A yolo v5l green pepper intelligent detection method based on NSGA-II pruning

By using the YOLOv5l model method based on NSGA-II pruning, the problems of large size, high computational cost, and low accuracy of green pepper detection models in field environments are solved, realizing efficient and accurate intelligent detection of green peppers, which is suitable for green pepper phenotypic monitoring and robotic harvesting.

CN115761728BActive Publication Date: 2026-06-09YANCHENG INST OF TECH
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN ยท China
Patent Type
Patents(China)
Current Assignee / Owner
YANCHENG INST OF TECH
Filing Date
2022-09-08
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing green pepper detection models are large in size, computationally intensive, and have low detection accuracy, making it difficult to meet the needs of accurate detection of green pepper fruits in field environments.

Method used

We employ an NSGA-II-based pruning method for the YOLOv5l model. This method uses a multi-objective optimization algorithm to prune the YOLOv5l model, reducing the number of model parameters and computational cost while maintaining or restoring detection accuracy. This includes steps such as dataset labeling, model training, pruning, and fine-tuning.

Benefits of technology

It significantly reduces the number of parameters and computational load of the green pepper detection model, improves the detection speed, and maintains high detection accuracy, making it suitable for green pepper phenotypic monitoring and robotic harvesting.

โœฆ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115761728B_ABST
    Figure CN115761728B_ABST
Patent Text Reader

Abstract

The application discloses a YOLOv5l green pepper intelligent detection method based on NSGA-II pruning, which comprises the following steps: collecting field green pepper image data; marking green pepper fruits in the image by using LabelIMG; preparing a green pepper dataset and dividing the dataset into a training set and a test set; obtaining a green pepper detection model by adopting a pruning YOLOv5l model method based on NSGA-II; detecting the reliability and accuracy of the pruning YOLOv5l green pepper detection model of NSGA-II in the field, and evaluating the model; if the reliability and accuracy meet the set requirements, the next step is carried out; if the reliability and accuracy do not meet the set requirements, the above steps are re-performed, and the number of green pepper images is increased and the network is modified; and the model is deployed to a green pepper phenotype monitoring and green pepper robot picking production application scene, so that the parameter quantity, model size and GFlops of the green pepper detection model are greatly reduced, and the detection speed is significantly improved.
Need to check novelty before this filing date? Find Prior Art