feat: 新增 OCRmyPDF 轉換引擎 (v0.1.14)
## 新功能 - OCRmyPDF 轉換引擎：將掃描版 PDF 轉換為可搜尋 PDF - 支援 7 種語言：en, zh-TW, zh, ja, ko, de, fr - 與 PDFMathTranslate 風格一致的 UI 格式 (pdf-<lang>) - 自動偵測頁面方向並旋轉 - 自動校正傾斜 - 跳過已有文字層的頁面 - 詳細的 5 階段處理進度輸出 ## 建置 - Dockerfile：安裝 ocrmypdf 與 Tesseract OCR 語言包 ## 文件 - 更新 OCR 功能文件 - 文件目錄結構改為中文名稱 ## 測試 - 修復 BabelDOC 和 PDFMathTranslate 測試的 OCR mock - 所有 345 個測試通過
This commit is contained in:
Your Name
parent
f24eec070c
commit
a06df23b1d
53 changed files with 1427 additions and 675 deletions
|
|
@ -11,6 +11,7 @@ import {
|
|||
import { join, basename, dirname } from "node:path";
|
||||
import { ExecFileFn } from "./types";
|
||||
import { getArchiveFileName } from "../transfer";
|
||||
import { ensureSearchablePdf, cleanupOcrTempFile } from "../helpers/pdfOcr";
|
||||
|
||||
/**
|
||||
* BabelDOC Content Engine
|
||||
|
|
@ -275,7 +276,19 @@ export async function convert(
|
|||
_options?: unknown,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<string> {
|
||||
let ocrTempFile: string | undefined;
|
||||
|
||||
try {
|
||||
// 0. 自動偵測掃描版 PDF 並進行 OCR 處理
|
||||
console.log(`[BabelDOC] Checking if PDF needs OCR...`);
|
||||
const ocrResult = await ensureSearchablePdf(filePath, execFile);
|
||||
const inputPdf = ocrResult.path;
|
||||
ocrTempFile = ocrResult.tempFile;
|
||||
|
||||
if (ocrResult.wasOcred) {
|
||||
console.log(`[BabelDOC] ✅ Scanned PDF detected and OCR'd automatically`);
|
||||
}
|
||||
|
||||
// 1. 檢查資源（警告但不阻止）
|
||||
checkResourcesExist();
|
||||
|
||||
|
|
@ -300,8 +313,8 @@ export async function convert(
|
|||
// 5. 設定 BabelDOC 輸出路徑（依輸出格式決定副檔名）
|
||||
const translatedFilePath = join(tempDir, `${inputFileName}-translated.${outputExt}`);
|
||||
|
||||
// 6. 執行 babeldoc 翻譯
|
||||
await runBabelDoc(filePath, translatedFilePath, targetLang, outputFormat, execFile);
|
||||
// 6. 執行 babeldoc 翻譯（使用 OCR 處理後的 PDF）
|
||||
await runBabelDoc(inputPdf, translatedFilePath, targetLang, outputFormat, execFile);
|
||||
|
||||
// 7. 複製翻譯後的檔案到封裝目錄
|
||||
const translatedDest = join(archiveDir, `translated-${targetLang}.${outputExt}`);
|
||||
|
|
@ -345,8 +358,13 @@ export async function convert(
|
|||
// 10. 清理臨時目錄
|
||||
removeDir(tempDir);
|
||||
|
||||
// 11. 清理 OCR 暫存檔案
|
||||
cleanupOcrTempFile(ocrTempFile);
|
||||
|
||||
return "Done";
|
||||
} catch (error) {
|
||||
// 確保清理 OCR 暫存檔案
|
||||
cleanupOcrTempFile(ocrTempFile);
|
||||
throw new Error(`BabelDOC error: ${error}`);
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -31,6 +31,7 @@ import {
|
|||
properties as propertiesPDFMathTranslate,
|
||||
} from "./pdfmathtranslate";
