Smart farming application using knowledge embedded-graph convolutional neural network (KEGCNN) for banana quality detection

Abstract

The appearance of fruits is crucial in their quality grading and consumer choices. Colour, texture, size, and shape determine fruit quality. Existing computer vision systems have been implemented for external quality control, relying on observations for fruit grading and classification. Banana quality detection systems, which employ advanced algorithms and sensors to evaluate the ripeness and general quality of bananas throughout their life cycle, are an innovative application of smart farming technology. In this proposed system, Knowledge Embedded-Graph Convolutional Neural Networks (KEGCNNs) are employed to classify and grade banana fruit. The approach aims to detect banana fruit quality by converting banana images into a knowledge graph, applying knowledge embedding to transform them into a continuous vector space, and using Graph Convolution Neural Networks (GCNNs) to analyze the graph structure and make accurate detections. KEGCNNs are especially useful for detecting the quality of banana fruits because they provide a form for capturing the contextual interactions between distinct nodes. KEGCNNs can learn from the data within the graph in an unsupervised manner, allowing them to use the knowledge inherent in the graph structure. KEGCNNs enable more accurate and efficient diagnosis of banana quality as they can discover patterns in data that conventional machine learning algorithms cannot. The suggested technique demonstrates an impressive performance score, indicating its suitability for detecting the quality or grade of banana fruit.

Keywords

• Knowledge graph
• Graph convolutional neural network
• Smart farming
• Banana
• Quality detection