圖形辨識筆記-OPEN CV (haarcascades )

Haar Cascade classifier

• OpenCV 官方 Github：https://github.com/opencv/opencv/tree/4.x/data
• 人臉特徵模型：haarcascade_frontalface_default.xml

pip install opencv-python

偵測人臉

import cv2

def detectFace(img):

    filename = img.split(".")[0] # 取得檔案名稱(不添加副檔名)

    img = cv2.imread(img) # 讀取圖檔

    grayImg = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 透過轉換函式轉為灰階影像

    color = (0, 255, 0)  # 定義框的顏色

    # OpenCV 人臉識別分類器

    face_classifier = cv2.CascadeClassifier(cv2.data.haarcascades + "haarcascade_frontalface_default.xml")

    # 調用偵測識別人臉函式

    faceRects = face_classifier.detectMultiScale(

        grayImg, scaleFactor=1.2, minNeighbors=3, minSize=(32, 32))

    # 大於 0 則檢測到人臉

    if len(faceRects):  

        # 框出每一張人臉

        for faceRect in faceRects: 

            x, y, w, h = faceRect

            cv2.rectangle(img, (x, y), (x + h, y + w), color, 2)

    # 將結果圖片輸出

    cv2.imwrite(filename + "_face.jpg", img)

detectFace('123.jpg')

import cv2

img = cv2.imread('a.jpg')

gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)   # 將圖片轉成灰階

face_cascade = cv2.CascadeClassifier("haarcascade_frontalface_default.xml")   # 載入人臉模型

faces = face_cascade.detectMultiScale(gray)    # 偵測人臉

for (x, y, w, h) in faces:

    cv2.rectangle(img, (x, y), (x+w, y+h), (0, 255, 0), 2)    # 利用 for 迴圈，抓取每個人臉屬性，繪製方框

cv2.imshow('test', img)

cv2.waitKey(0) # 按下任意鍵停止

cv2.destroyAllWindows()

偵測多人臉

import cv2

name=input("輸入圖片檔案:")

def detect(filename):

    face_cascade = cv2.CascadeClassifier("haarcascade_frontalface_default.xml")

    img = cv2.imread(filename)

    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    #傳遞引數是scaleFactor和minNeighbors,分別表示人臉檢測過程中每次迭代時影象的壓縮率以及每個人臉矩形保留近鄰數目的最小值

    #檢測結果返回人臉矩形陣列

    face = face_cascade.detectMultiScale(gray)

    for (x, y, w, h) in face:

        img = cv2.rectangle(img, (x, y), (x + w, y + h), (255, 0, 0), 2)

    cv2.namedWindow("Face Detected!")

    cv2.imshow("Face Detected!", img)

    cv2.imwrite("./images/Face.jpg", img)

    cv2.waitKey(0)

    cv2.destroyAllWindows()

detect(name)

即時偵測

import cv2

cap = cv2.VideoCapture(0)

face_cascade = cv2.CascadeClassifier("haarcascade_frontalface_default.xml")

#faces = face_cascade.detectMultiScale(gray)

if not cap.isOpened():

    print("Cannot open camera")

    exit()

while True:

    ret, frame = cap.read()

    if not ret:

        print("Cannot receive frame")

        break

    frame = cv2.resize(frame,(540,320))              # 縮小尺寸，避免尺寸過大導致效能不好

    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)   # 將鏡頭影像轉換成灰階

    faces = face_cascade.detectMultiScale(gray)      # 偵測人臉

    for (x, y, w, h) in faces:

        cv2.rectangle(frame, (x, y), (x+w, y+h), (0, 255, 0), 2)   # 標記人臉

    cv2.imshow('oxxostudio', frame)

    if cv2.waitKey(1) == ord('q'):

        break

cap.release()

cv2.destroyAllWindows()

