Vous êtes ici : Accueil » Kiosque » Annonce


Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?


23 janvier 2020

Computer vision system for automated mushroom picking

Catégorie : Post-doctorant

Duration : 12 months
Framework : CHAMPIBOT (Start'AIRR : Isite, Région Hauts de France)
Starting date : January 2020 Location : CRISTtAL laboratory, Villeneuve d'Ascq

The interested candidates should contact rapidly either :

Ass. Pr. Duvieubourg Luc Pr. Cabestaing François or Pr. Macaire Ludovic

with a full CV and a reference letter :
luc.duvieubourg@univ-lille.fr - francois.cabestaing@univ-lille.fr - ludovic.macaire@univ-lille.fr


This project is part of a collaboration between the Gontière farm (mushroom industry) and the CRIStAL laboratory of the University of Lille.
The industrial production of mushrooms is currently penalized by the picking phase. The realization of this phase is carried out by manual, repetitive and precise tasks. The problem for manufacturers is to find a skilled workforce to perform this task. Based on this observation, this collaborative project aims to provide flexibility to the industry by developing a prototype machine for automatic mushroom picking. Although there are prototypes of industrial machines for picking fruits and vegetables that work properly, the performance of these machines remains poor in the case of the mushroom.
This project will be led by two teams from the CRIStAL laboratory. A first team will be in charge of the robotic part and the second of the vision part. The postdoctoral fellow will intervene in the vision part.



The main objective of the vision part is to determine the mushrooms to pick and indicate their precise location to the control system of the robot. The information needed to control the robot is the position and the rientation of the mushroom. The disorderly growth of mushrooms in their compost bed makes it difficult to accurately recognize their position by a 2D image. Tomechanize the harvest, the 3D structure information of the mushroom bed is needed. It will therefore be, in this preliminary study, to propose a 3D acquisition system. There are many ways to get 3D information from a target using one or more cameras. For this application, we are considering a 3D scanner based on a high definition color camera and a laser projecting a line. This 3D acquisition technique requires a prior calibration procedure to obtain the relative positions between the camera and the laser plane. The calibration will establish a relationship between the 3D points, in the same frame as the robot, and the 2D points of the image acquired by the camera.
From the calibrated system, the crates (compost bed) of mushrooms will pass under the acquisition system. At each image acquisition, the laser line will be extracted and then concatenated with the previous ones in order to obtain the 3D reconstruction of the mushrooms. When the whole crate of mushrooms is passed in front of the camera, an overall analysis will make it possible to obtain a complete cartography of the crate.
In the framework of CHAMPIBOT project, the teams 'image and BCI' are seeking for a talented postdoctoral research fellow. The hosting laboratory possesses expertise on the development of 3D vision systems (from the sensors to the data processing). The candidate will both implement the laser scanning three-dimensional digitiser system proposed and investigate relevant and efficiency algorithms.
The Postdoc fellow will be in charge of :
- Software integration on the acqusition system.
- Processing of data collected.
- Study of the relevance and efficiency of the approach in real conditions.

Candidates profile :
- Doctor in image processing, signal processing or similar.
- Strong background in vision system and data processing.
- Skills in instrumentation, applied mathematics, numerical computation (C, Python, OpenCV) are preferred.
- Applicants must be self-driven, higly motivated and able to work autonomously towards the objectives of the project.


Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2020.