import cv2 import asyncio import av import torch import json import time import os import sys import numpy as np from . import api from quart import jsonify, request,render_template, Response,websocket from aiortc import RTCPeerConnection, RTCSessionDescription, VideoStreamTrack from aiortc.contrib.media import MediaBlackhole, MediaPlayer, MediaRecorder from fractions import Fraction from shapely.geometry import Point, Polygon import threading import logging # 配置日志 logging.basicConfig(level=logging.INFO) logger = logging.getLogger(__name__) rtsp_url = 'rtsp://127.0.0.1/live1' #-------------------基于WEBRTC实现拉流 pcs = set() #创建一个空的集合，去重复且无序 ''' --------------基础信息--后续封装在函数里--------------- ''' #print(f"Current working directory (video.py): {os.getcwd()}") # 自定义模型文件的路径 model_path = 'model/mode_test/yolov5s.pt' # 假设模型文件名为 yolov5s.pt 并与执行文件在同一个目录 # 本地YOLOv5仓库路径 repo_path = 'model/base_model/yolov5' # 加载自定义模型 model = torch.hub.load(repo_path, 'custom', path=model_path, source='local') # 定义监控区域（多边形顶点） region_points = [(100, 100), (500, 100), (500, 400), (100, 400)] polygon = Polygon(region_points) # 打开摄像头 def get_camera(): cap = cv2.VideoCapture(0) if not cap.isOpened(): raise IOError("Cannot open webcam") return cap # cap = get_camera() ''' ---------------------传输-------------------------- ''' class VideoTransformTrack(VideoStreamTrack): kind = "video" def __init__(self,strUrl,itype=1): #0-usb,1-RTSP,2-海康SDK super().__init__() self.cap = cv2.VideoCapture(strUrl) if not self.cap.isOpened(): raise Exception("Unable to open video source") # Set desired resolution and frame rate # self.cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1280) # Set width to 1280 # self.cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 720) # Set height to 720 # self.cap.set(cv2.CAP_PROP_FPS, 30) # Set frame rate to 30 FPS self.frame_rate = int(self.cap.get(cv2.CAP_PROP_FPS)) or 15 # Default to 30 if unable to get FPS self.time_base = Fraction(1, self.frame_rate) async def recv(self): start_time = time.time() # 开始时间 ret, frame = self.cap.read() if not ret: raise Exception("Unable to read frame") # # Convert frame to RGB # frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) # # # Convert to VideoFrame # video_frame = av.VideoFrame.from_ndarray(frame_rgb, format="rgb24") # # Set timestamp # video_frame.pts = int(self.cap.get(cv2.CAP_PROP_POS_FRAMES)) # video_frame.time_base = self.time_base img = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) #img = frame.to_ndarray(format="bgr24") # 使用 Yolov5 进行目标检测 results = model(img) detections = results.pandas().xyxy[0].to_dict(orient="records") # 绘制检测结果 for det in detections: if det['name'] == 'person': # 只检测人员 x1, y1, x2, y2 = int(det['xmin']), int(det['ymin']), int(det['xmax']), int(det['ymax']) cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2) cv2.putText(img, f"{det['name']} {det['confidence']:.2f}", (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2) # 将检测结果图像转换为帧 new_frame = av.VideoFrame.from_ndarray(img, format="rgb24") new_frame.pts = int(self.cap.get(cv2.CAP_PROP_POS_FRAMES)) new_frame.time_base = self.time_base end_time = time.time() # 结束时间 print(f"Processing time: {end_time - start_time} seconds") return new_frame @api.route('/offer', methods=['POST']) async def offer(): #接收客户端的连接请求 params = await request.json url_type = params["url_type"] url = params["url"] #提取客户端发来的SDP，生成服务器端的SDP offer = RTCSessionDescription(sdp=params["sdp"], type=params["type"]) # # 配置 STUN 服务器 # ice_servers = [{"urls": ["stun:stun.voipbuster.com"]}] # pc = RTCPeerConnection(configuration={"iceServers": ice_servers}) pc = RTCPeerConnection() #实例化一个rtcp对象 pcs.add(pc) #集合中添加一个对象，若已存在则不添加 @pc.on("datachannel") def on_datachannel(channel): @channel.on("message") def on_message(message): print(f'Message received: {message}') # 处理接收到的数据 if message == 'pause': # 执行暂停操作 pass elif message == 'resume': # 执行继续操作 pass #监听RTC连接状态 @pc.on("iconnectionstatechange") #当ice连接状态发生变化时 async def iconnectionstatechange(): if pc.iceConnectionState == "failed": await pc.close() pcs.discard(pc) #移除对象 # 添加视频轨道 video_track = VideoTransformTrack(url,url_type) pc.addTrack(video_track) # @pc.on('track') --- stream.getTracks().forEach(track => pc.addTrack(track, stream)); 猜测是这里触发的添加了摄像头通道 # def on_track(track): # if track.kind == 'video': # local_video = VideoTransformTrack(track) # pc.addTrack(local_video) # 记录客户端 SDP await pc.setRemoteDescription(offer) # 生成本地 SDP answer = await pc.createAnswer() # 记录本地 SDP await pc.setLocalDescription(answer) print("返回sdp") return jsonify({ "sdp": pc.localDescription.sdp, "type": pc.localDescription.type }) @api.route('/shutdown', methods=['POST']) async def shutdown(): coros = [pc.close() for pc in pcs] await asyncio.gather(*coros) pcs.clear() return 'Server shutting down...' ''' ----------------------实现方式二，网页显示分析画面--基础版 ''' def is_point_in_polygon(point, polygon): return polygon.contains(Point(point)) frame_buffer = [] def camera_thread(): global cap, frame_buffer while True: try: ret, frame = cap.read() if not ret: logger.warning("Camera disconnected. Reconnecting...") cap.release() cap = get_camera() continue frame = cv2.resize(frame, (640, 480)) # 降低视频分辨率 frame_buffer.append(frame) if len(frame_buffer) > 10: # 保持缓冲区大小不超过10帧 frame_buffer.pop(0) logger.debug("drop one frame!") except Exception as e: logger.error(f"Error in camera thread: {e}") continue #threading.Thread(target=camera_thread, daemon=True).start() async def generate_frames(): global frame_buffer while True: frame = None buffer = None try: if frame_buffer: frame = frame_buffer.pop(0) # 进行推理 results = model(frame) detections = results.pandas().xyxy[0] # 绘制监控区域 cv2.polylines(frame, [np.array(region_points, np.int32)], isClosed=True, color=(0, 255, 0), thickness=2) for _, row in detections.iterrows(): if row['name'] == 'person': # 获取人员检测框的中心点 x_center = (row['xmin'] + row['xmax']) / 2 y_center = (row['ymin'] + row['ymax']) / 2 if is_point_in_polygon((x_center, y_center), polygon): # 触发报警 cv2.putText(frame, 'Alert: Intrusion Detected', (50, 50), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2) #logger.info('Alert: Intrusion Detected') # 绘制检测框 cv2.rectangle(frame, (int(row['xmin']), int(row['ymin'])), (int(row['xmax']), int(row['ymax'])), (255, 0, 0), 2) cv2.circle(frame, (int(x_center), int(y_center)), 5, (0, 0, 255), -1) ret, buffer = cv2.imencode('.jpg', frame) if not ret: continue frame_bytes = buffer.tobytes() yield (b'--frame\r\n' b'Content-Type: image/jpeg\r\n\r\n' + frame_bytes + b'\r\n') await asyncio.sleep(0.2) except Exception as e: logger.error(f"Error in generate_frames: {e}") finally: if frame is not None: del frame if buffer is not None: del buffer @api.route('/video_feed') async def video_feed(): #return jsonify(1) return Response(generate_frames(), mimetype='multipart/x-mixed-replace; boundary=frame') ''' ----------------------实现方式三，网页显示分析画面--WEBRTC ''' class VideoTransformTrack_new(VideoStreamTrack): def __init__(self, track): super().__init__() self.track = track async def recv(self): frame = await self.track.recv() img = frame.to_ndarray(format="bgr24") # 使用 Yolov5 进行目标检测 results = model(img) detections = results.pandas().xyxy[0].to_dict(orient="records") # 绘制检测结果 for det in detections: if det['name'] == 'person': # 只检测人员 x1, y1, x2, y2 = int(det['xmin']), int(det['ymin']), int(det['xmax']), int(det['ymax']) cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2) cv2.putText(img, f"{det['name']} {det['confidence']:.2f}", (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2) # 将检测结果图像转换为帧 new_frame = frame.from_ndarray(img, format="bgr24") return new_frame pcs = set() @api.websocket('/ws') async def ws(): params = await websocket.receive_json() offer = RTCSessionDescription(sdp=params['sdp'], type=params['type']) pc = RTCPeerConnection() pcs.add(pc) @pc.on('iceconnectionstatechange') async def on_iceconnectionstatechange(): if pc.iceConnectionState == 'failed': await pc.close() pcs.discard(pc) @pc.on('track') def on_track(track): if track.kind == 'video': local_video = VideoTransformTrack(track) pc.addTrack(local_video) await pc.setRemoteDescription(offer) answer = await pc.createAnswer() await pc.setLocalDescription(answer) await websocket.send(json.dumps({ 'sdp': pc.localDescription.sdp, 'type': pc.localDescription.type })) try: while True: await asyncio.sleep(1) finally: await pc.close() pcs.discard(pc)