3.[人脸识别] python-opencv 人脸特征采集与录入

目录

1.环境 

2.描述

3.代码

4.效果

---

1.环境 

1.python: 3.6.6 [64bit]
 
2.python packages:
    1). opencv-python==3.4.1.15
    2). opencv-contrib-python==4.4.0.42
    3). sklearn==0.0
    
3.python包下载安装出错，可以尝试一下方式：
    pip install -i https://pypi.tuna.tsinghua.edu.cn/simple --trusted-host pypi.tuna.tsinghua.edu.cn <+python包>
 
4.探测器 Haar级联文件下载：
    https://github.com/opencv/opencv/tree/master/data

2.描述

1.文件目录结构：
    project：
            ----image_photograph        # 人脸图片文件夹
                    |----lihua              # 名字叫lihua的多张人脸图片
                    |----xiaoming           # 名字叫xiaoming的多张人脸图片
                    |----zhangfei           # 名字叫zhangfei的多张人脸图片
            ----facefeatures            # 存储人脸特征的文件夹
            ----haarcascade             # 人脸探测器的haar级联文件
            ----main.py                 # 主程序

2.流程：
    (1).获取每一张图片路径，对应id名字(文件夹作为名字)
    (2).将每张图片的人脸部分采集出来
    (3).id名字进行标签编码处理LabelEncoder
    (4).局部二值模式直方图创建人脸识别器
    (5).训练人脸识别器
    (5).将所有人脸特征与名字信息写入到yml文件中
        注意：以后需要用到yml文件时，如果图片名字要与原来一致，还需要原图片名字作为标签解码用
              （代码最后return就是名字列表）

3.代码

"""处理类标签编码"""
class LabelEncoder(object):
    def __init__(self):
        self.le = preprocessing.LabelEncoder()

    def encode_labels(self, label_words):
        self.le.fit(label_words)

    def word_to_num(self, label_word):
        return int(self.le.transform([label_word])[0])

    def num_to_word(self, label_num):
        return self.le.inverse_transform([label_num])[0]


"""采集人脸信息生成yml"""
class FaceFeatures:
    def __init__(self, photo_path):
        self.face_cascade = cv2.CascadeClassifier('haarcascade\\haarcascade_frontalface_alt.xml')
        self.data_transformation = LabelEncoder()
        self.photo_path = photo_path
        self.information = None

    """获取所有图片路径，存储图片的父文件夹名字"""
    def get_images_information(self):
        images = []
        information = []
        for root, dirs, files in os.walk(self.photo_path):
            root = root.replace("/", "\\")
            for filename in [x for x in files if x.endswith(".jpg")]:
                images.append(os.path.join(os.getcwd(), root, filename))
                information.append(root.split('\\')[-1])
        self.data_transformation.encode_labels(information)
        self.information = information
        return images, information

    """采集所有人脸信息，名字标签编码处理"""
    def get_features(self, images, information):
        face_images_info = []
        face_name_info = []
        for num in range(len(images)):
            ret = cv2.imread(images[num], 0)
            faces = self.face_cascade.detectMultiScale(ret, 1.1, 2, minSize=(100, 100))
            for (x, y, w, h) in faces:
                face_images_info.append(ret[y:y + h, x:x + w])
             face_name_info.append(self.data_transformation.word_to_num(information[num]))
        return face_images_info, face_name_info

    """人脸信息录入yml文件, 返回图片对应的名字"""
    def info_enter(self):
        recognizer = cv2.face.LBPHFaceRecognizer_create()
        images, information = self.get_images_information()
        face_images_info, face_name_info = self.get_features(images, information)
        recognizer.train(face_images_info, np.array(face_name_info))
        recognizer.write("facefeatures/trainer.yml")
        return self.information


if __name__ == '__main__':
    FaceFeatures(os.path.join(os.getcwd(), "image_photograph")).info_enter()

4.效果

参考：Python-opencv实现摄像头实时进行人脸识别(点击)