完成初版选点提取坐标

biz/base_detector.py

@@ -0,0 +1,62 @@
+from collections import deque
+from typing import Optional
+import numpy as np
+
+
+class BaseDetector:
+    """
+    检测器基类
+    提供通用的帧回溯缓存功能，子类可按需使用
+    """
+
+    def __init__(self):
+        # 帧回溯缓存（子类需要时调用 init_frame_buffer 初始化）
+        self._frame_buffer: Optional[deque] = None
+
+    def init_frame_buffer(self, buffer_seconds: float, fps: float):
+        """
+        初始化帧回溯缓存队列
+
+        Args:
+            buffer_seconds: 需要缓存的时间长度（秒）
+            fps: 视频帧率
+        """
+        maxlen = int(buffer_seconds * fps)
+        self._frame_buffer = deque(maxlen=maxlen)
+
+    def append_frame(self, frame: np.ndarray, timestamp: float):
+        """
+        将当前帧入队缓存
+
+        Args:
+            frame: 当前帧图像
+            timestamp: 当前帧的时间戳
+        """
+        if self._frame_buffer is not None:
+            self._frame_buffer.append({
+                'timestamp': timestamp,
+                'frame': frame.copy(),
+            })
+
+    def find_target_frame(self, target_time_sec: float) -> Optional[np.ndarray]:
+        """
+        在帧缓存中找到最接近目标时间的帧
+
+        Args:
+            target_time_sec: 目标时间戳
+
+        Returns:
+            最接近目标时间的帧图像，缓存为空则返回 None
+        """
+        if self._frame_buffer is None or len(self._frame_buffer) == 0:
+            return None
+
+        target_frame = None
+        min_time_diff = float('inf')
+        for buffered in self._frame_buffer:
+            time_diff = abs(buffered['timestamp'] - target_time_sec)
+            if time_diff < min_time_diff:
+                min_time_diff = time_diff
+                target_frame = buffered['frame']
+
+        return target_frame

biz/checkpoint/checkpoint_biz.py

@@ -1,16 +1,14 @@
 import cv2
 import numpy as np
 from typing import Dict, Any
-import threading
-import queue
-from collections import deque
 
 
 from biz.base_frame_processor import BaseFrameProcessorWorker
+from biz.base_detector import BaseDetector
 
 # -------------------------- Kadian 检测相关导入 --------------------------
 from algorithm.common.npu_yolo_onnx_person_car_phone import YOLOv8_ONNX          # 主检测模型（人/车/后备箱/手机）
-from algorithm.common.npu_yolo_pose_onnx import YOLOv8_Pose_ONNX                 # Pose 专用模型
+# from algorithm.common.npu_yolo_pose_onnx import YOLOv8_Pose_ONNX                 # Pose 专用模型
 from yolox.tracker.byte_tracker import BYTETracker
 
 from utils.logger import get_logger
@@ -55,8 +53,9 @@ PERSON_CAR_INPUT_SIZE = 640
 RTSP_TARGET_FPS = 10.0
 
 # ========================= Kadian TrafficMonitor（精简版，专为服务设计） =========================
-class KadianDetector:
+class KadianDetector(BaseDetector):
     def __init__(self, params=None):
+        super().__init__()
         # 摄像头额外参数
         self.params = params if params is not None else {}
 
@@ -147,8 +146,8 @@ class KadianDetector:
         self.nobody_frames = 0  # 累计无人在场帧数
         self.only_one_frames = 0  # 累计单人在场帧数
 
-        self.max_car_frames = int((15.0 + self.TIME_TOLERANCE_CAR) * self.fps)  # 
-        self.frame_buffer_ignore_untrunk = deque(maxlen=self.max_car_frames)
+        buffer_seconds = 15.0 + self.TIME_TOLERANCE_CAR
+        self.init_frame_buffer(buffer_seconds, self.fps)
 
         self.untrunk_rollback_time = 12.0    # 未检查后备箱需要回溯的时间
         self.ignored_rollback_time = 12.0    # 漏检需要回溯的时间
@@ -219,21 +218,6 @@ class KadianDetector:
         x1, y1, x2, y2 = box
         return x1 < px < x2 and y1 < py < y2
 
-    def find_target_frame(self, target_time_sec):
-
-        target_frame = None
-        min_time_diff = float('inf')
-        for buffered in self.frame_buffer_ignore_untrunk:
-            time_diff = abs(buffered['timestamp'] - target_time_sec)
-            if time_diff < min_time_diff:
-                min_time_diff = time_diff
-                target_frame = buffered['frame']
-        # 如果没找到，返回最早的帧
-        if target_frame is None and len(self.frame_buffer_ignore_untrunk) > 0:
-            target_frame = self.frame_buffer_ignore_untrunk[0]['frame']
-
-        return target_frame
-
     def process_frame(self, frame, camera_id: int, timestamp: float) -> Dict[str, Any]:
         h, w = frame.shape[:2]
         self.width, self.height = w, h
@@ -402,11 +386,7 @@ class KadianDetector:
             cv2.putText(frame, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 2)
 
         # 每帧保存到缓存（移到循环外，确保每帧只写入一次）
-        self.frame_buffer_ignore_untrunk.append({
-            'frame_idx': self.current_frame_idx,
-            'timestamp': current_time_sec,
-            'frame': frame.copy(),
-        })
+        self.append_frame(frame, current_time_sec)
 
         # ==========================================
         # 关联分析: 哪个后备箱属于哪辆车?

biz/prison/indoor_biz.py

@@ -1,7 +1,10 @@
 import cv2
 import numpy as np
 import time
-import requests
+# import requests
+from collections import deque
+
+from biz.base_detector import BaseDetector
 from biz.base_frame_processor import BaseFrameProcessorWorker
 from algorithm.common.npu_yolo_onnx_person_car_phone import YOLOv8_ONNX
 from yolox.tracker.byte_tracker import BYTETracker
@@ -12,21 +15,26 @@ DETECT_MODEL_PATH = 'YOLO_Weight/kanshousuo.onnx'  # 犯人检测onnx模型路
 INPUT_SIZE = 640  # 模型输入尺寸
 RTSP_FPS = 10  # 视频流目标FPS
 ALERT_PUSH_INTERVAL = 5  # 相同报警5秒内仅推送1次
-ALERT_PUSH_URL = "http://123.57.151.210:10000/picenter/websocket/test/process"
+# ALERT_PUSH_URL = "http://123.57.151.210:10000/picenter/websocket/test/process"
 # 消失判定：中心点在ROI内消失后，持续无检测的帧数（1.0秒，可微调）
-ROI_LOST_FRAMES_THRESH = int(0.5 * RTSP_FPS)
+ROI_LOST_FRAMES_THRESH = int(0.5 * RTSP_FPS)  # todo: 从frame改为时间
 
 # ========================= 默认ROI区域配置（当config.yaml未配置时使用） =========================
 DEFAULT_DOOR_ROIS = {
-    "left_door_1": {
-        "points": [[0.195, 0.242], [0.265, 0.17], [0.3, 0.63], [0.248, 0.8]],
+    "left": {
+        "points": [[0.195, 0.245], [0.42, 0], [0.421, 0.185], [0.248, 0.8]],
+        "color": [255, 0, 0]
+    },
+    "right": {
+        "points": [[0.575, 0.], [0.81, 0.22], [0.78, 0.8], [0.575, 0.185]],
         "color": [255, 0, 0]
     }
 }
 
