600字范文 > （03更新）海康威视网络摄像头sdk python版（报错排查）移动物体检测

（03更新）海康威视网络摄像头sdk python版（报错排查）移动物体检测

时间：2022-07-05 05:18:28

海康威视网络摄像头Python SDK+移动物体位置检测

文末附海康威视网络摄像头Python SDK 下载链接

本文参考博客

/c1052/article/details/95082377

/pengpengloveqiaoqiao/article/details/89487049

注意：摄像头要在镜头上贴红外滤光片过滤掉可见光

效果展示，已经过滤到大部分波长可见光，手在镜头前来回移动，红框即检测到移动物体的位置，并框出来（海康威视的摄像头可见光好像不能完全被滤光片过滤掉，但是可以过滤到投影机的画面或者屏幕的画面，不影响使用）

1.将海康威视的C++ sdk 转化为 Python sdk

2.调用海康威视Python SDK

3.调用多个海康威视摄像头

4. 将多个摄像头画面拼接成一个

5.挨个摄像头透视变换

6. 透视变换完后再在拼接后的画面上做动态物体检测，参考博客

1.将海康威视的C++ sdk 转化为 Python sdk (本文已经处理完了)

如何将海康威视的C++ sdk 转化为 Python 可用的sdk ，请参考博主ciky奇的文章，链接如下

/c1052/article/details/95082377 如果最后一步Python调用时遇到报错，尝试将所有相关的文件全部复制到Python工程的路径下

环境：

1. windows10 64位

2. opecv

3. python3

4.海康威视网络摄像头64位 C++ sdk

/cn/download_61.html

下载SDK_Win64

5.其他库文件：

5.1 opencv-python

5.2 numpy

5.3multiprocessing （多进程使用）

本文已将海康威视C++ sdk转为Python sdk处理好，可以直接下载使用，文末网址分享

如果HKIPCamera 一直报错，找不到模块，首先 python 和 opencv要安装好，并且设置好环境变量，如果还是报错：

一般是缺少VC++库的运行时组件，两个方案：

1. 从微软官网下载VC++库的运行时组件：

此组件也可以解决其他项目上的很多疑难杂症

官网链接：/zh-hans/downloads/

2、下载一个DirectX修复工具

链接：/s/1thfI2SyCpQoToBxuwFnxuA
提取码：8aza

打开后点击工具----> 选项---->扩展---->开始扩展

扩展完了以后，此时会默认选中同时更新C++，点击检测并修复，

如果缺少组件就会自动安装，如果已经安装了但是不是最新的，会自动更新，

等修复完了，再尝试运行Python程序

如果还是一直存在问题，请回头参考博文/c1052/article/details/95082377一步一步重新操作

2.调用海康威视Python SDK

import cv2import numpy as npimport HKIPcameradef get_cam_frame(ip, account, password):ip = str(ip) # 摄像头IP地址，要和本机IP在同一局域网name = str(account) # 管理员用户名pw = str(password) # 管理员密码HKIPcamera.init(ip, name, pw)while(True): fram = HKIPcamera.getframe()frame = np.array(fram)cv2.imshow('frame',frame)c = cv2.waitKey(10)if c == 27:break;HKIPcamera.release()if __name__ == '__main__':get_cam_frame('192.168.1.64', 'admin', '123456')

3.调用多个海康威视摄像头

调用多个摄像头需要用到多进程，多线程不起作用

在第一步的代码基础上修改一下

import cv2import numpy as npimport HKIPcamerafrom multiprocessing import Process, Queuedef get_cam_frame(frameName, ip, account, password):ip = str(ip) # 摄像头IP地址，要和本机IP在同一局域网name = str(account) # 管理员用户名pw = str(password) # 管理员密码HKIPcamera.init(ip, name, pw)while True:fram = HKIPcamera.getframe()frame = np.array(fram)cv2.imshow(frameName, frame)c = cv2.waitKey(10)if c == 27:break;HKIPcamera.release()if __name__ == '__main__':p0 = Process(target=get_cam_frame, args=('frame0', '192.168.1.64', 'admin', '123456',))p0.start()p1 = Process(target=get_cam_frame, args=('frame1', '192.168.1.65', 'admin', '123456',))p1.start()# 以此类推 p2 p3 ... pn

