Computer vision-based prototype robotic picking cum grading system for fruits

Abstract

Huge harvest and post-harvest losses in conventional processes and methods demands precise applications of agricultural inputs. Computer vision based robotic picking is one of such systems that reduces time consumption, labor requirement; and enhances fruits picking and grading efficiency. In this study, an all-in-one prototype robotic picking cum grading system is developed that integrates robotic picking and fruit quality recognition based grading processes. Therein, computer vision algorithms are used for fruit detection followed by picking and placing through controlled robotic arm. For fruit detection, the steps includes color scheme conversion, masking of normal/fresh skin, masking of defects, and morpological dilation operations; and opencv library is used for implementation in Python Language on Anaconda Spyder Integrated Development Environment (IDE). The detection information is then communicated to robotic arm through microcontroller. The fabricated arm is of 4 Degrees of Freedom (4-DOF) and therein, four servo motors aid the robotic arm. The time taken by proposed computer vision algotihms scheme is as follows: 0.0035136 s, 0.0081954 s, 0.0005433 s, and 0.0002408 s for color scheme conversion, segmentation, morphological operations, and bitwise operations respectively; and total time for complete scheme is 0.0125 s. Overall, the developed system took average time of 15 s for one grading cycle in case of poor quality fruits. For this case, no action is carried out and the fruit passes the conveyor directly to storage bin. However, the time required for good quality fruits is high i.e. 21 s as the robotic arm picks and places the fruits for this case.