使用Python和OpenCV實現樹莓派/PC實時攝像頭資料共享，主要分為伺服器和客戶端兩部分。

伺服器Demo如下：

#伺服器端

import socket
import threading
import struct
import time
import cv2
import numpy

class Carame_Accept_Object:
    def __init__(self,S_addr_port=("",8880)):
        self.resolution=(640,480)       #解析度
        self.img_fps=15                 #每秒傳輸多少幀數
        self.addr_port=S_addr_port
        self.Set_Socket(self.addr_port)

    #設定套接字
    def Set_Socket(self,S_addr_port):
        self.server=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
        self.server.setsockopt(socket.SOL_SOCKET,socket.SO_REUSEADDR,1) #埠可複用
        self.server.bind(S_addr_port)
        self.server.listen(5)
        #print("the process work in the port:%d" % S_addr_port[1])


def check_option(object,client):
    #按格式解碼，確定幀數和解析度
    info=struct.unpack('lhh',client.recv(8))
    if info[0]>888:
        object.img_fps=int(info[0])-888          #獲取幀數
        object.resolution=list(object.resolution)
        # 獲取解析度
        object.resolution[0]=info[1]
        object.resolution[1]=info[2]
        object.resolution = tuple(object.resolution)
        return 1
    else:
        return 0

def RT_Image(object,client,D_addr):
    if(check_option(object,client)==0):
        return
    camera=cv2.VideoCapture(0)                                #從攝像頭中獲取視訊
    img_param=[int(cv2.IMWRITE_JPEG_QUALITY),object.img_fps]  #設定傳送影象格式、幀數
    while(1):
        time.sleep(0.1)             #推遲執行緒執行0.1s
        _,object.img=camera.read()  #讀取視訊每一幀

        object.img=cv2.resize(object.img,object.resolution)     #按要求調整影象大小(resolution必須為元組)
        _,img_encode=cv2.imencode('.jpg',object.img,img_param)  #按格式生成圖片
        img_code=numpy.array(img_encode)                        #轉換成矩陣
        object.img_data=img_code.tostring()                     #生成相應的字串
        try:
            #按照相應的格式進行打包傳送圖片
            client.send(struct.pack("lhh",len(object.img_data),object.resolution[0],object.resolution[1])+object.img_data)
        except:
            camera.release()        #釋放資源
            return

if __name__ == '__main__':
    camera=Carame_Accept_Object()
    while(1):
        client,D_addr=camera.server.accept()
        clientThread=threading.Thread(None,target=RT_Image,args=(camera,client,D_addr,))
        clientThread.start()
 

 客戶端Demo如下：

#客戶端


import socket
import cv2
import threading
import struct
import numpy

class Camera_Connect_Object:
    def __init__(self,D_addr_port=["",8880]):
        self.resolution=[640,480]
        self.addr_port=D_addr_port
        self.src=888+15                 #雙方確定傳輸幀數，（888）為校驗值
        self.interval=0                 #圖片播放時間間隔
        self.img_fps=100               #每秒傳輸多少幀數

    def Set_socket(self):
        self.client=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
        self.client.setsockopt(socket.SOL_SOCKET,socket.SO_REUSEADDR,1)

    def Socket_Connect(self):
        self.Set_socket()
        self.client.connect(self.addr_port)
        print("IP is %s:%d" % (self.addr_port[0],self.addr_port[1]))

    def RT_Image(self):
        #按照格式打包傳送幀數和解析度
        self.name=self.addr_port[0]+" Camera"
        self.client.send(struct.pack("lhh", self.src, self.resolution[0], self.resolution[1]))
        while(1):
            info=struct.unpack("lhh",self.client.recv(8))
            buf_size=info[0]                    #獲取讀的圖片總長度
            if buf_size:
                try:
                    self.buf=b""                #代表bytes型別
                    temp_buf=self.buf
                    while(buf_size):            #讀取每一張圖片的長度
                        temp_buf=self.client.recv(buf_size)
                        buf_size-=len(temp_buf)
                        self.buf+=temp_buf      #獲取圖片
                        data = numpy.fromstring(self.buf, dtype='uint8')    #按uint8轉換為影象矩陣
                        self.image = cv2.imdecode(data, 1)                  #影象解碼
                        gray = cv2.cvtColor(self.image, cv2.COLOR_BGR2GRAY)
                        cv2.imshow(self.name, self.image)                   #展示圖片
                except:
                    pass;
                finally:
                    if(cv2.waitKey(10)==27):        #每10ms重新整理一次圖片，按‘ESC’（27）退出
                        self.client.close()
                        cv2.destroyAllWindows()
                        break

    def Get_Data(self,interval):
        showThread=threading.Thread(target=self.RT_Image)
        showThread.start()

if __name__ == '__main__':
    camera=Camera_Connect_Object()
    camera.addr_port[0]=input("Please input IP:")
    camera.addr_port=tuple(camera.addr_port)
    camera.Socket_Connect()
    camera.Get_Data(camera.interval)
 

執行效果還是比較流暢的，因為傳輸網路視訊流，所以畫質欠佳。

樹莓派視訊傳輸改進方式見：https://blog.csdn.net/m0_38106923/article/details/82628061