To achieve greater efficiency at harvesting tasks, the mechanization of such task is unavoidable. Apart from the mechanical aspects, the harvesting systems needs software that can locate the fruit to be harvested. 

The use of machine learning and deep learning techniques to achieve such software was studied in this thesis. The results showed that an accuracy similar to other studies is feasible with a limited number of training samples using deep learning techniques. 

From this thesis we conclude that the mechanization of the harvesting labour is possible, at least from the software point of view, while the crop estimation application may need some more work before being feasible.