Python与OpenCV实时人脸识别

Yourtion 创作于：2013-07-30     全文约 1562 字， 预计阅读时间为 5 分钟

刚刚开始使用Python写OpenCV的东西，发现关于使用Python写OpenCV的还是比较少的，先整了一个人脸识别的最简单实例，与大家共享！

环境：Python 2.7.4、OpenCV 2.4.6、LinuxMint 15

最简单代码如下：

import os,sys,cv2
pth = os.path.dirname(sys.argv[0])
cv2.namedWindow("camera")
det = cv2.CascadeClassifier("./haarcascade_frontalface_alt.xml")
cam = cv2.VideoCapture(0)
while cam.isOpened():
    flag, frame = cam.read();
    if flag == True:
        tmp = cv2.cvtColor(frame, cv2.cv.CV_RGB2GRAY)
        objs = det.detectMultiScale(tmp, 1.1, 4, 0,(frame.shape[1] / 10, frame.shape[0] / 10))
    for rct in objs:
        cv2.rectangle(frame, (rct[0], rct[1]),(rct[0] + rct[2], rct[1] + rct[3]),(0, 0, 255), 2, cv2.CV_AA)
    cv2.imshow("camera", frame)
    if cv2.waitKey(10) == 27:
        cam.release()
        break
cv2.destroyWindow("camera")

接下来我还增加输出统计人数的功能，还是最简单的方法，统计找到的obj数目，只有在变化的时候才输出，代码如下：

import os,sys,cv2

pth = os.path.dirname(sys.argv[0])
cv2.namedWindow("camera")
det = cv2.CascadeClassifier("./haarcascade_frontalface_alt.xml")
cam = cv2.VideoCapture(0)
people = 0
while cam.isOpened():
    flag, frame = cam.read();
    if flag == True:
        tmp = cv2.cvtColor(frame, cv2.cv.CV_RGB2GRAY)
        objs = det.detectMultiScale(tmp, 1.1, 4, 0,(frame.shape[1] / 10, frame.shape[0] / 10))
    if len(objs)!=people:
        people = len(objs)
        if people!=0:
            print people
    for rct in objs:
        cv2.rectangle(frame, (rct[0], rct[1]),(rct[0] + rct[2], rct[1] + rct[3]),(0, 0, 255), 2, cv2.CV_AA)
    cv2.imshow("camera", frame)
    if cv2.waitKey(10) == 27:
        cam.release()
        break
cv2.destroyWindow("camera")

希望对你有帮助，接下来会继续更新相关的东西

原文链接：https://blog.yourtion.com/python-opencv-realtime-face-recognition.html

Twitter Facebook WeiBo
Python