Coconut Internal Structure Feature Detection and Developmental Quality Inspection Based on YOLOv7

Coconut is one of the largest seeded fruits in the world. While its hard shell provides some protection for the growth and development of coconuts, it also hinders the observation of internal structural changes and quality inspection during early seedling selection. The cultivation of coconut seedlings takes about a year, and it takes several years for them to grow into mature coconut trees. Therefore, early selection of coconuts is crucial. To address the issue of difficult observation of internal coconut features, we utilized CT scans for non-destructive scanning to obtain internal images of coconuts at different stages. We annotated the internal structures, such as coconut water, haustorium, germ and internal cracks, which exhibit significant changes during coconut growth and development. This led to the establishment of a dataset called Internal Structure Feature of Coconuts (ISFC), providing data resources for coconut image detection. In this paper, we propose an improved model based on YOLOv7 object detection specifically designed for the ISFC dataset, aiming to detect notable changes in internal coconut features and cracks. Experimental results demonstrate that this method accurately identifies and detects the targets, achieving an average precision of up to 97.4%. It satisfies the requirements for high-precision detection and provides a basis for observing internal coconut features and conducting quality inspections during development.