台灣因地勢陡峭，許多具有高災害潛勢的山區河川難以透過傳統泥沙粒徑調查方法取得準確的粒徑資料。然而，河道底床粒徑分布資料為評估河道輸砂能力、極端事件後沖刷潛勢及土石流發生條件之關鍵依據。隨著影像分析技術的快速發展，數位影像分析已成為突破傳統粒徑調查限制的重要工具，自Detert與Weitbrecht (2012) 開發出整合式自動化礫石粒徑判釋工具—BASEGRAIN軟體，該系統已成為當前影像篩分析領域中應用最為普及且穩定性高的自動顆粒判釋工具之一。
本研究延伸Chen (2023)之研究基礎，利用BASGRAIN V2.3針對非均勻粒徑組成條件下之表層泥沙粒徑進行判釋，並提出適用於不同複雜環境之參數調整建議。本研究採用兩種不同的分析方法：一為BASEGRAIN軟體內建的「fig60.14」選項，以Fehr線採樣法作為理論基礎；另一種則透過「fully detected elements」工作表，設定於分析區域內的均勻採樣點，於±120像素(px)範圍內搜尋最接近採樣點的顆粒。兩種分析方法與人工採樣法所得結果相比較，發現方法一粗粒徑區段分析結果出現高估、細粒徑區段出現低估的現象，是因為該模組模擬Fehr(1987)線採樣方法，會沿水平與垂直虛擬線掃描影像，並將與線相交的顆粒全部記錄。然而，在顆粒尺寸差異較大的情況下，較大顆粒因覆蓋面積較廣，容易被多條虛擬線重複計入，導致其權重相對放大，進而高估粗粒徑比例；相對地，細顆粒因交集次數少而被低估。相比之下，本研究提出的Python輔助模擬網格採樣法透過建立固定距離的121點網格採樣，確保每顆顆粒最多只被計算一次，並維持樣本在空間上的均勻代表性，因此在粗粒徑區段語系粒徑區段的分布曲線更為平滑且接近人工採樣結果，整體相對誤差控制在±10%之內，在細粒徑區段更是達到±5%精確度，整體分析效果明顯優於內建模組。
應用Basegrain軟體進行非均勻粒徑分析，有助於快速且精確地掌握河川底床顆粒粒徑分布資訊，並可作為洪水與土石流災害之預警評估、風險管理及防災減災規劃之參考依據。

In Taiwan, due to the steep terrain, traditional sediment grain sizes survey method often struggle to obtain accurate data in mountainous rivers with high disaster potential. Nevertheless, information on the distribution of riverbed sediment grain sizes is essential for evaluating sediment transport capacity, post-event scouring potential, and the conditions under which debris flows may occur. With the rapid advancement of image analysis technology, digital image processing has emerged as a critical tool for overcoming the limitations of conventional grain size surveys. Since Detert and Weitbrecht (2012) developed the integrated and automated gravel grain size analysis software known as BASEGRAIN, it has become one of the most widely adopted and reliable tools in the field of image-based grain size analysis.
Building on the foundation of Chen (2023), this study utilizes BASEGRAIN V2.3 to interpret surface sediment grain sizes under non-uniform grain-size composition conditions and proposes parameter adjustment strategies suitable for various complex environments. Two analytical approaches are adopted. The first relies on the built-in “fig60.14” option, which is based on Fehr’s (1987) line-sampling method. The second utilizes the “fully detected elements” worksheet, where evenly distributed sampling points are set within the analysis area, and the nearest particle within ±120 pixels is identified. When compared with manual sampling, the first approach tends to overestimate the coarse fraction and underestimate the fine fraction. This bias occurs because the module simulates line sampling by scanning the image with horizontal and vertical virtual lines, recording all intersected particles. Larger grains, due to their broader surface coverage, are more likely to be intersected by multiple lines, thus inflating their weights, whereas smaller grains are undercounted due to fewer intersections.
In contrast, the Python-assisted grid-sampling approach employs a fixed 121-point grid, ensuring each particle is counted only once while maintaining spatial uniformity. This method yields smoother grain-size distribution curves that align more closely with manual sampling, with overall relative error within ±10% and fine fractions accurate to ±5%. These results show that the proposed approach outperforms the built-in module and that applying BASEGRAIN to non-uniform grain-size analysis provides a rapid and reliable basis for sediment characterization, hazard assessment, and disaster risk management.