|
||||
import { convert as convertBabelDoc, properties as propertiesBabelDoc } from "./babeldoc";
|
||||
import { convert as convertOcrMyPdf, properties as propertiesOcrMyPdf } from "./ocrmypdf";
|
||||
|
||||
// This should probably be reconstructed so that the functions are not imported instead the functions hook into this to make the converters more modular
|
||||
|
||||
|
|
@ -156,6 +157,10 @@ const properties: Record<
|
|||
properties: propertiesBabelDoc,
|
||||
converter: convertBabelDoc,
|
||||
},
|
||||
OCRmyPDF: {
|
||||
properties: propertiesOcrMyPdf,
|
||||
converter: convertOcrMyPdf,
|
||||
},
|
||||
};
|
||||
|
||||
function chunks<T>(arr: T[], size: number): T[][] {
|
||||
|
|
|
|||
271
src/converters/ocrmypdf.ts
Normal file
271
src/converters/ocrmypdf.ts
Normal file
|
|
@ -0,0 +1,271 @@
|
|||
import { execFile as execFileOriginal } from "node:child_process";
|
||||
import { existsSync, mkdirSync, copyFileSync, statSync } from "node:fs";
|
||||
import { dirname, basename } from "node:path";
|
||||
import type { ExecFileFn } from "./types";
|
||||
|
||||
/**
|
||||
* OCRmyPDF Content Engine
|
||||
*
|
||||
* 將掃描版 PDF 轉換為可搜尋 PDF（加入文字層）
|
||||
* 使用 Tesseract OCR 進行文字辨識
|
||||
*
|
||||
* 支援的 OCR 語言（Docker 內建）：
|
||||
* - eng: 英文
|
||||
* - chi_tra: 繁體中文
|
||||
* - chi_sim: 簡體中文
|
||||
* - jpn: 日文
|
||||
* - kor: 韓文
|
||||
* - deu: 德文
|
||||
* - fra: 法文
|
||||
*
|
||||
* UI 顯示格式（與 PDFMathTranslate 風格一致）：
|
||||
* pdf-en, pdf-zh-TW, pdf-ja 等
|
||||
*/
|
||||
|
||||
// 內建支援的 OCR 語言（與 PDFMathTranslate 風格一致）
|
||||
const SUPPORTED_LANGUAGES = [
|
||||
"en", // English
|
||||
"zh-TW", // 繁體中文
|
||||
"zh", // 簡體中文
|
||||
"ja", // 日本語
|
||||
"ko", // 한국어
|
||||
"de", // Deutsch
|
||||
"fr", // Français
|
||||
] as const;
|
||||
|
||||
// Tesseract 語言代碼映射（UI 代碼 → Tesseract 代碼）
|
||||
const LANG_MAP: Record<string, string> = {
|
||||
"en": "eng",
|
||||
"zh-TW": "chi_tra",
|
||||
"zh": "chi_sim",
|
||||
"ja": "jpn",
|
||||
"ko": "kor",
|
||||
"de": "deu",
|
||||
"fr": "fra",
|
||||
};
|
||||
|
||||
export const properties = {
|
||||
from: {
|
||||
document: ["pdf"],
|
||||
},
|
||||
to: {
|
||||
document: SUPPORTED_LANGUAGES.map((lang) => `pdf-${lang}`),
|
||||
},
|
||||
};
|
||||
|
||||
/**
|
||||
* 從 convertTo 格式中提取 OCR 語言
|
||||
* @param convertTo 格式如 "pdf-en" 或 "pdf-zh-TW"
|
||||
* @returns Tesseract OCR 語言代碼
|
||||
*/
|
||||
function extractOcrLanguage(convertTo: string): string {
|
||||
// convertTo 格式: pdf-<lang>
|
||||
const match = convertTo.match(/^pdf-(.+)$/);
|
||||
if (!match || !match[1]) {
|
||||
