Artificial Marker to Predict (Banganapalle) Mango Fruit Size at Multi-Targets of an Image Using Semantic Segmentation

Abstract

Getting the size of any fruit on a tree is not an easy task especially mango fruit, because of its irregular shape, it is not easy to model with its shape. The size of the fruit is expected in length and width. Objective: Horticulture farmers need to engage in extra activities to obtain better yields, such as trying to know the fruit shape and size at the time of maturity, or before plucking the fruits from the tree, which will help farmers obtain as per their predicted price while selling the fruits to the market. Methods: Researchers applied a deep learning model called YOLOv7, semantic segmentation, to obtain fruit size using an aruco marker and proposed a technique to help farmers as detect markers and the fruits in images and predict the size of the fruit at multiple targets. A custom dataset was created by collecting mango fruit frames from an on-tree-mango-360° recorded video. After training and validating the model, its performance is tested using a test dataset. Results: The contributions of this study are as follows: The researcher developed a procedure to obtain mango size from an image. The researcher implemented and tested a model to detect mangoes in different challenging situations using YOLOv7 with semantic segmentation. This model achieves excellent results for fruit size estimation. The training and testing results of YOLOv7-SS-AM show that the Aruco marker-based model is superior to manual size prediction, with good accuracy.

Keywords

• ArUco
• aruco-marker
• semantic-segmentation
• shape
• YOLOv7