Composite defect detection method based on semantic segmentation and target detection fusion model
A technology of semantic segmentation and target detection, applied in neural learning methods, character and pattern recognition, biological neural network models, etc., can solve the problems of unreasonable, mixed, and undetectable division of positive and negative samples
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0051] The existing mobile phone screen surface defect detection technology usually uses some popular target detection networks such as YoLo V3 single-stage target detection network and Faster R-CNN two-stage target detection network. Ordinary target detection networks can achieve better results in the detection of single-type defects and multiple types of defect images, but for overlapping images of multiple types of defects and problems in the detection process, ordinary detection networks cannot solve them well. To this end, the present invention proposes a composite defect detection method based on semantic segmentation and target detection fusion model, such as figure 1 shown. Include the following steps:
[0052] S1: Obtain the original image and use the image chroma value analysis method to calculate the ratio of the number of low chroma value pixels in the original image, and obtain the original image with normal range chroma value;
[0053] S2: Preprocessing the ori...
Embodiment 2
[0064] More specifically, on the basis of Example 1, such as figure 2 as shown, figure 2 A schematic flow chart showing steps S1-S4.
[0065] Wherein, in the step S2, the composite defect image preprocessing includes scratch defect image preprocessing, air bubble defect image preprocessing, tin ash defect image preprocessing and pinhole defect image preprocessing; wherein, the scratch Defect image preprocessing can highlight the scratch defect in the original image and get SP-1 defect image; bubble defect image preprocessing can highlight the bubble defect in the original image and get BP-1 defect image; tin ash defect image preprocessing It can realize the highlighting of tin dust defects in the original image, and obtain the TP-1 defect image; the pinhole defect image preprocessing can realize the image of the pinhole defect in the original image, and obtain the PP-1 defect image.
[0066] In the specific implementation process, the methods of scratch defect image prepro...
Embodiment 3
[0082] More specifically, the composite defect detection method based on semantic segmentation and target detection fusion model also includes the following steps:
[0083] S10: Using the intersection-union ratio and similarity synthesis method, calculate according to the prediction frame obtained by the Faster R-CNN network, and realize the evaluation of the performance of the Faster R-CNN network model.
[0084] More specifically, in the step S10, the intersection-over-union ratio and similarity synthesis method includes an intersection-over-union ratio calculation part and a similarity calculation part using a twin neural network. Through the comprehensive calculation of these two parts, the detection of Faster R- The number of true examples, false negative examples, false positive examples, and true negative examples in the CNN network model, so as to evaluate the performance of the Faster R-CNN network model; specifically:
[0085] Obtain the position, size and category i...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap