Three-dimensional segmentation of air-void system in hardened concrete using photometric stereo and artificial intelligence methods
dc.contributor.advisor | Wang, Feng | |
dc.contributor.author | Tao, Jueqiang | |
dc.contributor.committeeMember | Torres, Anthony | |
dc.contributor.committeeMember | Gao, Byron | |
dc.contributor.committeeMember | Shi, Xijun | |
dc.contributor.committeeMember | Huang, Yaxiong | |
dc.date.accessioned | 2022-08-31T15:09:05Z | |
dc.date.available | 2022-08-31T15:09:05Z | |
dc.date.issued | 2022-08 | |
dc.description.abstract | A well-distributed air-void system inside a hardened concrete can help protect the concrete structure from being damaged by freeze-thaw cycles. The concrete air-void petrographic test, which is specified in ASTM C457, is the standard test procedure for characterizing the air-void system and evaluating the freeze-thaw performance of hardened concrete samples. Specifically, the linear-traverse method (Procedure A) and the modified point-count method (Procedure B) are two manual methods described in ASTM C457. These two methods are based on manually microscopical observation. In addition, considering the fact that air voids in hardened concrete surfaces are difficult to be observed with the naked eye due to the low contrast between air voids and hardened cement paste, the identification of air voids is difficult and subjective. Consequently, Procedures A and B are error-prone and labor-intensive. The contrast enhanced method (Procedure C) was introduced in ASTM C457 as a computer-aided air-void system measurement method. Procedure C requires the aid of contrast enhancement, in which the air voids are manually highlighted with color on the concrete surface by an experienced human rater. Then, the air-void system can be automatically measured with the assistance of computer-aided image processing techniques. However, during the contrast enhancement procedure, voids in aggregates and cracks can also be filled with white powders, which is not wanted and needs to be carefully checked and avoided. As a result, although Procedure C reduces human labor to some degree, it is still labor-intensive and time-intensive. Therefore, a more efficient air-void measurement method that can automatically identify the air voids from hardened concrete surfaces needs to be developed. This dissertation aims to investigating the detection of the air-void system in hardened concrete surfaces using three-dimensional (3D) reconstruction and Artificial Intelligence (AI) techniques. The proposed method can be considered as an extension of Procedure C but is free from the manual contrast enhancement procedures. In this dissertation, a new air-void detection method was proposed to automatically segment the air-void system from the solid phase in an automated manner. The proposed method includes: 1) a 3D image reconstruction system based on two-dimensional (2D) image data collection, and a delicate and novel engineering design for hardware; 2) an automated air-void segmentation method without using contrast enhancement preprocessing. According to our knowledge, the method is innovative and has not been attempted before. Unlike other existing methods that have been used in the research and industry, the method we proposed has the following potential advantages: less labor/time-intensity, higher cost-effectiveness, and higher accuracy. The research results showed that the basic photometric stereo method is able to contrast the air voids in hardened concrete surfaces to some extent using the 3D nature of air voids. It took 10-15 seconds for the basic photometric stereo method to reconstruct the surface normal image for each concrete sample. However, the polished concrete surface cannot be considered as an ideal Lambertian surface, and some air-void like noises can be generated due to the bias of the basic photometric stereo method. The deep learning based image segmentation method provided good robustness to differentiate most of the noises from true positive air voids. The experimental results showed that the deep learning based methods can accurately distinguish air voids from hardened concrete images with the detection accuracy of over 0.9 in only less than a minute. The accuracy rates for air content, specific surface, and spacing factor were 0.92, 0.91, and 0.89, respectively. In addition, considering the limitations of using 2D air-void segmentation for concrete petrographic analysis, the reliability of using the Saltykov method to restore the 3D air-void radius was also evaluated. The recovered spatial air-void distribution can be used to simulate the actual air-void system inside hardened concrete and then provide insights into the percentage of the hardened concrete paste that is protected by the air-void system. In this research, the unfolding results of both in-section air voids and out-section air voids are evaluated using the Minkowski Distance metric. The research results showed that the up-to-date methods can accurately unfold the size distribution of in-section air voids, while the methods failed to achieve an accurate estimation for out-section air voids. | |
dc.description.department | Materials Science, Engineering, and Commercialization | |
dc.format | Text | |
dc.format.extent | 156 pages | |
dc.format.medium | 1 file (.pdf) | |
dc.identifier.citation | Tao, J. (2022). Three-dimensional segmentation of air-void system in hardened concrete using photometric stereo and artificial intelligence methods (Unpublished dissertation). Texas State University, San Marcos, Texas. | |
dc.identifier.uri | https://hdl.handle.net/10877/16094 | |
dc.language.iso | en | |
dc.subject | hardened concrete | |
dc.subject | air-void systems | |
dc.subject | ASTM C457 | |
dc.subject | image segmentation | |
dc.subject | convolution neural network | |
dc.subject | stereological analysis | |
dc.title | Three-dimensional segmentation of air-void system in hardened concrete using photometric stereo and artificial intelligence methods | |
dc.type | Dissertation | |
thesis.degree.department | Materials Science, Engineering, and Commercialization Program | |
thesis.degree.discipline | Materials Science, Engineering, and Commercialization | |
thesis.degree.grantor | Texas State University | |
thesis.degree.level | Doctoral | |
thesis.degree.name | Doctor of Philosophy |
Files
Original bundle
1 - 1 of 1