双引擎OCR并发: Windows OCR + EasyOCR同时跑最多等10秒，谁先出用谁

之前: Windows OCR串行等完→再串行等EasyOCR(微信端甚至无超时)。 EasyOCR子进程挂起=OCR整条链路都卡死。改为 _ocr_dual_engine 方法: - Windows OCR + EasyOCR 通过 ThreadPoolExecutor 同时跑 - concurrent.futures.wait(timeout=10.0, ALL_COMPLETED) 等最多10秒 - 两引擎都在10秒内出结果→_pick_ocr_text质量对比选优 - 只有一个在10秒内出结果→直接用那个 - 都超时→返回空 - Windows OCR bad allocation 自动降级到原图重试(在worker线程内处理) 两个平台调用点(_ocr_read_qq / _ocr_read_wechat)都改为调_ocr_dual_engine。微信端同时补上im_original备份(之前缺bad allocation降级)。
2026-07-29 17:51:52 +08:00
parent 68d364a0c6
commit 722d2a5438
1 changed files with 84 additions and 91 deletions
@@ -1049,6 +1049,85 @@ class MessageMonitor:
        )
        return (win_text, chosen) if chosen == "windows" else (easy_text, chosen)

+    def _ocr_dual_engine(self, im_processed: Image.Image, im_original: Image.Image, platform: str) -> tuple:
+        """双引擎并发 OCR：Windows OCR + EasyOCR 同时跑，最多等 10 秒。
+
+        哪個先出用哪個，兩個都出時 _pick_ocr_text 質量對比選優。
+
+        im_processed: 预处理后的图（2x resize + sharpen + contrast, RGBA）
+        im_original:  原始截图（bad allocation 降级用）
+
+        返回 (raw_text, engine_name) 含義同 _pick_ocr_text。
+        """
+        import concurrent.futures
+
+        engine = self._ocr_ensure_engine()
+        if not engine:
+            return ("", "")
+
+        win_text: list[str] = [""]
+        easy_text: list[str] = [""]
+
+        def _win_worker():
+            try:
+                im = im_processed
+                if im.mode != 'RGBA':
+                    im = im.convert('RGBA')
+                sb = SoftwareBitmap(BitmapPixelFormat.RGBA8, im.width, im.height)
+                sb.copy_from_buffer(im.tobytes())
+                text = engine.recognize_async(sb).get().text.strip()
+                win_text[0] = text
+            except (RuntimeError, OSError) as e:
+                estr = str(e)
+                if "bad allocation" in estr or "memory" in estr.lower():
+                    try:
+                        im2 = im_original
+                        if im2.mode != 'RGBA':
+                            im2 = im2.convert('RGBA')
+                        sb2 = SoftwareBitmap(BitmapPixelFormat.RGBA8, im2.width, im2.height)
+                        sb2.copy_from_buffer(im2.tobytes())
+                        text2 = engine.recognize_async(sb2).get().text.strip()
+                        win_text[0] = text2
+                    except Exception:
+                        pass
+                # 其他异常静默处理
+
+        def _easy_worker():
+            if not _HAS_EASYOCR or self._easyocr_broken:
+                return
+            img_np = _np.array(im_processed)
+            text = self._call_easyocr_safe(img_np, timeout=10)
+            if text:
+                easy_text[0] = text
+
+        with concurrent.futures.ThreadPoolExecutor(max_workers=2) as ex:
+            win_fut = ex.submit(_win_worker)
+            easy_fut = ex.submit(_easy_worker)
+            done, not_done = concurrent.futures.wait(
+                [win_fut, easy_fut], timeout=10.0,
+                return_when=concurrent.futures.ALL_COMPLETED
+            )
+            for f in not_done:
+                f.cancel()
+
+        wt = win_text[0]
+        et = easy_text[0]
+
+        # 日志（原始输出逐行记录，用於長期對比兩引擎能力）
+        if wt:
+            for i, l in enumerate(wt.split('\n')):
+                if l.strip():
+                    logger.info(f"[ocr] [windows/原始] line[{i}]: {l.strip()[:120]!r}")
+        if et:
+            for i, l in enumerate(et.split('\n')):
+                if l.strip():
+                    logger.info(f"[ocr] [easyocr/原始] line[{i}]: {l.strip()[:120]!r}")
+
+        if not wt and not et:
+            return ("", "")
+
+        return self._pick_ocr_text(wt, et, im_processed)
+
    def _ocr_read_qq(self) -> list:
        """OCR 截屏 QQ 消息区域 -> 双引擎（Windows OCR + EasyOCR）提取文本。
        包含窗口验证 + 垃圾过滤 + 连续垃圾清缓存。
@@ -1179,64 +1258,8 @@ class MessageMonitor:
            except Exception as e:
                logger.debug(f"[ocr] 预处理失败(跳过): {e}")

