The evolution of remote sensing applications in precision agriculture may lead to a reduction in phytochemical use. The development of data processing techniques allows speeding up the treatment of the information then making it possible to apply the right amount of chemical at the right place and time. In this study, the use of spectral information from remote sensing data using camera sensors coupled with artificial intelligence techniques at the field scale to detect wheat diseases is evaluated. In this study a Region based Convolutional Neural Network (R-CNN) model was evaluated for wheat disease detection. Besides, the methodology considered environmental variability during data collection to simulate real conditions during data acquisition in the field. The preliminary results of the model show a mean average precision of 0.45 and a precision between 0.2 to 0.6 for each image. Considering limitations in terms of data annotations for training the model, these results look promising on the way to consider further improvement of the model to reach higher accuracies and precisions. Furthermore, the study aims to develop a methodology for treatments reduction based on the enhanced version of the R-CNN model. Therefore, wheat disease detection and localized treatment in the field have a strong potential to reduce the use of chemicals. The developed methodology will optimize the costs and the use of chemicals over agricultural fields.