# -*- coding: utf-8 -*-
 import os
 import cv2
 import numpyas np
 from sklearnimport preprocessing
 # 为图像做标记，并将文字转化为数字，再训练
 class LabelEncoder(object):
     # 编码：文字到数字
     def encode_labels(self, label_words):
         self.le =  preprocessing.LabelEncoder()
         self.le.fit(label_words)
     # 数字转换成文字
      def word_to_num(self, label_word):
         return int(self.le.transform([label_word])[0])
     # 数学到文字的转换
     def num_to_word(self, label_num):
         return self.le.inverse_transform([label_num])[0]
 # 根据路径获取图片
 def get_images_and_labels(input_path):
     label_words = []
     # 循环读取所有图片
     for root, dirs, files in os.walk(input_path):
         for filename in (x for x in files if x.endswith('.jpg')):
             filepath =  os.path.join(root, filename)
             label_words.append(filepath.split('\\')[-2])
     # 编码
     images = []
     le = LabelEncoder()
     le.encode_labels(label_words)
     labels = []
     # 解析目录结构
     for root, dirs, files in os.walk(input_path):
         for filename in (x for x in files if x.endswith('.jpg')):
             filepath =  os.path.join(root, filename)
             # 读入灰度图
             image = cv2.imread(filepath, 0)
             # 获取标记
             name = filepath.split('\\')[-2]
             # 检测是否有人脸，并获得人脸数据
             faces = faceCascade.detectMultiScale(image, 1.1, 2, minSize=(100,100))
             # 输入每个人脸
             for (x, y, w, h) in faces:
                  images.append(image[y:y+h, x:x+w])
                  labels.append(le.word_to_num(name))
     return images, labels,  le
 if __name__=='__main__':
     cascade_path = "cascade_files/haarcascade_frontalface_alt.xml"
     path_train = 'faces_dataset/train'
     path_test = 'faces_dataset/test'
     # 人脸检测训练结果读取
     faceCascade = cv2.CascadeClassifier(cascade_path)
     # 人脸识别方法初始化
     recognizer =  cv2.face.LBPHFaceRecognizer_create()
     # 获取训练数据集
     images, labels, le  = get_images_and_labels(path_train)
     # 模型训练
     print("\n使用训练集对模型进行训练...")
     print(labels)
     recognizer.train(images,  np.array(labels))
     # 识别测试数据集
     print('\n识别图像中的人脸...')
     stop_flag = False
     for root, dirs, files in os.walk(path_test):
         for filename in (x for x in files if x.endswith('.jpg')):
             filepath =  os.path.join(root, filename)
             predict_image = cv2.imread(filepath,0)
             # 人脸检测
             faces = faceCascade.detectMultiScale(predict_image, 1.1,
                     2, minSize=(100,100))
             # 人脸识别
             for (x, y, w, h) in faces:
                 # 识别
                 predicted_index,  conf = recognizer.predict(
                          predict_image[y:y+h, x:x+w])
                 # 文字到数字的转换
                 predicted_person  = le.num_to_word(predicted_index)
                 # 显示结果（彩色）
                 predict_image =  cv2.imread(filepath)
                 cv2.putText(predict_image,  predicted_person,
                         (10,60),  cv2.FONT_HERSHEY_SIMPLEX, 2, (0,255,255), 6)
                 cv2.imshow("result", predict_image)
                 cv2.waitKey(0)
                 stop_flag = True
                 break
         if stop_flag:
             break

# 模型训练
 print("\n使用训练集对模型进行训练...")
 recognizer.train(images, np.array(labels))
 recognizer.save('my_LBPHFaceRecognizer.xml')

 print('\n读取模型...')
 recognizer.read('my_LBPHFaceRecognizer.xml')

cv2.imwrite("result-"+predicted_person+'.jpg', predict_image) 

 predict_image = cv2.imread(filepath)
 cv2.rectangle(predict_image, (x, y), (x + w, y + h), (0, 0, 255), 2)
 cv2.rectangle(predict_image, (x, y + h - 35), (x + w, y +  h), (0, 0, 255), 2)
 cv2.putText(predict_image, predicted_person,
         (x + 6,y + h - 6),cv2.FONT_HERSHEY_DUPLEX,  1.5, (255,255,255), 2)
 cv2.imshow("result", predict_image)
 cv2.imwrite("result-" + predicted_person + '.jpg', predict_image)