A method for detecting defects of engine blades based on machine vision

By constructing an image pyramid through multi-scale Gaussian blurring and downsampling, and combining it with adaptive enhancement and a multi-scale detection network, the problem of low accuracy in engine blade detection is solved, and efficient defect identification is achieved.

CN122199536APending Publication Date: 2026-06-12CHENGDU AERONAUTIC POLYTECHNIC
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHENGDU AERONAUTIC POLYTECHNIC
Filing Date
2026-05-14
Publication Date
2026-06-12

AI Technical Summary

⚠Technical Problem

Existing machine vision inspection technology cannot effectively separate defects and texture information of different scales in engine blade inspection, resulting in low detection accuracy and easy to miss or misdetect.

⚗Method used

An image pyramid is constructed using multi-scale Gaussian blur and downsampling. The image is then processed with adaptive enhancement coefficients to improve the detail at different scales. Finally, a multi-scale detail detection network is used for defect identification.

🎯Benefits of technology

It significantly improves the accuracy and robustness of engine blade defect detection, reduces the false negative and false positive rates, and can accurately distinguish between normal textures and real defects on complex surfaces.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122199536A_ABST
    Figure CN122199536A_ABST
Patent Text Reader

Abstract

The application discloses a kind of engine blade defect detection methods based on machine vision, belong to image processing technical field.The application is carried out to engine blade image continuous multiple times Gaussian blur and down-sampling, obtain multiscale smooth image and multiscale down-sampling image;To the thinnest scale down-sampling image is up-sampled, obtains blade structure image;Again based on multiscale smooth image and multiscale down-sampling image, extract small, medium, large three kinds of scale detail images;Blade structure image is fused with three kinds of scale detail images respectively, obtain three detail structure images;Through for each scale detail image setting corresponding neighborhood window, according to the pixel difference between center and neighborhood window determines enhancement coefficient, and utilize the coefficient to three detail structure images are carried out defect enhancement processing;Finally using multiscale detail detection network is handled to the image after enhancement, obtains defect recognition result.The application effectively improves the precision of engine blade defect detection.
Need to check novelty before this filing date? Find Prior Art