openvino yolov5/ssd 实时推流目标检测在html上显示

安装ffmepg并添加到环境变量中，流媒体使用m7s

运行效果 SSD：检测在10ms左右，yolov5在100ms左右

app.py

#!/usr/local/bin/python3
# encodin: utf-8import subprocess
import threading
import time
import cv2
import osfrom OpenVinoYoloV5Detector import OpenVinoYoloV5Detector
from OpenVinoSSDDetector import OpenVinoSSDDetectorclass RTSCapture(cv2.VideoCapture):_cur_frame = None_reading = False_frame1 = None# 视频播放质量 值越大越模糊# _quality = 2.5  # 1;1.5;2;2.5# _interval_check = 1000  # 检测间隔 0 不跳帧 1秒10帧_quality = 1_interval_check = 0_recvtime = 0@staticmethoddef get_even(num):if (num % 2) != 0:num = num + 1return num@staticmethoddef create(srcstream, dststream):rtscap = RTSCapture(srcstream)rtscap.frame_receiver = threading.Thread(target=rtscap.recv_frame, daemon=True)rtscap.frame_processer = threading.Thread(target=rtscap.prcoess_frame, daemon=True)rtscap.frame_sender = threading.Thread(target=rtscap.send_frame, daemon=True)fps = int(rtscap.get(cv2.CAP_PROP_FPS))rtscap.width = rtscap.get_even(int(rtscap.get(cv2.CAP_PROP_FRAME_WIDTH) / rtscap._quality))rtscap.height = rtscap.get_even(int(rtscap.get(cv2.CAP_PROP_FRAME_HEIGHT) / rtscap._quality))rtscap.command = ['G:\\ZLMediaKit_Release\\ffmpeg.exe','-y','-f', 'rawvideo','-vcodec', 'rawvideo','-pix_fmt', 'bgr24','-s', "{}x{}".format(rtscap.width, rtscap.height),'-r', str(fps),'-i', '-','-c:v', 'libx264','-pix_fmt', 'yuv420p','-preset', 'ultrafast','-f', 'rtsp',  # flv rtsp'-rtsp_transport', 'udp',  # 使用TCP推流，linux中一定要有这行dststream]# 管道配置rtscap._reading = True# 加载算法rtscap.box_color = (0, 255, 0)# yolov5# conf = {#     # "weight_file": "weights/yolov5n_openvino_model/yolov5n.xml",#     "weight_file": "weights/yolov5n_openvino_model/yolov5n.xml",#     "device": "CPU"# }# rtscap.detector = OpenVinoYoloV5Detector(IN_conf=conf)#ssdconf = {"model_xml": "./weights/ssdlite_mobilenet_v2/FP16/ssdlite_mobilenet_v2.xml","model_bin": "./weights/ssdlite_mobilenet_v2/FP16/ssdlite_mobilenet_v2.bin","device": "CPU"}rtscap.detector = OpenVinoSSDDetector(IN_conf=conf)return rtscapdef get_milsecond(self):t = time.time()return (int(round(t * 1000)))def isOK(self):status = self.isOpened() and self._readingreturn statusdef recv_frame(self):while self.isOK():time.sleep(0.01)ok, frame = self.read()if not ok:breakself._cur_frame = frame#self.p.stdin.write(frame.tostring())self._reading = Falsedef prcoess_frame(self):while self.isOK():time.sleep(0.01)try:ok, frame = self.read_latest_frame()if not ok:continue# 算法解析# 跳帧检测if self.get_milsecond() - self._recvtime > self._interval_check:starttime = self.get_milsecond()detect_num, detect_data = self.detector.detect(frame)if len(detect_data):for m in detect_data:classname = m.get('class_name')score = m.get('score')location = m.get('location')box_l, box_t = int(location.get('x1')), int(location.get('y1'))box_r, box_b = int(location.get('x2')), int(location.get('y2'))frame = cv2.rectangle(frame, (box_l, box_t), (box_r, box_b), self.box_color, 2)frame = cv2.putText(frame, classname, (box_l, box_t + 15), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)self._recvtime = self.get_milsecond()endtime  = self.get_milsecond();print('检测耗时:' + str(endtime - starttime))self._frame1 = cv2.resize(frame, (self.width, self.height)) # 图像压缩#self.p.stdin.write(frame1.tostring())except Exception as e:ss = str(e)def send_frame(self):timeout = 3self.p = subprocess.Popen(self.command, shell=False, stdin=subprocess.PIPE)while self.isOK():time.sleep(0.01)try:if  self._frame1 is not None:self.p.stdin.write(self._frame1.tostring())except Exception as e:ss = str(e)def read2(self):frame = self._cur_frameself._cur_frame = Nonereturn frame is not None, framedef start_read(self):self.frame_receiver.start()self.frame_processer.start()self.frame_sender.start()self.read_latest_frame = self.read2 if self._reading else self.readdef stop_read(self):self._reading = Falseif self.frame_receiver.is_alive():self.frame_receiver.join()if self.frame_processer.is_alive():self.frame_processer.join()if self.frame_sender.is_alive():self.frame_sender.join()if __name__ == '__main__':url = src = '0'if src.isdigit():url = int(src)rtscap1 = RTSCapture.create(url, "rtsp://127.0.0.1:554/live/test0")  # test.mp4 0rtscap1.start_read()os.system('pause')