ReceiptTracker 统一技术设计文档 (方案A: 本地离线端侧 OCR - 多语言国际化版)

apps/
└── ReceiptTracker/
    ├── src/
    │   ├── android/             # Android 原生壳工程与 ML Kit 桥接
    │   │   ├── app/src/main/
    │   │   │   ├── java/com/appmatrix/receipttracker/OCRBridgeModule.kt   # 支持中日英动态 OCR 的桥接代码
    │   │   │   └── AndroidManifest.xml   # 声明 Camera 权限与硬件需求
    │   │   └── build.gradle             # 引入 ML Kit 中文与日文识别依赖包
    │   └── ios/                 # iOS 原生壳工程与 Vision.framework 桥接
    │       ├── ReceiptTracker/
    │       │   ├── OCRBridgeModule.swift  # 支持中日英 Vision OCR 桥接代码
    │       │   └── Info.plist            # NSCameraUsageDescription 声明
    │       └── ReceiptTracker.xcodeproj
    └── shared/                  # 双端共享的 React Native (Expo) 业务代码
        ├── src/
        │   ├── components/      # 共享 UI（如 Canvas 趋势图折线组件）
        │   ├── screens/         # 记账历史、扫描核对、物价大盘三大页面
        │   └── utils/
        │       ├── alignLines.ts # Y 轴文本行聚类核心算法
        │       ├── database.ts   # 支持多币种与语言标志的 SQLite 封装
        │       └── i18n.ts       # 轻量级 UI 多语言国际化字典
        └── package.json

[拍摄原始大图] ──(Native C++/WebGL)──> [1. 灰度化] ──> [2. 对比度增强 (拉伸)] ──> [3. 二值化/去噪] ──> [生成临时干净图片]

import Vision
import UIKit

@objc(OCRBridgeModule)
class OCRBridgeModule: NSObject {
    
    @objc
    func recognizeReceiptText(_ imagePath: String, resolver resolve: @escaping RCTPromiseResolveBlock, rejecter reject: @escaping RCTPromiseRejectBlock) {
        guard let image = UIImage(contentsOfFile: imagePath),
              let cgImage = image.cgImage else {
            reject("ERR_IMAGE_LOAD", "无法加载缓存小票图片", nil)
            return
        }
        
        let requestHandler = VNImageRequestHandler(cgImage: cgImage, options: [:])
        
        let request = VNRecognizeTextRequest { (request, error) in
            if let error = error {
                reject("ERR_OCR_FAILED", "iOS Vision OCR 失败: \(error.localizedDescription)", error)
                return
            }
            
            guard let observations = request.results as? [VNRecognizedTextObservation] else {
                resolve([])
                return
            }
            
            var recognizedBlocks: [[String: Any]] = []
            for observation in observations {
                guard let candidate = observation.topCandidates(1).first else { continue }
                
                let boundingBox = observation.boundingBox
                recognizedBlocks.append([
                    "text": candidate.string,
                    "confidence": candidate.confidence * 100,
                    "x": boundingBox.origin.x,
                    "y": 1.0 - boundingBox.origin.y - boundingBox.height, // 翻转 Y 轴对齐常规坐标系
                    "width": boundingBox.width,
                    "height": boundingBox.height
                ])
            }
            resolve(recognizedBlocks)
        }
        
        // 设置 Vision 参数 - 启用中日英兼容识别
        request.recognitionLevel = .accurate
        request.usesLanguageCorrection = true
        // 关键更新：支持 简体中文、繁体中文、日文、英文
        request.recognitionLanguages = ["zh-Hans", "zh-Hant", "ja-JP", "en-US"] 
        
        do {
            try requestHandler.perform([request])
        } catch {
            reject("ERR_OCR_PERFORM", "执行 OCR 失败", error)
        }
    }
}

interface OCRBlock {
  text: string;
  x: number;      // 坐标范围 0 ~ 图像宽或归一化 0~1
  y: number;      // 坐标范围 0 ~ 图像高或归一化 0~1
  width: number;
  height: number;
}

/**
 * 基于 Y 轴位置的高斯/区间聚类算法，还原文本物理行
 */
export function alignOCRBlocksToLines(blocks: OCRBlock[]): string[][] {
  if (blocks.length === 0) return [];

  // 1. 按照 Y 轴坐标从上到下排序
  const sortedBlocks = [...blocks].sort((a, b) => a.y - b.y);
  
  const lines: OCRBlock[][] = [];
  
  for (const block of sortedBlocks) {
    let placed = false;
    
    // 2. 遍历已有的行，检查该文本块是否在某一行的 Y 轴高度容差范围内
    for (const line of lines) {
      const avgHeight = line.reduce((sum, b) => sum + b.height, 0) / line.length;
      const avgY = line.reduce((sum, b) => sum + b.y, 0) / line.length;
      
