Desarrollo de un sistema de visión artificial para la detección de hongos en plantas de cannabis medicinal mediante redes neuronales en ambientes controlados.

Abstract

This project is focused on implementing an automatic artificial vision system to detect three types of fungi such as powdery mildew, mildew and rust as they are the most common in this class of crops, using a camera installed inside a greenhouse, previously built, capture images in real time of the changes that occur in the leaves, these can be type: change of color, change of shape or imperfections depending on the class of fungus. The system will be developed through 4 main phases which are: adaptation of the development site, data acquisition, image processing and pattern recognition. In the adequacy of the development site will be carried out the construction of a controlled environment that will meet the basic standards to ensure the environment for cultivation, as it is a plant that requires care, in this environment you can visualize the humidity and temperature key parameters for growth and development. In the data acquisition phase, photographic information will be collected on the different types of fungi that the plant may suffer, among them: oidium, botrytis, rust, Pythium, Fusarium, among others. This database will be obtained from bibliographic consultations and web pages, which will contribute to the subsequent training of the different neuronal networks to be used. Once the database with the images of the fungi is determined, we proceed to phase 3, which consists in the processing of these, where the patterns that correspond to specific characteristics related to each disease and pest will be extracted. All this will be done by means of a computer with a reduced plate known as raspberry pi. Finally, phase 4 will be started, which basically consists on the use of a camera (raspicam) installed in the mini-greenhouse that will make a tour through the crop. The objective is to recognize patterns that identify the presence of some imperfect present in the crop, when detecting some anomaly it will send a text message automatically to the student through a sim808 module reporting the presence of the fungus and sending its current position.