throw new Error(`Invalid convertTo format: ${convertTo}. Expected pdf-<lang>`);
|
||||
}
|
||||
|
||||
const uiLang = match[1];
|
||||
// 轉換為 Tesseract 語言代碼
|
||||
const tessLang = LANG_MAP[uiLang] || uiLang.replace(/-/g, "_");
|
||||
return tessLang;
|
||||
}
|
||||
|
||||
/**
|
||||
* 格式化檔案大小
|
||||
*/
|
||||
function formatFileSize(bytes: number): string {
|
||||
if (bytes < 1024) return `${bytes} B`;
|
||||
if (bytes < 1024 * 1024) return `${(bytes / 1024).toFixed(1)} KB`;
|
||||
return `${(bytes / (1024 * 1024)).toFixed(2)} MB`;
|
||||
}
|
||||
|
||||
/**
|
||||
* 執行 ocrmypdf 進行 PDF OCR 處理
|
||||
*
|
||||
* @param inputPath 輸入 PDF 路徑
|
||||
* @param outputPath 輸出 PDF 路徑
|
||||
* @param lang OCR 語言
|
||||
* @param execFile execFile 函數
|
||||
*/
|
||||
function runOcrMyPdf(
|
||||
inputPath: string,
|
||||
outputPath: string,
|
||||
lang: string,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<string> {
|
||||
return new Promise((resolve, reject) => {
|
||||
console.log(`[OCRmyPDF] ========================================`);
|
||||
console.log(`[OCRmyPDF] 🔍 開始 PDF OCR 處理`);
|
||||
console.log(`[OCRmyPDF] ========================================`);
|
||||
|
||||
// 階段 1：驗證輸入檔案
|
||||
console.log(`[OCRmyPDF] 📋 階段 1/5：驗證輸入檔案`);
|
||||
if (!existsSync(inputPath)) {
|
||||
reject(new Error(`Input file not found: ${inputPath}`));
|
||||
return;
|
||||
}
|
||||
const inputStats = statSync(inputPath);
|
||||
console.log(`[OCRmyPDF] ✅ 輸入檔案: ${basename(inputPath)}`);
|
||||
console.log(`[OCRmyPDF] ✅ 檔案大小: ${formatFileSize(inputStats.size)}`);
|
||||
|
||||
// 階段 2：準備 OCR 參數
|
||||
console.log(`[OCRmyPDF] 📋 階段 2/5：準備 OCR 參數`);
|
||||
console.log(`[OCRmyPDF] ✅ OCR 語言: ${lang}`);
|
||||
|
||||
const args = [
|
||||
"-l", lang,
|
||||
"--skip-text", // 跳過已有文字的頁面
|
||||
"--optimize", "1", // 輕度優化
|
||||
"--deskew", // 自動校正傾斜
|
||||
"--rotate-pages", // 自動偵測頁面方向
|
||||
"--jobs", "2", // 使用 2 個並行處理
|
||||
inputPath,
|
||||
outputPath,
|
||||
];
|
||||
console.log(`[OCRmyPDF] ✅ 參數: --skip-text --optimize 1 --deskew --rotate-pages`);
|
||||
|
||||
// 階段 3：執行 OCR
|
||||
console.log(`[OCRmyPDF] 📋 階段 3/5：執行 Tesseract OCR...`);
|
||||
console.log(`[OCRmyPDF] ⏳ 處理中（大型 PDF 可能需要數分鐘）...`);
|
||||
const startTime = Date.now();
|
||||
|
||||
execFile("ocrmypdf", args, (error: Error | null, stdout: string, stderr: string) => {
|
||||
const elapsed = ((Date.now() - startTime) / 1000).toFixed(1);
|
||||
|
||||
if (error) {
|
||||
// 階段 3 失敗處理
|
||||
console.log(`[OCRmyPDF] ❌ OCR 處理失敗（耗時 ${elapsed}s）`);
|
||||
|
||||
// 檢查特定錯誤類型
|
||||
if (stderr && stderr.includes("PriorOcrFoundError")) {
|
