区域识别——基于python语言

1.水域识别

2.模型介绍

3.文件框架

4.代码示例

4.1 data_preprocess.py

4.2 model1.0.py

4.3 train2.0.py

4.4 predict.py

4.5 运行结果

5.总结

1.水域识别

人眼看见河道可以直接分辨出这是河道，但是如何让计算机也能识别出这是河道呢？

这里就用到了深度学习之图像处理技术。

想要识别河道中的河水，对拍摄的河道图片进行编辑，可以创建两个图片文件的集合，一个文件中存放具体的河道图片，另一个存放带有标注的河道水域的分割掩码（标签），然后进行模型训练，训练一个可以识别河道水域的深度学习模型。

原始图像

掩码图像

2.模型介绍

使用深度学习框架TensorFlow和Keras进行图像分割任务。

模型的训练首先要进行数据处理，然后再进行模型训练。

3.文件框架

data2.0中的三个数据文件夹分别是训练集，验证集和测试集的图片。

data_preprocess.py是图片预处理代码。

model1.0.py是定义了一个基于U-Net架构的卷积神经网络模型，用来做图像分割任务。

train2.0.py是训练程序，最重要的是要运行这个。

predict.py是调用训练后的模型的代码。

4.代码示例

4.1 data_preprocess.py

import os
import cv2
import numpy as np
from tensorflow.keras.preprocessing.image import img_to_array, load_img

# 图像大小和通道数
IMG_HEIGHT = 256
IMG_WIDTH = 256
IMG_CHANNELS = 3

# 路径设置
image_dir = './River Water Segmentation/data2.0/train/images'
mask_dir = './River Water Segmentation/data2.0/train/masks'

def preprocess_images(image_dir, target_size=(IMG_HEIGHT, IMG_WIDTH)):
    images = []
    for filename in os.listdir(image_dir):
        img_path = os.path.join(image_dir, filename)
        img = load_img(img_path, target_size=target_size)
        img_array = img_to_array(img)
        images.append(img_array)
    return np.array(images)

def preprocess_masks(mask_dir, target_size=(IMG_HEIGHT, IMG_WIDTH)):
    masks = []
    for filename in os.listdir(mask_dir):
        mask_path = os.path.join(mask_dir, filename)
        mask = load_img(mask_path, target_size=target_size, color_mode="grayscale")
        mask_array = img_to_array(mask) / 255.0  # 将掩码值标准化为0和1
        masks.append(mask_array)
    return np.array(masks)

if __name__ == "__main__":
    images = preprocess_images(image_dir)
    masks = preprocess_masks(mask_dir)
    print(f"Processed {len(images)} images and {len(masks)} masks.")

4.2 model1.0.py

import tensorflow as tf
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Conv2DTranspose, concatenate, Input, Dropout
from tensorflow.keras.models import Model

def build_unet(input_shape):
    inputs = Input(input_shape)
    
    # 编码器部分 (下采样)
    c1 = Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(inputs)
    c1 = Dropout(0.1)(c1)
    c1 = Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c1)
    p1 = MaxPooling2D((2, 2))(c1)

    c2 = Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(p1)
    c2 = Dropout(0.1)(c2)
    c2 = Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c2)
    p2 = MaxPooling2D((2, 2))(c2)

    c3 = Conv2D(256, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(p2)
    c3 = Dropout(0.2)(c3)
    c3 = Conv2D(256, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c3)
    p3 = MaxPooling2D((2, 2))(c3)

    c4 = Conv2D(512, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(p3)
    c4 = Dropout(0.3)(c4)
    c4 = Conv2D(512, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c4)
    p4 = MaxPooling2D((2, 2))(c4)

    # 底层
    c5 = Conv2D(1024, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(p4)
    c5 = Dropout(0.4)(c5)
    c5 = Conv2D(1024, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c5)

    # 解码器部分 (上采样)
    u6 = Conv2DTranspose(512, (2, 2), strides=(2, 2), padding='same')(c5)
    u6 = concatenate([u6, c4])
    c6 = Conv2D(512, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(u6)
    c6 = Dropout(0.3)(c6)
    c6 = Conv2D(512, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c6)

    u7 = Conv2DTranspose(256, (2, 2), strides=(2, 2), padding='same')(c6)
    u7 = concatenate([u7, c3])
    c7 = Conv2D(256, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(u7)
    c7 = Dropout(0.2)(c7)
    c7 = Conv2D(256, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c7)

    u8 = Conv2DTranspose(128, (2, 2), strides=(2, 2), padding='same')(c7)
    u8 = concatenate([u8, c2])
    c8 = Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(u8)
    c8 = Dropout(0.1)(c8)
    c8 = Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c8)

    u9 = Conv2DTranspose(64, (2, 2), strides=(2, 2), padding='same')(c8)
    u9 = concatenate([u9, c1], axis=3)
    c9 = Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(u9)
    c9 = Dropout(0.1)(c9)
    c9 = Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_normal', padding='same')(c9)

    outputs = Conv2D(1, (1, 1), activation='sigmoid')(c9)  # 单通道输出用于二分类分割

    model = Model(inputs=[inputs], outputs=[outputs])
    
    return model

if __name__ == "__main__":
    # 输入形状 (256, 256, 3) 用于 RGB 图像
    model = build_unet((256, 256, 3))
    model.summary()

4.3 train2.0.py

import numpy as np
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from model import build_unet
from data_preprocess import preprocess_images, preprocess_masks

# 设置路径
train_images_path = './River Water Segmentation/data2.0/train/images'
train_masks_path = './River Water Segmentation/data2.0/train/masks'
val_images_path = './River Water Segmentation/data2.0/train/images'
val_masks_path = './River Water Segmentation/data2.0/train/masks'

# 加载数据
train_images = preprocess_images(train_images_path)
train_masks = preprocess_masks(train_masks_path)
val_images = preprocess_images(val_images_path)
val_masks = preprocess_masks(val_masks_path)

# 数据增强
data_gen_args = dict(rotation_range=0.2,
                     width_shift_range=0.05,
                     height_shift_range=0.05,
                     shear_range=0.05,
                     zoom_range=0.05,
                     horizontal_flip=True,
                     fill_mode='nearest')

image_datagen = ImageDataGenerator(**data_gen_args)
mask_datagen = ImageDataGenerator(**data_gen_args)

image_datagen.fit(train_images)
mask_datagen.fit(train_masks)


# 编译模型
model.compile(optimizer=Adam(learning_rate=1e-4), loss='binary_crossentropy', metrics=['accuracy'])

# 设置回调函数
checkpoint = ModelCheckpoint('models/unet_model.h5', save_best_only=True, monitor='val_loss', mode='min')
early_stopping = EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True)
reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2, patience=5, min_lr=1e-6)

# 训练模型
history = model.fit(image_datagen.flow(train_images, train_masks, batch_size=16),
                    validation_data=(val_images, val_masks), 
                    epochs=50, 
                    callbacks=[checkpoint, early_stopping, reduce_lr])

# 保存最终模型
model.save('./River Water Segmentation/models/6.0model.h5')

4.4 predict.py

import numpy as np
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
import matplotlib.pyplot as plt
import cv2

# 加载模型
model = load_model('./River Water Segmentation/models/unet_final_model.h5')

# 预测函数
def predict_image(image_path):
    img = image.load_img(image_path, target_size=(256, 256))
    img_array = image.img_to_array(img) / 255.0
    img_array = np.expand_dims(img_array, axis=0)
    
    # 预测掩码
    pred_mask = model.predict(img_array)
    pred_mask = (pred_mask > 0.5).astype(np.uint8)  # 二值化处理
    
    # 显示原图和预测掩码
    plt.figure(figsize=(10, 5))
    plt.subplot(1, 2, 1)
    plt.imshow(img)
    plt.title('Original')
    
    plt.subplot(1, 2, 2)
    plt.imshow(pred_mask[0, :, :, 0], cmap='gray')
    plt.title('Predicted')
    plt.show()

# 示例预测
predict_image('./River Water Segmentation/data/test/river/water_body_27.jpg')