前提

*1、注册百度AI
2、得到自己的token凭证
3、安装推流工具 ffemeg
4、配置 nginx - rtmp 流媒体服务器
*

依赖库

import cv2
import base64
from PIL import Image
from io import BytesIO
import requests
import json
import numpy as np
import threading
import time
import subprocess as sp

code

1、access_token:添加自己的
2、RTMP_HOST = ’ ’ 自己流媒体服务器的地址

shared_image = (np.ones((540, 960, 3), dtype=np.uint8) * 255).astype(np.uint8)
process_image = (np.ones((540, 960, 3), dtype=np.uint8) * 255).astype(np.uint8)

# 左上角
# 右上角
# 左下角
# 右下角
# area 代表的是要检测的区域范围
def getResult(img):
    params = {
        "area": "1,1,     590,1,    590,190,     1,190",
        "case_id": 1,
        "case_init": "false",
        "dynamic": "true",
        "image": img,
        "show": "true", }
    # 请求参数
    request_url = "https://aip.baidubce.com/rest/2.0/image-classify/v1/body_tracking"
    access_token = '***********' 
    request_url = request_url + "?access_token=" + access_token
    headers = {'content-type': 'application/x-www-form-urlencoded'}

    response = requests.post(request_url, data=params, headers=headers)

    if response:
        # 将 Python 字典类型转换为 JSON 对象
        # print(response.json())
        json_str = json.dumps(response.json())
        # 转化为 json字符串
        d = json.loads(json_str)

        # print(d['image'])
        # print(d['person_count'])
        person_count = d['person_count']
        in_num = person_count['in']
        out_num = person_count['out']
        print("in: ", in_num)
        print("out: ", out_num)
        # 再把加密后的结果解码， 结果为二进制数据
        temp = base64.b64decode(d['image'])
        # print(temp)
        # 二进制数据流转np.ndarray [np.uint8: 8位像素]
        img = cv2.imdecode(np.frombuffer(temp, np.uint8), cv2.IMREAD_COLOR)
        #   将bgr转为rbg
        rgb_img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
        # print(rgb_img)
        # cv2.imshow("res", rgb_img)
        # cv2.waitKey()
        return rgb_img

def frame2base64(frame):
    img = Image.fromarray(frame)  # 将每一帧转为Image
    output_buffer = BytesIO()  # 创建一个BytesIO
    img.save(output_buffer, format='JPEG')  # 写入output_buffer
    byte_data = output_buffer.getvalue()  # 在内存中读取
    base64_data = base64.b64encode(byte_data)  # 转为BASE64
    return base64_data  # 转码成功 返回base64编码

def generate():
    camera = cv2.VideoCapture(0)
    try:
        while True:
            ret, frame = camera.read()
            gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
            cv2.imshow("camera", frame)
            base64_data = frame2base64(frame)

            getResult(base64_data)

            # print(base64_data)

            if cv2.waitKey(1) & 0xFF == ord("q"):
                break

    except Exception as e:
        print(e)
    finally:
        # 释放资源
        camera.release()
        cv2.destroyAllWindows()
    return  base64_data

# 定义CV线程， 提取视频流 抽帧
class CvThread(threading.Thread):
    def __init__(self):
        super(CvThread, self).__init__()  # 注意：一定要显式的调用父类的初始化函数。
        # self.arg=arg
    def run(self):  # 定义每个线程要运行的函数
        print('CvThread thread is run!')
        global shared_image
        camera = cv2.VideoCapture(0)
        if (camera.isOpened()):
            print('Open camera 1')
        else:
            print('Fail to open camera 1!')
            time.sleep(0.05)
        camera.set(cv2.CAP_PROP_FRAME_WIDTH, 864)  # 2560x1920 2217x2217 2952×1944 1920x1080
        camera.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)
        camera.set(cv2.CAP_PROP_FPS, 5) # 抽帧参数
        size = (int(camera.get(cv2.CAP_PROP_FRAME_WIDTH)), int(camera.get(cv2.CAP_PROP_FRAME_HEIGHT)))
        sizeStr = str(960) + 'x' + str(540)
        fps = camera.get(cv2.CAP_PROP_FPS)  # 30p/self
        fourcc = cv2.VideoWriter_fourcc(*'XVID')
        out = cv2.VideoWriter('res_mv.avi', fourcc, fps, size)
        while True:
            ret, frame = camera.read()  # 逐帧采集视频流
            if frame is not None:
                image = Image.fromarray(frame)
                image = image.resize((960, 540))
                frame = np.array(image)
                # cv2.imwrite('F:/renliu/chouzhen/'+ str(i) +'.jpg',frame)
                # print(frame)
                shared_image = frame

# 定义检测线程， 检测视频
class BaiDuThread(threading.Thread):
    def __init__(self):
        super(BaiDuThread, self).__init__()  # 注意：一定要显式的调用父类的初始化函数。

    def run(self):  # 定义每个线程要运行的函数
        print('BaiDuThread thread is run!')
        global shared_image
        global process_image
        while True:
            # 把图片编码
            img = Image.fromarray(shared_image)  # 将每一帧转为Image
            output_buffer = BytesIO()  # 创建一个BytesIO
            img.save(output_buffer, format='JPEG')  # 写入output_buffer
            byte_data = output_buffer.getvalue()  # 在内存中读取
            base64_data = base64.b64encode(byte_data)  # 转为BASE64
            res = getResult(base64_data)
            process_image = res
            time.sleep(0.05)

RTMP_HOST = ' '
rtmpUrl = 'rtmp://' + RTMP_HOST + ':1935/live/test'
command = ['ffmpeg',
    '-y',
    '-f', 'rawvideo',
    '-vcodec','rawvideo',
    '-pix_fmt', 'bgr24',
    '-s', '960x540',
    '-r', str(5),
    '-i', '-',
    '-c:v', 'libx264',
    '-pix_fmt', 'yuv420p',
    '-preset', 'ultrafast',
    '-f', 'flv',
    rtmpUrl]

global pipe
pipe = sp.Popen(command, stdin=sp.PIPE)

class PushThread(threading.Thread):
    def __init__(self):
        super(PushThread, self).__init__()  # 注意：一定要显式的调用父类的初始化函数。
        # self.arg=arg
    def run(self):  # 定义每个线程要运行的函数
        print('PushThread thread is run!')
        global process_image
        while True:
            pipe.stdin.write(process_image.tostring())  # 存入管道
            cv2.imwrite('1.jpg',process_image)
            time.sleep(0.198)

CvThread = CvThread()
CvThread.start()

BaiDuThread = BaiDuThread()
BaiDuThread.start()

PushThread = PushThread()
PushThread.start()

示意图：

参考：

https://blog.csdn.net/zhangpan929/article/details/85410690?depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromBaidu-4&utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromBaidu-4