Automatic vision guided intelligent fruits and vegetables processing system and method

A method for vision-guided automated removal of an unwanted portion of a produce, comprising:

• entering a plurality of produces into a loading station;
• operatively coupling a vision station to said loading station, equipping said vision station with images acquisition sub-system, and acquiring images of said plurality of produces received from said loading station;
• operatively coupling a cutting station to said vision station, establishing a cutting mechanism at said cutting station for removal of an unwanted portion of said plurality of produces, said cutting station being configured to separate the unwanted portions from said plurality of produces according to separation trajectories;
• operatively coupling an unloading station to said cutting station for unloading said plurality of produces devoid of said unwanted portion;
• operatively coupling a conveyor sub-system to said loading, vision, cutting and unloading stations and transporting said plurality of produces therealong;
• operatively coupling a visual data processor sub-system at least to said vision, cutting and unloading stations and said conveyor sub-system, configuring said visual processor sub-system with (a) a convolutional neural network, (b) a regression deep neural network integrated in said convolutional neural network, and (c) a synchronization processor unit configured with at least one shaft encoder;
• operatively coupling a network training mechanism to said visual processor sub-system, and feeding said network training mechanism with a set of training produces images, each produce image being provided with a corresponding ground truth cutline;
• training said convolutional neural network, by operatively interacting with said network training mechanism, to identify at least a first produce from said plurality of produces to be cut and at least a second produce from said plurality of produces to be rejected from cutting;
• training said regression deep neural network, by operatively interacting with said network training mechanism, to compute a cutline trajectory for said at least first and second produce based on the images of said plurality of produces acquired by said images acquisition sub-system at said vision station;
• tracking, by said shaft encoder, locations of said unwanted portions of said plurality of produces based on the images of said plurality of produces acquired by said images acquisition sub-system at said vision station;
• synchronizing, by said synchronization processor unit, displacement of the conveyor sub-system with actuation of said cutting mechanism;
• cutting said at least first produce along a respective cutline trajectory computed for said at least first produce;
• and
• refraining said cutting mechanism from an interaction with said at least second produce.