Fig 1.
Common types of pavement cracks on urban asphalt roads (Source from: The author filmed it themselves; Chengdu, Jinxiu Avenue, longitude 103 ° 41 ′ −103 ° 55 ′ E, latitude 30 ° 36 ′ −30 ° 52 ′ N).
Fig 2.
Flowchart of DL principle.
Fig 3.
Schematic diagram of U-net structure.
Fig 4.
The operational structure of the U-ResNet.
Fig 5.
The operation structure of the DBCCN.
Fig 6.
The operating structure of ResNeXt classification network.
Fig 7.
The urban road surface crack detection method.
Table 1.
Types of urban road surface damage and weight calculation results in the dataset.
Fig 8.
Performance comparison results of three methods.
Fig 9.
Classification performance of ResNeXt on four types of pavement cracks in the Crack500 dataset.
Fig 10.
Classification performance of ResNeXt for 4 kinds of pavement cracks in CFD.
Fig 11.
Segmentation and detection results of road surface crack images (Image source: The author filmed it themselves; Chengdu, Jinxiu Avenue, longitude 103 ° 41 ′ −103 ° 55 ′ E, latitude 30 ° 36 ′ −30 ° 52 ′ N).
Fig 12.
Classification results of ResNeXt classification network for road crack images (Image source: The author filmed it themselves; Chengdu, Jinxiu Avenue, longitude 103 ° 41 ′ −103 ° 55 ′ E, latitude 30 ° 36 ′ −30 ° 52 ′ N).
Fig 13.
Simulation results of detection and classification using three methods (Source from: The author filmed it themselves; Chengdu, Jinxiu Avenue, longitude 103 ° 41 ′ −103 ° 55 ′ E, latitude 30 ° 36 ′ −30 ° 52 ′ N).
Table 2.
Statistical test results.
Table 3.
Ablation experiment.
Table 4.
Comparison of the overall performance of different advanced methods across various deployment platforms.