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 # -------------------------- 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 yolox.tracker.byte_tracker import BYTETracker from utils.logger import get_logger from common.constants import MODEL_ROOT_PATH logger = get_logger(__name__) DETECT_MODEL_PATH = 'YOLO_Weight/Kadian.onnx' # 默认相对ROI（与原文件一致） #ROI_RELATIVE = np.array([ # [0.10989583333333333, 0.006481481481481481], # [0.421875, 0.005555555555555556], # [0.9921875, 0.9888888888888889], # [0.3411458333333333, 0.9861111111111112] #]) # ROI_RELATIVE=np.array([ # [0.12,0.0], # [0.3,0.0], # [0.5,0.2], # [1.0, 0.95], # [1.0,1.0], # [0.42,1.0] # ]) ROI_RELATIVE=np.array([ [0.15,0.15], [0.37,0.15], [0.55,0.2], [0.9,0.85], [0.35,0.85] ]) ALERT_PUSH_INTERVAL = 5.0 # 输入尺寸 PERSON_CAR_INPUT_SIZE = 640 #POSE_INPUT_SIZE = 640 RTSP_TARGET_FPS = 10.0 # ========================= Kadian TrafficMonitor（精简版，专为服务设计） ========================= class KadianDetector: def __init__(self, params=None): # 摄像头额外参数 self.params = params if params is not None else {} # 模型路径：从 params 读取，未配置则使用默认值 DETECT_MODEL_PATH model_path = self.params.get('model_path') if model_path: full_model_path = f"{MODEL_ROOT_PATH}/{model_path}" else: full_model_path = DETECT_MODEL_PATH logger.info(f"Loading model from: {full_model_path}") # 模型加载 self.detector = YOLOv8_ONNX( full_model_path, conf_threshold=0.6, iou_threshold=0.65, input_size=PERSON_CAR_INPUT_SIZE ) # 跟踪器配置 class TrackerArgs: track_thresh = 0.61 # 必须大于等于yolo的conf_threshold track_buffer = 40 match_thresh = 0.85 mot20 = True self.fps = RTSP_TARGET_FPS self.tracker = BYTETracker(TrackerArgs(), frame_rate=self.fps) self.track_role = {} # 跟踪ID到类别的映射 # ROI 处理：优先从 params 获取，否则使用默认值 ROI_RELATIVE roi_points = self.params.get('roi_points', ROI_RELATIVE) self.roi_points = np.array(roi_points, dtype=np.float64) if roi_points is not None else None # ===================== 超参数设置 (仅保留车/后备箱相关) ===================== # 后备箱检查判定阈值 self.TIME_THRESHOLD_TRUNK_OPEN = 0.3 # 车辆最小停留时间阈值 (小于此时间视为无人检查/直接通过) self.TIME_THRESHOLD_CAR_MIN_DURATION = 3.0 # Car 丢帧/ID维持缓冲 self.TIME_TOLERANCE_CAR = 2.0 # police丢失阈值 self.TIME_TOLERANCE_POLICE = 3.0 # police状态判定阈值 (累计秒数) self.TIME_THRESHOLD_NOBODY = 10.0 self.TIME_THRESHOLD_ONLY_ONE = 10.0 # --- 计算对应的帧数阈值 --- self.frame_thresh_trunk_valid = int(self.TIME_THRESHOLD_TRUNK_OPEN * self.fps) self.frame_thresh_car_min_duration = int(self.TIME_THRESHOLD_CAR_MIN_DURATION * self.fps) self.frame_buffer_limit_car = int(self.TIME_TOLERANCE_CAR * self.fps) self.frame_buffer_limit_police = int(self.TIME_TOLERANCE_POLICE * self.fps) self.frame_thresh_nobody = int(self.TIME_THRESHOLD_NOBODY * self.fps) self.frame_thresh_only_one = int(self.TIME_THRESHOLD_ONLY_ONE * self.fps) # 显示相关阈值 self.ignore_show_seconds = 0.2 # 未检测的警告显示时长 self.openTrunk_show_seconds = 0.2 # 打开后备箱的警告显示时长 self.police_show_seconds = 0.2 # 警察在场警告显示时长 # 状态变量初始化 self.current_frame_idx = 0 self.width = 0 self.height = 0 # 车辆注册表 (字典) self.roi_car_registry = {} # 违规车辆记录 self.unchecked_trunk_alerts = {} # 后备箱未检 self.fast_pass_alerts = {} # 通过过快 # 警察注册表 (字典) self.roi_police_registry = {} # 警察在场告警记录 self.nobody_alerts = {} # 无人在场 self.only_one_alerts = {} # 单人在场 # 累计帧数计数器 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((3.0 + self.TIME_TOLERANCE_CAR) * self.fps) # 