 # ==================================================================================
-class PrisonerDoorDetector:
+class PrisonerDoorDetector(BaseDetector):
     def __init__(self, params=None):
+        super().__init__()
         self.params = params or {}
 
         # 0. 从params解析ROI配置，无则使用默认值
@@ -45,26 +53,48 @@ class PrisonerDoorDetector:
 
         self.detector = YOLOv8_ONNX(
             full_model_path,
-            conf_threshold=0.5,  # 置信度阈值，可根据模型精度调整
-            iou_threshold=0.45,  # IOU阈值
+            conf_threshold=0.7,  # 置信度阈值，可根据模型精度调整
+            iou_threshold=0.4,  # IOU阈值
             input_size=INPUT_SIZE
         )
 
         # 2. 初始化ByteTracker跟踪器（适配走廊单/多犯人跟踪）
         class TrackerArgs:
-            track_thresh = 0.25
-            track_buffer = 20  # 减小缓冲避免跟踪漂移
-            match_thresh = 0.75
+            track_thresh = 0.65
+            track_buffer = 60  # 减小缓冲避免跟踪漂移
+            match_thresh = 0.5
             mot20 = False
         self.tracker = BYTETracker(TrackerArgs(), frame_rate=RTSP_FPS)
 
         # 3. 状态变量初始化
-        self.last_alert_time = 0.0  # 最后报警时间（防重复推送）
+        # self.last_alert_time = 0.0  # 最后报警时间（防重复推送）
         # 犯人跟踪信息：{track_id: {'is_cx_in_roi': 中心点是否在ROI, 'lost_frames': 消失帧数, 'lost_roi': 消失的ROI名称, 'last_cxcy': 最后中心点坐标}}
         self.prisoner_track_info = {}
         self.frame_width = 0  # 帧宽度（动态获取）
         self.frame_height = 0  # 帧高度（动态获取）
         self.roi_abs_cache = {}  # ROI绝对坐标缓存：{roi_name: np.int32数组}
+        self.entry_frame_cache = {}
+
+        # 基于位置的跟踪状态管理
+        self.active_targets = {}  # {target_id: {...}}
+        self.next_target_id = 0
+        self.position_history = {}  # {target_id: deque of positions}
+
+        # 距离阈值（用于匹配检测框和已有目标）
+        self.distance_threshold = 100  # 像素距离
+
+        buffer_seconds = 3  # 最大回溯3秒
+        self.init_frame_buffer(buffer_seconds, RTSP_FPS)
+        self.detect_rollback_time = 0.9 # 警报帧回溯时间（秒）
+
+
+    def compute_center_distance(self, box1, box2):
+        """计算两个框中心点的欧氏距离"""
+        cx1 = (box1[0] + box1[2]) / 2
+        cy1 = (box1[1] + box1[3]) / 2
+        cx2 = (box2[0] + box2[2]) / 2
+        cy2 = (box2[1] + box2[3]) / 2
+        return np.sqrt((cx1 - cx2) ** 2 + (cy1 - cy2) ** 2)
 
     def compute_iou(self, boxA, boxB):
         """IOU计算：匹配跟踪框与犯人检测框，过滤非犯人目标"""
@@ -96,32 +126,81 @@ class PrisonerDoorDetector:
                 return (True, roi_name)
         return (False, "outside")
 
-    # def push_alert(self, camera_id, track_id, lost_roi, last_cxcy, timestamp):
-    #     """报警推送：带频率限制，携带消失ROI、最后中心点坐标"""
+    def match_detection_to_target(self, detection_box, detection_conf):
+        """
+        【核心】将检测框匹配到已有目标
+        返回: (matched_target_id, match_score)
+        """
+        best_match_id = None
+        best_match_score = 0
+
+        det_center = np.array([(detection_box[0] + detection_box[2]) / 2,
+                               (detection_box[1] + detection_box[3]) / 2])
+
+        for target_id, target_info in self.active_targets.items():
+            # 计算与目标最后已知位置的距离
+            last_box = target_info['last_box']
+            last_center = np.array([(last_box[0] + last_box[2]) / 2,
+                                    (last_box[1] + last_box[3]) / 2])
+
+            distance = np.linalg.norm(det_center - last_center)
+
+            # 计算IOU（如果目标最近刚更新）
+            time_since_update = time.time() - target_info['last_update_time']
+            iou_score = self.compute_iou(detection_box, last_box) if time_since_update < 1.0 else 0
+
+            # 综合评分：距离近 + IOU高
+            distance_score = max(0, 1 - distance / self.distance_threshold)
+            match_score = 0.3 * distance_score + 0.7 * iou_score
+
+            # 考虑位置预测（如果目标在移动中）
+            if target_id in self.position_history and len(self.position_history[target_id]) >= 2:
+                # 简单的线性预测
+                hist = list(self.position_history[target_id])
+                if len(hist) >= 2:
+                    velocity = hist[-1] - hist[-2]
+                    predicted_pos = last_center + velocity
+                    pred_distance = np.linalg.norm(det_center - predicted_pos)
+                    pred_score = max(0, 1 - pred_distance / self.distance_threshold)
+                    match_score = 0.7 * match_score + 0.3 * pred_score
+
+            if match_score > best_match_score and match_score > 0.3:  # 阈值可调
+                best_match_score = match_score
+                best_match_id = target_id
+
+        return best_match_id, best_match_score
+
+    # def push_alert(self, camera_id, target_id, lost_roi, last_cxcy, timestamp, entry_frame):
+    #     """报警推送"""
     #     current_time = time.time()
     #     if current_time - self.last_alert_time < ALERT_PUSH_INTERVAL:
     #         return False
-    #     # 构造报警信息（可根据平台要求扩展字段）
+    #
+    #     _, frame_encoded = cv2.imencode('.jpg', entry_frame)
+    #     frame_base64 = frame_encoded.tobytes()
+    #
     #     alert_info = {
     #         "camera_id": camera_id,
     #         "alert_type": "prisoner_cx_disappear_in_roi",
-    #         "prisoner_track_id": track_id,
+    #         "prisoner_track_id": target_id,
     #         "disappear_roi": lost_roi,
     #         "last_cx": round(last_cxcy[0], 2),
     #         "last_cy": round(last_cxcy[1], 2),
     #         "timestamp": timestamp,
-    #         "details": f"犯人框中心点在{lost_roi}区域内消失，触发报警"
+    #         "entry_frame_base64": frame_base64,
+    #         "details": f"犯人框中心点在{lost_roi}区域内消失"
     #     }
-    #     # 推送报警请求
+    #
     #     try:
     #         requests.post(ALERT_PUSH_URL, json=alert_info, timeout=3)
-    #         print(f"[报警成功] {alert_info}")
+    #         print(f"[报警成功] target_id={target_id}, roi={lost_roi}")
     #         self.last_alert_time = current_time
     #         return True
     #     except Exception as e:
-    #         print(f"[报警失败] 原因：{str(e)}")
+    #         print(f"[报警失败] {str(e)}")
     #         return False
 
+
     def process_frame(self, frame, camera_id: int, timestamp: float) -> dict:
         """
         核心帧处理：
@@ -130,6 +209,8 @@ class PrisonerDoorDetector:
         """
         self.frame_height, self.frame_width = frame.shape[:2]
         current_frame_alerts = []  # 本帧报警信息
+        frame_copy = frame.copy()
+        current_time = time.time()
 
         # ========================= 1. 初始化ROI绝对坐标并绘制ROI =========================
         self.roi_abs_cache.clear()
@@ -147,98 +228,260 @@ class PrisonerDoorDetector:
 