      // Y 轴容差设为行高的 60%
      const tolerance = avgHeight * 0.6;
      
      if (Math.abs(block.y - avgY) <= tolerance) {
        line.push(block);
        placed = true;
        break;
      }
    }
    
    // 3. 如果与所有已有行都对不齐，说明是一行新开始的文本
    if (!placed) {
      lines.push([block]);
    }
  }
  
  // 4. 将每一行内部的文本块按照 X 轴从左到右排序，还原文字顺序并拼接
  return lines.map(line => {
    const sortedLine = line.sort((a, b) => a.x - b.x);
    return sortedLine.map(b => b.text.trim());
  });
}

-- 1. 小票主表（扩展多语言与币种字段）
CREATE TABLE IF NOT EXISTS receipts (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    merchant TEXT NOT NULL,
    date TEXT NOT NULL,
    total_amount REAL NOT NULL,
    lang TEXT NOT NULL,         -- 'zh', 'ja', 'en'
    currency TEXT NOT NULL      -- 'CNY', 'JPY', 'USD'
);

-- 2. 小票条目表
CREATE TABLE IF NOT EXISTS receipt_items (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    receipt_id INTEGER NOT NULL,
    product_name TEXT NOT NULL,
    unit_price REAL NOT NULL,
    quantity INTEGER NOT NULL,
    category TEXT NOT NULL,
    FOREIGN KEY(receipt_id) REFERENCES receipts(id) ON DELETE CASCADE
);

-- 3. 商品统一视图（用于物价分类追踪）
CREATE TABLE IF NOT EXISTS products (
    product_name TEXT PRIMARY KEY,
    category TEXT NOT NULL,
    first_detected_date TEXT NOT NULL,
    lang TEXT NOT NULL          -- 标记该商品隶属于哪个语言类别，防止多语言混杂
);

-- 4. 商品历史价格走势点
CREATE TABLE IF NOT EXISTS price_history (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    product_name TEXT NOT NULL,
    date TEXT NOT NULL,
    price REAL NOT NULL,
    receipt_id INTEGER NOT NULL,
    FOREIGN KEY(product_name) REFERENCES products(product_name) ON DELETE CASCADE,
    FOREIGN KEY(receipt_id) REFERENCES receipts(id) ON DELETE CASCADE
);

<key>NSCameraUsageDescription</key>
<string>ReceiptTracker需要使用您的摄像头来拍摄购物小票以进行记账文本识别。</string>
<key>NSPhotoLibraryUsageDescription</key>
<string>ReceiptTracker需要访问您的相册以导入保存的小票图片进行记账。</string>

<uses-permission android:name="android.permission.CAMERA" />
<uses-permission android:name="android.permission.READ_EXTERNAL_STORAGE" />
<uses-permission android:name="android.permission.WRITE_EXTERNAL_STORAGE" android:maxSdkVersion="28" />
<uses-feature android:name="android.hardware.camera" android:required="false" />

export const i18nDictionary = {
  zh: {
    navTitle: "小票扫码记账",
    currencySymbol: "¥",
    lblTotal: "实付总计",
    lblSave: "确认记账",
    categories: ["全部", "食品饮料", "日用百货", "个人护理", "数码配件"]
  },
  ja: {
    navTitle: "レシートスキャン",
    currencySymbol: "¥",
    lblTotal: "合計金額",
    lblSave: "記帳する",
    categories: ["すべて", "食品飲料", "生活雑貨", "パーソナルケア", "デジタル機器"]
  },
  en: {
    navTitle: "RECEIPT SCANNER",
    currencySymbol: "$",
    lblTotal: "Total Amount",
    lblSave: "Confirm Save",
    categories: ["All", "Food & Beverage", "Households", "Personal Care", "Electronics"]
  }
};

+------------------------------------------+
|  [ 返回 ]        扫描小票                  | <-- 顶部导航栏 (50px)
+------------------------------------------+
|                                          |
|         +----------------------+         |
|         |                      |         |
|         |                      |         |
|         |    Scan Guide Box    |         | <-- 竖向扫描框 (72%宽 x 58%高)
|         |                      |         |     (指示小票在取景区正中)
|         |                      |         |
|         +----------------------+         |
|                                          |
+------------------------------------------+
|  [ 相册 ]       ( 快门按钮 )      [闪光灯]  | <-- 底部面板栏 (150px高，横排居中)
+------------------------------------------+

+------------------------------------+-----+
| (返回)                             | [闪]| <-- 顶/底控制小按钮
|                                    | [光]|
|      +-----------------------+     |     |
|      |                       |     |  快 |
|      |    Scan Guide Box     |     |  门 | <-- 右侧控制栏 (100px宽)
|      |    (65%宽 x 78%高)    |     |  按 |     (三按钮竖排，方便右手拇指)
|      |                       |     |  钮 |
|      +-----------------------+     |     |
|                                    | [相]|
|                                    | [册]|
+------------------------------------+-----+
  ^                                     ^
相机取景区 (Flex: 1 撑满)             右侧控制栏