A simple and efficient method for automatic strawberry shape and size estimation and classification

In strawberry production farms, shape and size classification of harvested strawberry fruits is very important phase before packing and sending to the market. However, it is not only very labour-intensive but also time-consuming task for farmers. Computer vision-based automatic strawberry grading systems are capable to overcome this labour-intensive and time-consuming process. In this work, a simple and efficient image processing algorithm for automatic strawberry shape and size estimation and classification is presented. Being different from other existing methods in literature, the current method is based on the geometrical properties of 'right kite' and 'simple kite' which resemble to strawberry fruit shape. The proposed method is used to estimate diameter, length and apex angle from two-dimensional images of strawberry fruits. Then, these parameters are used as input data to a 3-layer neural network for class-A, B, C and D classification. The performance of proposed method is tested for a total of 337 strawberry samples with and without calyx occlusion. The results show that the accuracies for diameter and length estimations are 94% and 93% respectively for strawberries without calyx occlusion and 94% and 89% for that with calyx occlusion. The classification accuracy is between 94 and 97% and the average processing time for one strawberry (one piece) is below 0.45-0.5 s. (C) 2018 IAgrE. Published by Elsevier Ltd. All rights reserved.