Deep Learning-Based Crack Damage Detection Using Convolutional Neural Networks

Proposes a CNN classifier that detects concrete cracks without hand-crafted features, trained on about 40k 256x256 image patches and applied to large photos with a sliding window. Tested on high-resolution photos of a structure not used for training, including glare, shadows and hairline cracks, it outperformed Canny and Sobel edge detection. The abstract does not describe UAV imagery; it is a ground-camera study that later UAV crack-detection work builds on.

Key findings

• About 98% accuracy on 256x256 patches (per abstract)
• Tested on 55 images of 5,888x3,584 px from a structure not used in training/validation
• Clearly better than Canny and Sobel edge detectors under glare, shadow and thin-crack conditions

Limitations

• Not UAV-specific per the abstract; images are not described as aerial
• Patch-level classification, not pixel-level crack width measurement
• Concrete cracks only; other damage types not covered