EasyOCR GPU crash隔离：子进程隔离GPU推理，崩溃只杀子进程不冻电脑；新增easyocr_worker.py常驻worker；主进程降级到Windows OCR

easyocr_worker.py

@@ -0,0 +1,92 @@
+#!/usr/bin/env python3
+"""EasyOCR 子进程 worker。
+
+通过 stdin 读 base64 图片、stdout 写 JSON 结果。GPU crash 只杀此进程，不冻主进程。
+协议：
+  stdin:  {"img":"<base64_png_bytes>"}
+  stdout: {"ok":true,"text":"识别结果"}
+  stdout: {"ok":false,"error":"错误描述"}
+
+初始化慢（首次 3-10s），初始化完成后常驻处理。
+"""
+
+import json
+import sys
+import base64
+import io
+import os
+
+# ── 线程/进程限制：避免 MKL/OpenMP 在子进程多线程冲突 ──
+os.environ.setdefault("OMP_NUM_THREADS", "1")
+os.environ.setdefault("MKL_NUM_THREADS", "1")
+
+from PIL import Image
+import numpy as np
+import easyocr
+
+
+def _init_reader():
+    """初始化 EasyOCR reader（GPU 优先）。"""
+    try:
+        import torch
+        use_gpu = torch.cuda.is_available()
+    except Exception:
+        use_gpu = False
+
+    try:
+        reader = easyocr.Reader(["ch_sim", "en"], gpu=use_gpu)
+        mode = "GPU" if use_gpu else "CPU"
+        sys.stderr.write(f"[easyocr_worker] Reader 就绪（{mode}模式）\n")
+        sys.stderr.flush()
+        return reader
+    except Exception as e:
+        # GPU 不可用 → CPU 兜底
+        sys.stderr.write(f"[easyocr_worker] GPU 初始化失败({e})，尝试 CPU\n")
+        sys.stderr.flush()
+        reader = easyocr.Reader(["ch_sim", "en"], gpu=False)
+        sys.stderr.write("[easyocr_worker] Reader 就绪（CPU 模式）\n")
+        sys.stderr.flush()
+        return reader
+
+
+def _process_image(img_b64: str, reader) -> str:
+    """解码 base64 图片 → EasyOCR 识别 → 返回空格拼接文本。"""
+    raw = base64.b64decode(img_b64)
+    img = Image.open(io.BytesIO(raw))
+    # 保持与 monitor.py 一致：RGBA → numpy
+    if img.mode != "RGBA":
+        img = img.convert("RGBA")
+    img_np = np.array(img)
+    results = reader.readtext(img_np)
+    return " ".join([r[1] for r in results]).strip()
+
+
+def _main():
+    reader = _init_reader()
+    sys.stderr.write("[easyocr_worker] 就绪，等待 stdin 输入\n")
+    sys.stderr.flush()
+
+    for line in sys.stdin:
+        line = line.strip()
+        if not line:
+            continue
+
+        try:
+            payload = json.loads(line)
+            img_b64 = payload.get("img", "")
+            if not img_b64:
+                raise ValueError("缺少 img 字段")
+            text = _process_image(img_b64, reader)
+            out = json.dumps({"ok": True, "text": text}, ensure_ascii=False)
+            sys.stdout.write(out + "\n")
+            sys.stdout.flush()
+        except Exception as e:
+            out = json.dumps({"ok": False, "error": str(e)}, ensure_ascii=False)
+            sys.stdout.write(out + "\n")
+            sys.stdout.flush()
+            sys.stderr.write(f"[easyocr_worker] 处理失败: {e}\n")
+            sys.stderr.flush()
+
+
+if __name__ == "__main__":
+    _main()

monitor.py

@@ -37,6 +37,8 @@ import win32process
 import re
 import ctypes
 import difflib
+import subprocess
+import json
 import psutil
 import os
 from datetime import datetime
@@ -182,6 +184,9 @@ class MessageMonitor:
         self._ocr_engine = None
         # EasyOCR 备用引擎（lazy init）
         self._easyocr_reader = None
+        # EasyOCR 子进程（GPU crash 隔离，崩溃只杀子进程不冻电脑）
+        self._easyocr_proc: Optional['subprocess.Popen'] = None
+        self._easyocr_broken: bool = False
         # 引擎选择计数器
         self._ocr_engine_wins: dict = {"windows": 0, "easyocr": 0}
         # 上次 OCR 原始文本（用于增量检测）
@@ -208,11 +213,7 @@ class MessageMonitor:
         if self._running:
             return
         self._running = True
-        # 在主线程预加载 EasyOCR，避免 daemon 线程 torch 段错误
-        if _HAS_EASYOCR:
-            import torch
-            torch.set_num_threads(1)
-            self._init_easyocr()
+        # EasyOCR 已在子进程隔离（easyocr_worker.py），不需在主线程预加载
         self._thread = threading.Thread(target=self._run_loop, daemon=True, name="monitor")
         self._thread.start()
         logger.info("消息监控已启动")
@@ -766,6 +767,82 @@ class MessageMonitor:
             logger.error(f"[ocr] EasyOCR 初始化失败: {e}")
             self._easyocr_reader = None
 