-            # ── 双引擎 OCR ──────────────────────────────────────
-            # Windows OCR
-            if im.mode != 'RGBA':
-                im = im.convert('RGBA')
-            logger.debug("[ocr] Windows OCR 开始...")
-            try:
-                sb = SoftwareBitmap(BitmapPixelFormat.RGBA8, im.width, im.height)
-                sb.copy_from_buffer(im.tobytes())
-                win_result = engine.recognize_async(sb).get()
-                win_text = win_result.text.strip()
-                logger.debug(f"[ocr] Windows OCR 完成: {len(win_text)} chars")
-            except (RuntimeError, OSError) as e:
-                estr = str(e)
-                if "bad allocation" in estr or "memory" in estr.lower():
-                    logger.debug(f"[ocr] Windows OCR bad allocation({im.width}x{im.height})，降级到原图重试")
-                    # 用原图（跳过 2x resize）重试
-                    try:
-                        im2 = im_original
-                        if im2.mode != 'RGBA':
-                            im2 = im2.convert('RGBA')
-                        sb2 = SoftwareBitmap(BitmapPixelFormat.RGBA8, im2.width, im2.height)
-                        sb2.copy_from_buffer(im2.tobytes())
-                        win_result2 = engine.recognize_async(sb2).get()
-                        win_text = win_result2.text.strip()
-                        im = im2  # EasyOCR 也用原图
-                        logger.debug(f"[ocr] Windows OCR 降级重试完成: {len(win_text)} chars")
-                    except Exception as e2:
-                        logger.debug(f"[ocr] Windows OCR 降级也失败: {e2}")
-                        win_text = ""
-                else:
-                    logger.debug(f"[ocr] Windows OCR 异常: {e}")
-                    win_text = ""
-
-            # EasyOCR 备用（GPU crash 隔离：通过子进程调用）
-            easy_text = ""
-            if _HAS_EASYOCR and not self._easyocr_broken:
-                logger.debug("[ocr] EasyOCR 子进程...")
-                img_np = _np.array(im)
-                easy_text = self._call_easyocr_safe(img_np, timeout=8)
-                if easy_text is None:
-                    # None = 子进程崩溃或超时；超时已自动杀子进程，下次自动重启
-                    # 只有真正的崩溃（BrokenPipeError / OSError）才永久禁用
-                    easy_text = ""
-                    logger.debug("[ocr] EasyOCR 无结果（超时或崩溃），使用 Windows OCR 结果")
-                elif easy_text:
-                    logger.debug(f"[ocr] EasyOCR 识别完成: {len(easy_text)} chars")
-
-            # 引擎原始输出逐行标记日志（用于长期对比两引擎能力）
-            if win_text:
-                for i, l in enumerate(win_text.split('\n')):
-                    if l.strip():
-                        logger.info(f"[ocr] [windows/原始] line[{i}]: {l.strip()[:120]!r}")
-            if easy_text:
-                for i, l in enumerate(easy_text.split('\n')):
-                    if l.strip():
-                        logger.info(f"[ocr] [easyocr/原始] line[{i}]: {l.strip()[:120]!r}")
-
-            raw_text, chosen_engine = self._pick_ocr_text(win_text, easy_text, im)
+            # ── 双引擎并发 OCR（Windows OCR + EasyOCR 同时跑，最多等 10 秒）──
+            raw_text, chosen_engine = self._ocr_dual_engine(im, im_original, "qq")
            if not raw_text:
                return []
            self._last_chosen_engine["qq"] = chosen_engine
@@ -1325,6 +1348,7 @@ class MessageMonitor:
            self._ocr_unchanged.pop("wechat", None)

            # 图像预处理（同 QQ）
+            im_original = im  # 保留原图，bad allocation 时降级
            try:
                im = im.resize((im.width * 2, im.height * 2), Image.LANCZOS)
                im = ImageEnhance.Sharpness(im).enhance(1.8)
@@ -1332,39 +1356,8 @@ class MessageMonitor:
            except Exception as e:
                logger.debug(f"[wechat] OCR 预处理失败(跳过): {e}")

-            # ── 双引擎 OCR ──
-            if im.mode != 'RGBA':
-                im = im.convert('RGBA')
-            logger.debug("[wechat] Windows OCR 开始...")
-            sb = SoftwareBitmap(BitmapPixelFormat.RGBA8, im.width, im.height)
-            sb.copy_from_buffer(im.tobytes())
-            win_result = engine.recognize_async(sb).get()
-            win_text = win_result.text.strip()
-            logger.debug(f"[wechat] Windows OCR 完成: {len(win_text)} chars")
-
-            easy_text = ""
-            if _HAS_EASYOCR and not self._easyocr_broken:
-                logger.debug("[wechat] EasyOCR 子进程...")
-                img_np = _np.array(im)
-                easy_text = self._call_easyocr_safe(img_np)
-                if easy_text is None:
-                    self._easyocr_broken = True
-                    easy_text = ""
-                    logger.warning("[wechat] EasyOCR 子进程崩溃，已禁用（本次会话不再尝试）")
-                elif easy_text:
-                    logger.debug(f"[wechat] EasyOCR 识别完成: {len(easy_text)} chars")
-
-            # 引擎原始输出逐行标记日志（用于长期对比两引擎能力）
-            if win_text:
-                for i, l in enumerate(win_text.split('\n')):
-                    if l.strip():
-                        logger.info(f"[wechat] [windows/原始] line[{i}]: {l.strip()[:120]!r}")
-            if easy_text:
-                for i, l in enumerate(easy_text.split('\n')):
-                    if l.strip():
-                        logger.info(f"[wechat] [easyocr/原始] line[{i}]: {l.strip()[:120]!r}")
-
-            raw_text, chosen_engine = self._pick_ocr_text(win_text, easy_text, im)
+            # ── 双引擎并发 OCR（Windows OCR + EasyOCR 同时跑，最多等 10 秒）──
+            raw_text, chosen_engine = self._ocr_dual_engine(im, im_original, "wechat")
            if not raw_text:
                return []
            self._last_chosen_engine["wechat"] = chosen_engine