UAV-Based Image and LiDAR Fusion for Pavement Crack Segmentation

Area of Science:

• Civil Engineering
• Computer Vision
• Remote Sensing

Background:

• Manual pavement inspection is time-consuming and labor-intensive.
• Existing unmanned aerial system (UAS) methods use only image or LiDAR data, missing complementary information.
• Developing automated, accurate pavement distress detection is crucial for road safety and maintenance.

Purpose of the Study:

• To explore the feasibility of fusing UAS-based imaging and low-cost LiDAR data for enhanced pavement crack segmentation.
• To evaluate the performance of a deep convolutional neural network (DCNN) model using fused data.
• To investigate the impact of different data fusion combinations (RGB, intensity, elevation) on crack and sealed crack detection.

Main Methods:

• Collected three datasets using two UASs at varying altitudes.
• Investigated two pavement distress types: cracks and sealed cracks.
• Employed a modified U-net DCNN with residual blocks for segmentation.
• Compared four fusion combinations: RGB, RGB + intensity, RGB + elevation, RGB + intensity + elevation.

Main Results:

• The DCNN model demonstrated superior accuracy and generalizability compared to state-of-the-art networks.
• Fusing elevation data with RGB images increased recall by 2% for crack detection.
• LiDAR intensity data fusion reduced precision, recall, and F-measure due to sensor limitations.
• LiDAR data improved sealed crack segmentation by 4-7% across datasets.
• Higher resolution LiDAR data at lower altitudes improved crack detail detection but decreased precision.

Conclusions:

• Fusing UAS-based imaging and LiDAR data, particularly elevation, can enhance pavement crack segmentation.
• The effectiveness of LiDAR data fusion depends on data quality and the type of pavement distress.
• The developed DCNN model shows significant potential for automated pavement condition assessment.
• Further research with higher-quality LiDAR data could overcome current limitations and improve fusion benefits.