50帧 self.frame_buffer_ignore_untrunk = deque(maxlen=self.max_car_frames) self.untrunk_rollback_time = 3.0 # 未检查后备箱需要回溯的时间 self.ignored_rollback_time = 1.0 # 漏检需要回溯的时间 def _get_roi_points(self, frame_width: int, frame_height: int): """ 每帧动态计算正确的 ROI 绝对坐标，并确保类型为 np.int32 用于 pointPolygonTest 和 polylines """ if self.roi_points is None: raise ValueError("ROI points must be provided; cannot be None.") if self.roi_points.max() <= 1.0: # 相对坐标 → 转换为绝对 roi_abs = self.roi_points * np.array([frame_width, frame_height]) else: # 绝对坐标，直接使用 roi_abs = self.roi_points.copy() # 强制转为 int32（关键！解决 OpenCV 断言错误） return roi_abs.astype(np.int32) def check_point_in_roi(self, roi_points, point): """判断点是否在ROI内""" return cv2.pointPolygonTest(roi_points, point, False) >= 0 def compute_iou(self, boxA, boxB): """计算两个框的IOU""" # box = [x1, y1, x2, y2] xA = max(boxA[0], boxB[0]) yA = max(boxA[1], boxB[1]) xB = min(boxA[2], boxB[2]) yB = min(boxA[3], boxB[3]) interW = max(0, xB - xA) interH = max(0, yB - yA) interArea = interW * interH boxAArea = (boxA[2] - boxA[0]) * (boxA[3] - boxA[1]) boxBArea = (boxB[2] - boxB[0]) * (boxB[3] - boxB[1]) unionArea = boxAArea + boxBArea - interArea if unionArea == 0: return 0.0 return interArea / unionArea def draw_alert(self, frame, text, color=(0, 0, 255), sub_text=None, offset_y=0): """在右上角绘制警告文字 (支持垂直偏移，防止文字重叠)""" font_scale = 1.5 thickness = 3 font = cv2.FONT_HERSHEY_SIMPLEX (text_w, text_h), _ = cv2.getTextSize(text, font, font_scale, thickness) x = self.width - text_w - 20 y = 50 + text_h + offset_y # 增加 Y 轴偏移 cv2.rectangle(frame, (x - 10, y - text_h - 10), (x + text_w + 10, y + 10), (0, 0, 0), -1) cv2.putText(frame, text, (x, y), font, font_scale, color, thickness) if sub_text: cv2.putText(frame, sub_text, (x, y + 40), font, 0.7, (200, 200, 200), 2) def is_point_in_box(self, point, box): """判断点是否在框内""" px, py = point 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['frame_idx'] - 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 self.current_frame_idx += 1 # ========= 每帧动态获取正确的 ROI（int32）========= roi_points_int32 = self._get_roi_points(w, h) # shape: (4, 2), dtype: int32 roi_points_draw = roi_points_int32.reshape((-1, 1, 2)) # shape: (4, 1, 2) 用于绘制 current_time_sec = timestamp # print('current_time_sec:', current_time_sec) # ========= 主检测（删除pose检测）========= detections = self.detector(frame) dets_xyxy = [] dets_roles = [] dets_for_tracker = [] # ========= 当前帧所有警告列表 ========== current_frame_alerts = [] # 每帧清空，重新收集 if detections: for det in detections: x1, y1, x2, y2, conf, cls_id = det # x1,y1:左上角，x2,y2:右下角 dets_xyxy.append([x1, y1, x2, y2]) dets_for_tracker.append([x1, y1, x2, y2, conf]) # 更新类别映射：0=Car,1=OpenTrunk,2=Passerby,3=Police if cls_id == 0: dets_roles.append("car") elif cls_id == 1: dets_roles.append("opentrunk") elif cls_id == 2: dets_roles.append("passerby") # 路人 elif cls_id == 3: dets_roles.append("police") # 警察 dets = np.array(dets_for_tracker, dtype=np.float32) if len(dets_for_tracker) else np.empty((0, 5)) # 跟踪器更新 tracks = self.tracker.update( dets, [self.height, self.width], [self.height, self.width] ) # logger.debug("tracks: {}".format(tracks)) # ========= 绘制 ROI ========= cv2.polylines(frame, [roi_points_draw], isClosed=True, color=(255, 0, 0), thickness=3) # ========= 单帧统计变量 ========= current_roi_trunk_count = 0 # 仅保留后备箱统计 current_roi_police_count = 0 # ROI内警察数量 # 临时存储本帧的目标，用于后续关联分析 current_cars = [] # {'id':, 'box':} current_trunks = [] # (cx, cy) # ========= 处理跟踪结果 ========= for t in tracks: tid = t.track_id REVALIDATE_FRAME_INTERVAL = 2 # 定期重新匹配跟踪ID的类别 if (self.current_frame_idx % REVALIDATE_FRAME_INTERVAL == 0) or (tid not in self.track_role): best_iou = 0 best_role = "unknown" t_box = list(map(float, t.tlbr)) # [x1,y1,x2,y2] for i, box in enumerate(dets_xyxy): iou_val = self.compute_iou(t_box, box) if iou_val > best_iou: best_iou = iou_val best_role = dets_roles[i] if best_iou > 0.1: self.track_role[tid] = best_role else: self.track_role[tid] = "unknown" role = self.track_role.get(tid, "unknown") x1, y1, x2, y2 = map(int, t.tlbr) cx, cy = (x1 + x2) // 2, (y1 + y2) // 2 # 定义不同类别的颜色（仅标框，不告警） if role == "car": color = (0, 255, 0) # 绿色 label = f"Car:{tid}" # 仅处理ROI内的车辆 if self.check_point_in_roi(roi_points_int32, (cx, cy)): current_cars.append({'id': tid, 'box': [x1, y1, x2, y2]}) # 车辆注册表初始化 if tid not in self.roi_car_registry: self.roi_car_registry[tid] = { 'first_seen': self.current_frame_idx, 'last_seen': self.current_frame_idx, # 'first_seen': current_time_sec, # 'last_seen': current_time_sec, 'trunk_frames': 0, 'is_checked': False, #'frame_buffer': deque(maxlen=self.max_car_frames), # 新增 } else: self.roi_car_registry[tid]['last_seen'] = self.current_frame_idx #self.roi_car_registry[tid]['last_seen'] = current_time_sec label += " IN" elif role == "opentrunk": color = (255, 165, 0) # 橙色 label = "OpenTrunk" if self.check_point_in_roi(roi_points_int32, (cx, cy)): current_roi_trunk_count += 1 current_trunks.append((cx, cy)) label += " IN" elif role == "passerby": color = (255, 255, 0) # 黄色（仅标框，不告警） label = "Passerby" elif role == "police": color = (0, 255, 255) # 青色 label = "Police" if self.check_point_in_roi(roi_points_int32, (cx, cy)): current_roi_police_count += 1 # 警察注册表初始化 if tid not in self.roi_police_registry: self.roi_police_registry[tid] = { 'first_seen': self.current_frame_idx, 'last_seen': self.current_frame_idx, #'first_seen': current_time_sec, #'last_seen': current_time_sec, } else: #self.roi_police_registry[tid]['last_seen'] = current_time_sec self.roi_police_registry[tid]['last_seen'] = self.current_frame_idx label += " IN" else: color = (255, 255, 255) # 白色 label = "Unknown" # 绘制检测框和标签（所有类别都标框，仅车/后备箱有逻辑） cv2.rectangle(frame, (x1, y1), (x2, y2), color, 2) cv2.putText(frame, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 2) # 每帧保存到缓存 # self.roi_car_registry[tid]['frame_buffer'].append({ # 'frame_idx': current_time_sec, # 'frame': frame.copy(), # }) self.frame_buffer_ignore_untrunk.append({ #'frame_idx': current_time_sec, 'frame_idx': self.current_frame_idx, 'frame': frame.copy(), }) # ========================================== # 关联分析: 哪个后备箱属于哪辆车? # ========================================== for car_info in current_cars: c_id = car_info['id'] c_box = car_info['box'] trunk_found_for_this_car = False for t_pt in current_trunks: if self.is_point_in_box(t_pt, c_box): trunk_found_for_this_car = True break if trunk_found_for_this_car: self.roi_car_registry[c_id]['trunk_frames'] += 1 if self.roi_car_registry[c_id]['trunk_frames'] >= self.frame_thresh_trunk_valid: self.roi_car_registry[c_id]['is_checked'] = True # ========================================== # 维护车辆注册表 & 生成离场报警 # ========================================== active_car_ids = [] cars_to_remove = [] for car_id, info in self.roi_car_registry.items(): last_seen = info['last_seen'] if (self.current_frame_idx - last_seen) <= self.frame_buffer_limit_car: #if (current_time_sec - last_seen) <= self.frame_buffer_limit_car: active_car_ids.append(car_id) else: cars_to_remove.append(car_id) unchecked_trunk_frame = None # 保存未检查后备箱的帧 ignored_trunk_frame = None # 保存漏检的帧 # 处理离场车辆，生成违规告警 for car_id in cars_to_remove: car_info = self.roi_car_registry[car_id] duration_frames = car_info['last_seen'] - car_info['first_seen'] # 情况1：通过时间太短 -> Ignore (Too Fast) if duration_frames < self.frame_thresh_car_min_duration: logger.info(f"ALARM: Car {car_id} passed too fast -> Regarded as Ignore Checked!") self.fast_pass_alerts[car_id] = self.current_frame_idx + int(self.ignore_show_seconds * self.fps) #self.fast_pass_alerts[car_id] = current_time_sec + int(self.ignore_show_seconds * self.fps) target_time_sec = car_info['last_seen'] - self.ignored_rollback_time ignored_trunk_frame = self.find_target_frame(target_time_sec) # 情况2：时间够长，但没检查后备箱 -> Unchecked Trunk elif not car_info['is_checked']: logger.info(f"ALARM: Car {car_id} left without checking trunk!") self.unchecked_trunk_alerts[car_id] = self.current_frame_idx + int(self.openTrunk_show_seconds * self.fps) #self.unchecked_trunk_alerts[car_id] = current_time_sec + int(self.openTrunk_show_seconds * self.fps) target_time_sec = car_info['last_seen'] - self.untrunk_rollback_time unchecked_trunk_frame = self.find_target_frame(target_time_sec) del self.roi_car_registry[car_id] effective_car_count = len(active_car_ids) # ========================================== # 维护警察注册表 # ========================================== active_police_ids = [] polices_to_remove = [] for police_id, info in self.roi_police_registry.items(): last_seen = info['last_seen'] if (self.current_frame_idx - last_seen) <= self.frame_buffer_limit_police: #if (current_time_sec - last_seen) <= self.frame_buffer_limit_police: active_police_ids.append(police_id) else: polices_to_remove.append(police_id) for police_id in polices_to_remove: del self.roi_police_registry[police_id] effective_police_count = len(active_police_ids) # ========================================== # 显示调试信息和报警 (仅保留车/后备箱相关) # ========================================== # 调试信息 debug_info = f"Cars: {len(active_car_ids)} | Trunk: {current_roi_trunk_count} | Police: {effective_police_count} | Nobody:{self.nobody_frames}/{self.frame_thresh_nobody} | OnlyOne:{self.only_one_frames}/{self.frame_thresh_only_one}" cv2.putText(frame, debug_info, (20, 40), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 0), 2) # 报警偏移量（防止重叠） alert_offset = 0 # A. 显示 Trunk Checked (车辆已检查后备箱) # for car_id in active_car_ids: # if car_id in self.roi_car_registry and self.roi_car_registry[car_id]['is_checked']: # current_frame_alerts.append({ # 'time': current_time_sec, # 'action': "Trunk Checked", # }) # self.draw_alert(frame, "Trunk Checked!!", (0, 255, 0), offset_y=alert_offset) # alert_offset += 100 # break # 只显示一次 # B. 