1. 程式人生 > >opencv計算機視覺學習筆記五

opencv計算機視覺學習筆記五

第六章 影象檢索以及基於影象描述符的搜尋

通過提取特徵進行影象的匹配與搜尋

1 特徵檢測演算法

常見的特徵和提取演算法:

Harris 檢測角點

Sift 檢測斑點(blob) 有專利保護

Surf 檢測斑點   有專利保護

Fast 檢測角點

Brief 檢測斑點

Orb  帶方向的fast演算法和具有旋轉不變性的brief演算法

特徵的定義

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time    : 2016/12/5 12:30
# @Author  : Retacn
# @Site    : 檢測影象的角點
# @File    : cornerHarris.py
# @Software: PyCharm
__author__ = "retacn" __copyright__ = "property of mankind." __license__ = "CN" __version__ = "0.0.1" __maintainer__ = "retacn" __email__ = "[email protected]" __status__ = "Development" import cv2 import numpy as np # 讀入影象 img = cv2.imread('../test1.jpg') # 轉換顏色空間 gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) gray = np.float32(gray)
# 檢測影象角點 dst = cv2.cornerHarris(gray,                        2,                        23# sobel運算元的中孔,3-31之間的奇數                        0.04) # 將檢測到有角點標記為紅色 img[dst > 0.01 * dst.max()] = [0, 0, 255] while (True):     cv2.imshow("corners", img)     if cv2.waitKey(33) & 0xFF == ord('q'):         break
cv2.destroyAllWindows()

使用dog和sift進行特徵提取和描述

示例程式碼如下:

import cv2
import sys
import numpy as py

# 讀入影象
# imgpath=sys.argv[1]
imgpath = '../test1.jpg'
img = cv2.imread(imgpath)
# 更換顏色空間
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# 建立sift物件,計算灰度影象,會使用dog檢測角點
sift = cv2.xfeatures2d.SIFT_create()
keypoints, descriptor = sift.detectAndCompute(gray, None)

# print(keypoints)
# 關鍵點有以下幾個屬性
# angle 表示特徵的方向
# class_id 關鍵點的id
# octave 特徵所在金字塔的等級
# pt 影象中關鍵點的座標
# response 表示關鍵點的強度
# size  表示特徵的直徑
img = cv2.drawKeypoints(image=img,
                        outImage=img,
                        keypoints=keypoints,
                        color=(51, 163, 236),
                        flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)

# 顯示影象
cv2.imshow('sift_keypoints', img)
while (True):
    if cv2.waitKey(int(1000 / 12)) & 0xFF == ord('q'):
        break
cv2.destroyAllWindows()

使用心有快速hessian演算法和SURF來提取特徵

示例程式碼發如下:

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time    : 2016/12/10 17:30
# @Author  : Retacn
# @Site    : sift用於檢測斑點
# @File    : sift.py
# @Software: PyCharm
__author__ = "retacn"
__copyright__ = "property of mankind."
__license__ = "CN"
__version__ = "0.0.1"
__maintainer__ = "retacn"
__email__ = "[email protected]"
__status__ = "Development"

import cv2
import sys
import numpy as py

# 讀入影象
# imgpath=sys.argv[1]
# alg=sys.argv[2]
# threshold=sys.argv[3]

imgpath = '../test1.jpg'
img = cv2.imread(imgpath)
# alg = 'SURF'
alg = 'SIFT'
# threshold = '8000'
# 閾值越小特徵點越多
threshold = '4000'


def fd(algorithm):
    if algorithm == 'SIFT':
        return cv2.xfeatures2d.SIFT_create()
    if algorithm == 'SURF':
        # return cv2.xfeatures2d.SURF_create(float(threshold) if len(sys.argv) == 4 else 4000)
        return cv2.xfeatures2d.SURF_create(float(threshold))


# 更換顏色空間
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# 建立sift物件,計算灰度影象,會使用dog檢測角點
fd_alg = fd(alg)
keypoints, descriptor = fd_alg.detectAndCompute(gray, None)

# print(keypoints)
# 關鍵點有以下幾個屬性
# angle 表示特徵的方向
# class_id 關鍵點的id
# octave 特徵所在金字塔的等級
# pt 影象中關鍵點的座標
# response 表示關鍵點的強度
# size  表示特徵的直徑
img = cv2.drawKeypoints(image=img,
                        outImage=img,
                        keypoints=keypoints,
                        color=(51, 163, 236),
                        flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)

# 顯示影象
cv2.imshow('keypoints', img)
while (True):
    if cv2.waitKey(int(1000 / 12)) & 0xFF == ord('q'):
        break
cv2.destroyAllWindows()

基於ORB的特徵檢測和特徵匹配

ORB是基於

FAST(featuresfrom accelerated segment test)關鍵點檢測技術

在畫素周圍繪製一個圓,包含16個畫素

BRIEF(binaryrobust independent elementary features) 描述符

暴力(brute-force)匹配法

比較兩個描述符,併產生匹配結果

ORB特徵匹配

示例程式碼如下:

import numpy as np
import cv2
from matplotlib import pyplot as plt

cv2.ocl.setUseOpenCL(False)
# 讀入灰度影象
img1 = cv2.imread('../test2_part.jpg', cv2.IMREAD_GRAYSCALE)
img2 = cv2.imread('../test2.jpg', cv2.IMREAD_GRAYSCALE)

