Embedded system for real time analysis of thermal images for prevention of water stress on plants

摘要

Given the evolution of electronics technology and electronics equipment, their price has fallen, thus making them affordable for the use in new branches: industry, security, agriculture. In this paper we propose a complete integrated analysis system that can learn - through deep learning algorithm - several templates for one parameter (spatial temperature evolution) for the occurrence of water stress and can then recognize these preconditions in the real environment. The system can then generate alarms to farmers / researchers about the occurrence of water stress. The system uses a FLIR camera sensor and capture the spatial arrangement of temperature at plant level, make an local intelligent analysis and return different levels of alarm. Thus, instead of transmitting the image to the central station on a conventional channel, the system will perform an on-site analysis over a period, identify potential water stress by recognizing templates, and will only convey indications of the crop state.