         # ========================= 2. 模型推理：仅提取犯人检测框 =========================
         detect_results = self.detector(frame)
-        prisoner_dets_xyxy = []  # 仅存犯人检测框 [x1,y1,x2,y2]
-        dets_for_tracker = []    # 跟踪器输入 [x1,y1,x2,y2,conf]
+        prisoner_detections = []
+
         if detect_results:
             for det in detect_results:
                 x1, y1, x2, y2, conf, cls_id = det
-                dets_for_tracker.append([x1, y1, x2, y2, conf])
-                # 替换为你模型中「犯人」的实际类别ID，此处默认cls_id=1
-                if cls_id == 1:
-                    prisoner_dets_xyxy.append([x1, y1, x2, y2])
+                # 确保坐标在图像范围内
+                x1 = max(0, min(x1, self.frame_width - 1))
+                y1 = max(0, min(y1, self.frame_height - 1))
+                x2 = max(0, min(x2, self.frame_width - 1))
+                y2 = max(0, min(y2, self.frame_height - 1))
 
-        # ========================= 3. 目标跟踪：更新犯人跟踪结果 =========================
-        dets_np = np.array(dets_for_tracker, dtype=np.float32) if dets_for_tracker else np.empty((0, 5))
-        track_results = self.tracker.update(dets_np, [self.frame_height, self.frame_width],
-                                            [self.frame_height, self.frame_width])
+                if cls_id == 1 and x2 > x1 and y2 > y1 and (x2 - x1) * (y2 - y1) > 100:  # 过滤太小的框
+                    prisoner_detections.append([x1, y1, x2, y2, conf, cls_id])
+
+        # ========================= 3. ByteTracker跟踪 =========================
+        prisoner_det_boxes = np.array(
+            [[x1, y1, x2, y2, conf] for x1, y1, x2, y2, conf, cls_id in prisoner_detections],
+            dtype=np.float32) if prisoner_detections else np.empty((0, 5))
+
+        if len(prisoner_det_boxes) > 0:
+            track_results = self.tracker.update(
+                prisoner_det_boxes,
+                [self.frame_height, self.frame_width],
+                [self.frame_height, self.frame_width]
+            )
+        else:
+            track_results = []
+
+
+        # ========================= 4. 【核心改进】融合跟踪和检测 =========================
+        # 4.1 先处理跟踪结果
+        tracked_detections = {}  # {track_id: detection_box}
+        used_det_indices = set()
 
-        # ========================= 4. 遍历跟踪结果：判定犯人中心点是否在ROI =========================
-        current_prisoner_tids = set()  # 本帧存在的犯人track_id
         for track in track_results:
             track_id = track.track_id
-            track_box = list(map(float, track.tlbr))  # 跟踪框 [x1,y1,x2,y2]
-            # IOU匹配：过滤非犯人目标，仅保留真正的犯人
-            is_prisoner = False
-            for p_box in prisoner_dets_xyxy:
-                if self.compute_iou(track_box, p_box) > 0.3:
-                    is_prisoner = True
-                    break
-            if not is_prisoner:
-                continue
+            t_box = [float(x) for x in track.tlbr]
 
-            # 计算犯人框**中心点坐标**（核心判定依据）
-            cx = (track_box[0] + track_box[2]) / 2
-            cy = (track_box[1] + track_box[3]) / 2
-            # 判定中心点是否在ROI内，返回(是否在ROI, 所在ROI名称)
-            is_cx_in_roi, current_roi = self.is_cxcy_in_roi(cx, cy)
-            # 更新犯人跟踪信息：记录中心点状态、所在ROI、最后坐标，重置消失帧数
-            self.prisoner_track_info[track_id] = {
-                "is_cx_in_roi": is_cx_in_roi,
-                "lost_frames": 0,
-                "lost_roi": current_roi,
-                "last_cxcy": (cx, cy)
-            }
-            current_prisoner_tids.add(track_id)
+            # 寻找匹配的检测框
+            best_iou = 0.0  # 最低阈值
+            best_det_idx = -1
 
-            # 绘制犯人框+中心点+状态标签（可视化调试）
-            x1, y1, x2, y2 = map(int, track_box)
-            cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 2)  # 红色犯人框
-            cv2.circle(frame, (int(cx), int(cy)), 5, (0, 255, 255), -1)  # 黄色中心点
-            cv2.putText(frame, f"Prisoner_{track_id}({current_roi})", (x1, y1 - 10),
-                        cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 255), 2)
+            for det_idx, det in enumerate(prisoner_detections):
+                if det_idx in used_det_indices:
+                    continue
+                iou = self.compute_iou(t_box, det[:4])
+                if iou > best_iou:
+                    best_iou = iou
+                    best_det_idx = det_idx
 
-        # ========================= 5. 核心判定：中心点在ROI内消失则报警 =========================
-        for track_id in list(self.prisoner_track_info.keys()):
-            if track_id not in current_prisoner_tids:
-                # 犯人本帧消失，获取其最后状态
-                track_info = self.prisoner_track_info[track_id]
-                # 仅处理「**中心点原本在ROI内**」的消失情况
-                if track_info["is_cx_in_roi"]:
-                    track_info["lost_frames"] += 1  # 累计消失帧数
-                    # 消失帧数达到阈值，触发报警
-                    if track_info["lost_frames"] >= ROI_LOST_FRAMES_THRESH:
+            if best_det_idx != -1:
+                # 跟踪框有对应的检测框，使用检测框（更准确）
+                tracked_detections[f"track_{track_id}"] = {
+                    'box': prisoner_detections[best_det_idx][:4],
+                    'conf': prisoner_detections[best_det_idx][4],
+                    'source': 'tracked'
+                }
+                used_det_indices.add(best_det_idx)
+            else:
+                # 跟踪框没有对应的检测框，但仍保留跟踪框
+                tracked_detections[f"track_{track_id}"] = {
+                    'box': t_box,
+                    'conf': 0.5,  # 给个中等置信度
+                    'source': 'track_only'
+                }
+
+        # 4.2 处理未被跟踪的检测框
+        for det_idx, det in enumerate(prisoner_detections):
+            if det_idx not in used_det_indices:
+                tracked_detections[f"det_{det_idx}"] = {
+                    'box': det[:4],
+                    'conf': det[4],
+                    'source': 'det_only'
+                }
+
+        # ========================= 5. 匹配到已有目标 =========================
+        current_target_ids = set()
+        matched_det_keys = set()
+
+        for det_key, det_info in tracked_detections.items():
+            det_box = det_info['box']
+            det_conf = det_info['conf']
+
+            # 计算中心点
+            cx = (det_box[0] + det_box[2]) / 2
+            cy = (det_box[1] + det_box[3]) / 2
+
+            # 匹配到已有目标
+            matched_target_id, match_score = self.match_detection_to_target(det_box, det_conf)
+
+            if matched_target_id is not None and match_score > 0.3:
+                # 更新已有目标
+                target_id = matched_target_id
+                target_info = self.active_targets[target_id]
+
+                # 更新位置历史
+                if target_id not in self.position_history:
+                    self.position_history[target_id] = deque(maxlen=10)
+                self.position_history[target_id].append(np.array([cx, cy]))
+
+                # 判断是否在ROI内
+                is_cx_in_roi, current_roi = self.is_cxcy_in_roi(cx, cy)
+
+                # 首次进入ROI缓存帧
+                if not target_info.get('in_roi', False) and is_cx_in_roi:
+                    self.entry_frame_cache[target_id] = frame_copy.copy()
+                    target_info['lost_frames'] = 0
+
+                # 更新目标信息
+                target_info.update({
+                    'last_box': det_box,
+                    'last_cxcy': (cx, cy),
+                    'last_conf': det_conf,
+                    'last_update_time': current_time,
+                    'in_roi': is_cx_in_roi,
+                    'current_roi': current_roi if is_cx_in_roi else target_info.get('current_roi', 'outside'),
+                    'detection_source': det_info['source']
+                })
+
+                current_target_ids.add(target_id)
+                matched_det_keys.add(det_key)
+
+            else:
+                # 创建新目标
+                target_id = self.next_target_id
+                self.next_target_id += 1
+
+                is_cx_in_roi, current_roi = self.is_cxcy_in_roi(cx, cy)
+
+                self.active_targets[target_id] = {
+                    'first_seen': current_time,
+                    'last_box': det_box,
+                    'last_cxcy': (cx, cy),
+                    'last_conf': det_conf,
+                    'last_update_time': current_time,
+                    'in_roi': is_cx_in_roi,
+                    'current_roi': current_roi if is_cx_in_roi else 'outside',
+                    'lost_frames': 0,
+                    'detection_source': det_info['source']
+                }
+
+                self.position_history[target_id] = deque(maxlen=10)
+                self.position_history[target_id].append(np.array([cx, cy]))
+
+                if is_cx_in_roi:
+                    self.entry_frame_cache[target_id] = frame_copy.copy()
+
+                current_target_ids.add(target_id)
+                matched_det_keys.add(det_key)
+
+        # ========================= 6. 处理消失和报警 =========================
+        for target_id in list(self.active_targets.keys()):
+            target_info = self.active_targets[target_id]
+
+            if target_id not in current_target_ids:
+                # 目标在当前帧未出现
+                if target_info['in_roi']:
+                    # 在ROI内消失
+                    target_info['lost_frames'] += 1
+
+                    if target_info['lost_frames'] >= ROI_LOST_FRAMES_THRESH:
+                        # 触发报警
+                        # entry_frame = self.entry_frame_cache.get(target_id, frame_copy)
                         # self.push_alert(
                         #     camera_id=camera_id,
-                        #     track_id=track_id,
-                        #     lost_roi=track_info["lost_roi"],
-                        #     last_cxcy=track_info["last_cxcy"],
-                        #     timestamp=timestamp
+                        #     target_id=target_id,
+                        #     lost_roi=target_info['current_roi'],
+                        #     last_cxcy=target_info['last_cxcy'],
+                        #     timestamp=timestamp,
+                        #     entry_frame=entry_frame
                         # )
-                        # 记录本帧报警信息
+                        alert_frame = self.find_target_frame(timestamp - self.detect_rollback_time)
+
                         current_frame_alerts.append({
                             "time": timestamp,
                             "camera_id": camera_id,
                             "action": "Indoor Violation",
-                            "prisoner_track_id": track_id,
-                            "disappear_roi": track_info["lost_roi"],
-                            "last_cx": round(track_info["last_cxcy"][0], 2),
-                            "last_cy": round(track_info["last_cxcy"][1], 2)
+                            'image': alert_frame,
+                            "prisoner_track_id": target_id,
+                            "disappear_roi": target_info['current_roi'],
+                            "last_cx": round(target_info['last_cxcy'][0], 2),
+                            "last_cy": round(target_info['last_cxcy'][1], 2)
                         })
-                        del self.prisoner_track_info[track_id]  # 报警后清除状态，避免重复触发
-                else:
-                    del self.prisoner_track_info[track_id]  # 中心点不在ROI的消失，直接清除
 
-        # ========================= 6. 绘制辅助信息（摄像头ID、在押犯人数） =========================
+                        # 清理
+                        del self.active_targets[target_id]
+                        if target_id in self.position_history:
+                            del self.position_history[target_id]
+                        if target_id in self.entry_frame_cache:
+                            del self.entry_frame_cache[target_id]
+                else:
+                    # 不在ROI内消失，直接清理
+                    del self.active_targets[target_id]
+                    if target_id in self.position_history:
+                        del self.position_history[target_id]
+                    if target_id in self.entry_frame_cache:
+                        del self.entry_frame_cache[target_id]
+            else:
+                # 目标仍在，但可能已离开ROI
+                if not target_info['in_roi']:
+                    target_info['lost_frames'] = 0
+
+        # ========================= 7. 清理超时目标 =========================
+        timeout_threshold = 5.0  # 5秒无更新就清理
+        for target_id in list(self.active_targets.keys()):
+            if current_time - self.active_targets[target_id]['last_update_time'] > timeout_threshold:
+                del self.active_targets[target_id]
+                if target_id in self.position_history:
+                    del self.position_history[target_id]
+                if target_id in self.entry_frame_cache:
+                    del self.entry_frame_cache[target_id]
+
+        # ========================= 8. 绘制可视化 =========================
+        for target_id, target_info in self.active_targets.items():
+            box = target_info['last_box']
+            cx, cy = target_info['last_cxcy']
+            in_roi = target_info['in_roi']
+            current_roi = target_info['current_roi']
+            source = target_info.get('detection_source', 'unknown')
+
+            # 根据状态选择颜色
+            if in_roi:
+                color = (0, 0, 255)  # 红色：在ROI内
+            else:
+                color = (0, 255, 0)  # 绿色：不在ROI内
+
+            # 根据来源选择线型
+            thickness = 3 if source == 'tracked' else 2
+
+            cv2.rectangle(frame, (int(box[0]), int(box[1])),
+                          (int(box[2]), int(box[3])), color, thickness)
+            cv2.circle(frame, (int(cx), int(cy)), 5, color, -1)
+
+            status = f"T{target_id}_{current_roi[:2]}"
+            if source == 'det_only':
+                status += "_DET"
+            cv2.putText(frame, status, (int(box[0]), int(box[1]) - 10),
+                        cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
+
+        # ========================= 9. 统计信息 =========================
         cv2.putText(frame, f"Camera: {camera_id}", (20, self.frame_height - 20),
                     cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 0), 2)
-        cv2.putText(frame, f"Prisoners: {len(current_prisoner_tids)}", (20, self.frame_height - 50),
+        cv2.putText(frame, f"Active Targets: {len(self.active_targets)}",
+                    (20, self.frame_height - 50),
                     cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 0), 2)
 
+        self.append_frame(frame, timestamp)
+
         return {"image": frame, "alerts": current_frame_alerts}
 
