled the determination of precise action points such as the barycenter of the plant surface. Discussion:Instance segmentation opens many possibilities for optimized weed removal actions. Weed electrification, for instance, could benefit from the targeted adjustment of the voltage, frequency, and location of the electrode to the plant. The results of this work will enable the evaluation of this type of weeding approach in the coming months.

SUBMITTER: Champ J

PROVIDER: S-EPMC7394709 | biostudies-literature | 2020 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Instance segmentation for the fine detection of crop and weed plants by precision agricultural robots.

Champ Julien J Mora-Fallas Adan A Goëau Hervé H Mata-Montero Erick E Bonnet Pierre P Joly Alexis A

Applications in plant sciences 20200728 7

<h4>Premise</h4>Weed removal in agriculture is typically achieved using herbicides. The use of autonomous robots to reduce weeds is a promising alternative solution, although their implementation requires the precise detection and identification of crops and weeds to allow an efficient action.<h4>Methods</h4>We trained and evaluated an instance segmentation convolutional neural network aimed at segmenting and identifying each plant specimen visible in images produced by agricultural robots. The ...[more]

PMID: 32765972

Similar Datasets

Project description:In many parts of Eastern India that have a very high prevalence of rural poverty and food insecurity, the prevailing rice establishment practice of 'beushening' is characterized by low yields and modest profitability, while labor and energy inputs are high. Beushening consists of broadcasting ungerminated rice seed at high rates (>100 kg ha-1) prior to the onset of monsoon rain, followed by ploughing at 4-6 weeks after crop emergence to control weeds with subsequent manual gap filling through seedling redistribution to ensure stand uniformity. Dry-direct seeding of rice (DSR), both drill-DSR and precision broadcast-DSR in combination with integrated weed management (IWM) may offer a pathway for simultaneously reducing costs and markedly increasing productivity. On-farm trials were conducted from 2016 to 2018 in four districts of Odisha (Mayurbhanj, Cuttack, Bhadrak, and Puri) to evaluate the yield and economic performance of dry-DSR (drill and precision broadcast), coupled with herbicide-based IWM strategies, in comparison with conventional beushening. Drill-DSR with IWM increased grain yield by 1.7 t ha-1 in Mayurbhanj and 1.3 t ha-1 in Cuttack, but not in Bhadrak, compared to beushening. The combination of increased yield and lower variable cost in drill-DSR increased net benefit by 550, 395, and 166 US$ ha-1 in Mayurbhanj, Cuttack, and Bhadrak, respectively. For farmers without access to seed drills, precision broadcast-DSR with IWM increased yields by 0.91, 1.22 and 0.60 t ha-1, and net benefits by 270, 312, and 188 US$ ha-1 in Mayurbhanj, Puri, and Bhadrak, respectively. Among the IWM practices evaluated in dry-DSR, application of pretilachlor + safener @ 500 g ai ha-1 as pre-emergence, followed by bispyribac-sodium @ 20 g ai ha-1 at 15-25 days after sowing as post-emergence, and then one spot hand weeding at 30-35 days after sowing was effective in controlling weeds. These results suggest that rice yield gaps in eastern India can be reduced, and farmers' income from rice can be increased by more than 50 % by replacing beushening with drill-DSR or precision broadcast-DSR. The results could be applicable to approximately 6.4 million ha of lowland rice where beushening is currently practiced in Eastern India.

Dataset Information

Instance segmentation for the fine detection of crop and weed plants by precision agricultural robots.

Publications

Instance segmentation for the fine detection of crop and weed plants by precision agricultural robots.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets