import cv2
import os

# 數(shù)據(jù)預(yù)處理示例：調(diào)整圖像尺寸
def resize_images(input_folder, output_folder, target_size):
    if not os.path.exists(output_folder):
        os.makedirs(output_folder)
    
    for filename in os.listdir(input_folder):
        img_path = os.path.join(input_folder, filename)
        img = cv2.imread(img_path)
        img_resized = cv2.resize(img, target_size)
        output_path = os.path.join(output_folder, filename)
        cv2.imwrite(output_path, img_resized)

# 示例調(diào)用
input_folder = 'raw_data'
output_folder = 'preprocessed_data'
target_size = (224, 224)
resize_images(input_folder, output_folder, target_size)

import cv2
from skimage.feature import hog
from skimage import exposure

# 使用HOG特征提取
def extract_hog_features(img):
    # 轉(zhuǎn)換為灰度圖像
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    
    # 計(jì)算HOG特征
    fd, hog_image = hog(gray, orientations=9, pixels_per_cell=(8, 8),
                        cells_per_block=(2, 2), visualize=True, multichannel=False)
    
    # 可選：對(duì)HOG圖像進(jìn)行直方圖均衡化
    hog_image_rescaled = exposure.rescale_intensity(hog_image, in_range=(0, 10))

    return fd, hog_image_rescaled

# 示例調(diào)用
img_path = 'preprocessed_data/example.jpg'
img = cv2.imread(img_path)
features, hog_image = extract_hog_features(img)

# 可視化HOG圖像
cv2.imshow('HOG Image', hog_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

import numpy as np
from sklearn.model_selection import train_test_split
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense

# 加載數(shù)據(jù)集
X = np.load('preprocessed_data/images.npy')  # 銀行卡圖像數(shù)據(jù)
y = np.load('preprocessed_data/labels.npy')  # 對(duì)應(yīng)標(biāo)簽

# 數(shù)據(jù)集劃分
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, random_state=42)

# 構(gòu)建模型
model = Sequential()
model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D((2, 2)))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D((2, 2)))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(Flatten())
model.add(Dense(64, activation='relu'))
model.add(Dense(10, activation='softmax'))

# 編譯模型
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

# 訓(xùn)練模型
model.fit(X_train, y_train, epochs=10, validation_data=(X_val, y_val))

# 模型保存
model.save('bank_card_recognition_model.h5')

from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from keras.models import load_model

# 加載測(cè)試集數(shù)據(jù)
X_test = np.load('preprocessed_data/test_images.npy')
y_test = np.load('preprocessed_data/test_labels.npy')

# 加載訓(xùn)練好的模型
model = load_model('bank_card_recognition_model.h5')

# 模型評(píng)估
y_pred = model.predict(X_test)
y_pred_classes = np.argmax(y_pred, axis=1)

accuracy = accuracy_score(y_test, y_pred_classes)
precision = precision_score(y_test, y_pred_classes, average='macro')
recall = recall_score(y_test, y_pred_classes, average='macro')
f1 = f1_score(y_test, y_pred_classes, average='macro')

print(f'Accuracy: {accuracy}')
print(f'Precision: {precision}')
print(f'Recall: {recall}')
print(f'F1 Score: {f1}')

# 性能優(yōu)化（示例：調(diào)參）
# 可以通過調(diào)整模型超參數(shù)、增加數(shù)據(jù)樣本、使用更復(fù)雜的模型等方法來優(yōu)化模型性能

import cv2
import numpy as np
from keras.models import load_model

# 加載訓(xùn)練好的模型
model = load_model('bank_card_recognition_model.h5')

# 加載測(cè)試圖像
img = cv2.imread('test_image.jpg')

# 圖像預(yù)處理
img = cv2.resize(img, (224, 224))
img = np.expand_dims(img, axis=0)

# 預(yù)測(cè)銀行卡類型
pred = model.predict(img)
card_type = np.argmax(pred)

print(f'Card Type: {card_type}')