Neural network for apple fruit recognition and classification

Abstract

The article proposes a method for monitoring industrial gardens based on artificial intelligence and machine learning. To identify apple fruits on the crown of a tree using a robotic platform moving in the garden areas with a camera attached to it, a neural network was developed, the VGG-16 model and SSD architecture were used, which detect the output space and generate bounding rectangles in images with different aspect ratios. To count the number of fruits relative to each row of plantings, a method is proposed for stitching a series of photographs of fruit trees in a row into a cylindrical panorama. To assess the quality of the developed neural network when working with 6 classes, the multi-classification task was applied. Analysis of the results of the conducted research has shown that the developed neural network model has high performance and high quality of ordering class objects. The developed neural network allows processing at least 200 requests, identifying healthy apple fruits and apple fruits affected by diseases on tree crown images, and counting their number.