||||
console.log(`[OCRmyPDF] 📋 階段 4/5：PDF 已有文字層，跳過 OCR`);
|
||||
console.log(`[OCRmyPDF] ✅ 複製原始檔案...`);
|
||||
try {
|
||||
copyFileSync(inputPath, outputPath);
|
||||
console.log(`[OCRmyPDF] 📋 階段 5/5：完成`);
|
||||
console.log(`[OCRmyPDF] ✅ 輸出: ${basename(outputPath)}`);
|
||||
console.log(`[OCRmyPDF] ========================================`);
|
||||
console.log(`[OCRmyPDF] ✅ OCR 完成（PDF 已有文字層）`);
|
||||
console.log(`[OCRmyPDF] ========================================`);
|
||||
resolve(outputPath);
|
||||
return;
|
||||
} catch (copyError) {
|
||||
reject(new Error(`Failed to copy already-OCRed PDF: ${copyError}`));
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
if (stderr && stderr.includes("EncryptedPdfError")) {
|
||||
console.log(`[OCRmyPDF] ❌ PDF 已加密，無法處理`);
|
||||
reject(new Error("PDF is encrypted. Please decrypt it first."));
|
||||
return;
|
||||
}
|
||||
|
||||
if (stderr && stderr.includes("InputFileError")) {
|
||||
console.log(`[OCRmyPDF] ❌ PDF 檔案無效或損壞`);
|
||||
reject(new Error("Invalid or corrupted PDF file."));
|
||||
return;
|
||||
}
|
||||
|
||||
console.log(`[OCRmyPDF] 錯誤訊息: ${stderr || error.message}`);
|
||||
reject(new Error(`ocrmypdf error: ${error}\nstderr: ${stderr}`));
|
||||
return;
|
||||
}
|
||||
|
||||
// 階段 3 成功
|
||||
console.log(`[OCRmyPDF] ✅ OCR 引擎處理完成（耗時 ${elapsed}s）`);
|
||||
|
||||
if (stdout) {
|
||||
console.log(`[OCRmyPDF] stdout: ${stdout}`);
|
||||
}
|
||||
|
||||
// 階段 4：驗證輸出
|
||||
console.log(`[OCRmyPDF] 📋 階段 4/5：驗證輸出檔案`);
|
||||
if (!existsSync(outputPath)) {
|
||||
console.log(`[OCRmyPDF] ❌ 輸出檔案未生成`);
|
||||
reject(new Error(`OCR output file not found: ${outputPath}`));
|
||||
return;
|
||||
}
|
||||
|
||||
const outputStats = statSync(outputPath);
|
||||
console.log(`[OCRmyPDF] ✅ 輸出檔案: ${basename(outputPath)}`);
|
||||
console.log(`[OCRmyPDF] ✅ 檔案大小: ${formatFileSize(outputStats.size)}`);
|
||||
|
||||
// 階段 5：完成
|
||||
console.log(`[OCRmyPDF] 📋 階段 5/5：處理完成`);
|
||||
console.log(`[OCRmyPDF] ========================================`);
|
||||
console.log(`[OCRmyPDF] ✅ OCR 處理成功完成！`);
|
||||
console.log(`[OCRmyPDF] 📥 輸入: ${formatFileSize(inputStats.size)}`);
|
||||
console.log(`[OCRmyPDF] 📤 輸出: ${formatFileSize(outputStats.size)}`);
|
||||
console.log(`[OCRmyPDF] ⏱️ 耗時: ${elapsed}s`);
|
||||
console.log(`[OCRmyPDF] ========================================`);
|
||||
|
||||
resolve(outputPath);
|
||||
});
|
||||
});
|
||||
}
|
||||
|
||||
/**
|
||||
* 主要轉換函數
|
||||
*
|
||||
* @param filePath 輸入 PDF 檔案路徑
|
||||
* @param fileType 檔案類型（應為 "pdf"）
|
||||
* @param convertTo 目標格式（如 "pdf-en"、"pdf-zh-TW"）
