Building surface crack detection method based on U-Net
A surface crack and detection method technology, applied in the field of detection, can solve problems such as false detection and difficulty in ensuring data, and achieve the effect of improving segmentation accuracy, ensuring robustness and invariance, and powerful real-time data enhancement
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment Construction
[0028] The technical solution of the present invention will be specifically described below in conjunction with the accompanying drawings.
[0029] The present invention first constructs a small data set. A total of 30 raw images containing different types of cracks were obtained under various conditions to enhance the adaptability of the neural network. After annotating the original images, they are cropped into small patch images with a resolution of 512×512 pixels for training, validation and testing processes. Utilizes the improved U-net's powerful network architecture to detect cracks on building surfaces. The input image size of our network is 512×512 pixels, and the output is a mask image showing each pixel class. Using the sliding window technique, the model can handle images of larger size. The proposed method without post-processing achieves acceptable accuracy in various complex backgrounds, and outperforms traditional edge detection methods.
[0030] Concrete t...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


