The application discloses a kind of based on YOLO series model and
confidence threshold determination Defect
Momordica fruit recognition method and
system belong to agricultural
product processing target detection field.This method includes: collecting
Momordica fruit multi-view image;It is preprocessed by size normalization, brightness adjustment,
noise suppression etc.;Input YOLO series model outputs whole fruit detection frame, good / bad class and confidence;Final determination is completed by non-maximum suppression and
confidence threshold;Output result is used for display, storage or drive sorting.
System is composed of
image acquisition, image preprocessing, YOLO identification,
confidence threshold determination, result output module.This application covers a variety of surface defects, reduces environmental and texture interference, outputs stable determination
signal, can be directly connected to automatic sorting equipment, with high precision, strong adaptability, deployment flexible etc.Advantages, suitable for
Momordica fruit large-scale
processing quality detection and grading.