Shulou(Shulou.com)06/02 Report--

This article is about how to use Python to implement simple Face Recognition features. Xiaobian thinks it is quite practical, so share it with everyone for reference. Let's follow Xiaobian and have a look.

1. First of all

Sort out the steps needed to implement Face Recognition:

The process is roughly like this. Before this, we must first let the face be accurately found, that is, the classifier that can accurately distinguish the face. Here we can use the classifier that has been trained. The online category is more complete and the classification accuracy is higher. We can also save time in this area.

Since Python is used, it is naturally necessary to use the package. Before looking at the code, we will list the packages needed for the whole project:

CV2(Opencv): Image recognition, camera call

· OS: file manipulation

Numpy: NumPy(Numerical Python) is an extension library for Python that supports a large number of dimensional arrays and matrix operations, as well as a large number of mathematical libraries for array operations.

PIL: Python Imaging Library, Python platform is in fact the standard library for image processing

2. Next 1. Get face samples by comparing faces #----import cv2 #Call notebook built-in camera, parameter is 0, if there are other cameras, adjust parameter is 1,2cap = cv2.VideoCapture(0)#Call face classifier, adjust 3face_detector = cv2.CascadeClassifier(r'X: /Users/73950/Desktop/FaceRec/haarccascade_frontface_default. xml') #To be changed #Mark an idface_id = input ('\n User data input,Look at the camera and wait... ')#sampleNum used to count the number of samples count = 0 while True: #Read pictures from camera success,img = cap.read() #Turn to grayscale picture, reduce program compliance, improve recognition if success is True: gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) else: break #Face detection, bring the data recorded by each frame camera into OpenCv, and let Classifier judge the face #where gray is the grayscale image to be detected, 1.3 is the proportion of each image size reduction, and 5 is minNeighbors faces = face_detector.detectMultiScale(gray, 1.3, 5) #Frame faces, for loop to ensure a detectable real-time dynamic video stream for (x, y, w, h) in faces: #xy is the coordinates of the upper left corner,w is width, h is height, and rectangle is used to mark the frame of the face. cv2.rectangle(img, (x, y), (x+w, y+w), (255, 0, 0)) #Successful box selection increases sample size count += 1 #Save the image and treat the gray image as a two-dimensional array to detect the face area #(Here is the folder where data is created, of course, it can also be set to other paths or call the database) cv2.imwrite("data/User. "+str(face_id)+'. '+str(count)+'.jpg',gray[y:y+h,x:x+w]) #Show pictures cv2.imshow('image',img) #Keep the picture continuous. waitkey method can bind the key to ensure the display of the screen, through the q key to exit the camera k = cv2.waitKey(1) if k == '27': break #Or quit shooting after getting 800 samples. Here, the data amount can be modified according to the actual situation. The effect of 800 samples after actual test is ideal. elif count >= 800: break #Turn off camera, free resources cap.realease() cv2.destrucyAllWindows ()

After the blogger tests, in execution

"face_detector = cv2.CascadeClssifier(r'C:\Users\admin\Desktop\python\data\haarcascade_frontface_default.xml')" In this statement, try not to have Chinese characters in the directory name in the actual path, otherwise it is easy to report errors.

So your computer can see you!

2. Control model established by algorithm

The algorithm used this time is the algorithm included in opencv. In the newer version of opencv (I use 2.4.8), a FaceRecognizer class is provided, which has some related algorithms and function interfaces for Face Recognition, including three Face Recognition algorithms (we use the third one).

1.eigenface

2.fisherface

3.LBPHFaceRecognizer

LBP is a feature extraction method that can extract local texture features of an image. The initial LBP operator is to take the pixel value of the center pixel as the threshold value in a 3X3 window and compare it with the pixel values of eight surrounding pixels. If the pixel value of the pixel is greater than the threshold value, the pixel is marked as 1, otherwise it is marked as 0. In this way, we can get an octal binary code, which is converted to decimal, that is, LBP code, so we get the LBP value of this window, and use this value to reflect the texture information in this window.

LBPH is an improvement on the original LBP. With opencv support, we can directly call the function to create a LBPH Face Recognition model.

We created a Python file named trainner.py in the same directory as the previous section to write the dataset generation script. In the same directory, create a folder called trainer to store our trained recognizers.

#----Build Model, Create Data Set----#---Build Model, Create Data Set---- import osimport cv2import numpy as npfrom PIL import Image#import pillow library, Used to process images #Set previously collected data file path = 'data' #Initialize recognition method recog = cv2.face.LBPHFaceRecognizer_create() #Call familiar face classifier detector = cv2.CascadeClassifier ('haarcascade_frontface_default.xml')#Create a function to get training images from the dataset folder and get id#Note that the naming format for images is User. id. sampleNumdef get_images_and_labels(path): image_paths = [os.path.join(path,f) for f in os.listdir(path)] #Create a list to store face_samples = [] ids = [] #Traverse the image path, import the image and id to add to the list for image_path in image_paths: #Convert it to grayscale via image path img = Image.open(image_path).convert('L') #Convert images to arrays img_np = np.array(img,'uint8') if os.path.split(image_path)[-1].split(". ")[-1] != 'jpg': continue #To get id, split picture and path and get id = int(os.path.split(image_path)[-1].split(". ")[1]) faces = detector.detectMultiScale(img_np) #Add images and id to list for(x,y,w,h) in faces: face_samples.append(img_np[y:y+h,x:x+w]) ids.append(id) return face_samples,ids #Call function and feed data to recognizer Training print ('Training... ')faces,ids = get_images_and_labels(path)#train model recog.train(faces,np.array(ids))#save model recog. save ('trainer/trainer. yml') 3. identify

Detection, verification, and output are actually the processes of identification. Unlike the first two processes, this is a process involving actual use, so we integrate them into a unified file.

#----Detect, verify and output results----import cv2 #Prepare recognition method recognizer = cv2.face.LBPHFaceRecognizer_create() #Use previously trained model recognizer. read ('trainer/trainer. yml')#invoke face classifier cascade_path = "haarcascade_frontface_default.xml" face_cascade = cv2.Cascade Classifier again (cascade_path) #Load a font for identification, mark the name of the object on the picture font = cv2.FONT_HERSHEY_SIMPLEX idnum = 0#Set the user name corresponding to the ID number as follows, such as 0 corresponding to the initial names ="'initial ',' admin','user1',' user2','user 3']#Call camera cam = cv2.VideoCapture(0)minW = 0.1*cam.get(3)minH = 0.1*cam.get(4) while True: ret,img = cam.read() gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) #Recognize faces faces = face_cascade.detectMultiScale( gray, scaleFactor = 1.2, minNeighbors = 5, minSize = (int(minW),int(minH)) ) #Perform verification for(x,y,w,h) in faces: cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2) idnum,confidence = recognizer.predict(gray[y:y+h,x:x+w]) #Calculate a test result if confidence

< 100: idum = names[idnum] confidence = "{0}%",format(round(100-confidence)) else: idum = "unknown" confidence = "{0}%",format(round(100-confidence)) #输出检验结果以及用户名 cv2.putText(img,str(idum),(x+5,y-5),font,1,(0,0,255),1) cv2.putText(img,str(confidence),(x+5,y+h-5),font,1,(0,0,0),1) #展示结果 cv2.imshow('camera',img) k = cv2.waitKey(20) if k == 27: break #释放资源cam.release()cv2.destroyAllWindows() 现在，你的电脑就能识别出你来啦！ 通过其他组合也可以实现开机检测等多种功能，你学会了吗？ 下面是博主审稿时的测试结果以及出现的一些问题哦~希望对大家有帮助（呲牙.jpg） 测试结果：

Questions that arise during the blogger's review test:

(1) Version issues

Solution: After numerous failures of the blogger, it is suggested that it is best to install python 2.7, you can directly use pip install numpy and pip install opencv-python to install numpy and opencv corresponding to python version.

(If you are using Anaconda2, pip commands can be entered in Anaconda Prompt under the Anaconda2 folder in the Start menu)

Click the link given in the tweet, download the file from github and put it under the folder where the compiled file is located, and change the relevant directory in the code.

(2) If the prompt "module' object has no attribute ' face'"

Solution: you can enter pip install opencv-contrib-python solution, if prompted to commission, you can add--user, that is, pip install opencv-contrib-python -user

Thank you for reading! About "how to use Python to achieve simple Face Recognition function" this article is shared here, I hope the above content can be of some help to everyone, so that everyone can learn more knowledge, if you think the article is good, you can share it to let more people see it!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.