# 建立orb特徵檢測器和描述符
orb = cv2.ORB_create()
kp1, des1 = orb.detectAndCompute(img1, None)
kp2, des2 = orb.detectAndCompute(img2, None)

# 暴力匹配
bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
matches = bf.match(des1, des2)
matches = sorted(matches, key=lambda x: x.distance)

# 顯示影象
img3 = cv2.drawMatches(img1, kp1, img2, kp2, matches[:40], img2, flags=2)
plt.imshow(img3), plt.show()

報如下錯誤:

cv2.error: D:\Build\OpenCV\opencv-3.1.0\modules\python\src2\cv2.cpp:163:error: (-215) The data should normally be NULL! in functionNumpyAllocator::allocate

解決辦法,新增如下程式碼 :

 cv2.ocl.setUseOpenCL(False)

k最鄰近配匹

import numpy as np
import cv2
from matplotlib import pyplot as plt

cv2.ocl.setUseOpenCL(False)
# 讀入灰度影象
img1 = cv2.imread('../test2_part.jpg', cv2.IMREAD_GRAYSCALE)
img2 = cv2.imread('../test2.jpg', cv2.IMREAD_GRAYSCALE)

# 建立orb特徵檢測器和描述符
orb = cv2.ORB_create()
kp1, des1 = orb.detectAndCompute(img1, None)
kp2, des2 = orb.detectAndCompute(img2, None)

# knn匹配,返回k個匹配
bf = cv2.BFMatcher(cv2.NORM_L1, crossCheck=False)
matches = bf.knnMatch(des1, des2, k=2)

# 顯示影象
img3 = cv2.drawMatchesKnn(img1, kp1, img2, kp2, matches, img2, flags=2)
plt.imshow(img3), plt.show()

Flann匹配法

Fast library for approximate nearestneighbors  近似最近鄰的快速庫

import numpy as np
import cv2
from matplotlib import pyplot as plt

# 讀入影象
queryImage = cv2.imread('../test2_part.jpg', cv2.IMREAD_GRAYSCALE)
trainingImage = cv2.imread('../test2.jpg', cv2.IMREAD_GRAYSCALE)

# 建立sift物件
sift = cv2.xfeatures2d.SIFT_create()
kp1, des1 = sift.detectAndCompute(queryImage, None)
kp2, des2 = sift.detectAndCompute(trainingImage, None)

FLANN_INDEX_KDTREE = 0
# 建立字典引數
indexParams = dict(algorithm=FLANN_INDEX_KDTREE, trees=5# 處理索引
searchParams = dict(checks=50# 建立物件,用來指定索引樹的遍歷次數

flann = cv2.FlannBasedMatcher(indexParams, searchParams)

matches = flann.knnMatch(des1, des2, k=2)

matchesMask = [[0, 0] for i in range(len(matches))]

for i, (m, n) in enumerate(matches):
    if m.distance < 0.7 * n.distance:
        matchesMask[i] = [1, 0]
drawParams = dict(matchColor=(0, 255, 0),
                  singlePointColor=(255, 0, 0),
                  matchesMask=matchesMask,
                  flags=0)

resultImage = cv2.drawMatchesKnn(queryImage, kp1, trainingImage, kp2, matches, None, **drawParams)
plt.imshow(resultImage), plt.show()

執行結果如下:

Flann單應性匹配

示例程式碼如下:

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time    : 2016/12/11 12:02
# @Author  : Retacn
# @Site    : flann的單應性匹配
# @File    : flann_homography.py
# @Software: PyCharm
__author__ = "retacn"
__copyright__ = "property of mankind."
__license__ = "CN"
__version__ = "0.0.1"
__maintainer__ = "retacn"
__email__ = "[email protected]"
__status__ = "Development"

import numpy as np
import cv2
from matplotlib import pyplot as plt

MIN_MATCH_COUNT = 10

# 讀入影象
img1 = cv2.imread('../test3_part.jpg', cv2.IMREAD_GRAYSCALE)
img2 = cv2.imread('../test3.jpg', cv2.IMREAD_GRAYSCALE)

# 建立sift物件
sift = cv2.xfeatures2d.SIFT_create()
# 查詢特徵點和描述符
kp1, des1 = sift.detectAndCompute(img1, None)
kp2, des2 = sift.detectAndCompute(img2, None)

FLANN_INDEX_KDTREE = 0
# 建立字典引數
indexParams = dict(algorithm=FLANN_INDEX_KDTREE, trees=5# 處理索引
searchParams = dict(checks=50# 建立物件,用來指定索引樹的遍歷次數

flann = cv2.FlannBasedMatcher(indexParams, searchParams)

matches = flann.knnMatch(des1, des2, k=2)

good = []
for m, n in matches:
    if m.distance < 0.7 * n.distance:
        good.append(m)

if len(good) > MIN_MATCH_COUNT:
    # 在原始影象和訓練影象中查詢特徵點
    src_pts = np.float32([kp1[m.queryIdx].pt for m in good]).reshape(-1, 1, 2)
    dst_pts = np.float32([kp2[m.trainIdx].pt for m in good]).reshape(-1, 1, 2)

    # 單應性
    M, mask = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, 5.0)
    matchesMask = mask.ravel().tolist()

    # 對第二張圖片計算相對於原始影象的投影畸變,並繪製邊框
    h, w = img1.shape
    pts = np.float32([[0, 0], [0, h - 1], [w - 1, h - 1], [w - 1, 0]]).reshape(-1, 1, 2)
    dst = cv2.perspectiveTransform(pts, M)
    img2 = cv2.polylines(img2, [np.int32(dst)], True, 255, 3, cv2.LINE_AA)
else:
    print("Not enough matches are found -%d/%d" % (len(good), MIN_MATCH_COUNT))
    matchesMask = None

# 顯示影象
draw_params = dict(matchColor=(0, 255, 0)# 綠線
                   singlePointColor=None,
                   matchesMask=matchesMask,
                   flags=2)
img3 = cv2.drawMatches(img1, kp1, img2, kp2, good, None, **draw_params)
plt.imshow(img3, 'gray'), plt.show()

執行結果如下:

基於紋身取證的應用程式示例

A 將影象描述符儲存到檔案中

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time    : 2016/12/11 13:52
# @Author  : Retacn
# @Site    : 將影象描述符儲存到檔案中
# @File    : generate_descriptors.py
# @Software: PyCharm
__author__ = "retacn"
__copyright__ = "property of mankind."
__license__ = "CN"
__version__ = "0.0.1"
__maintainer__ = "retacn"
__email__ = "[email protected]"
__status__ = "Development"

import cv2
import numpy as np
from os import walk
from os.path import joinimport sys


# 建立描述符
def create_descriptors(folder):
    files = []
    for (dirpath, dirnames, filenames) in walk(folder):
        files.extend(filenames)
    for f in files:
        save_descriptor(folder, f, cv2.xfeatures2d.SIFT_create())


# 儲存描述符
def save_descriptor(folder, image_path, feature_detector):
    print("reading %s" % image_path)
    if image_path.endswith("npy") or image_path.endswith("avi"):
        return
    img = cv2.imread(join(folder, image_path), cv2.IMREAD_GRAYSCALE)
    keypoints, descriptors = feature_detector.detectAndCompute(img, None)
    descriptor_file = image_path.replace("jpg", "npy")
    np.save(join(folder, descriptor_file), descriptors)


# 從執行引數中取得檔案目錄
dir = sys.argv[1]
create_descriptors(dir)

B 掃描匹配

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time    : 2016/12/11 14:05
# @Author  : Retacn
# @Site    : 掃描匹配
# @File    : scan4matches.py
# @Software: PyCharm
__author__ = "retacn"
__copyright__ = "property of mankind."
__license__ = "CN"
__version__ = "0.0.1"
__maintainer__ = "retacn"
__email__ = "[email protected]"
__status__ = "Development"

from os.path import join
from os import walk
import numpy as np
import cv2
from sys import argv
from matplotlib import pyplot as plt

# 建立檔名陣列
folder = argv[1]
query = cv2.imread(join(folder, 'part.jpg'), cv2.IMREAD_GRAYSCALE)

# 建立全域性的檔案,圖片,描述符
files = []
images = []
descriptors = []
for (dirpath, dirnames, filenames) in walk(folder):
    files.extend(filenames)
    for f in files:
        if f.endswith('npy') and f != 'part.npy':
            descriptors.append(f)
    print(descriptors)

# 建立sift檢測器
sift = cv2.xfeatures2d.SIFT_create()
query_kp, query_ds = sift.detectAndCompute(query, None)

# 建立flann匹配
FLANN_INDEX_KDTREE = 0
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50)
flann = cv2.FlannBasedMatcher(index_params, search_params)

# 最小匹配數
MIN_MATCH_COUNT = 10

potential_culprits = {}
print(">> Initiating picture scan...")
for d in descriptors:
    print("--------- analyzing %s for matches ------------" % d)
    matches = flann.knnMatch(query_ds, np.load(join(folder, d)), k=2)
    good = []
    for m, n in matches:
        if m.distance < 0.7 * n.distance:
            good.append(m)
    if len(good) > MIN_MATCH_COUNT:
        print('%s is a match! (%d)' % (d, len(good)))
    else:
        print('%s is not a match ' % d)
    potential_culprits[d] = len(good)

max_matches = None
potential_suspect = None
for culprit, matches in potential_culprits.items():
    if max_matches == None or matches > max_matches:
        max_matches = matches
        potential_suspect = culprit
print("potential suspect is %s" % potential_suspect.replace("npy", "").upper())

相關推薦

no