開發環境:win10 anaconda python3.6 opencv3
1、四種特徵提取算法簡介(需要深入的同學可以參考其他博主的博客)
SIFT特徵主要提取圖像的局部特徵,對平移、旋轉、尺度縮放、亮度變化、遮擋和噪聲等具有很好的不變性,對視覺變化、仿射變換也保持一定程度的穩定性。SURF可以說是從SIFT發展而來的,速度上優於SIFT,快了一個數量級(10倍),並且穩定性要高於SIFT。而FAST算法提取了大量的特徵點,但在計算時間上,比SIFT要快兩個數量級 比SURF快一個數量級。ORB算法基於FAST算法,但是提取的特徵點質量比FAST高,特徵點數目比FAST較少。
實驗數據在這篇博客:點擊打開鏈接
計算速度: ORB>>SURF>>SIFT(各差一個量級)
旋轉魯棒性: SURF>ORB~SIFT(表示差不多)
模糊魯棒性: SURF>ORB~SIFT
尺度變換魯棒性: SURF>SIFT>ORB(ORB並不具備尺度變換性)
2、代碼+效果對比
# -*- coding: utf-8 -*- """ Spyder Editor This is a temporary script file. """ import cv2 img = cv2.imread(r'C:\Users\Pictures\Camera Roll/test.jpg') gray= cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) # one SIFT #sift = cv2.xfeatures2d.SIFT_create() #kp1 = sift.detect(gray,None) #img1=cv2.drawKeypoints(gray,kp1,img) #cv2.imshow("sift_Image", img1) #cv2.imwrite(r'C:\Users\Pictures\Camera Roll/sift_test.jpg',img1) #Compare four alogorithms #Another SIFT sift = cv2.xfeatures2d.SIFT_create() (kps, descs) = sift.detectAndCompute(gray, None) img1=cv2.drawKeypoints(gray, kps, img, flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS) cv2.imshow('SIFT_Algorithm', img1) # SURF surf = cv2.xfeatures2d.SURF_create() (kps2, descs2) = surf.detectAndCompute(gray, None) img2=cv2.drawKeypoints(gray, kps2, img, flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS) cv2.imshow('SURF_Algorithm', img2) # FAST fast = cv2.FastFeatureDetector_create() kps3 = fast.detect(gray, None) img3=cv2.drawKeypoints(gray, kps3, img, flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS) cv2.imshow('FAST_Algorithm', img3) # ORB orb = cv2.ORB_create() kps4 = orb.detect(gray, None) (kps4, des4) = orb.compute(gray, kps4) img4=cv2.drawKeypoints(gray, kps4, img, flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS) cv2.imshow('ORB_Algorithm', img4)
最終的圖像對比爲: