ARQUITETURA DE SOFTWARE PARA OTIMIZAÇÃO DO USO DE AERONAVES REMOTAMENTE PILOTADAS NA AGRICULTURA DE PRECISÃO UTILIZANDO RACIOCÍNIO BASEADO EM CASOS

Abstract

The use of remotely piloted aircrafts (RPA) in precision agriculture (PA) is constantly evolving being a process comprised of the following steps: objective determination, capture and processing of images and the analisys of the obtained data. Although there is software and hardware made for those steps, the challange remains being the integration of the collected data, its reliability, and the intepretation of the resulting data for decision making. This work proposes a software architeture to make use of RPA on PA, allowing to perform all steps of the process using the Domain-Driven Design and Command and Query Responsability Segregation architectural patterns. The architecture, when used by researchers, allows the integration of new image processing modules, making use of case-based reasoning (CBR) for its evaluation. The architecture was evaluated in a case study under the following aspects: the reliability of the image processing methods and the adequacy of the CBR method when determining the best image processing algorithm for the specified objective. The results of the image processing algorithm for estimate the Normalized Difference Vegetation Index (NDVI) were compared to the results obtained by the field equipment (Greenseeker) and by processing made with the Pix4D software. Both the software and the equipment produced similar results, with a difference of about 7%. When doing the simulation for the best algorithm to be automaticaly used by the end user, the software correctly selected the algorithm with the highest probability of generating a correct outcome. In addition to the technical benefits, the developed architecture allows the results of field experiments analysis, associated with the algorithms used, to feed the knowledge base of the software so that it generates better parametrizations in the executions made by the end user. In addition to the technical benefits, the developed architecture allows the results of field experiments, associated with the used algorithms, to feed the knowledge base of the softwares that it generates better parametrizations when executed by the end user. The developed architecture in this work made possible integrating the several steps when using RPAs in PA, making easier its use by resaerchers and end users. The joint use by researchers and end users in the same environment could be an interesting alterative to publish new and better methods for image processing, giving to the end users more reliable results and more information about the crops.