-# ========================= 帧处理线程（对接原有框架，直接复用） =========================
+
+# ========================= 帧处理线程 =========================
 class FrameProcessorWorker(BaseFrameProcessorWorker):
-    """看守所走廊犯人检测 - 5ROI+中心点消失判定"""
+    """看守所走廊犯人检测 - 增强跟踪版"""
     DETECTOR_FACTORY = lambda params: PrisonerDoorDetector(params)
-    POST_TYPE = 3  # 与原有业务区分，自定义即可
+    POST_TYPE = 3
     TARGET_FPS = RTSP_FPS

live_catalog/index.html

@@ -150,6 +150,63 @@
             overflow: hidden;
         }
 
+        /* 视频控制栏 */
+        .video-controls {
+            position: absolute;
+            top: 0;
+            left: 0;
+            right: 0;
+            padding: 12px 16px;
+            background: linear-gradient(to bottom, rgba(0,0,0,0.7), transparent);
+            display: flex;
+            align-items: center;
+            gap: 16px;
+            z-index: 10;
+        }
+
+        .video-controls label {
+            color: #fff;
+            font-size: 13px;
+        }
+
+        .video-controls select {
+            padding: 4px 8px;
+            border-radius: 4px;
+            border: 1px solid #444;
+            background: #333;
+            color: #fff;
+            font-size: 13px;
+            cursor: pointer;
+        }
+
+        /* 视频信息栏 */
+        .video-info {
+            position: absolute;
+            bottom: 0;
+            left: 0;
+            right: 0;
+            padding: 12px 16px;
+            background: linear-gradient(to top, rgba(0,0,0,0.8), transparent);
+            color: #ccc;
+            font-size: 12px;
+            z-index: 10;
+            font-family: monospace;
+            word-break: break-all;
+        }
+
+        .video-info .info-row {
+            margin: 4px 0;
+        }
+
+        .video-info .info-label {
+            color: #888;
+            margin-right: 8px;
+        }
+
+        .video-info .info-value {
+            color: #0f0;
+        }
+
         #video-container {
             width: 100%;
             height: 100%;
@@ -196,9 +253,28 @@
         </div>
     </div>
     <div class="video-area">
+        <!-- 视频控制栏 -->
+        <div class="video-controls" id="video-controls" style="display: none;">
+            <label>码流类型：</label>
+            <select id="stream-type">
+                <option value="0">主码流（高清）</option>
+                <option value="1">子码流（流畅）</option>
+            </select>
+        </div>
         <div id="video-container">
             <div class="placeholder">👈 从左侧选择一个直播源</div>
         </div>
+        <!-- 视频信息栏 -->
+        <div class="video-info" id="video-info" style="display: none;">
+            <div class="info-row">
+                <span class="info-label">摄像头ID：</span>
+                <span class="info-value" id="info-camera-id">-</span>
+            </div>
+            <div class="info-row">
+                <span class="info-label">播放地址：</span>
+                <span class="info-value" id="info-url">-</span>
+            </div>
+        </div>
     </div>
 </div>
 
@@ -209,6 +285,7 @@
     // 全局状态
     let currentVideoNode = null;       // 当前播放的节点对象
     let hls = null;                    // HLS 实例
+    let currentStreamType = 0;         // 当前码流类型
 
     // 缓存已加载的子节点数据: { parentId: [childrenNodes] }
     const childrenCache = new Map();
@@ -243,13 +320,12 @@
         }
     }
 
-    // 获取节点详情（这里主要用来获取视频流地址，但也可以直接用子节点数据，不过为符合 API 设计，单独调用 stream 接口）
-    async function fetchStreamUrl(nodeId) {
+    // 获取视频流地址
+    async function fetchStreamUrl(nodeId, streamType = 0) {
         try {
-            const res = await fetch(`${API_BASE}/stream/${nodeId}`);
+            const res = await fetch(`${API_BASE}/stream/${nodeId}?stream_type=${streamType}`);
             if (!res.ok) throw new Error(`HTTP ${res.status}`);
-            const data = await res.json();
-            return data.url;
+            return await res.json();
         } catch (err) {
             console.error(`获取节点 ${nodeId} 的视频地址失败:`, err);
             return null;
@@ -257,20 +333,26 @@
     }
 
     // 播放视频
-    async function playVideo(node) {
+    async function playVideo(node, streamType = 0) {
         if (!node || !node.is_leaf) return;
 
         // 显示加载占位
         const container = document.getElementById('video-container');
         container.innerHTML = '<div class="placeholder">📡 正在加载直播流...</div>';
+        
+        // 隐藏信息和控制栏
+        document.getElementById('video-controls').style.display = 'none';
+        document.getElementById('video-info').style.display = 'none';
 
         // 获取流地址
-        const streamUrl = await fetchStreamUrl(node.id);
-        if (!streamUrl) {
+        const streamData = await fetchStreamUrl(node.id, streamType);
+        if (!streamData || !streamData.url) {
             container.innerHTML = '<div class="placeholder">❌ 无法获取视频地址，请稍后重试</div>';
             return;
         }
 
+        const streamUrl = streamData.url;
+
         // 清理旧播放器
         if (hls) {
             hls.destroy();
@@ -313,6 +395,14 @@
             });
         }
 
+        // 更新并显示视频信息
+        document.getElementById('info-camera-id').textContent = streamData.cameraIndexCode;
+        document.getElementById('info-url').textContent = streamUrl;
+        document.getElementById('video-info').style.display = 'block';
+        
+        // 显示控制栏
+        document.getElementById('video-controls').style.display = 'flex';
+        
         currentVideoNode = node;
     }
 
@@ -388,7 +478,7 @@
             // 高亮当前选中的节点
             clearActiveHighlight();
             contentDiv.classList.add('active');
-            await playVideo(node);
+            await playVideo(node, currentStreamType);
             return;
         }
 
@@ -442,7 +532,14 @@
 
     // 初始化视频区域（可选：尝试自动播放需用户交互）
     function initVideoArea() {
-        // 预留
+        // 监听码流类型切换
+        document.getElementById('stream-type').addEventListener('change', async (e) => {
+            currentStreamType = parseInt(e.target.value);
+            // 如果当前有播放的视频，重新加载
+            if (currentVideoNode) {
+                await playVideo(currentVideoNode, currentStreamType);
+            }
+        });
     }
 
     // 启动

live_catalog/live_catalog_backend.py

@@ -3,49 +3,89 @@ import urllib.parse
 import socket
 import json
 from http.server import ThreadingHTTPServer, SimpleHTTPRequestHandler
+import sys
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+from utils.hikvision_cam_utils import get_organization_list, get_final_list, get_camera_preview_url
 
-# ========== 硬编码的树形数据 ==========
-# 节点结构:
-# {
-#   "id": 1,
-#   "name": "河北省",
-#   "parent_id": None,    # None 表示根节点
-#   "is_leaf": False,
-#   "stream_url": None    # 叶子节点才会有值
-# }
+# ========== 海康威视 API 配置 ==========
+ROOT_PARENT_INDEX_CODE = "4fa15af07b6b400f94af1e35d8235c30"
 
-nodes = {
-    1: {"id": 1, "name": "河北省", "parent_id": None, "is_leaf": False, "stream_url": None},
-    2: {"id": 2, "name": "河南省", "parent_id": None, "is_leaf": False, "stream_url": None},
-    3: {"id": 3, "name": "石家庄市", "parent_id": 1, "is_leaf": False, "stream_url": None},
-    4: {"id": 4, "name": "保定市", "parent_id": 1, "is_leaf": False, "stream_url": None},
-    5: {"id": 5, "name": "郑州市", "parent_id": 2, "is_leaf": False, "stream_url": None},
-    6: {"id": 6, "name": "长安区", "parent_id": 3, "is_leaf": True,
-        "stream_url": "http://localhost:8355/stream.m3u8"},
-    7: {"id": 7, "name": "桥西区", "parent_id": 3, "is_leaf": True,
-        "stream_url": "https://example.com/live/qiaoxi.m3u8"},
-    8: {"id": 8, "name": "竞秀区", "parent_id": 4, "is_leaf": True,
-        "stream_url": "https://example.com/live/jingxiu.m3u8"},
-    9: {"id": 9, "name": "莲池区", "parent_id": 4, "is_leaf": True,
-        "stream_url": "https://example.com/live/lianchi.m3u8"},
-    10: {"id": 10, "name": "中原区", "parent_id": 5, "is_leaf": True,
-         "stream_url": "https://example.com/live/zhongyuan.m3u8"},
-}
+def transform_org_node(item, parent_id=None):
+    """将海康威视组织节点转换为前端期望的格式"""
+    return {
+        "id": item["indexCode"],
+        "name": item["name"],
+        "parent_id": parent_id or item.get("parentIndexCode"),
+        "is_leaf": False,  # 组织机构节点不是叶子节点
+        "stream_url": None
+    }
+
+def transform_camera_node(item, parent_id=None):
+    """将海康威视摄像头节点转换为前端期望的格式（叶子节点）"""
+    return {
+        "id": item["cameraIndexCode"],
+        "name": item["name"],
+        "parent_id": parent_id,
+        "is_leaf": True,  # 摄像头是叶子节点
+        "stream_url": None
+    }
 
