We present SEM2DXF, an open-source Python pipeline that converts scanning electron microscope (SEM) images of biological nanostructures into electron-beam lithography (EBL)-ready closed-polygon DXF pattern files, enabling systematic cloning of natural surface architectures. Applied to cicada (Cicada sp.) wing nanopillar arrays imaged at two magnifications — ×19,000 (5.29 nm px−1) and ×50,000 (2.00 nm px−1) — the pipeline automatically detects the SEM annotation band, calibrates the pixel scale from the scale bar, segments individual pillars by a hybrid Hough-circle/contour algorithm, completes truncated boundary features by ellipse fitting, and exports each pillar as a closed-polygon LWPOLYLINE entity in DXF format. Because EBL vector systems fill closed polygons as solid regions, this strategy reduces the number of beam positions by more than 8-fold compared with raster-scan exposure, directly cutting write time. Using ×50,000-derived patterns as the absolute-dimensional reference (individual pillars resolved; D = 115 nm, pitch = 155 nm), we characterise the dose-response of PMMA nanopillar arrays across 16 dose levels (200–700 µC cm−2). Atomic force microscopy (AFM) at eight dose levels provides height measurements and defines a three-dimensional process window (380–490 µC cm−2) that simultaneously optimizes pillar diameter, height (~200 nm), and aspect ratio relative to the cicada reference. Pillar-by-pillar registration of a fabricated array against the biological template (N = 1,174 interior pairs, boundary features excluded) yields a median positional mismatch of 32 nm — less than half the nearest-neighbour pitch — and a mean systematic displacement of (-8, +2) nm, confirming faithful spatial transcription. Manual ImageJ measurements confirm that EBL-fabricated PMMA pillars are approximately cylindrical (top ≈ base ≈ 100–110 nm), whereas cicada pillars are conical (base ~130 nm, top ~50 nm); this intrinsic morphological difference is discussed as an expected consequence of positive-tone resist processing.