opencv3
python3 win7环境
有了surf和sfit前提，特征点匹配就有基础。opencv提供了BFMatcher和FlannBasedMatcher两种方法进行匹配，本文先介绍BFMatcher。BFMatcher：所有可能的匹配，寻找最佳。FlannBasedMatcher：最近邻近似匹配，不是最佳匹配。代码片段：导入图片，其中是翻转过的图片imageA = cv.imread('c:\\haitun.png')cv.imshow('imageA', imageA)imageB = cv.imread('c:\\haitun1.png')cv.imshow('imageB', imageB)
初始化SIFT, 此处xfeatures2d.SIFT_create使用了参数min_hessian 阈值min_hessian = 1000sift = cv.xfeatures2d.SIFT_create(min_hessian)# 分别计算特征点和特征描述符，此处采用sift方法keypointsA, featuresA = sift.detectAndCompute(grayA,None)keypointsB, featuresB = sift.detectAndCompute(grayB,None)画特征点kpImgA=cv.drawKeypoints(grayA,keypointsA,imageA)kpImgB=cv.drawKeypoints(grayB,keypointsB,imageB)cv.imshow('kpImgA', kpImgA)cv.imshow('kpImgB', kpImgB)采用BFMatcher 寻找最佳匹配bf = cv.BFMatcher()使用knnMatch匹配处理，并返回匹配matches matches = bf.knnMatch(featuresA, featuresB, k=2)print(matches)
针对knnMatch匹配方法，创建一个列表，保存符合要求的描述。good = []for m,n in matches: if m.distance < 0.75*n.distance: good.append([m]) print([m])用drawMatchesKnn画出匹配状态并将结果输出resultImg 一个输出灰度图 一个输出彩图resultImg = cv.drawMatchesKnn(grayA, keypointsA, grayB, keypointsB, good,None, flags=2)resultImg1 = cv.drawMatchesKnn(imageA, keypointsA, imageB, keypointsB, good,None, flags=2)plt.imshow(resultImg),plt.show()cv.imshow('resultImg', resultImg)cv.imshow('resultImg1', resultImg1)
小结：完成寻找特征点---匹配特征点过程。并验证旋转不变性的特点。mport numpy as npimport cv2 as cvimport copyfrom matplotlib import pyplot as pltimageA = cv.imread('c:\\haitun.png')cv.imshow('imageA', imageA)imageB = cv.imread('c:\\haitun1.png')cv.imshow('imageB', imageB)grayA = cv.cvtColor(imageA, cv.COLOR_BGR2GRAY)cv.imshow('grayA', grayA)grayB = cv.cvtColor(imageB, cv.COLOR_BGR2GRAY)cv.imshow('grayB', grayB)min_hessian = 1000sift = cv.xfeatures2d.SIFT_create(min_hessian)keypointsA, featuresA = sift.detectAndCompute(grayA,None)keypointsB, featuresB = sift.detectAndCompute(grayB,None)kpImgA=cv.drawKeypoints(grayA,keypointsA,imageA)kpImgB=cv.drawKeypoints(grayB,keypointsB,imageB)cv.imshow('kpImgA', kpImgA)cv.imshow('kpImgB', kpImgB)bf = cv.BFMatcher()matches = bf.knnMatch(featuresA, featuresB, k=2)print(matches)good = []for m,n in matches: if m.distance < 0.75*n.distance: good.append([m]) print([m])resultImg = cv.drawMatchesKnn(grayA, keypointsA, grayB, keypointsB, good,None, flags=2)resultImg1 = cv.drawMatchesKnn(imageA, keypointsA, imageB, keypointsB, good,None, flags=2)plt.imshow(resultImg),plt.show()cv.imshow('resultImg', resultImg)cv.imshow('resultImg1', resultImg1)cv.waitKey(0)cv.destroyAllWindows()
注意good列表的处理，只对knnMatch处理有效。通过调整参数可改变匹配数量
完成寻找特征点---匹配特征点过程