|
||||
* @param targetPath 輸出路徑
|
||||
* @param _options 額外選項
|
||||
* @param execFile execFile 函數（用於測試注入）
|
||||
*/
|
||||
export async function convert(
|
||||
filePath: string,
|
||||
fileType: string,
|
||||
convertTo: string,
|
||||
targetPath: string,
|
||||
_options?: unknown,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<string> {
|
||||
console.log(``);
|
||||
console.log(`[OCRmyPDF] ╔════════════════════════════════════════╗`);
|
||||
console.log(`[OCRmyPDF] ║ OCRmyPDF - PDF OCR 處理引擎 ║`);
|
||||
console.log(`[OCRmyPDF] ╚════════════════════════════════════════╝`);
|
||||
console.log(``);
|
||||
|
||||
try {
|
||||
// 步驟 1：解析參數
|
||||
console.log(`[OCRmyPDF] 🔧 解析轉換參數...`);
|
||||
const ocrLang = extractOcrLanguage(convertTo);
|
||||
console.log(`[OCRmyPDF] 輸入格式: ${fileType}`);
|
||||
console.log(`[OCRmyPDF] 目標格式: ${convertTo}`);
|
||||
console.log(`[OCRmyPDF] OCR 語言: ${ocrLang}`);
|
||||
|
||||
// 步驟 2：確保輸出目錄存在
|
||||
console.log(`[OCRmyPDF] 📁 準備輸出目錄...`);
|
||||
const outputDir = dirname(targetPath);
|
||||
if (!existsSync(outputDir)) {
|
||||
mkdirSync(outputDir, { recursive: true });
|
||||
console.log(`[OCRmyPDF] ✅ 建立目錄: ${outputDir}`);
|
||||
} else {
|
||||
console.log(`[OCRmyPDF] ✅ 目錄已存在`);
|
||||
}
|
||||
|
||||
// 步驟 3：執行 OCR
|
||||
console.log(`[OCRmyPDF] 🚀 開始 OCR 處理...`);
|
||||
await runOcrMyPdf(filePath, targetPath, ocrLang, execFile);
|
||||
|
||||
console.log(``);
|
||||
console.log(`[OCRmyPDF] ╔════════════════════════════════════════╗`);
|
||||
console.log(`[OCRmyPDF] ║ ✅ PDF OCR 處理完成！ ║`);
|
||||
console.log(`[OCRmyPDF] ╚════════════════════════════════════════╝`);
|
||||
console.log(``);
|
||||
|
||||
return "Done";
|
||||
} catch (error) {
|
||||
console.log(``);
|
||||
console.log(`[OCRmyPDF] ╔════════════════════════════════════════╗`);
|
||||
console.log(`[OCRmyPDF] ║ ❌ PDF OCR 處理失敗 ║`);
|
||||
console.log(`[OCRmyPDF] ╚════════════════════════════════════════╝`);
|
||||
console.log(`[OCRmyPDF] 錯誤: ${error}`);
|
||||
console.log(``);
|
||||
throw new Error(`OCRmyPDF error: ${error}`);
|
||||
}
|
||||
}
|
||||
|
|
@ -2,6 +2,7 @@ import { execFile as execFileOriginal } from "node:child_process";
|
|||
import { mkdirSync, existsSync, readdirSync, unlinkSync, rmdirSync, copyFileSync } from "node:fs";
|
||||
import { join, basename, dirname } from "node:path";
|
||||
import { getArchiveFileName } from "../transfer";
|
||||
import { ensureSearchablePdf, cleanupOcrTempFile } from "../helpers/pdfOcr";
|
||||
import type { ExecFileFn } from "./types";
|
||||
|
||||
// 翻譯服務優先順序（自動 fallback）
|
||||
|
|