 def get_children(parent_id):
-    """返回父节点下的直接子节点列表"""
-    return [node for node in nodes.values() if node["parent_id"] == parent_id]
+    """返回父节点下的直接子节点列表（从海康威视 API 获取）
+    
+    逻辑：
+    1. 先调用 get_organization_list 获取子组织
+    2. 如果返回 list 为空，则调用 get_final_list 获取摄像头（叶子节点）
+    """
+    if parent_id is None:
+        parent_id = ROOT_PARENT_INDEX_CODE
+    
+    try:
+        # 先尝试获取子组织
+        result = get_organization_list(parent_id)
+        if result.get("code") != "0":
+            print(f"海康威视 API 返回错误: {result.get('msg')}")
+            return []
+        
+        items = result.get("data", {}).get("list", [])
+        
+        # 如果有子组织，返回组织节点
+        if items:
+            return [transform_org_node(item, parent_id) for item in items]
+        
+        # 如果没有子组织，尝试获取摄像头列表（叶子节点）
+        print(f"组织 {parent_id} 无下级组织，尝试获取摄像头列表...")
+        final_result = get_final_list(parent_id)
+        if final_result.get("code") != "0":
+            print(f"获取摄像头列表失败: {final_result.get('msg')}")
+            return []
+        
+        camera_items = final_result.get("data", {}).get("list", [])
+        print(f"获取到 {len(camera_items)} 个摄像头")
+        return [transform_camera_node(item, parent_id) for item in camera_items]
+        
+    except Exception as e:
+        print(f"调用海康威视 API 失败: {e}")
+        return []
 
-def get_node(node_id):
-    """根据 id 获取节点详情"""
-    return nodes.get(node_id)
-
-def get_stream_url(node_id):
-    """获取叶子节点的视频流地址"""
-    node = nodes.get(node_id)
-    if node and node["is_leaf"]:
-        return node["stream_url"]
-    return None
+def get_stream_url(node_id, stream_type=0):
+    """获取摄像头的视频流地址
+    
+    Args:
+        node_id: 摄像头的 cameraIndexCode
+        stream_type: 码流类型，0=主码流，1=子码流
+    """
+    try:
+        result = get_camera_preview_url(node_id, stream_type)
+        if result.get("code") != "0":
+            print(f"获取视频流地址失败: {result.get('msg')}")
+            return None
+        
+        url = result.get("data", {}).get("url")
+        return url
+    except Exception as e:
+        print(f"调用 get_camera_preview_url 失败: {e}")
+        return None
 
 # ========== HTTP 处理器 ==========
 class APIHandler(SimpleHTTPRequestHandler):
@@ -99,42 +139,36 @@ class APIHandler(SimpleHTTPRequestHandler):
                 return
 
             elif path.startswith('/api/children/'):
-                # GET /api/children/3
-                try:
-                    node_id = int(path.split('/')[-1])
-                except (ValueError, IndexError):
+                # GET /api/children/21020000
+                node_id = path.split('/')[-1]
+                if not node_id:
                     self.send_error_json("Invalid node id", 400)
                     return
                 children = get_children(node_id)
                 self.send_json_response(children)
                 return
 
-            elif path.startswith('/api/node/'):
-                # GET /api/node/3
-                try:
-                    node_id = int(path.split('/')[-1])
-                except (ValueError, IndexError):
-                    self.send_error_json("Invalid node id", 400)
-                    return
-                node = get_node(node_id)
-                if node is None:
-                    self.send_error_json("Node not found", 404)
-                    return
-                self.send_json_response(node)
-                return
-
             elif path.startswith('/api/stream/'):
-                # GET /api/stream/6
-                try:
-                    node_id = int(path.split('/')[-1])
-                except (ValueError, IndexError):
+                # GET /api/stream/21020000?stream_type=0
+                node_id = path.split('/')[-1]
+                if not node_id:
                     self.send_error_json("Invalid node id", 400)
                     return
-                url = get_stream_url(node_id)
+                
+                # 解析 stream_type 参数
+                params = urllib.parse.parse_qs(query)
+                stream_type = int(params.get('stream_type', ['0'])[0])
+                
+                url = get_stream_url(node_id, stream_type)
                 if url is None:
                     self.send_error_json("Stream not found or node is not a leaf", 404)
                     return
-                self.send_json_response({"url": url})
+                # 返回完整信息
+                self.send_json_response({
+                    "cameraIndexCode": node_id,
+                    "url": url,
+                    "stream_type": stream_type
+                })
                 return
 
             # 静态文件服务（与原逻辑一致）
@@ -221,8 +255,7 @@ def run():
     print(f'Server running on http://localhost:{port}')
     print('API endpoints:')
     print('  GET /api/roots                - 获取所有根节点')
-    print('  GET /api/children/<id>        - 获取指定节点的子节点')
-    print('  GET /api/node/<id>            - 获取节点详情')
+    print('  GET /api/children/<id>        - 获取指定节点的子节点（自动判断组织/摄像头）')
     print('  GET /api/stream/<id>          - 获取视频流地址')
     print('静态文件服务: 访问 / 或 /index.html')
     print('按 Ctrl+C 停止服务器')