+    # ── EasyOCR 子进程隔离 ────────────────────────────────────
+    # 直接 GPU 推理 crash（access violation）会杀整个进程 + 冻电脑。
+    # 子进程方案：GPU crash 只杀子进程，主进程降级到 Windows OCR。
+
+    def _ensure_easyocr_subprocess(self) -> bool:
+        """确保 EasyOCR 子进程在运行。返回 True = 可用。"""
+        if self._easyocr_broken:
+            return False
+        if self._easyocr_proc is not None:
+            if self._easyocr_proc.poll() is None:
+                return True  # 还在跑
+            logger.warning("[ocr] EasyOCR 子进程已退出，重启...")
+            self._easyocr_proc = None
+
+        worker_path = os.path.join(
+            os.path.dirname(os.path.abspath(__file__)),
+            'easyocr_worker.py'
+        )
+        if not os.path.exists(worker_path):
+            logger.error(f"[ocr] EasyOCR worker 不存在: {worker_path}")
+            self._easyocr_broken = True
+            return False
+
+        try:
+            self._easyocr_proc = subprocess.Popen(
+                [sys.executable, '-u', worker_path],
+                stdin=subprocess.PIPE,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.DEVNULL,   # 防 stderr 管道满阻塞
+                text=True,
+                bufsize=1,  # 行缓冲
+            )
+            logger.debug("[ocr] EasyOCR 子进程已启动")
+            return True
+        except Exception as e:
+            logger.error(f"[ocr] 启动 EasyOCR 子进程失败: {e}")
+            self._easyocr_broken = True
+            self._easyocr_proc = None
+            return False
+
+    def _call_easyocr_safe(self, img_np) -> Optional[str]:
+        """通过子进程安全调用 EasyOCR。返回识别文本，失败/崩溃时返回 None。"""
+        import io
+        import base64
+
+        if not self._ensure_easyocr_subprocess():
+            return None
+
+        try:
+            # numpy → PIL → PNG bytes → base64
+            img_pil = Image.fromarray(img_np)
+            buf = io.BytesIO()
+            img_pil.save(buf, format='PNG')
+            img_b64 = base64.b64encode(buf.getvalue()).decode()
+
+            self._easyocr_proc.stdin.write(json.dumps({"img": img_b64}) + "\n")
+            self._easyocr_proc.stdin.flush()
+
+            line = self._easyocr_proc.stdout.readline()
+            if not line:
+                raise ConnectionError("EasyOCR 子进程 stdout 关闭")
+            result = json.loads(line)
+            if result.get("ok"):
+                return result["text"]
+            else:
+                logger.debug(f"[ocr] EasyOCR worker 错误: {result.get('error', '?')[:100]}")
+                return None
+        except (BrokenPipeError, ConnectionError, OSError) as e:
+            logger.debug(f"[ocr] EasyOCR 子进程通信断开: {e}")
+            self._easyocr_broken = True
+            self._easyocr_proc = None
+            return None
+        except Exception as e:
+            logger.debug(f"[ocr] EasyOCR 子进程异常: {e}")
+            return None
+
     @staticmethod
     def _score_ocr_text(text: str) -> float:
         """为 OCR 结果质量打分（越高越好）。对比引擎用。"""
@@ -924,21 +1001,18 @@ class MessageMonitor:
             win_text = win_result.text.strip()
             logger.debug(f"[ocr] Windows OCR 完成: {len(win_text)} chars")
 
-            # EasyOCR 备用
+            # EasyOCR 备用（GPU crash 隔离：通过子进程调用）
             easy_text = ""
-            if _HAS_EASYOCR:
-                logger.debug("[ocr] EasyOCR 检查/初始化...")
-                self._init_easyocr()
-                logger.debug("[ocr] EasyOCR reader 就绪状态")
-                if self._easyocr_reader:
-                    try:
-                        logger.debug("[ocr] EasyOCR 识别开始...")
-                        img_np = _np.array(im)
-                        easy_results = self._easyocr_reader.readtext(img_np)
-                        easy_text = ' '.join([r[1] for r in easy_results]).strip()
-                        logger.debug(f"[ocr] EasyOCR 识别完成: {len(easy_text)} chars")
-                    except Exception as e:
-                        logger.debug(f"[ocr] EasyOCR 识别失败: {e}")
+            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)
+                if easy_text is None:
+                    self._easyocr_broken = True
+                    easy_text = ""
+                    logger.warning("[ocr] EasyOCR 子进程崩溃，已禁用（本次会话不再尝试）")
+                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:
@@ -1041,18 +1115,16 @@ class MessageMonitor:
             logger.debug(f"[wechat] Windows OCR 完成: {len(win_text)} chars")
 
             easy_text = ""
-            if _HAS_EASYOCR:
-                logger.debug("[wechat] EasyOCR 检查/初始化...")
-                self._init_easyocr()
-                logger.debug("[wechat] EasyOCR reader 就绪状态")
-                if self._easyocr_reader:
-                    try:
-                        logger.debug("[wechat] EasyOCR 识别开始...")
-                        img_np = _np.array(im)
-                        easy_results = self._easyocr_reader.readtext(img_np)
-                        easy_text = ' '.join([r[1] for r in easy_results]).strip()
-                    except Exception as e:
-                        logger.debug(f"[wechat] EasyOCR 识别失败: {e}")
+            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: