類似度 - メモ

ensekitt.hatenablog.com

!pip install pdfplumber

"""再起動"""

!apt update

!wget "https://www.city.kumamoto.jp/common/UploadFileDsp.aspx?c_id=5&id=4645&sub_id=18&flid=239935" -O data.pdf

!apt install libmagickwand-dev ghostscript

# Commented out IPython magic to ensure Python compatibility.
# %%writefile /etc/ImageMagick-6/policy.xml
# <?xml version="1.0" encoding="UTF-8"?>
# <!DOCTYPE policymap>
# 
# <policymap>
# 
#   <policy domain="resource" name="memory" value="256MiB"/>
#   <policy domain="resource" name="map" value="512MiB"/>
#   <policy domain="resource" name="width" value="16KP"/>
#   <policy domain="resource" name="height" value="16KP"/>
#   <policy domain="resource" name="area" value="128MB"/>
#   <policy domain="resource" name="disk" value="1GiB"/>
# 
#   <policy domain="delegate" rights="none" pattern="URL"/>
#   <policy domain="delegate" rights="none" pattern="HTTPS"/>
#   <policy domain="delegate" rights="none" pattern="HTTP"/>
# 
#   <policy domain="path" rights="none" pattern="@*"/>
#   <policy domain="cache" name="shared-secret" value="passphrase" stealth="true"/>
# 
#   <policy domain="coder" rights="none" pattern="PS"/>
#   <policy domain="coder" rights="none" pattern="PS2"/>
#   <policy domain="coder" rights="none" pattern="PS3"/>
#   <policy domain="coder" rights="none" pattern="EPS"/>
#   <policy domain="coder" rights="read|write" pattern="PDF" />
#   <policy domain="coder" rights="none" pattern="XPS"/>
# </policymap>

import pdfplumber
import pandas as pd

pdf = pdfplumber.open("data.pdf")

page = pdf.pages[1]

bboxs = [
    [49, 172, 418, 509],
    [455, 172, 824, 509],
    [49, 712, 418, 1049],
    [455, 712, 824, 1049],
]

n = 1

bbox = bboxs[n]

# cropでテキスト取得
crop = page.within_bbox(bbox)

# PDF確認
im = crop.to_image(resolution=300)
im

im.save("pdf.png", format="PNG")

"""# 画像認識"""

import cv2
import numpy as np

from google.colab.patches import cv2_imshow

from matplotlib import pyplot as plt

img = cv2.imread("pdf.png")

cv2_imshow(img)

# 行数
rows = 5

# 列数
cols = 6

chunks = []

for row_img in np.array_split(img, rows, axis=0):
    for chunk in np.array_split(row_img, cols, axis=1):
        chunks.append(chunk[5:-5, 5:-5, :])

print(len(chunks))

from pathlib import Path

output_dir = Path("output")
output_dir.mkdir(exist_ok=True)

for i, chunk in enumerate(chunks):
    save_path = output_dir / f"chunk_{i:02d}.png"
    cv2.imwrite(str(save_path), chunk)

!zip -r output.zip output

"""# 画像確認"""

day1, day2 = 6, 24

chunks[day1].shape

img1 = chunks[day1][:, 90: ,:]

img1g = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)

cv2_imshow(img1)

chunks[day2].shape

img2 = chunks[day2][:, 90: ,:]

img2g = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)

cv2_imshow(img2)

"""# ヒストグラム比較

+ https://ensekitt.hatenablog.com/entry/2018/07/09/200000
"""

hist_g_1 = cv2.calcHist([img1], [0], None, [256], [0, 256])

hist_g_2 = cv2.calcHist([img2], [0], None, [256], [0, 256])

comp_hist = cv2.compareHist(hist_g_1, hist_g_2, cv2.HISTCMP_CORREL)

comp_hist

def img2hist(img):

    histrgb = []

    color = ("b", "g", "r")

    for i, col in enumerate(color):
        histrgb.append(cv2.calcHist([img], [i], None, [256], [0, 256]))
        plt.plot(histrgb[i], color=col)
        plt.xlim([0, 256])
        
    plt.show()

    return histrgb

histrgb_1 = img2hist(img1)

histrgb_2 = img2hist(img2)

histarray = np.array(histrgb_1)
histvec_1 = histarray.reshape(histarray.shape[0] * histarray.shape[1], 1)

histvec_1.shape

histarray = np.array(histrgb_2)
histvec_2 = histarray.reshape(histarray.shape[0] * histarray.shape[1], 1)

histvec_2.shape

print(cv2.compareHist(histvec_1, histvec_2, 0))

"""# akaze"""

akaze = cv2.AKAZE_create()

kp1, des1 = akaze.detectAndCompute(img1, None)
kp2, des2 = akaze.detectAndCompute(img2, None)

# 特徴量のマッチングを実行
bf = cv2.BFMatcher() # 総当たりマッチング(Brute-Force Matcher)生成

# 特徴量ベクトル同士をBrute-ForceとKNN(kth-nearest neighbor)でマッチング
matches = bf.knnMatch(des1, des2, k=2)

ratio = 0.5
good = []
for m, n in matches:
    if m.distance < ratio * n.distance:
        good.append([m])

# 対応する特徴点同士を描画
img_3 = cv2.drawMatchesKnn(img1, kp1, img2, kp2, good, None, flags=2)

# 画像表示
cv2_imshow(img_3)

len(good)

"""# 類似度

https://zenn.dev/kazuhito/articles/1dc73eeb7e1297
"""

# Average Hash（入力画像の平均ハッシュ値を計算）
hash_func = cv2.img_hash.AverageHash_create()

# Block Mean Hash（ブロック平均に基づいて画像のハッシュを計算）
hash_func = cv2.img_hash.BlockMeanHash_create()

# Color Moment Hash（カラーモーメントに基づく画像ハッシュを計算）
hash_func = cv2.img_hash.ColorMomentHash_create()

# Marr Hildreth Hash（Marr-Hildreth Operator Based Hash，最も遅いが識別性が高い）
hash_func = cv2.img_hash.MarrHildrethHash_create()

# Perceptual hash（pHash）
hash_func = cv2.img_hash.PHash_create()

# Radial Variance Hash（ラドン変換に基づく画像ハッシュを計算）
hash_func = cv2.img_hash.RadialVarianceHash_create()

"""## 比較"""

hash1 = hash_func.compute(img1)
hash2 = hash_func.compute(img2)

result = hash_func.compare(hash1, hash2)
result