Multiclass Fruit Packing System Based on Computer Vision

Abstract

The traditional human-based fruit packing method is inefficient and costly. In response to the requirement of packing multiple fruits automatically in the present market, a multiclass fruit packing system based on computer vision is designed. Firstly, the image acquisition module captures images and passes them to the recognition module, which converts the RGB space of the image to the HSV color space and extracts the H and S components. Secondly, by using the morphological method to process the target areas, the class, size and position of the fruit are calculated. Finally, in the fruit packing control module, the manipulator automatically grasps different kinds of fruits into the different packing boxes according to the recognition results. The experimental results show that the accuracy of recognition is 93.73 %, the average recognition time of each image is 0.0402s. The accuracy of packing is 95.67% and the injury rate of fruit is 1.67%, which lay the foundation for fully automatic production of fruit.