Algorithm to detect nitrogen foliar deficiency in bean crops applying digital image processing and data mining

The determination of the Nitrogen correct application in bean crops, aiming to increase productivity and reduce costs with fertilizers, is a great challenge, which has a solution based on the use of precision farming tools. This work proposes the use of digital image processing techniques coupled with a data mining process aiming at the construction of a decision tree based algorithm for the definition of dose for the maximum production Nitrogen dosage to be applied by means of cover fertilization at focal points of the beans cultures. From the digital images collection of the bean plants leaves, these are processed obtaining the color channels RGB. From these images, the histograms are extracted, to be compared with the reference histograms already obtained from other witnesses’ sheets with various degrees of known nitrogen deficiency from a set of previously prepared and studied portions. The database generated through the comparison between these histograms was processed in a data mining tool using the classification algorithm J48 (C4.5) and, the process allowed the construction of the classification tree. In the obtained results it was possible to classify the Nitrogen dosage to be applied to the focal points of the cultures under analysis, with 76.33% accuracy.