1、surf是sift的加强版本，具有计算简单，速度更快的特点。应用比较广。本文实现surf特征点检测。先看结果，原图使用湖南祁阳的颜真卿 大唐中兴碑真迹为例。
2、import cv2 as cvimport copyimage = cv.imread('c:\\lzy.jpg')cv.imshow("image", image)image1 = copy.copy(image)image2 = copy.copy(image)
3、gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)cv.imshow("gray", gray)转成灰度图
4、SURF特征计算使用xfeatures2d.SURF_create完成S诔罨租磊IFT特征点初始化surf = cv.xfeatures2d.SURF_cr髫潋啜缅eate()surf.detectAndCompute实现特征点和特征点描述分别输出。keypoints, features = surf.detectAndCompute(gray, None)print(keypoints)坐标点print(keypoints[0].pt[0], keypoints[0].pt[1])特征点描述。print(features)
5、image1=cv.蟠校盯昂drawKeypoints(gray,keypoints,image1)#绘制关键点drawKeyPoints()参数1：image：原始图；参数2：keypoints，特征点向量，元素为KeyPoint对象，包含特征点信息；参数3：outImage；参数4：color：特征点颜色；参数5：flags：设置特征点需要不要画 DEFAULT：只绘制特征点的坐标点,显示小圆点 DRAW_OVER_OUTIMG： NOT_DRAW_SINGLE_POINTS：单点特征点不绘制 DRAW_RICH_KEYPOINTS：带有方向的圆,显示坐标,大小和方向。keypoints：angle：角度，特征点方向，对特征点点邻域梯度计算，获得方向， 缺省-1。class_id：对每个特征点进行区分，缺省-1octave：从金字塔哪层提取的得到的数据。pt：特征点坐标response：该点角点的程度。size：直径
6、for key in keypoints: cv.circle(image2, (int(key.pt[0]), int(key.pt[1])), 1, (255, 255, 0), -1)采用自己绘制方式画特征点 半径=1
7、修改drawKeypoints 增加flags=cv.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS 增加方向image1=cv.drawKeypoints(gray,keypoints,image1,flags=cv.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
8、小结：import cv2 as cvimport copyimage = cv.imread(争犸禀淫'c:\\lzy.jpg')cv.imshow("i罪焐芡拂mage", image)image1 = copy.copy(image)image2 = copy.copy(image)gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)cv.imshow("gray", gray)surf = cv.xfeatures2d.SURF_create()keypoints, features = surf.detectAndCompute(gray, None)# keypoints = surf.detect(gray, None)print(keypoints)print(keypoints[0].pt[0], keypoints[0].pt[1])print(features)image1=cv.drawKeypoints(gray,keypoints,image1)image1=cv.drawKeypoints(gray,keypoints,image1,flags=cv.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)for key in keypoints: cv.circle(image2, (int(key.pt[0]), int(key.pt[1])), 1, (255, 255, 0), -1)cv.imshow("image1", image1)cv.imshow("image2", image2)cv.waitKey(0)cv.destroyAllWindows()