显示 Unchecked Trunk (离场未检查后备箱) expired_alerts = [cid for cid, end_frame in self.unchecked_trunk_alerts.items() if self.current_frame_idx > end_frame] #expired_alerts = [cid for cid, end_frame in self.unchecked_trunk_alerts.items() if current_time_sec > end_frame] for cid in expired_alerts: del self.unchecked_trunk_alerts[cid] if len(self.unchecked_trunk_alerts) > 0: alert_text = f"Unchecked Trunk! (ID:{list(self.unchecked_trunk_alerts.keys())})" current_frame_alerts.append({ 'time': current_time_sec, 'action': "Unchecked Trunk", 'image': unchecked_trunk_frame }) #self.draw_alert(frame, alert_text, (0, 0, 255), offset_y=alert_offset) alert_offset += 100 # C. 显示 Ignore (通过过快) expired_fast_alerts = [cid for cid, end_frame in self.fast_pass_alerts.items() if self.current_frame_idx > end_frame] #expired_fast_alerts = [cid for cid, end_frame in self.fast_pass_alerts.items() if current_time_sec > end_frame] for cid in expired_fast_alerts: del self.fast_pass_alerts[cid] if len(self.fast_pass_alerts) > 0: alert_text = f"Ignore: (ID:{list(self.fast_pass_alerts.keys())})" current_frame_alerts.append({ 'time': current_time_sec, 'action': "Ignore", 'image': ignored_trunk_frame }) self.draw_alert(frame, alert_text, (0, 0, 255), offset_y=alert_offset) alert_offset += 100 # D. 显示警察在场状态 (Nobody/Only One) # 清理过期的 Nobody 告警 expired_nobody = [k for k, v in self.nobody_alerts.items() if self.current_frame_idx > v] #expired_nobody = [k for k, v in self.nobody_alerts.items() if current_time_sec > v] for k in expired_nobody: del self.nobody_alerts[k] # 清理过期的 Only One 告警 expired_only_one = [k for k, v in self.only_one_alerts.items() if self.current_frame_idx > v] # expired_only_one = [k for k, v in self.only_one_alerts.items() if current_time_sec > v] for k in expired_only_one: del self.only_one_alerts[k] if effective_car_count > 0: # 更新累计帧数 if effective_police_count == 0: self.nobody_frames += 1 self.only_one_frames = 0 elif effective_police_count == 1: self.only_one_frames += 1 self.nobody_frames = 0 else: self.nobody_frames = 0 self.only_one_frames = 0 else: self.nobody_frames = 0 self.only_one_frames = 0 if effective_police_count == 0 and self.nobody_frames >= self.frame_thresh_nobody: alert_text = "Nobody" if "Nobody" not in self.nobody_alerts: self.nobody_alerts["Nobody"] = self.current_frame_idx + int(self.police_show_seconds * self.fps) # self.nobody_alerts["Nobody"] = current_time_sec + int(self.police_show_seconds * self.fps) current_frame_alerts.append({ 'time': current_time_sec, 'action': "Nobody", }) # self.draw_alert(frame, alert_text, (0, 0, 255), offset_y=alert_offset) alert_offset += 100 elif effective_police_count == 1 and self.only_one_frames >= self.frame_thresh_only_one: alert_text = "Only One" if "Only One" not in self.only_one_alerts: self.only_one_alerts["Only One"] = self.current_frame_idx + int(self.police_show_seconds * self.fps) # self.only_one_alerts["Only One"] = current_time_sec + int(self.police_show_seconds * self.fps) current_frame_alerts.append({ 'time': current_time_sec, 'action': "Only One", }) # self.draw_alert(frame, alert_text, (255, 165, 0), offset_y=alert_offset) alert_offset += 100 return { "image": frame, "alerts": current_frame_alerts, } # ========================= 帧处理线程 ========================= class FrameProcessorWorker(BaseFrameProcessorWorker): """卡点检测帧处理线程""" # 子类配置 DETECTOR_FACTORY = lambda params: KadianDetector(params) POST_TYPE = 1 TARGET_FPS = RTSP_TARGET_FPS