web_page_2/coordinate.html

@@ -0,0 +1,416 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+  <meta charset="UTF-8" />
+  <title>坐标提取工具</title>
+  <style>
+    * { margin: 0; padding: 0; box-sizing: border-box; }
+    html, body { width: 100%; height: 100%; overflow: hidden; background: #111827; color: #e5e7eb; }
+    body { font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; display: flex; }
+
+    .left-panel {
+      width: 320px; min-width: 320px; background: #020617; border-right: 1px solid #1f2937;
+      display: flex; flex-direction: column; height: 100vh;
+    }
+    .left-header {
+      padding: 12px 16px; border-bottom: 1px solid #1f2937; font-size: 14px; color: #9ca3af;
+      font-weight: 500; display: flex; align-items: center; justify-content: space-between; flex-shrink: 0;
+    }
+    .coord-list {
+      flex: 1; overflow-y: auto; padding: 8px;
+    }
+    .coord-group {
+      margin-bottom: 10px; background: #0f172a; border: 1px solid #1f2937; border-radius: 6px;
+      overflow: hidden;
+    }
+    .coord-group-header {
+      display: flex; align-items: center; justify-content: space-between;
+      padding: 8px 12px; border-bottom: 1px solid #1f2937; font-size: 13px;
+    }
+    .coord-group-color {
+      display: inline-block; width: 10px; height: 10px; border-radius: 50%; margin-right: 8px;
+    }
+    .coord-group-actions { display: flex; gap: 6px; }
+    .coord-group-actions button {
+      background: none; border: none; cursor: pointer; font-size: 13px; padding: 2px 4px;
+      border-radius: 3px; transition: background 0.15s;
+    }
+    .coord-group-actions .copy-btn { color: #60a5fa; }
+    .coord-group-actions .copy-btn:hover { background: rgba(96,165,250,0.15); }
+    .coord-group-actions .del-btn { color: #f87171; }
+    .coord-group-actions .del-btn:hover { background: rgba(248,113,113,0.15); }
+    .coord-group pre {
+      margin: 0; padding: 8px 12px; font-family: 'Courier New', monospace; font-size: 12px;
+      color: #d1d5db; white-space: pre; overflow-x: auto; line-height: 1.6;
+    }
+
+    .main-area {
+      flex: 1; display: flex; flex-direction: column; min-width: 0;
+    }
+    .toolbar {
+      padding: 10px 16px; background: #0f172a; border-bottom: 1px solid #1f2937;
+      display: flex; align-items: center; gap: 12px; flex-shrink: 0;
+    }
+    .upload-btn {
+      padding: 6px 16px; background: #3b82f6; color: #fff; border: none; border-radius: 4px;
+      font-size: 13px; cursor: pointer; transition: background 0.2s;
+    }
+    .upload-btn:hover { background: #2563eb; }
+    .toolbar-info { font-size: 12px; color: #9ca3af; }
+    .toolbar-hint {
+      margin-left: auto; font-size: 12px; color: #6b7280;
+    }
+    .toolbar-hint kbd {
+      padding: 1px 5px; background: #1f2937; border: 1px solid #374151;
+      border-radius: 3px; font-size: 11px; font-family: inherit;
+    }
+
+    .canvas-area {
+      flex: 1; position: relative; overflow: hidden; background: #000;
+      display: flex; align-items: center; justify-content: center;
+    }
+    #imageCanvas { cursor: crosshair; }
+    .upload-placeholder {
+      position: absolute; display: flex; flex-direction: column; align-items: center;
+      gap: 12px; color: #6b7280; font-size: 14px;
+    }
+    .upload-placeholder svg { opacity: 0.3; }
+
+    .toast {
+      position: fixed; top: 20px; right: 20px; padding: 10px 18px;
+      background: #10b981; color: #fff; border-radius: 6px; font-size: 13px;
+      opacity: 0; transition: opacity 0.3s; pointer-events: none; z-index: 999;
+    }
+    .toast.show { opacity: 1; }
+
+    #fileInput { display: none; }
+  </style>
+</head>
+<body>
+  <aside class="left-panel">
+    <div class="left-header">
+      <span>坐标数据</span>
+      <button id="copyAllBtn" class="copy-btn" style="background:none;border:none;color:#60a5fa;cursor:pointer;font-size:12px;">复制全部</button>
+    </div>
+    <div id="coordList" class="coord-list"></div>
+  </aside>
+
+  <div class="main-area">
+    <div class="toolbar">
+      <button class="upload-btn" id="uploadBtn">上传图片</button>
+      <input type="file" id="fileInput" accept=".jpg,.jpeg,.png" />
+      <span class="toolbar-info" id="imageInfo"></span>
+      <span class="toolbar-hint">
+        <kbd>点击</kbd> 标记点 &nbsp;
+        <kbd>Backspace</kbd> 撤销 &nbsp;
+        <kbd>Enter</kbd> 完成当前组
+      </span>
+    </div>
+    <div class="canvas-area" id="canvasArea">
+      <canvas id="imageCanvas"></canvas>
+      <div class="upload-placeholder" id="placeholder">
+        <svg width="64" height="64" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5">
+          <rect x="3" y="3" width="18" height="18" rx="2" ry="2"/>
+          <circle cx="8.5" cy="8.5" r="1.5"/>
+          <polyline points="21 15 16 10 5 21"/>
+        </svg>
+        <span>点击「上传图片」开始</span>
+      </div>
+    </div>
+  </div>
+
+  <div class="toast" id="toast"></div>
+
+  <script>
+    const GROUP_COLORS = [
+      '#ef4444', '#3b82f6', '#10b981', '#f59e0b', '#8b5cf6',
+      '#ec4899', '#06b6d4', '#f97316', '#6366f1', '#14b8a6'
+    ];
+
+    let groups = [];        // [{points: [[x,y], ...]}, ...]
+    let currentPoints = []; // current group being edited
+    let image = null;       // HTMLImageElement
+    let canvasScale = 1;
+    let canvasOffsetX = 0;
+    let canvasOffsetY = 0;
+
+    const canvas = document.getElementById('imageCanvas');
+    const ctx = canvas.getContext('2d');
+    const canvasArea = document.getElementById('canvasArea');
+    const placeholder = document.getElementById('placeholder');
+    const fileInput = document.getElementById('fileInput');
+    const imageInfo = document.getElementById('imageInfo');
+    const coordList = document.getElementById('coordList');
+    const toast = document.getElementById('toast');
+
+    // --- Upload ---
+    document.getElementById('uploadBtn').addEventListener('click', () => fileInput.click());
+    fileInput.addEventListener('change', (e) => {
+      const file = e.target.files[0];
+      if (!file) return;
+      const ext = file.name.toLowerCase().split('.').pop();
+      if (!['jpg', 'jpeg', 'png'].includes(ext)) {
+        showToast('仅支持 JPG/PNG 格式', true);
+        return;
+      }
+      const reader = new FileReader();
+      reader.onload = (ev) => {
+        image = new Image();
+        image.onload = () => {
+          imageInfo.textContent = `${image.width} x ${image.height}`;
+          placeholder.style.display = 'none';
+          fitCanvas();
+          redraw();
+        };
+        image.src = ev.target.result;
+      };
+      reader.readAsDataURL(file);
+      // Reset groups on new image
+      groups = [];
+      currentPoints = [];
+      renderCoordList();
+    });
+
+    // --- Canvas sizing ---
+    function fitCanvas() {
+      if (!image) return;
+      const areaW = canvasArea.clientWidth;
+      const areaH = canvasArea.clientHeight;
+      const imgAspect = image.width / image.height;
+      const areaAspect = areaW / areaH;
+
+      let drawW, drawH;
+      if (imgAspect > areaAspect) {
+        drawW = areaW;
+        drawH = areaW / imgAspect;
+      } else {
+        drawH = areaH;
+        drawW = areaH * imgAspect;
+      }
+      canvasScale = drawW / image.width;
+      canvasOffsetX = (areaW - drawW) / 2;
+      canvasOffsetY = (areaH - drawH) / 2;
+
+      canvas.width = areaW;
+      canvas.height = areaH;
+      canvas.style.width = areaW + 'px';
+      canvas.style.height = areaH + 'px';
+    }
+
+    window.addEventListener('resize', () => { fitCanvas(); redraw(); });
+
+    // --- Drawing ---
+    function redraw() {
+      ctx.clearRect(0, 0, canvas.width, canvas.height);