@ -339,7 +340,19 @@ export async function convert(
|
|||
_options?: unknown,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<string> {
|
||||
let ocrTempFile: string | undefined;
|
||||
|
||||
try {
|
||||
// 0. 自動偵測掃描版 PDF 並進行 OCR 處理
|
||||
console.log(`[PDFMathTranslate] Checking if PDF needs OCR...`);
|
||||
const ocrResult = await ensureSearchablePdf(filePath, execFile);
|
||||
const inputPdf = ocrResult.path;
|
||||
ocrTempFile = ocrResult.tempFile;
|
||||
|
||||
if (ocrResult.wasOcred) {
|
||||
console.log(`[PDFMathTranslate] ✅ Scanned PDF detected and OCR'd automatically`);
|
||||
}
|
||||
|
||||
// 1. 檢查模型（警告但不阻止，因為 pdf2zh 可能會自動下載）
|
||||
checkModelsExist();
|
||||
|
||||
|
|
@ -361,8 +374,8 @@ export async function convert(
|
|||
const archiveDir = join(tempDir, "archive");
|
||||
mkdirSync(archiveDir, { recursive: true });
|
||||
|
||||
// 5. 執行 pdf2zh 翻譯
|
||||
const { monoPath, dualPath } = await runPdf2zh(filePath, tempDir, targetLang, execFile);
|
||||
// 5. 執行 pdf2zh 翻譯（使用 OCR 處理後的 PDF）
|
||||
const { monoPath, dualPath } = await runPdf2zh(inputPdf, tempDir, targetLang, execFile);
|
||||
|
||||
// 6. 複製翻譯後的檔案到封裝目錄
|
||||
// PDFMathTranslate 輸出：
|
||||
|
|
@ -440,8 +453,13 @@ export async function convert(
|
|||
// 9. 清理臨時目錄
|
||||
removeDir(tempDir);
|
||||
|
||||
// 10. 清理 OCR 暫存檔案
|
||||
cleanupOcrTempFile(ocrTempFile);
|
||||
|
||||
return "Done";
|
||||
} catch (error) {
|
||||
// 確保清理 OCR 暫存檔案
|
||||
cleanupOcrTempFile(ocrTempFile);
|
||||
throw new Error(`PDFMathTranslate error: ${error}`);
|
||||
}
|
||||
}
|
||||
|
|
|
|||
202
src/helpers/pdfOcr.ts
Normal file
202
src/helpers/pdfOcr.ts
Normal file
|
|
@ -0,0 +1,202 @@
|
|||
import { execFile as execFileOriginal } from "node:child_process";
|
||||
import { existsSync, unlinkSync, readFileSync } from "node:fs";
|
||||
import { join, dirname, basename } from "node:path";
|
||||
import type { ExecFileFn } from "../converters/types";
|
||||
|
||||
/**
|
||||
* PDF OCR 輔助工具
|
||||
*
|
||||
* 自動偵測掃描版 PDF 並使用 ocrmypdf 進行 OCR 處理
|
||||
* 確保翻譯引擎可以正確提取文字
|
||||
*/
|
||||
|
||||
// OCR 偵測閾值（文字字元數低於此值視為掃描版）
|
||||
const SCANNED_PDF_TEXT_THRESHOLD = 100;
|
||||
|
||||
// 預設 OCR 語言（可透過環境變數覆蓋）
|
||||
const DEFAULT_OCR_LANG = process.env.OCR_LANG || "eng+chi_tra+chi_sim+jpn";
|
||||
|
||||
/**
|
||||
* 使用 pdftotext 提取 PDF 中的文字
|
||||
*
|
||||
* @param pdfPath PDF 檔案路徑
|
||||
* @param execFile execFile 函數
|
||||
* @returns 提取的文字內容
|
||||
*/
|
||||
function extractTextFromPdf(
|
||||
pdfPath: string,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<string> {
|
||||
