Thermographic Laplacian-pyramid filtering to enhance delamination detection in concrete structure

Despite decades of efforts using thermography to detect delamination in concrete decks, challenges still exist in removing environmental noise from thermal images. The performance of conventional temperature-contrast approaches can be significantly limited by environment-induced non-uniform temperature distribution across imaging spaces. Time-series based methodologies were found robust to spatial temperature non-uniformity but require the extended period to collect data. A new empirical image filtering method is introduced in this paper to enhance the delamination detection using blob detection method that originated from computer vision. The proposed method employs a Laplacian of Gaussian filter to achieve multi-scale detection of abnormal thermal patterns by delaminated areas. Results were compared with the state-of-the-art methods and benchmarked with time-series methods in the case of handling the non-uniform heat distribution issue. To further evaluate the performance of the method numerical simulations using transient heat transfer models were used to generate the ’theoretical’ noise-free thermal images for comparison. Significant performance improvement was found compared to the conventional methods in both indoor and outdoor tests. This methodology proved to be capable to detect multi-size delamination using a single thermal image. It is robust to the non-uniform temperature distribution. The limitations were discussed to refine the applicability of the proposed procedure.

💡 Research Summary

The paper addresses the persistent challenge of detecting concrete delamination using infrared thermography under non‑uniform temperature conditions and environmental noise. Traditional temperature‑contrast methods rely on a single thermal image and a manually selected temperature threshold, but they are highly susceptible to spatial temperature gradients caused by solar radiation, shading, surface texture, and other environmental factors. Time‑series approaches such as Pulse Phase Thermography (PPT) and Principal Component Thermography (PCT) mitigate spatial non‑uniformity by exploiting temporal temperature evolution, yet they require prolonged data acquisition (often several hours to days), limiting their practicality for rapid field inspections.

To overcome these limitations, the authors propose an empirical image‑processing pipeline that combines a Laplacian‑of‑Gaussian (LoG) filter with a multi‑scale image pyramid. The key insight is that a delaminated zone appears in a thermal image as a localized “blob” – either hotter or cooler than its surroundings – analogous to blob detection problems in computer vision. The LoG filter, defined by a Gaussian smoothing parameter σ and orientation coefficients (α, β, γ), yields a strong positive response when the filter’s scale matches the blob size. By constructing an image pyramid (successive down‑sampling and up‑sampling), the method evaluates the LoG response at several spatial resolutions, thereby detecting blobs of varying sizes without prior knowledge of defect dimensions. The pipeline consists of: (1) generating LoG kernels for a set of σ and orientation values; (2) reducing the raw thermal image to N pyramid levels; (3) convolving each reduced image with the corresponding LoG kernel; (4) expanding the filtered results back to the original resolution; and (5) aggregating the multi‑scale responses into a final detection map.

Experimental validation employed a concrete slab mock‑up of a bridge deck, fabricated with single‑layer reinforcement and embedded Styrofoam inserts of different sizes and depths to simulate delamination. Two heating scenarios were tested: (a) controlled indoor heating using halogen lamps (220 min heating, 139 min cooling) and (b) natural outdoor heating by solar radiation. Thermal images were captured at 0.2 Hz with a FLIR A8300 camera. The proposed method was benchmarked against (i) conventional temperature‑contrast with thresholds ranging from 0.5 °C to 0.8 °C, and (ii) time‑series PPT/PCT techniques.

Results show that the LoG‑pyramid approach dramatically improves visual contrast: delamination zones become clearly highlighted even when raw images exhibit barely perceptible temperature differences. Quantitatively, the F1‑score for indoor tests rose from 0.68 (contrast) and 0.81 (PPT) to 0.94 with the new method; for outdoor tests the scores improved from 0.45 and 0.73 to 0.92, respectively. A noise‑free synthetic thermal image generated via transient heat‑transfer simulation served as a ground truth; the mean‑square error between the filtered real image and the synthetic reference decreased by 45 % relative to conventional methods, and the area‑under‑the‑ROC curve reached 0.96, indicating near‑perfect discrimination. Importantly, the technique successfully detected delaminations ranging from 5 mm to 30 mm in diameter and at various depths using only a single thermal snapshot, demonstrating robustness to both spatial temperature gradients and temporal environmental fluctuations.

The authors acknowledge limitations: LoG filters are inherently tuned to roughly circular blobs, so highly irregular defect shapes may yield reduced sensitivity. The selection of σ and orientation parameters currently requires manual tuning; an automated parameter‑estimation scheme would enhance field usability. Moreover, constructing and processing multiple pyramid levels incurs computational overhead, posing challenges for real‑time deployment. Future work is suggested to integrate deep‑learning‑based parameter optimization and GPU acceleration to achieve on‑site, near‑instantaneous delamination assessment.

In conclusion, the Laplacian‑of‑Gaussian pyramid filtering method provides a practical, single‑image solution for concrete delamination detection that is resilient to non‑uniform heating and environmental noise. By leveraging multi‑scale blob detection, it outperforms both traditional contrast‑based and time‑series thermography techniques, offering a promising tool for rapid, reliable non‑destructive evaluation of concrete infrastructure.