模型訓練

import cv2

import numpy as np

detector = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')  # 載入人臉追蹤模型

recog = cv2.face.LBPHFaceRecognizer_create()      # 啟用訓練人臉模型方法

faces = []   # 儲存人臉位置大小的串列

ids = []     # 記錄該人臉 id 的串列

for i in range(1,5):

    img = cv2.imread(f'D:\\test\\face01\\{i}.jpg')           # 依序開啟每一張第一位的照片

    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)  # 色彩轉換成黑白

    img_np = np.array(gray,'uint8')               # 轉換成指定編碼的 numpy 陣列

    face = detector.detectMultiScale(gray)        # 擷取人臉區域

    for(x,y,w,h) in face:

        faces.append(img_np[y:y+h,x:x+w])         # 記錄第一位人臉的位置和大小內像素的數值

        ids.append(1)                             # 記錄第一位人臉對應的 id，只能是整數，都是 1 表示第一位的 id 為 1

for i in range(1,2):

    img = cv2.imread(f'D:\\test\\face02\\{i}.jpg')           # 依序開啟每一張第二位的照片

    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)  # 色彩轉換成黑白

    img_np = np.array(gray,'uint8')               # 轉換成指定編碼的 numpy 陣列

    face = detector.detectMultiScale(gray)        # 擷取人臉區域

    for(x,y,w,h) in face:

        faces.append(img_np[y:y+h,x:x+w])         # 記錄第二位人臉的位置和大小內像素的數值

        ids.append(2)                             # 記錄第二位人臉對應的 id，只能是整數，都是 1 表示第二位的 id 為 1

print('camera...')                                # 提示啟用相機

cap = cv2.VideoCapture(0)                         # 啟用相機

if not cap.isOpened():

    print("Cannot open camera")

    exit()

while True:

    ret, img = cap.read()                         # 讀取影片的每一幀

    if not ret:

        print("Cannot receive frame")

        break

    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)  # 色彩轉換成黑白

    img_np = np.array(gray,'uint8')               # 轉換成指定編碼的 numpy 陣列

    face = detector.detectMultiScale(gray)        # 擷取人臉區域

    for(x,y,w,h) in face:

        faces.append(img_np[y:y+h,x:x+w])         # 記錄自己人臉的位置和大小內像素的數值

        ids.append(3)                             # 記錄自己人臉對應的 id，只能是整數，都是 1 表示川普的 id

    cv2.imshow('TEST', img)                      # 顯示攝影機畫面

    if cv2.waitKey(100) == ord('q'):              # 每一毫秒更新一次，直到按下 q 結束

        break

print('training...')                              # 提示開始訓練

recog.train(faces,np.array(ids))                  # 開始訓練

recog.save('face.yml')                            # 訓練完成儲存為 face.yml

print('ok!')

透過訓練的模型即時辨識

import cv2

recognizer = cv2.face.LBPHFaceRecognizer_create()         # 啟用訓練人臉模型方法

recognizer.read('face.yml')                               # 讀取人臉模型檔

cascade_path = "haarcascade_frontalface_default.xml"  # 載入人臉追蹤模型

face_cascade = cv2.CascadeClassifier(cascade_path)        # 啟用人臉追蹤

cap = cv2.VideoCapture(0)                                 # 開啟攝影機

if not cap.isOpened():

    print("Cannot open camera")

    exit()

while True:

    ret, img = cap.read()

    if not ret:

        print("Cannot receive frame")

        break

    img = cv2.resize(img,(540,300))              # 縮小尺寸，加快辨識效率

    gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)  # 轉換成黑白

    faces = face_cascade.detectMultiScale(gray)  # 追蹤人臉 ( 目的在於標記出外框 )

    # 建立姓名和 id 的對照表

    name = {

        '1':'One',

        '2':'Two',

        '3':'Joker'

    }

    # 依序判斷每張臉屬於哪個 id

    for(x,y,w,h) in faces:

        cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)            # 標記人臉外框

        idnum,confidence = recognizer.predict(gray[y:y+h,x:x+w])  # 取出 id 號碼以及信心指數 confidence

        if confidence < 60:

            text = name[str(idnum)]                               # 如果信心指數小於 60，取得對應的名字

        else:

            text = '???'                                          # 不然名字就是 ???

        # 在人臉外框旁加上名字

        cv2.putText(img, text, (x,y-5),cv2.FONT_HERSHEY_SIMPLEX, 1, (0,255,0), 2, cv2.LINE_AA)

    cv2.imshow('oxxostudio', img)

    if cv2.waitKey(5) == ord('q'):

        break    # 按下 q 鍵停止

cap.release()

cv2.destroyAllWindows()