'commit'

Apps/XinDianTu/Crawler.py

@@ -279,8 +279,7 @@ async def get_station_list(d, service, uploader, max_stations_count=MAX_STATIONS
                     loop.run_in_executor(None, uploader.upload_file, detail_object_key, detail_path)
 
                     # --- 抓取价格时段信息 ---
-                    # 【优化】梯级识别策略：优先模板匹配 (qbsd.jpg)，失败则降级为 VL 识别
-                    template_qbsd = os.path.join(os.path.dirname(__file__), "Template", "qbsd.jpg")
+                    # 【优化】使用 OCR 识别“全部时段”按钮，替代之前的模板匹配 (qbsd.jpg)
                     
                     # 记录点击前的页面特征，用于验证是否成功进入三级页面
                     before_click_md5 = Kit.get_image_content_md5(detail_path)
@@ -289,28 +288,20 @@ async def get_station_list(d, service, uploader, max_stations_count=MAX_STATIONS
                     click_pos = None
                     
                     try:
-                        # 1. 尝试模板匹配
-                        match_res = d.image.match(template_qbsd)
-                        if match_res:
-                            # 兼容 uiautomator2 不同版本的返回结果 (Match对象或dict)
-                            if hasattr(match_res, 'point') and match_res.point:
-                                click_pos = match_res.point
-                            elif isinstance(match_res, dict):
-                                if 'point' in match_res and match_res['point']:
-                                    click_pos = match_res['point']
-                                elif 'x' in match_res and 'y' in match_res:
-                                    click_pos = (match_res['x'], match_res['y'])
-                            elif isinstance(match_res, (list, tuple)) and len(match_res) >= 2:
-                                click_pos = (match_res[0], match_res[1])
-                            
-                            if click_pos:
-                                logger.info(f"通过 qbsd.jpg 成功找到！坐标: {click_pos}")
-                            else:
-                                logger.warning(f"模板匹配成功但无法解析坐标: {match_res}")
+                        # 1. 尝试 OCR 识别
+                        logger.info("正在使用 OCR 识别 '全部时段' 按钮...")
+                        ocr_res = await ReadImageKit.find_price_entrance_ocr(detail_path)
                         
-                        # 如果模板匹配未成功解析坐标，或者匹配直接失败，则降级到 VL
+                        if ocr_res.get("found"):
+                            p = ocr_res["point"]
+                            # 归一化坐标转像素坐标
+                            w, h = d.window_size()
+                            click_pos = (int(p[0] * w / 1000), int(p[1] * h / 1000))
+                            logger.info(f"OCR 成功找到 '全部时段' 按钮: {click_pos}")
+                        
+                        # 2. 如果 OCR 未找到，则降级到 VL
                         if not click_pos:
-                            logger.info("模板匹配未找到或解析失败，降级使用 VL 识别...")
+                            logger.info("OCR 未找到，降级使用 VL 识别...")
                             res = await ReadImageKit.find_all_time_button_coordinate(detail_path, device_info=device_info)
                             if res.get("uia_center_x"):
                                 click_pos = (res.get("uia_center_x"), res.get("uia_center_y"))
@@ -327,13 +318,50 @@ async def get_station_list(d, service, uploader, max_stations_count=MAX_STATIONS
                             if before_click_md5 != after_click_md5:
                                 entered_price_page = True
                                 price_screenshots = []
-                                for p_idx in range(1, 4):
+                                
+                                # 1. 向下滚动到底 (根据用户反馈：只有不断向下滚动，才能看到00点的)
+                                logger.info("正在向下滚动价格列表到底部...")
+                                max_scroll_down = 6
+                                for i in range(max_scroll_down):
+                                    before_scroll_path = take_screenshot(d, f"scroll_dn_{i}", save_dir=TEMP_IMAGE_DIR)
+                                    before_scroll_md5 = Kit.get_image_content_md5(before_scroll_path)
+                                    
+                                    d.swipe_ext("up", scale=0.8)
+                                    await asyncio.sleep(1.2)
+                                    
+                                    after_scroll_path = take_screenshot(d, f"scroll_dn_after_{i}", save_dir=TEMP_IMAGE_DIR)
+                                    after_scroll_md5 = Kit.get_image_content_md5(after_scroll_path)
+                                    
+                                    # 清理临时截图
+                                    if os.path.exists(before_scroll_path): os.remove(before_scroll_path)
+                                    if os.path.exists(after_scroll_path): os.remove(after_scroll_path)
+                                    
+                                    if before_scroll_md5 == after_scroll_md5:
+                                        logger.info(f"价格列表已到达底部 (滚动次数: {i})")
+                                        break
+                                
+                                # 2. 向上滚动并逐页截图 (从底向上抓取)
+                                logger.info("正在向上滚动价格列表并逐页截图...")
+                                max_scroll_up = 8
+                                for p_idx in range(1, max_scroll_up + 1):
+                                    # 截图当前页
                                     p_uuid = f"{hashlib.md5(station_name.encode('utf-8')).hexdigest()}_p_{p_idx}"
                                     p_path = take_screenshot(d, p_uuid, save_dir=TEMP_IMAGE_DIR)
                                     price_screenshots.append(p_path)
-                                    if p_idx < 3:
-                                        d.swipe_ext("up", scale=0.7)
-                                        await asyncio.sleep(1.0)
+                                    
+                                    # 检查是否还能向上滚动
+                                    before_up_md5 = Kit.get_image_content_md5(p_path)
+                                    d.swipe_ext("down", scale=0.85) # 向上滚动 = 手势向下
+                                    await asyncio.sleep(1.5)
+                                    
+                                    # 检查是否还有新内容
+                                    check_up_path = take_screenshot(d, f"check_up_{p_idx}", save_dir=TEMP_IMAGE_DIR)
+                                    after_up_md5 = Kit.get_image_content_md5(check_up_path)
+                                    if os.path.exists(check_up_path): os.remove(check_up_path)
+                                    
+                                    if before_up_md5 == after_up_md5:
+                                        logger.info(f"价格列表已到达顶部 (共抓取页数: {p_idx})")
+                                        break
                                 