return new Promise((resolve, reject) => {
|
||||
// pdftotext -layout <input.pdf> -
|
||||
// -layout: 保持排版
|
||||
// -: 輸出到 stdout
|
||||
execFile("pdftotext", ["-layout", pdfPath, "-"], (error, stdout, stderr) => {
|
||||
if (error) {
|
||||
// pdftotext 失敗可能是因為 PDF 沒有文字層
|
||||
console.warn(`[PDF-OCR] pdftotext failed: ${stderr}`);
|
||||
resolve(""); // 返回空字串，視為掃描版
|
||||
return;
|
||||
}
|
||||
resolve(stdout || "");
|
||||
});
|
||||
});
|
||||
}
|
||||
|
||||
/**
|
||||
* 偵測 PDF 是否為掃描版（無文字層）
|
||||
*
|
||||
* @param pdfPath PDF 檔案路徑
|
||||
* @param execFile execFile 函數
|
||||
* @returns 是否為掃描版 PDF
|
||||
*/
|
||||
export async function isScannedPdf(
|
||||
pdfPath: string,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<boolean> {
|
||||
try {
|
||||
const text = await extractTextFromPdf(pdfPath, execFile);
|
||||
|
||||
// 移除空白字元後計算有效字元數
|
||||
const cleanText = text.replace(/\s+/g, "");
|
||||
const charCount = cleanText.length;
|
||||
|
||||
console.log(`[PDF-OCR] Extracted ${charCount} characters from PDF`);
|
||||
|
||||
// 如果字元數低於閾值，視為掃描版
|
||||
const isScanned = charCount < SCANNED_PDF_TEXT_THRESHOLD;
|
||||
|
||||
if (isScanned) {
|
||||
console.log(`[PDF-OCR] ⚠️ Detected scanned PDF (chars: ${charCount} < ${SCANNED_PDF_TEXT_THRESHOLD})`);
|
||||
} else {
|
||||
console.log(`[PDF-OCR] ✅ PDF has text layer (chars: ${charCount})`);
|
||||
}
|
||||
|
||||
return isScanned;
|
||||
} catch (error) {
|
||||
console.warn(`[PDF-OCR] Detection failed, assuming scanned: ${error}`);
|
||||
return true; // 偵測失敗時，保守起見視為掃描版
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 使用 ocrmypdf 對 PDF 進行 OCR 處理
|
||||
*
|
||||
* @param inputPath 輸入 PDF 路徑
|
||||
* @param outputPath 輸出 PDF 路徑（可搜尋）
|
||||
* @param lang OCR 語言（預設：eng+chi_tra+chi_sim+jpn）
|
||||
* @param execFile execFile 函數
|
||||
* @returns 處理後的 PDF 路徑
|
||||
*/
|
||||
export function runOcrMyPdf(
|
||||
inputPath: string,
|
||||
outputPath: string,
|
||||
lang: string = DEFAULT_OCR_LANG,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<string> {
|
||||
return new Promise((resolve, reject) => {
|
||||
// ocrmypdf 參數：
|
||||
// -l <lang>: OCR 語言（可用 + 連接多語言）
|
||||
// --skip-text: 跳過已有文字的頁面（避免重複 OCR）
|
||||
// --optimize 1: 輕度優化，平衡速度和品質
|
||||
// --deskew: 自動校正傾斜
|
||||
// --clean: 清理背景雜訊
|
||||
// --force-ocr: 強制對所有頁面進行 OCR（即使有文字層）
|
||||
|
||||
const args = [
|
||||
"-l", lang,
|
||||
"--skip-text",
|
||||
"--optimize", "1",
|
||||
"--deskew",
|
||||
inputPath,
|
||||
outputPath,
|
||||
];
|
||||
|
||||
console.log(`[PDF-OCR] Running: ocrmypdf ${args.join(" ")}`);
|
||||
|
||||
execFile("ocrmypdf", args, (error: Error | null, stdout: string, stderr: string) => {
|
||||
if (error) {
|
||||
// 檢查是否是「PDF 已經有文字」的警告
|
||||
if (stderr && stderr.includes("page already has text")) {
|
||||
console.log(`[PDF-OCR] PDF already has text layer, using original`);
|
||||
resolve(inputPath);
|
||||
return;
|
||||
}
|
||||
|
||||
reject(new Error(`ocrmypdf error: ${error}\nstderr: ${stderr}`));
|
||||
return;
|
||||
}
|
||||
|
||||
if (stdout) {
|
||||
console.log(`[PDF-OCR] stdout: ${stdout}`);
|
||||
}
|
||||
|
||||
if (stderr) {
|
||||
// ocrmypdf 的進度資訊會輸出到 stderr
|
||||
console.log(`[PDF-OCR] stderr: ${stderr}`);
|
||||
}
|
||||
|
||||
if (!existsSync(outputPath)) {
|
||||
reject(new Error(`OCR output file not found: ${outputPath}`));
|
||||
return;
|
||||
}
|
||||
|
||||
console.log(`[PDF-OCR] ✅ OCR completed: ${outputPath}`);
|
||||
resolve(outputPath);
|
||||
});
|
||||
});
|
||||
}
|
||||
|
||||
/**
|
||||
* 自動偵測並處理掃描版 PDF
|
||||
*
|
||||
* 如果 PDF 是掃描版（無文字層），自動執行 OCR
|
||||
* 如果 PDF 已有文字層，直接返回原始路徑
|
||||
*
|
||||
* @param inputPath 輸入 PDF 路徑
|
||||
* @param execFile execFile 函數
|
||||
* @returns 處理後的 PDF 路徑（可能是原始路徑或 OCR 後的新檔案）
|
||||
*/
|
||||
export async function ensureSearchablePdf(
|
||||
inputPath: string,
|
||||
execFile: ExecFileFn = execFileOriginal,
|
||||
): Promise<{ path: string; wasOcred: boolean; tempFile?: string }> {
|
||||
// 1. 偵測是否為掃描版
|
||||
const scanned = await isScannedPdf(inputPath, execFile);
|
||||
|
||||
if (!scanned) {
|
||||
// PDF 已有文字層，直接使用
|
||||
return { path: inputPath, wasOcred: false };
|
||||
}
|
||||
|
||||
// 2. 建立 OCR 輸出路徑
|
||||
const dir = dirname(inputPath);
|
||||
const name = basename(inputPath, ".pdf");
|
||||
const ocrOutputPath = join(dir, `${name}_ocr_${Date.now()}.pdf`);
|
||||
|
||||
// 3. 執行 OCR
|
||||
console.log(`[PDF-OCR] Starting OCR for scanned PDF...`);
|
||||
await runOcrMyPdf(inputPath, ocrOutputPath, DEFAULT_OCR_LANG, execFile);
|
||||
|
||||
return {
|
||||
path: ocrOutputPath,
|
||||
wasOcred: true,
|
||||
tempFile: ocrOutputPath, // 呼叫者需要在完成後清理此暫存檔
|
||||
};
|
||||
}
|
||||
|
||||
/**
|
||||
* 清理 OCR 產生的暫存檔案
|
||||
*
|
||||
* @param tempFile 暫存檔案路徑
|
||||
*/
|
||||
export function cleanupOcrTempFile(tempFile: string | undefined): void {
|
||||
if (tempFile && existsSync(tempFile)) {
|
||||
try {
|
||||
unlinkSync(tempFile);
|
||||
console.log(`[PDF-OCR] Cleaned up temp file: ${tempFile}`);
|
||||
} catch (error) {
|
||||
console.warn(`[PDF-OCR] Failed to cleanup temp file: ${error}`);
|
||||
}
|
||||
}
|
||||
}
|
||||
Loading…
Add table
Add a link
Reference in a new issue