A fruit damage class and region recognition method based on multi-source images

By fusing RGB and IR images and combining the loss functions Dice and Focal_loss, the problem of inaccurate fruit damage category and region identification in existing technologies is solved, thus achieving both accuracy and efficiency in fruit damage detection.

CN117576466BActive Publication Date: 2026-06-02ANHUI UNIV
1 Cites 0 Cited by

Patent Information

Application Number
CN202311560217.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-22
Publication Date
2026-06-02
Estimated Expiration
2043-11-22

AI Technical Summary

Technical Problem

Existing technologies cannot accurately determine the type and area of ​​fruit damage. Traditional manual inspection is time-consuming and labor-intensive, and computer vision methods are not effective in detecting potential damage.

Method used

By combining RGB and infrared images, multimodal image data is acquired through visible light and IR imaging systems. An improved ADS-DeepLabv3+ network model is used for fruit damage detection, and the Dice_Loss and Focal_loss loss functions are combined to improve detection accuracy.

Benefits of technology

This method achieves improved accuracy and efficiency in fruit damage detection. By fusing a feature pyramid network and a convolutional attention module, it enhances detection accuracy and robustness against interference, improves the ability to distinguish difficult-to-distinguish samples, and improves detection accuracy and generalization performance by combining a hybrid loss function of Dice_Loss and Focal_loss. This method effectively identifies fruit damage.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present application relates to a kind of fruit damage class and area identification method based on multi-source image, comprising: the visible light and IR imaging system is built;Through visible light and IR imaging system, fruit image acquisition and processing are carried out, obtain multi-modal image dataset, multi-modal image dataset is divided into training set, verification set and test set;DeepLabv3 deep learning model is improved, and ADS-DeepLabv3+ network model is obtained;ADS-DeepLabv3+ network model is trained using training set, the fruit image to be identified is input into the ADS-DeepLabv3+ network model after training, and the fruit damage type and area are obtained.The present application can better obtain the "anti- -trans" characteristics of fruit, improve detection accuracy, have better adaptability and anti-interference;ADS-DeepLabv3+ network model improves the recognition ability of small target damage by integrating feature pyramid network and convolution attention module, and the prediction ability for difficult-to-distinguish samples is improved by the mixed loss function obtained by combining Dice_Loss and Focal_loss loss function.
Need to check novelty before this filing date? Find Prior Art