4. 将多个摄像头画面拼接成一个

再在上述的代码基础上修改

5.挨个摄像头透视变换

如果不了解opencv-python透视变换函数的参数，请先看下此博客的透视变换函数介绍，链接如下：

/wsp_1138886114/article/details/83374333

import cv2import numpy as npimport HKIPcamerafrom multiprocessing import Process, Queueimport osimport signaldef get_cam_frame(frameName, ip, account, password, n):ip = str(ip) # 摄像头IP地址，要和本机IP在同一局域网name = str(account) # 管理员用户名pw = str(password) # 管理员密码HKIPcamera.init(ip, name, pw)while True:fram = HKIPcamera.getframe()frame = np.array(fram)# 由于摄像头画面是1920*1080，多个拼接的话，画面会非常大，所有重置一下每个摄像头的画面frame = cv2.resize(frame, (768, 540), interpolation=cv2.INTER_CUBIC)n.put(frame)cv2.imshow(frameName, frame)c = cv2.waitKey(10)if c == 27:break;HKIPcamera.release()#第一个摄像头透视变化的四个点frame0_x0, frame0_y0, frame0_x1, frame0_y1, frame0_x2, frame0_y2, frame0_x3, frame0_y3 = 0, 0, 0, 0, 0, 0, 0, 0#第二个摄像头透视变化的四个点frame1_x0, frame1_y0, frame1_x1, frame1_y1, frame1_x2, frame1_y2, frame1_x3, frame1_y3 = 0, 0, 0, 0, 0, 0, 0, 0#简单来写，按照从第一个摄像头开始，鼠标每点击一次，获取一个点#例如鼠标第一次点击，就等于给坐标frame0_x0, frame0_y0 赋值，第二次点击就是给坐标frame0_x1, frame0_y1赋值，以此类推clickNum = 0def draw_point(event, x, y, flags, param):global frame0_x0, frame0_y0, frame0_x1, frame0_y1, frame0_x2, frame0_y2, frame0_x3, frame0_y3global frame1_x0, frame1_y0, frame1_x1, frame1_y1, frame1_x2, frame1_y2, frame1_x3, frame1_y3, clickNumif event == cv2.EVENT_LBUTTONDOWN:if clickNum == 0:frame0_x0 = xframe0_y0 = yelif clickNum == 1:frame0_x1 = xframe0_y1 = yelif clickNum == 2:frame0_x2 = xframe0_y2 = yelif clickNum == 3:frame0_x3 = xframe0_y3 = yelif clickNum == 4:frame1_x0 = xframe1_y0 = yelif clickNum == 5:frame1_x1 = xframe1_y1 = yelif clickNum == 6:frame1_x2 = xframe1_y2 = yelif clickNum == 7:frame1_x3 = xframe1_y3 = yclickNum += 1# 透视变换def perspective_transform():global frame0, frame1, clickNumglobal frame0_x0, frame0_y0, frame0_x1, frame0_y1, frame0_x2, frame0_y2, frame0_x3, frame0_y3global frame1_x0, frame1_y0, frame1_x1, frame1_y1, frame1_x2, frame1_y2, frame1_x3, frame1_y3if clickNum==8:W_cols, H_rows = frame0.shape[:2]pts0_1 = np.float32([[frame0_x0, frame0_y0], [frame0_x1, frame0_y1],[frame0_x2, frame0_y2], [frame0_x3, frame0_y3]])pts0_2 = np.float32([[0, 0], [W_cols, 0], [0, H_rows], [W_cols, H_rows]])M0 = cv2.getPerspectiveTransform(pts0_1, pts0_2)frame0 = cv2.warpPerspective(frame0, M0, (W_cols, H_rows))# 这里的变换需要减去第一个摄像头的画面的宽度，768pts1_1 = np.float32([[frame1_x0-768, frame1_y0], [frame1_x1-768, frame1_y1],[frame1_x2-768, frame1_y2], [frame1_x3-768, frame1_y3]])pts1_2 = np.float32([[0, 0], [W_cols, 0], [0, H_rows], [W_cols, H_rows]])M1 = cv2.getPerspectiveTransform(pts1_1, pts1_2)frame1 = cv2.warpPerspective(frame1, M1, (W_cols, H_rows))if __name__ == '__main__':q0 = Queue(3)p0 = Process(target=get_cam_frame, args=('frame0', '192.168.1.64', 'admin', '123456', q0,))p0.start()q1 = Queue(3)p1 = Process(target=get_cam_frame, args=('frame1', '192.168.1.65', 'admin', '123456', q1,))p1.start()cv2.namedWindow('full_frame')# 鼠标点击事件cv2.setMouseCallback('full_frame', draw_point)while True:if p0.is_alive() and p1.is_alive():frame0 = q0.get()frame1 = q1.get()# 透视变换perspective_transform()# 透视变换完再拼接# 横着拼full_frame = np.hstack([frame0, frame1])# 竖着拼#full_frame = np.vstack([frame0, frame1])cv2.imshow('full_frame', full_frame)c = cv2.waitKey(10)if c==27:os.kill(p0.pid, signal.SIGTERM)os.kill(p1.pid, signal.SIGTERM)break;

