Vision-based Automatic Strawberry Shape and Size Estimation and Classification Using Raspberry Pi

Nowadays, single board small computers have been a great region of interest for their low cost and reasonable processing power in a compact board. In this paper, a vision-based automatic strawberry shape and size estimation for occluded and non-occluded calyx conditions will be implemented on Raspberry Pi 3 model B with Python programming for the purpose of embedded system implementation instead of software simulation. There are several research works for strawberry grading system but most of them are used with high performance CPUs which are more expensive and not as portable as Raspberry Pi. With machine vision technology, classification of strawberries can be time saving, reduce labour costs and avoid different judgements of each labour. A test image of strawberry is accquired with a webcam to lower the cost of system and a morphological operation is used to preserve the size and shape of strawberry after a segmentation task. Strawberries are classified into four classes and total number of 60 strawberry samples are tested. The average classification accuracy is 96%. The algorithm has been optimised to acquire high processing speed on Raspberry Pi development board regardless of its limited computing power. The average processing time for image processing and image analysis with capturing image of each strawberry sample is 0.6 seconds.