Image-Based Monitoring of Cracks: Effectiveness Analysis of an Open-Source Machine Learning-Assisted Procedure

Abstract

The proper inspection of a cracks pattern over time is a critical diagnosis step to provide a thorough knowledge of the health state of a structure. When monitoring cracks propagating on a planar surface, adopting a single-image-based approach is a more convenient (costly and logistically) solution compared to subjective operators-based solutions. Machine learning (ML)- based monitoring solutions offer the advantage of automation in crack detection; however, complex and time-consuming training must be carried out. This study presents a simple and automated ML-based crack monitoring approach implemented in open sources software that only requires a single image for training. The effectiveness of the approach is assessed conducting work in controlled and real case study sites. For both sites, the generated outputs are significant in terms of accuracy (~1 mm), repeatability (sub-mm) and precision (sub-pixel). The presented results highlight that the successful detection of cracks is achievable with only a straightforward ML-based training procedure conducted on only a single image of the multi-temporal sequence. Furthermore, the use of an innovative camera kit allowed exploiting automated acquisition and transmission fundamental for Internet of Things (IoTs) for structural health monitoring and to reduce user-based operations and increase safety.

Keywords: Ilastik; ImageJ; crack; image processing; machine learning; monitoring; segmentation.

Figure 1

Workflow of the proposed algorithm. Each module includes implemented (blue) and optional actions (light blue).

Figure 2

Bixion protective case and components of the acquisition kit including (1) padlock slot; (2) tightening screw; (3) shelf; (4) Bixicon unit; (5) camera slider; (6) DSLR camera; (7) crystal glass flatport ~105 mm UV filter and (8) grommet.

Figure 3

The linear features calculated by the Ridge Detection method include the crack edge lines (green), the maximum line (red) and multiple width segments (blue).

Figure 4

(a) The crack pattern generated in AutoCAD for the laboratory test (the red circles indicate the two targets used for scaling purposes, and letters ‘a’, ‘b’ and ‘c’ show the three cracks considered to assess the proposed approach). (b) The acquisition kit installed at the laboratory test site (the red circle shows the 4G modem).

Figure 5

(a) The acquisition kit installed indoor for the on-site test. (b) Zoom in showing the crack and the locations of the six RoI used to quantify the precision of the proposed approach.

Figure 6

Segmentation outputs generated by the CDM for the laboratory test dataset.

Figure 7

Width (red box) and length (black box) values (in cm) along the three representative cracks (a–c). Each box includes CrackID, values from the CAD drawing and values estimated with the proposed approach (in bold).

Figure 8

An example of not uniform illumination conditions on multi-temporal sequence of images acquired with natural daylight (a–c) and the respective optimized images (d–f) and segmentation outputs (g–i).