Construção de câmera multiespectral de baixo custo com análise em cultivo de café arábica (Coffea arabica)

Abstract

Precision Agriculture has become an important tool in the agricultural production system. In an attempt to achieve fast access to information in an efficient way, imaging sensors have been used in modern agriculture, mainly for monitoring cultivated areas, with real-time diagnosis of the general condition of the crop. This equipment is expensive on the market, but can be replaced by affordable alternatives. In particular, we have instruments for acquiring spectral information aimed at studying vegetation from wavelengths in the visible and near-infrared range, which make it possible to parameterize morphophysiological attributes of plant canopies, indicated by the Normalized Difference Vegetation Index (NDVI). With this in mind, aimed to build a low-cost multispectral camera, implemented on the Arduino Uno open source platform, with operational tests applied to arabica coffee varieties (Coffea arabica L.) in the experimental area of Chã do Jardim, Areia, Paraíba. The materials were purchased with the company's own resources, with an expenditure of R$ 708.17 reais to build the equipment. The low-cost multispectral camera was assembled, built and programmed using the Arduino IDE 2.3.3 software and the C-Arduino programming language. In the field, two visits were made to the area to take data from the manual camera, and then the images were processed using the computer program QGIS Deskstop 3.16.1. At the same time, mapping was carried out using the commercial multispectral camera DJI Mavic 3 Enterprise, in order to compare the results. From the NDVI averages of the Acaiá and Araraçu coffee varieties in the three treatments generated with the cameras, a mutual relationship was observed, where the Manual camera proved to be efficient in identifying NDVI variations in the crop. The statistical analysis showed that the results of the equipment differed from each other. This is quite possibly due to temporal variation, unevenness due to lighting conditions, angle of view, flight time and data collection, as well as the difference between the heights of the sensors. Filters used in the manual camera (M) may not have filtered out the desired bands. The DJI on-board camera (D) showed less instability, giving less variation in the averages. Therefore, the manual multispectral camera (M) can be used to identify NDVI in small agricultural areas, voids and crop problems, with efficiency and low cost in its production.