5:15 PM - 6:45 PM
[MGI28-P04] Establishment of rock shape analysis method using 3D scanner:
Application to the Analysis of Pyroclastic Flow Deposits from the 1929 Eruption of Hokkaido-Komagatake Volcano.
Keywords:3D Scanner, volume, Density, Principal Component Analysis, pyroclastic flow
The density of the individual particles that make up a pyroclastic flow is important to measure because it is the dominant factor in its flow deposition mechanism. The shape of the particles is also considered to reflect the flow mechanism (Taddeucci & Palladin, 2002) and is equally important. For these reasons, we are developing a simple method to measure the density and analyze the shape of pumice and other rocks.
In general, volume measurement is more complex than mass measurement when determining the density of porous rock, and various methods have been proposed in the past (e.g. Gardner et al., 1996; Houghton and Wilson, 1989; Sasaki & Katsui, 1981; Suzuki & Fujii, 2010).
In recent years, 3D scanners, which use laser technology to scan and digitize three-dimensional objects, have become widely used in reverse engineering. Among them, non-contact types allow objects to be measured without destruction or contact, and by analyzing the obtained models, volume and shape characteristics can be determined with high accuracy and convenience. However, the high cost of the equipment has limited its use in the natural sciences. Recently, however, 3D scanners costing tens of thousands of yen have appeared on the market, reducing the barriers to adoption.In this study, we established a procedure for 3D analysis of natural rocks using a relatively inexpensive 3D scanner and validated its measurement accuracy.
In this study, a commercially available low cost (less than $650 USD) POP2 (from Revopoint International Limited Inc., Hong Kong) is used to measure a 3D model of the rocks. The rock analysis procedure is as follows: First, the sample is placed on a turntable and point cloud data is acquired using the 3D scanner. Mesh data is generated from the point cloud data and 3D model viewing software is used to check for anomalies (gaps, overlaps, etc.) in the model. Bulk volume, surface area, major and minor axes are then determined using a shape analysis program developed by the authors. When calculating the major axis, which is important in shape analysis, we tried methods such as oriented bounding box (OBB) and principal component analysis (PCA: Cruz-Matías et al., 2019), an unsupervised machine learning algorithm. In addition, we aimed to reduce the computational time of these programs by using CPU parallel processing. Measurement accuracy was verified by repeating the measurement error ten times on 14 pumice samples (-3 phi to -8 phi) from the pyroclastic flow of the 1929 eruption of Hokkaido-Komagatake volcano. The bulk volume of the same samples was also determined using the water displacement method (Takeuchi et al., 2023,2024) for comparison.
The repeatability error of the bulk volume was less than 0.16%. Accuracy decreased for small pumice but remained below 0.55%. It was difficult to measure stones less than 1 cm in length, which is below the specifications of the 3D scanner. The coefficient of determination in a regression equation between this method and the water displacement method was 0.999. When searching for the major axis, there was scatter in the direction of the major axis with OBB, and the standard error rate of the major axis length was ~1.9%. However, when PCA was used, the directions were almost aligned and the standard error rate was mostly around 0.1%, with a maximum of 0.6%.
Thus, 3D scanner measurements can be carried out using the procedure established in this study, which is simple, reliable and allows measurements to be performed in the field, providing data with an accuracy that can withstand much discussion for various applications in the natural sciences.
In the future, this method will be used to analyze the pumice obtained to analyze the total grain size distribution of pyroclastic flow deposits from the 1929 eruption of Hokkaido-Komagatake volcano (Uesawa et al., 2024), and to study the dynamics of the pyroclastic flow interior.
In general, volume measurement is more complex than mass measurement when determining the density of porous rock, and various methods have been proposed in the past (e.g. Gardner et al., 1996; Houghton and Wilson, 1989; Sasaki & Katsui, 1981; Suzuki & Fujii, 2010).
In recent years, 3D scanners, which use laser technology to scan and digitize three-dimensional objects, have become widely used in reverse engineering. Among them, non-contact types allow objects to be measured without destruction or contact, and by analyzing the obtained models, volume and shape characteristics can be determined with high accuracy and convenience. However, the high cost of the equipment has limited its use in the natural sciences. Recently, however, 3D scanners costing tens of thousands of yen have appeared on the market, reducing the barriers to adoption.In this study, we established a procedure for 3D analysis of natural rocks using a relatively inexpensive 3D scanner and validated its measurement accuracy.
In this study, a commercially available low cost (less than $650 USD) POP2 (from Revopoint International Limited Inc., Hong Kong) is used to measure a 3D model of the rocks. The rock analysis procedure is as follows: First, the sample is placed on a turntable and point cloud data is acquired using the 3D scanner. Mesh data is generated from the point cloud data and 3D model viewing software is used to check for anomalies (gaps, overlaps, etc.) in the model. Bulk volume, surface area, major and minor axes are then determined using a shape analysis program developed by the authors. When calculating the major axis, which is important in shape analysis, we tried methods such as oriented bounding box (OBB) and principal component analysis (PCA: Cruz-Matías et al., 2019), an unsupervised machine learning algorithm. In addition, we aimed to reduce the computational time of these programs by using CPU parallel processing. Measurement accuracy was verified by repeating the measurement error ten times on 14 pumice samples (-3 phi to -8 phi) from the pyroclastic flow of the 1929 eruption of Hokkaido-Komagatake volcano. The bulk volume of the same samples was also determined using the water displacement method (Takeuchi et al., 2023,2024) for comparison.
The repeatability error of the bulk volume was less than 0.16%. Accuracy decreased for small pumice but remained below 0.55%. It was difficult to measure stones less than 1 cm in length, which is below the specifications of the 3D scanner. The coefficient of determination in a regression equation between this method and the water displacement method was 0.999. When searching for the major axis, there was scatter in the direction of the major axis with OBB, and the standard error rate of the major axis length was ~1.9%. However, when PCA was used, the directions were almost aligned and the standard error rate was mostly around 0.1%, with a maximum of 0.6%.
Thus, 3D scanner measurements can be carried out using the procedure established in this study, which is simple, reliable and allows measurements to be performed in the field, providing data with an accuracy that can withstand much discussion for various applications in the natural sciences.
In the future, this method will be used to analyze the pumice obtained to analyze the total grain size distribution of pyroclastic flow deposits from the 1929 eruption of Hokkaido-Komagatake volcano (Uesawa et al., 2024), and to study the dynamics of the pyroclastic flow interior.