[toc]

塑料瓶图像检测

目的：用于判断一张图片是否为塑料瓶；

条件：总计300多张图片分为70多类，同一类塑料瓶分别放置在同一个文件夹；

思路：选取每个文件夹的一张图片与目标图片对比计算返回相似度最高的值，通过多次试验确定阈值，超过阈值则判定为是塑料瓶，否则不是；

目录：

@[toc]

说明

• 所有测试结果均为程序第一次运行结果

• 每一组测试的第一个（第一个样本）都是同一张图片相互比对

• 每一组测试的第二三四五个测试样本均是与第一个测试样本比对

• 以下为选用的5个测试样本

• 图片放置在与程序同一位置的images文件夹

<center><img src="https://auto2dev.coding.net/p/ImageHostingService/d/ImageHostingService/git/raw/master/md/20200114104340189-1592736787820.jpg" width="40%"/></center>
<center><img src="https://auto2dev.coding.net/p/ImageHostingService/d/ImageHostingService/git/raw/master/md/20200114104446378-1592736787433.jpg" width="40%"/></center>

<center><img src="https://auto2dev.coding.net/p/ImageHostingService/d/ImageHostingService/git/raw/master/md/202001141044484-1592736787888.jpg" width="40%"/></center>

<center><img src="https://auto2dev.coding.net/p/ImageHostingService/d/ImageHostingService/git/raw/master/md/20200114104502540-1592736787890.jpg" width="40%"/></center>

<center><img src="https://auto2dev.coding.net/p/ImageHostingService/d/ImageHostingService/git/raw/master/md/20200114104509945-1592736787431.jpg" width="40%"/></center>

1.方法

• 直方图

• 互信息

• 余弦相似度

• 感知哈希算法

2.测试结果（单张图片比对）

直方图

from time import *
begin_time = time()

from PIL import Image


def make_regalur_image(img, size=(256, 256)):
    return img.resize(size).convert('RGB')


def hist_similar(lh, rh):
    assert len(lh) == len(rh)
    return sum(1 - (0 if l == r else float(abs(l - r)) / max(l, r)) for l, r in zip(lh, rh)) / len(lh)


def calc_similar(li, ri):
    return hist_similar(li.histogram(), ri.histogram())


if __name__ == '__main__':
    img1 = Image.open('images/WIN_20200111_21_56_10_Pro.jpg')
    img1 = make_regalur_image(img1)
    img2 = Image.open('images/WIN_20200111_21_56_52_Pro.jpg')
    img2 = make_regalur_image(img2)
    print(calc_similar(img1, img2))

end_time = time()
run_time = end_time-begin_time
print ('该程序运行时间：',run_time)

互信息

from time import *
begin_time = time()

from sklearn import metrics as mr
from scipy.misc import imread
import numpy as np
 
img1 = imread('1.jpg')
img2 = imread('2.jpg')
 
img2 = np.resize(img2, (img1.shape[0], img1.shape[1], img1.shape[2]))
 
img1 = np.reshape(img1, -1)
img2 = np.reshape(img2, -1)
print(img2.shape)
print(img1.shape)
mutual_infor = mr.mutual_info_score(img1, img2)
 
print(mutual_infor)

end_time = time()
run_time = end_time-begin_time
print ('该程序运行时间：',run_time)

余弦相似度

from time import *
begin_time = time()

from PIL import Image
from numpy import average, linalg, dot

def get_thumbnail(image, size=(1200, 750), greyscale=False):
    image = image.resize(size, Image.ANTIALIAS)
    if greyscale:
        image = image.convert('L')
    return image


def image_similarity_vectors_via_numpy(image1, image2):
    image1 = get_thumbnail(image1)
    image2 = get_thumbnail(image2)
    images = [image1, image2]
    vectors = []
    norms = []
    for image in images:
        vector = []
        for pixel_tuple in image.getdata():
            vector.append(average(pixel_tuple))
        vectors.append(vector)
        norms.append(linalg.norm(vector, 2))
    a, b = vectors
    a_norm, b_norm = norms
    res = dot(a / a_norm, b / b_norm)
    return res


image1 = Image.open('images/WIN_20200111_21_56_10_Pro.jpg')
image2 = Image.open('images/WIN_20200111_21_56_10_Pro.jpg')
cosin = image_similarity_vectors_via_numpy(image1, image2)

print(cosin)

end_time = time()
run_time = end_time-begin_time
print ('该程序运行时间：',run_time)

感知哈希算法

from time import *
begin_time = time()

import cv2
import numpy as np
import os

#感知哈希算法
def pHash(image):
    image = cv2.resize(image,(32,32), interpolation=cv2.INTER_CUBIC)
    image = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
#     cv2.imshow('image', image)
#     cv2.waitKey(0)
#     cv2.destroyAllWindows()
    # 将灰度图转为浮点型，再进行dct变换
    dct = cv2.dct(np.float32(image))
#     print(dct)
    # 取左上角的8*8，这些代表图片的最低频率
    # 这个操作等价于c++中利用opencv实现的掩码操作
    # 在python中进行掩码操作，可以直接这样取出图像矩阵的某一部分
    dct_roi = dct[0:8,0:8]
    avreage = np.mean(dct_roi)
    hash = []
    for i in range(dct_roi.shape[0]):
        for j in range(dct_roi.shape[1]):
            if dct_roi[i,j] > avreage:
                hash.append(1)
            else:
                hash.append(0)
    return hash

#计算汉明距离
def Hamming_distance(hash1,hash2):
    num = 0
    for index in range(len(hash1)):
        if hash1[index] != hash2[index]:
            num += 1
    return num

if __name__ == "__main__":

    image_file1 = 'images/WIN_20200111_21_56_10_Pro.jpg'
    image_file2 = 'images/1.jpg'

    img1 = cv2.imread(image_file1)
    img2 = cv2.imread(image_file2)
    hash1 = pHash(img1)
    hash2 = pHash(img2)
    dist = Hamming_distance(hash1, hash2)
    #将距离转化为相似度
    similarity = 1 - dist * 1.0 / 64
    print(dist)
    print(similarity)

end_time = time()

run_time = end_time-begin_time
print ('该程序运行时间：',run_time)

3.评价

• 直方图计算结果与直观视觉严重不符合
• 余弦相似度准确度较高，但太耗时，比对平均耗时19s
• 互信息的方法从耗时和准确度上粗略观察，介于直方图和余弦相似度之间
• 感知哈希算法耗时较为可接受，且比对结果较有区分度且符合直观视觉