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

Champ Julien, Mora-Fallas Adán, Goeau Hervé, Mata‐Montero Erik, Bonnet Pierre, Joly Alexis. 2020. Instance segmentation for the fine detection of crop and weed plants by precision agricultural robots. Applications in Plant Sciences, 8 (7), n.spéc. Machine Learning in Plant Biology: From Genomics to Field Studies:e11373, 10 p.

Journal article ; Article de recherche ; Article de revue à facteur d'impact Revue en libre accès total
Published version - Anglais
License Licence Creative Commons.

Télécharger (1MB) | Preview

Url - jeu de données :

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.

Mots-clés Agrovoc : Détermination des espèces, Automate, agriculture de précision, Désherbage, Mauvaise herbe, Plante de culture, Lutte culturale, Désherbage mécanique

Mots-clés complémentaires : désherbage électrique

Mots-clés libres : Autonomous robot, Convolutional neural network, Digital agriculture, Plant detection, Weed electrification

Classification Agris : H60 - Weeds and weed control
N20 - Agricultural machinery and equipment
U30 - Research methods

Champ stratégique Cirad : CTS 4 (2019-) - Santé des plantes, des animaux et des écosystèmes

Auteurs et affiliations

  • Champ Julien, INRIA (FRA)
  • Mora-Fallas Adán, Instituto Tecnológico de Costa Rica (CRI)
  • Goeau Hervé, CIRAD-BIOS-UMR AMAP (FRA)
  • Mata‐Montero Erik, Instituto Tecnológico de Costa Rica (CRI)
  • Bonnet Pierre, CIRAD-BIOS-UMR AMAP (FRA) ORCID: 0000-0002-2828-4389 - auteur correspondant
  • Joly Alexis, INRIA (FRA)

Source : Cirad-Agritrop (

View Item (staff only) View Item (staff only)

[ Page générée et mise en cache le 2021-03-02 ]