+      if (!image) return;
+
+      // Draw image
+      ctx.drawImage(image, canvasOffsetX, canvasOffsetY, image.width * canvasScale, image.height * canvasScale);
+
+      // Draw completed groups
+      groups.forEach((group, gi) => {
+        const color = GROUP_COLORS[gi % GROUP_COLORS.length];
+        drawGroup(group.points, color);
+      });
+
+      // Draw current group
+      if (currentPoints.length > 0) {
+        const color = GROUP_COLORS[groups.length % GROUP_COLORS.length];
+        drawGroup(currentPoints, color, true);
+      }
+    }
+
+    function drawGroup(points, color, isCurrent) {
+      if (points.length === 0) return;
+
+      // Draw lines
+      if (points.length > 1) {
+        ctx.beginPath();
+        ctx.strokeStyle = color;
+        ctx.lineWidth = 2;
+        const p0 = toCanvas(points[0]);
+        ctx.moveTo(p0.x, p0.y);
+        for (let i = 1; i < points.length; i++) {
+          const p = toCanvas(points[i]);
+          ctx.lineTo(p.x, p.y);
+        }
+        ctx.stroke();
+      }
+
+      // Draw points
+      points.forEach((pt, i) => {
+        const p = toCanvas(pt);
+        ctx.beginPath();
+        ctx.arc(p.x, p.y, 5, 0, Math.PI * 2);
+        ctx.fillStyle = color;
+        ctx.fill();
+        ctx.strokeStyle = '#fff';
+        ctx.lineWidth = 1.5;
+        ctx.stroke();
+
+        // Label
+        ctx.fillStyle = '#fff';
+        ctx.font = '11px sans-serif';
+        ctx.fillText((i + 1).toString(), p.x + 7, p.y - 5);
+      });
+
+      // "editing" indicator for current group
+      if (isCurrent) {
+        const last = toCanvas(points[points.length - 1]);
+        ctx.beginPath();
+        ctx.arc(last.x, last.y, 10, 0, Math.PI * 2);
+        ctx.strokeStyle = color;
+        ctx.lineWidth = 1.5;
+        ctx.setLineDash([3, 3]);
+        ctx.stroke();
+        ctx.setLineDash([]);
+      }
+    }
+
+    function toCanvas(pt) {
+      return {
+        x: pt[0] * image.width * canvasScale + canvasOffsetX,
+        y: pt[1] * image.height * canvasScale + canvasOffsetY
+      };
+    }
+
+    function toNorm(cx, cy) {
+      return [
+        Math.round(((cx - canvasOffsetX) / (image.width * canvasScale)) * 1000) / 1000,
+        Math.round(((cy - canvasOffsetY) / (image.height * canvasScale)) * 1000) / 1000
+      ];
+    }
+
+    // --- Click ---
+    canvas.addEventListener('click', (e) => {
+      if (!image) return;
+      const rect = canvas.getBoundingClientRect();
+      const cx = e.clientX - rect.left;
+      const cy = e.clientY - rect.top;
+
+      // Check within image bounds
+      const nx = (cx - canvasOffsetX) / (image.width * canvasScale);
+      const ny = (cy - canvasOffsetY) / (image.height * canvasScale);
+      if (nx < 0 || nx > 1 || ny < 0 || ny > 1) return;
+
+      const pt = toNorm(cx, cy);
+      currentPoints.push(pt);
+      redraw();
+      renderCoordList();
+    });
+
+    // --- Keyboard ---
+    document.addEventListener('keydown', (e) => {
+      if (e.key === 'Enter') {
+        if (currentPoints.length === 0) return;
+        groups.push({ points: [...currentPoints] });
+        currentPoints = [];
+        redraw();
+        renderCoordList();
+        showToast(`第 ${groups.length} 组坐标已完成`);
+      } else if (e.key === 'Backspace') {
+        if (currentPoints.length > 0) {
+          currentPoints.pop();
+          redraw();
+          renderCoordList();
+        }
+      }
+    });
+
+    // --- Coord list ---
+    function renderCoordList() {
+      coordList.innerHTML = '';
+      if (groups.length === 0 && currentPoints.length === 0) {
+        coordList.innerHTML = '<div style="padding:16px;color:#4b5563;font-size:13px;text-align:center;">暂无坐标数据</div>';
+        return;
+      }
+
+      groups.forEach((group, gi) => {
+        const color = GROUP_COLORS[gi % GROUP_COLORS.length];
+        const div = document.createElement('div');
+        div.className = 'coord-group';
+        div.innerHTML = `
+          <div class="coord-group-header">
+            <span><span class="coord-group-color" style="background:${color}"></span>第 ${gi + 1} 组 (${group.points.length} 点)</span>
+            <div class="coord-group-actions">
+              <button class="copy-btn" data-group="${gi}" title="复制">复制</button>
+              <button class="del-btn" data-group="${gi}" title="删除">删除</button>
+            </div>
+          </div>
+          <pre>${formatYAML(group.points)}</pre>
+        `;
+        coordList.appendChild(div);
+      });
+
+      // Current editing group
+      if (currentPoints.length > 0) {
+        const gi = groups.length;
+        const color = GROUP_COLORS[gi % GROUP_COLORS.length];
+        const div = document.createElement('div');
+        div.className = 'coord-group';
+        div.style.borderColor = color;
+        div.innerHTML = `
+          <div class="coord-group-header">
+            <span><span class="coord-group-color" style="background:${color}"></span>第 ${gi + 1} 组 (编辑中, ${currentPoints.length} 点)</span>
+          </div>
+          <pre>${formatYAML(currentPoints)}</pre>
+        `;
+        coordList.appendChild(div);
+      }
+
+      // Bind copy/delete
+      coordList.querySelectorAll('.copy-btn[data-group]').forEach(btn => {
+        btn.addEventListener('click', () => {
+          const gi = parseInt(btn.dataset.group);
+          copyText(formatYAML(groups[gi].points));
+          showToast('已复制到剪贴板');
+        });
+      });
+      coordList.querySelectorAll('.del-btn[data-group]').forEach(btn => {
+        btn.addEventListener('click', () => {
+          const gi = parseInt(btn.dataset.group);
+          groups.splice(gi, 1);
+          redraw();
+          renderCoordList();
+        });
+      });
+    }
+
+    function formatYAML(points) {
+      return points.map(pt => `- [${pt[0]}, ${pt[1]}]`).join('\n');
+    }
+
+    // --- Copy all ---
+    document.getElementById('copyAllBtn').addEventListener('click', () => {
+      if (groups.length === 0) { showToast('暂无数据', true); return; }
+      const all = groups.map((g, i) => `# 第 ${i + 1} 组\n${formatYAML(g.points)}`).join('\n\n');
+      copyText(all);
+      showToast('已复制全部坐标');
+    });
+
+    // --- Utils ---
+    function copyText(text) {
+      navigator.clipboard.writeText(text).catch(() => {
+        const ta = document.createElement('textarea');
+        ta.value = text;
+        document.body.appendChild(ta);
+        ta.select();
+        document.execCommand('copy');
+        document.body.removeChild(ta);
+      });
+    }
+
+    let toastTimer = null;
+    function showToast(msg, isError) {
+      toast.textContent = msg;
+      toast.style.background = isError ? '#ef4444' : '#10b981';
+      toast.classList.add('show');
+      clearTimeout(toastTimer);
+      toastTimer = setTimeout(() => toast.classList.remove('show'), 1800);
+    }
+
+    // Initial render
+    renderCoordList();
+  </script>
+</body>
+</html>

web_page_2/http_server.py

@@ -48,6 +48,8 @@ class APIHandler(SimpleHTTPRequestHandler):
             elif path == '/' or path == '/index.html':
                 # 默认访问使用 api=1
                 self.serve_file('index.html', query='api=1')
+            elif path == '/coords' or path == '/coordinate.html':
+                self.serve_file('coordinate.html')
             else:
                 # 处理静态文件请求
                 # 移除开头的 /
@@ -130,6 +132,7 @@ def run():
     httpd = ThreadingHTTPServer(server_address, APIHandler)
     print(f'Server running on http://localhost:{port}')
     print(f'支持的接口: /, /api/1, /api/2, /api/3, /api/4, /api/5, /api/6, /api/7, /api/11-16')
+    print(f'坐标提取工具: /coords')
     print('按 Ctrl+C 停止服务器')
     
     try: