python3_批量读取指定文件夹下的图像_图像预处理（切割、降维、特征提取）+ 决策树应用实例_python3批量读取图片文件

1.调用的库

# PIL - Image: python呈像库
from PIL import Image
import numpy as np
import os
import re

2.主函数思路

def myMain():
 
    # 1.获取图片名称
    path = "./images/images/"
    imgNames = getName(path)
 
    # 2.加载图片
    data, labels = load_Img(path, imgNames)
 
    np.save("water_data.npy", data)
    np.save("water_labels.npy", labels)

3.模块函数getName(path)

def getName(path):
    """
    批量读取图片名称
    :param imgDir: 图片所在路径
    :param imgName: 图片名称
    :return: 图片名称列表
    """
    filenames = os.listdir(path)
 
    # Return a list of all non-overlapping matches in the string
    # def findall(pattern, string, flags=0)
    # pattern:待查找文件名称
    # string:查找位置
    imgNames = []
 
    for i in filenames:
        name = re.findall(r"\b\d_\d+\.jpg", i)
        imgNames.extend(name)  # 直接添加元素
    return imgNames

4.模块函数imgToArray(r, g, b)

def imgToArray(r, g, b):
    """
    将图片转换为数组
    :param r: 红
    :param g: 绿
    :param b: 蓝
    :return: 数组
    """
    return [np.asarray(r), np.asarray(g), np.asarray(b)]

5.求图片的各阶颜色距Var(rd=None, gd=None, bd=None)

def Var(rd=None, gd=None, bd=None):
    """
    求图片的颜色距
    :param rd: 三基色
    :return: 三阶颜色距
    """
    r1 = np.mean(rd)  # 一阶颜色距
    r2 = np.mean(gd)  # 一阶颜色距
    r3 = np.mean(bd)  # 一阶颜色距
    s4 = np.std(rd)  # 二阶颜色距
    s5 = np.std(gd)  # 二阶颜色距
    s6 = np.std(bd)  # 二阶颜色距
    t1 = np.mean((rd - rd.mean())**3)
    t2 = np.mean((gd - gd.mean())**3)
    t3 = np.mean((bd - bd.mean())**3)
    return [r1, r2, r3, s4, s5, s6, np.sign(np.mean(t1)*abs(t1)**(1/3)), np.sign(np.mean(t2)*abs(t2)**(1/3)), np.sign(np.mean(t3)*abs(t3)**(1/3))]

6.模块函数load_Img(path, fileNameList)

def load_Img(path, fileNameList):
    """
    加载图片，并预处理
    :param fileNameList:
    :return:
    """
    n = len(fileNameList)      # 获取图片数量
    data = np.zeros([n, 9])    # 样本自变量
    labels = np.zeros([n])  # 样本标签
 
    # 遍历每一张图片
    imgList = []
    for i in range(n):
        img = Image.open(path+fileNameList[i])
        M, N = img.size
        img = img.crop((M/2-50,N/2-50,M/2+50,N/2+50))  # 选取图片中心位置
 
        r, g, b = img.split()                          # 将颜色通道分开
 
        img_list = imgToArray(r, g, b)                 # 将图片转换为数组
        for j in range(9):
            data[i, j] = Var(img_list[0], img_list[1], img_list[2])[j]
        labels[i] = fileNameList[i][0]                 # 存储图片标签
 
    return data, labels

6.应用实例（决策树）

import numpy as np
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import train_test_split
from sklearn import metrics
 
 
data = np.load("water_data.npy")
labels = np.load("water_labels.npy")
 
data_tr, data_te, target_tr, target_te = train_test_split(data, labels, test_size=0.2, random_state=10)
 
model = DecisionTreeClassifier(random_state=10).fit(data_tr, target_tr)
 
res = model.predict(data_te)
 
pre = model.score(data_te, target_te)
 
print(target_te)
print(res)
 
# 精度
print(sum(res==target_te)/len(res))
print(pre)
 
# 报告
print(metrics.classification_report(target_te, res))
print(metrics.confusion_matrix(target_te, res))