                                 # 【异步优化】后台处理价格
                                 logger.info(f"已启动后台分析价格页: {station_name}")
@@ -367,12 +395,13 @@ async def get_station_list(d, service, uploader, max_stations_count=MAX_STATIONS
                                     logger.warning(f"返回后状态未知 (ListMD5:{check_list_md5[:8]} vs {current_md5[:8]}, DetailMD5:{check_detail_md5[:8]} vs {before_click_md5[:8]})。")
                                     logger.info("再次尝试识别按钮以确认当前页面状态...")
                                     
-                                    # 优先尝试模板匹配确认
-                                    if d.image.match(template_qbsd):
-                                        logger.info("模板匹配确认当前仍处于详情页，执行返回。")
+                                    # 优先尝试 OCR 确认
+                                    ocr_check = await ReadImageKit.find_price_entrance_ocr(check_back_path)
+                                    if ocr_check.get("found"):
+                                        logger.info("OCR 确认当前仍处于详情页，执行返回。")
                                         should_back_to_list = True
                                     else:
-                                        # 模板匹配未发现，则用 VL 兜底确认
+                                        # OCR 未发现，则用 VL 兜底确认
                                         check_res = await ReadImageKit.find_all_time_button_coordinate(check_back_path, device_info=device_info)
                                         if check_res and check_res.get("uia_center_x"):
                                             logger.info("VL 确认当前仍处于详情页，执行返回。")

Apps/XinDianTu/Kit.py

@@ -137,6 +137,43 @@ def is_background_dimmed(image_path, threshold=80):
     return avg_brightness < threshold
 
 
+from Util.EasyOcrKit import get_easyocr_reader
+
+# 预加载 EasyOCR Reader (单例模式)
+def get_ocr_reader():
+    return get_easyocr_reader(gpu=True)
+
+
+def detect_price_info_container_cv(image_path):
+    """
+    使用 OCR 精准定位详情页中的价格入口文本（“全部时段”）。
+    返回: [x1, y1, x2, y2] 归一化坐标 (0-1000)，如果未找到则返回 None
+    """
+    img = read_image(image_path)
+    if img is None:
+        return None
+    h, w = img.shape[:2]
+    
+    keywords = ['全部时段']
+    
+    try:
+        reader = get_ocr_reader()
+        # 获取所有识别结果
+        results = reader.read_text(img)
+        
+        for (quad, text, prob) in results:
+            # 检查是否包含关键字
+            if any(kw in text for kw in keywords) and prob >= 0.5:
+                # 使用封装后的方法计算归一化矩形
+                res = reader.get_normalized_rect(quad, w, h)
+                logger.info(f"[OCR识别] 找到文本: '{text}', 置信度: {prob:.4f}, 归一化坐标: {res}")
+                return res
+    except Exception as e:
+        logger.error(f"OCR 识别发生异常: {e}")
+        
+    return None
+
+
 def detect_rabbit_ad_close(image_path, debug_dir=None):
     """
     通过图形学算法检测“新电兔AI”广告的关闭按钮

Apps/XinDianTu/ReadImageKit.py

@@ -9,7 +9,7 @@ import logging
 import base64
 from openai import AsyncOpenAI, BadRequestError
 from Apps.XinDianTu.Kit import (
-    read_image, save_image, detect_rabbit_ad_close
+    read_image, save_image, detect_rabbit_ad_close, detect_price_info_container_cv
 )
 from Config.Config import (
     ALY_LLM_API_KEY, VL_MODEL_NAME, VL_MODEL_NAME_AD, TEMP_IMAGE_DIR
@@ -77,6 +77,28 @@ class ReadImageKit:
         "严格返回纯JSON格式。"
     )
 
+    @classmethod
+    async def find_price_entrance_ocr(cls, image_path):
+        """
+        使用 OCR 在详情页寻找价格入口 (全部时段 >)
+        """
+        container_norm = detect_price_info_container_cv(image_path)
+        if container_norm:
+            # 找到矩形框 [x1, y1, x2, y2]
+            x1, y1, x2, y2 = container_norm
+            # 计算中心点
+            center_x = (x1 + x2) // 2
+            center_y = (y1 + y2) // 2
+            
+            return {
+                "found": True,
+                "reason": "OCR 精准定位到价格入口关键字 '全部时段'",
+                "point": [center_x, center_y]
+            }
+        
+        logger.warning(f"⚠️ [OCR识别失败] 在图片中未发现价格入口关键字 (全部时段): {image_path}")
+        return {"found": False}
+
     @classmethod
     async def detect_ad_popup(cls, image_path: str, device_info=None):
         """