1. 丁怡瑄. (2020). 「正射影像應用於河床質辨識與流路變化探討」，[Master’s thesis, National Chiao Tung University].
2. 寺田康人、藤田一郎、浅見佳世和渡辺豊. (2015).「UAV による撮影画像を用いた洪水前後の砂州上粒度分布の計測」，土木学会論文集 B1(水工学), 71(4), I_919–I_924.
3. 陳嘉欣、邵允銓、王驥魁和吳富春. (2008). 「河床質粒徑分布之數位影像光篩分析」.農業工程學報, 54(4), 16-32.
4. 陳祥偉. (2023). 應用 UAV空拍影像及BASEGRAIN軟體進行河道泥沙粒徑分析之研究 [Master’s thesis, National Cheng Kung University].
5. 傅志偉和廖志中. (2004). 「河床質調查的方法、位置與頻率-以頭前溪為例」，[Master’s thesis, National Chiao Tung University].
6. 鍾政良和廖志中(2006)，「河床質調查位置與數量的合理性探討」，[Master’s thesis, National Chiao Tung University].
7. Bunte, K., & Abt, S. R. (2001). Sampling surface and subsurface particle-size distributions in wadable gravel-and cobble-bed streams for analyses in sediment transport, hydraulics, and streambed monitoring. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station.
8. Butler, J. B., Lane, S. N., & Chandler, J. H. (2001). Characterization of the structure of river-bed gravels using two-dimensional fractal analysis. Mathematical Geology, 33(3), 301–330.
9. Canny, J. (1986). A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, (6), 679–698.
10. Church, M., Kellerhals, R., & Day, T. J. (1987). River bed gravels: Sampling and analysis. In Sediment Transport in Gravel-bed Rivers (pp. 43–88).
11. Chardon, V., Piasny, G., & Schmitt, L.（2022）。 “Comparison of software accuracy to estimate the bed grain size distribution from digital images: A test performed along the Rhine River”。River Research and Applications, 38(2), 358–367.
12. Diplas, P., & Sutherland, A. J. (1988). "Sampling techniques for gravel sized sediments." Journal of Hydraulic Engineering, 114(5), 484–501.
13. Detert, M., & Weitbrecht, V. (2012). Automatic object detection to analyze the geometry of gravel grains–a free standalone tool. River Flow 2012 – Murillo (Ed.), Taylor & Francis Group, London. pp. 595–600.
14. Fehr, R. (1987). "Einfache bestimmung der korngrössenverteilung von geschiebematerial mit Hilfe der Linienzahlanalyse." Schweizer Ingenieur und Architekt 105(38)， 1104-1109.
15. Graham, D. J., Reid, I., & Rice, S. P. (2005a). Automated sizing of coarse-grained sediments: Image-processing procedures. Mathematical Geology, 37(1), 1–28.
16. Graham, D. J., Rice, S. P., & Reid, I. (2005b). A transferable method for the automated grain sizing of river gravels. Water Resources Research, 41(7).
17. Hey, R. D., & Thorne, C. R. (1983). Accuracy of surface samples from gravel bed material. Journal of Hydraulic Engineering, 109(6), 842–851.
18. Hunter, J. D. (2007). Matplotlib: A 2D graphics environment. Computing in Science & Engineering, 9(3), 90–95.
19. Harris, C. R., Millman, K. J., van der Walt, S. J., Gommers, R., Virtanen, P., Cournapeau, D., Wieser, E., Taylor, J., Berg, S., Smith, N. J., Kern, R., Picus, M., Hoyer, S., & Oliphant, T. E. (2020). Array programming with NumPy. Nature, 585(7825), 357–362.
20. McKinney, W. (2010). Data structures for statistical computing in Python. Proceedings of the 9th Python in Science Conference, 51–56.
21. Otsu, N. (1979). A threshold selection method from gray-level histograms. IEEE Transactions on Systems, Man, and Cybernetics, 9(1), 62–66.
22. Python Software Foundation. (n.d.). Python (Version 3.x) [Computer software]. https://www.python.org
23. Serra, J. (1982). Image analysis and mathematical morphology. Academic Press.
24. Vincent, L., & Soille, P. (1991). Watersheds in digital spaces: An efficient algorithm based on immersion simulations. IEEE Transactions on Pattern Analysis and Machine Intelligence, 13(6), 583–598.
25. Wolman, M. G. (1954). A method of sampling coarse river-bed material. Eos, Transactions American Geophysical Union, 35(6), 951–956.