You can extract drawing dimensions manually, from native PDF text, with optical character recognition (OCR), or with a hybrid drawing-analysis workflow. The hard part is not collecting numbers; it is preserving the correct characteristic, including its tolerance, symbols, quantity, feature relationship, source location, and revision. Verify every extracted record against the controlled drawing before using it for inspection.
Decide what the record must preserve
Contents
A number by itself is not a useful inspection characteristic. A reviewable record should keep the source representation alongside any normalized fields.
| Field | Why keep it | Example |
|---|---|---|
| Characteristic ID | Links drawing and report | 05 |
| Sheet / zone / source region | Lets a reviewer find the origin | Sheet 1, upper-right pattern |
| Raw callout | Preserves the reviewed source wording | 4X Ø8.0 ±0.1 THRU |
| Normalized requirement | Supports sorting, calculations, or mapping | 4 holes; diameter 8.0 mm; ±0.1; through |
| Type / quantity / tolerance | Retains the rule and repeated-feature context | Diameter; quantity 4; upper/lower ±0.1 |
| Geometric dimensioning and tolerancing (GD&T) and datum context | Prevents separating a control from its references | Position ⌀0.30 | A | B | C |
| Review state and source record | Shows whether it was accepted, corrected, or added | Corrected by reviewer; source crop retained |
| Revision and tolerance source | Prevents stale data or unverified defaults | Drawing Rev B; tolerance note 1 |
Choose an extraction method
| Method | Good starting point | Watch for |
|---|---|---|
| Manual spreadsheet entry | Small, clear drawing and high reviewer control | Transcription errors and revision rework |
| Native PDF text extraction | Digital PDF with encoded text | Separate text objects, wrong reading order, or outlined characters |
| Raster OCR | Scanned, flattened, or outlined content after rendering | Symbols, decimals, rotated text, skew, and grouping errors |
| Hybrid, drawing-aware workflow | Mixed documents or many linked characteristics | Review, correction, and source traceability still required |
Technical research on engineering drawings describes why spatial grouping matters: a dimension is meaningful only when associated text and geometry remain connected. See the DigiEDraw study for one research approach; its results are specific to that system and evaluation.
Worked example: preserve the whole callout
Start with the synthetic drawing text 4X Ø8.0 ±0.1 THRU and a related position frame POSITION ⌀0.30 | A | B | C.
- Source: record the exact raw callout, sheet, and crop or location.
- Candidate: retain the extracted text and the associated position-frame candidate separately until grouping is reviewed.
- Group: connect quantity, hole size, tolerance, and through condition to the four-hole pattern; connect the geometric control to the correct datum references.
- Normalize: add structured fields for quantity, nominal, tolerance, units, control, and datums without replacing the raw strings.
- Review: confirm that the leader and source geometry point to the intended pattern and that the drawing’s general tolerance does not alter an explicitly toleranced value.
- Export: map the reviewed characteristic IDs and values into the report format required by the process.
Do not guess at tolerances
For a direct symmetric callout such as 25.00 ±0.10, the range is 24.90–25.10 in the drawing’s units. For a unilateral callout, calculate each side according to its shown sign and magnitude. A limit callout gives its two boundaries explicitly. Before recording any of these values, verify the symbol, decimal, unit, sign, and source location.
Do not infer a tolerance from a title block until you resolve which general rule applies to the characteristic and whether an exception or local tolerance overrides it. Preserve the tolerance source and ask for clarification when the drawing is ambiguous.
Keep extraction states visible
A useful review flow distinguishes states such as detected/unreviewed, accepted, corrected, rejected, manually added, and ambiguous/requires review. These are conceptual labels; software uses different terminology. The important point is to avoid silently treating a machine proposal as reviewer-approved data.
Also preserve both the source text and the normalized representation. If a reviewer corrects an OCR error, store the corrected value and retain a source record that shows what changed and where the original came from.
Revision changes need a delta review
| Revision comparison | Source record action | ID/report action |
|---|---|---|
| Unchanged requirement | Confirm same location and content on new revision | Retain ID if procedure allows; link prior review |
| Tolerance changed | Capture new raw callout and revision | Recalculate limits; do not reuse old acceptance range |
| Requirement deleted | Record removal against old revision | Retire or mark old ID per procedure; remove stale current row |
| New requirement | Capture new source region and raw text | Assign ID without creating ambiguity with prior records |
Before you export: integrity checklist
- ☐ Every accepted characteristic has a source location and current drawing revision.
- ☐ Raw callout and normalized fields are both retained where applicable.
- ☐ Quantity, units, tolerance signs, and datum/control context are resolved.
- ☐ The source for every general tolerance is documented or ambiguity is escalated.
- ☐ No duplicate IDs or report rows remain unresolved.
- ☐ No report row is orphaned from its drawing requirement.
- ☐ No machine candidate remains unreviewed if the process requires review.
- ☐ Exported PDF and spreadsheet use synchronized identifiers and revision data.
BalloonForge and drawing-to-Excel work
BalloonForge presents candidate characteristics for review, supports correction and manual recovery, and can export an editable ballooned PDF and configurable Excel report.
Frequently asked questions
Can I extract dimensions from a PDF automatically?
Yes, with native text extraction, OCR, or a hybrid tool. The results still need source and grouping verification before inspection use.
What if the PDF text is not selectable?
It may be outlined or raster. Render or inspect the page and use OCR or a drawing-aware process appropriate to the source.
Can I send drawing dimensions straight to Excel?
You can, but keep the original callout, source location, requirement, tolerance, units, and review state. A list of numbers alone is not enough.
Can the software find every characteristic?
Do not assume that. Reconcile the output against the drawing using a documented completeness review.
Further reading
Free engineering resource
Drawing Inspection Starter Kit
An editable Excel report template, a printable drawing-review checklist, and a guide to connecting drawing requirements with inspection results.
ZIP · Excel template and printable HTML guides
Enter your email and submit the form to unlock your download.
