16:30 〜 16:45
[PPS03-04] Boulders in the Tlanuwa Regio of Bennu: A View from CNN-based Automatic Rock Identification
キーワード：岩石粒子、自動識別、ベヌー
NASA’s Origins, Spectral Interpretation, Resource Identification and Security-Regolith Explorer (OSIRIS-REx) mission [1] has explored a near-earth asteroid Bennu. The remote-sensing data revealed that this B-type asteroid [2] has a spheroidal spinning-top shape [1] and is covered by dark (low albedo) surface materials [e.g., 3]. Despite its small size of <500 m in diameter, high-resolution images obtained by OSIRIS-REx Camera Suite (OCAMS) [4] resolved an interesting surface diversity of Bennu [4] with a variety of topographic/geologic features and, especially, fragmented rock particles (boulders, cobbles, and pebbles) [5, 6]. This suggests the influence of impact or thermal processes occurring on the surface of Bennu [7]. OCAMS images also showed the signs of other geological processes, including particle ejections from the surface due to its low escape velocity [8], and mass wasting, where the surface materials move to topographic lows. These would lead to the accumulations of clusters of rock particles, depletion of small craters, and/or infilling of large craters [9]. Analyzing the distributions of rock particles on Bennu can thus contribute to understanding the origins and evolutions of its active surface. However, boulder identification can be time-consuming and sometimes challenging because boulder outlines are difficult to distinguish from the background due to irregular particle shapes, overlapping particles, and image resolution limits.
Here, we develop a Convolutional Neural Network (CNN)-based computational approach to identify rock particles automatically. Twenty-four OCAMS images (~5 cm/pix) are selected as a training dataset. We manually extract >40,000 outlines of rocks, which are used for the model training. Our model can automatically identify rock particles with an accuracy of ~80%. We apply this method to eighty OCAMS images (~5 cm/pix) of Tlanuwa Regio, a large boulder-rich area on Bennu’s surface. The outlines of rock particles are mapped on a three-dimensional shape model derived from OSIRIS-REx Laser Altimeter (OLA) [10] data. As a result, we calculate the cumulative size-frequency distribution with a power-law index of -2.2. By measuring the largest (a) and second the largest (b) dimensions of rock particles, the ratio of b/a is derived, showing the mean ratio of 0.63. This result is consistent with the value of Eros, Itokawa, and Ryugu [11]. The orientations of the longest axis of rock particles are also obtained and show that most particles have orientations parallel to the equator. Some studies suggested that boulders’ longest axes can be preferentially oriented perpendicular to the slope direction in the gravel migration [12]. Considering a south-to-north slope direction in Tlanuwa Regio under the current Bennu’s gravitational potential [13], the EW orientation of rock particles may indicate that surface materials have dominantly moved from the midlatitude to the equator.
References
[1] Lauretta D. S. et al. (2017) Space Sci. Rev. 212, 925-984. [2] Hergenrother C. W. et al. (2013) Icarus 226, 663-670. [3] Lauretta D. S. et al. (2019) Nature 568, 55-60. [4] Rizk B. et al. (2018) Space Sci. Rev. 214, 26. [5] Rizk B. et al. (2018) Space Sci. Rev. 214, 26. [6] Walsh K. J. et al. (2019) Nat. Geosci. 12, 399. [7] DellaGiustina D. N. et al. (2019) Nat. Astron. 3, 341-351. [8] Lauretta D. S. et al. (2019) Science, 366, eaay3544. [9] Jawin E. R. et al. (2020) JGR 125, e2020JE006475. [10] Barnouin O. S. et al. (2020) PSS 180, 104764. [11] Michikami and Hagermann (2021) Icarus 357, 114282. [12] Miyamoto H. et al. (2007) Science 316, 1011-1014. [13] Scheeres D. J. et al. (2019) Nat. Astron. 3, 352-361.
Here, we develop a Convolutional Neural Network (CNN)-based computational approach to identify rock particles automatically. Twenty-four OCAMS images (~5 cm/pix) are selected as a training dataset. We manually extract >40,000 outlines of rocks, which are used for the model training. Our model can automatically identify rock particles with an accuracy of ~80%. We apply this method to eighty OCAMS images (~5 cm/pix) of Tlanuwa Regio, a large boulder-rich area on Bennu’s surface. The outlines of rock particles are mapped on a three-dimensional shape model derived from OSIRIS-REx Laser Altimeter (OLA) [10] data. As a result, we calculate the cumulative size-frequency distribution with a power-law index of -2.2. By measuring the largest (a) and second the largest (b) dimensions of rock particles, the ratio of b/a is derived, showing the mean ratio of 0.63. This result is consistent with the value of Eros, Itokawa, and Ryugu [11]. The orientations of the longest axis of rock particles are also obtained and show that most particles have orientations parallel to the equator. Some studies suggested that boulders’ longest axes can be preferentially oriented perpendicular to the slope direction in the gravel migration [12]. Considering a south-to-north slope direction in Tlanuwa Regio under the current Bennu’s gravitational potential [13], the EW orientation of rock particles may indicate that surface materials have dominantly moved from the midlatitude to the equator.
References
[1] Lauretta D. S. et al. (2017) Space Sci. Rev. 212, 925-984. [2] Hergenrother C. W. et al. (2013) Icarus 226, 663-670. [3] Lauretta D. S. et al. (2019) Nature 568, 55-60. [4] Rizk B. et al. (2018) Space Sci. Rev. 214, 26. [5] Rizk B. et al. (2018) Space Sci. Rev. 214, 26. [6] Walsh K. J. et al. (2019) Nat. Geosci. 12, 399. [7] DellaGiustina D. N. et al. (2019) Nat. Astron. 3, 341-351. [8] Lauretta D. S. et al. (2019) Science, 366, eaay3544. [9] Jawin E. R. et al. (2020) JGR 125, e2020JE006475. [10] Barnouin O. S. et al. (2020) PSS 180, 104764. [11] Michikami and Hagermann (2021) Icarus 357, 114282. [12] Miyamoto H. et al. (2007) Science 316, 1011-1014. [13] Scheeres D. J. et al. (2019) Nat. Astron. 3, 352-361.