老旧图表数字化咨询：无原始数据的化学列线图（硫酸焓列线图为例）数学化可行性探讨

Absolutely, your proposed approach is fully feasible—and you’ve come to the right forum to ask about this! Stack Overflow is a perfect spot for questions related to scientific data extraction and image processing workflows like this.

Here’s a breakdown of why this works and how to execute it:

Core Logic: Column Charts Are Mathematical Relationships in Visual Form #

Column charts (like your sulfuric acid enthalpy nomogram) are just visualizations of underlying mathematical links between variables. By mapping the image’s pixel coordinates to the known numerical scales on the chart’s grid, you can reverse-engineer raw data points directly from the curves.

Step-by-Step Workflow (Code or No-Code Options) #

• Image Preprocessing: First, clean up the scanned chart:
  • Convert to grayscale, apply noise reduction (e.g., Gaussian blur) to fix faded or scratched areas common in old documents.
  • Calibrate the grid: Mark key tick points on the axes (e.g., "0°C" at pixel (x1,y1), "100°C" at (x2,y2)) and build a mapping between pixel positions and actual numerical values. This mapping might be linear, logarithmic, or non-linear depending on the chart’s axis type—verify with multiple calibration points to ensure accuracy.
• Curve Extraction:
  • Use image processing libraries (like OpenCV in Python with cv2.findContours()) to detect curve pixels, or use dedicated scientific digitization tools (no coding needed) that auto-detect lines.
  • For faded or broken curves, you may need to manually trace short segments to fill gaps, but most tools handle minor imperfections smoothly.
• Data Conversion & Refinement:
  • Convert extracted pixel coordinates to real-world values using your calibrated grid mapping.
  • Use interpolation (e.g., spline interpolation via scipy.interpolate in Python) to generate a smooth, continuous dataset from discrete extracted points. If you know the underlying chemical model for sulfuric acid enthalpy, fitting the data to that model can further boost accuracy.

Potential Pitfalls to Watch For #

• Worn/Faded Charts: If scale ticks are partially illegible, use multiple overlapping calibration points to minimize error.
• Non-Linear Axes: Don’t assume all axes are linear—many scientific nomograms use logarithmic or custom scales, so your calibration step needs to account for that.
• Overlapping Curves: If multiple curves are close together, use color thresholding (if the original chart has colored lines) or manual selection to separate them.

内容的提问来源于stack exchange，提问作者Capt.Krusty