6. 透视变换完后再在拼接后的画面上做动态物体检测，参考博客

/pengpengloveqiaoqiao/article/details/89487049

import cv2import numpy as npimport HKIPcamerafrom multiprocessing import Process, Queueimport osimport signaldef get_cam_frame(frameName, ip, account, password, n):ip = str(ip) # 摄像头IP地址，要和本机IP在同一局域网name = str(account) # 管理员用户名pw = str(password) # 管理员密码HKIPcamera.init(ip, name, pw)while True:fram = HKIPcamera.getframe()frame = np.array(fram)# 由于摄像头画面是1920*1080，多个拼接的话，画面会非常大，所有重置一下每个摄像头的画面frame = cv2.resize(frame, (768, 540), interpolation=cv2.INTER_CUBIC)n.put(frame)cv2.imshow(frameName, frame)c = cv2.waitKey(10)if c == 27:break;HKIPcamera.release()#第一个摄像头透视变化的四个点frame0_x0, frame0_y0, frame0_x1, frame0_y1, frame0_x2, frame0_y2, frame0_x3, frame0_y3 = 0, 0, 0, 0, 0, 0, 0, 0#第二个摄像头透视变化的四个点frame1_x0, frame1_y0, frame1_x1, frame1_y1, frame1_x2, frame1_y2, frame1_x3, frame1_y3 = 0, 0, 0, 0, 0, 0, 0, 0#简单来写，按照从第一个摄像头开始，鼠标每点击一次，获取一个点#例如鼠标第一次点击，就等于给坐标frame0_x0, frame0_y0 赋值，第二次点击就是给坐标frame0_x1, frame0_y1赋值，以此类推clickNum = 0def draw_point(event, x, y, flags, param):global frame0_x0, frame0_y0, frame0_x1, frame0_y1, frame0_x2, frame0_y2, frame0_x3, frame0_y3global frame1_x0, frame1_y0, frame1_x1, frame1_y1, frame1_x2, frame1_y2, frame1_x3, frame1_y3, clickNumif event == cv2.EVENT_LBUTTONDOWN:if clickNum == 0:frame0_x0 = xframe0_y0 = yelif clickNum == 1:frame0_x1 = xframe0_y1 = yelif clickNum == 2:frame0_x2 = xframe0_y2 = yelif clickNum == 3:frame0_x3 = xframe0_y3 = yelif clickNum == 4:frame1_x0 = xframe1_y0 = yelif clickNum == 5:frame1_x1 = xframe1_y1 = yelif clickNum == 6:frame1_x2 = xframe1_y2 = yelif clickNum == 7:frame1_x3 = xframe1_y3 = yclickNum += 1# 透视变换def perspective_transform():global frame0, frame1, clickNumglobal frame0_x0, frame0_y0, frame0_x1, frame0_y1, frame0_x2, frame0_y2, frame0_x3, frame0_y3global frame1_x0, frame1_y0, frame1_x1, frame1_y1, frame1_x2, frame1_y2, frame1_x3, frame1_y3if clickNum==8:W_cols, H_rows = frame0.shape[:2]pts0_1 = np.float32([[frame0_x0, frame0_y0], [frame0_x1, frame0_y1],[frame0_x2, frame0_y2], [frame0_x3, frame0_y3]])pts0_2 = np.float32([[0, 0], [W_cols, 0], [0, H_rows], [W_cols, H_rows]])M0 = cv2.getPerspectiveTransform(pts0_1, pts0_2)frame0 = cv2.warpPerspective(frame0, M0, (W_cols, H_rows))# 这里的变换需要减去第一个摄像头的画面的宽度，768pts1_1 = np.float32([[frame1_x0-768, frame1_y0], [frame1_x1-768, frame1_y1],[frame1_x2-768, frame1_y2], [frame1_x3-768, frame1_y3]])pts1_2 = np.float32([[0, 0], [W_cols, 0], [0, H_rows], [W_cols, H_rows]])M1 = cv2.getPerspectiveTransform(pts1_1, pts1_2)frame1 = cv2.warpPerspective(frame1, M1, (W_cols, H_rows))if __name__ == '__main__':q0 = Queue(3)p0 = Process(target=get_cam_frame, args=('frame0', '192.168.1.64', 'admin', '123456', q0,))p0.start()q1 = Queue(3)p1 = Process(target=get_cam_frame, args=('frame1', '192.168.1.65', 'admin', '123456', q1,))p1.start()cv2.namedWindow('full_frame')# 鼠标点击事件cv2.setMouseCallback('full_frame', draw_point)kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5, 5))fgbg = cv2.createBackgroundSubtractorMOG2()while True:if p0.is_alive() and p1.is_alive():frame0 = q0.get()frame1 = q1.get()# 透视变换perspective_transform()# 透视变换完再拼接# 横着拼full_frame = np.hstack([frame0, frame1])# 竖着拼#full_frame = np.vstack([frame0, frame1])fgmask = fgbg.apply(full_frame)mask = cv2.morphologyEx(fgmask, cv2.MORPH_OPEN, kernel)contours, _ = cv2.findContours(fgmask.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)for c in contours:if 100 < cv2.contourArea(c) < 40000:x, y, w, h = cv2.boundingRect(c)cv2.rectangle(full_frame, (x, y), (x + w, y + h), (0, 0, 255))cv2.imshow('full_frame', full_frame)c = cv2.waitKey(10)if c==27:os.kill(p0.pid, signal.SIGTERM)os.kill(p1.pid, signal.SIGTERM)break;

最后再写个Json或者XML(我没写)保存一下透视变换的点位信息完整工程分享

链接：/s/1zHwWCW5dW-XR6a98RfEuew

提取码：6gfc

（03更新）海康威视网络摄像头sdk python版（报错排查）移动物体检测多通道海康威视网洛摄像头后台检测海康威视网络摄像头Python SDK

本内容不代表本网观点和政治立场，如有侵犯你的权益请联系我们处理。

网友评论

网友评论仅供其表达个人看法，并不表明网站立场。