insects, classification, interface, activity recognition
The proposed internship is within an Action of Collborative Research INRIA aiming at early detection of bio-agressors in green house pesticide-free cultures (detection of harmful insects on plant organs such as rose leaves). Currently, the vision systems being experienced in green houses ( using static imaging ) are limited by their sampling capabilities in both spatial and temporal dimensions. One of the goals of this Action is to define new methods to early in situ detection of bio-agressors based on analysis and interpretation of muliple captor video scenes.
Our primary goal is to devise a vision system capable of monitoring continuously the green house by installation of a network of Wifi communicating video captors (cameras) . Positionning, number and nature of the used cameras are essential to the obtention of an optimized sampling in terms of cost and accuracy. The major challenge is then to acheive a sufficient level of robustness for continuous survaillence, capable of adapting to lighting changes that occur in day time. We have already installed a network of five Wifi cameras within our experimental green house of roses which we endowed with an intelligent acquisition module capable of scheduling in a timely way the acquisiton and storage of the collected pre-processed video data. The next step consists in the detection and classification of the insects in the acquired videos. The results of the detection are of key importance for decision making (chemical treatment, introduction of auxiliary insects ...). It is then crucial to provide final users ( exploiters of the green house, agricultural scientists ...) with results as quickly as possible and in a synthetic manner.
|Work to be Done
The goal of the proposed internship is to develop a user interface that can display the evolution of insect popolations in a green house. One of its funtionalities will be the display of the sampling map and the growth curves of insect populations at different temporal scales (day, week, month, ...) by taking as inputs the results provided by the classification module.
Good knowledge of C++, knowledge of Qt notions in image processing.
PULSAR Team, INRIA-Sophia Antipolis.
5 to 6 months.
EPI PULSAR, INRIA-Sophia Antipolis
2004 route des Lucioles BP 93
06902 Sophia Antipolis Cedex
Email : Sabine.Moisan@sophia.inria.fr
Tel: (+33) (0)4 92 38 78 47