博客
关于我
基于OpenCV的透视图转化为不同平面
阅读量:329 次
发布时间:2019-03-04

本文共 2850 字,大约阅读时间需要 9 分钟。

??????????????????????????????????????????????????????????????????????????????????????????????????

???????????????????????????????????????????????????????????????

??????????????????????????????????????????????????????2D???????????????????????????

?????????3D???????????????????????????????????????????????????????????????????????????

????????????????3D???????2D????????????3D????3x4????????????2D???

?????3x4???????3D???????x?y?z????2D??????????u?v??????????2D???????3D??????????????z=0??

????

  • ???????

    ????????z????????????????????z??????3x4???????3x3????????????????????????????????????

  • ??????

    ??????????????????????????????????????????????????????????u?v????????????x?y???OpenCV??????cv2.findHomography??????????

  • ????

    import cv2import numpy as npdef get_inverse_perspective(perspective_matrix: np.array) -> np.array:    """ This method calculates the inverse of perspective matrix by homography. """    # Take 5 homogeneous points on the floor plane (Unit is in Meters)    pts_dst = np.array([[0, 0, 0, 1],                        [0, 1, 0, 1],                        [1, 0, 0, 1],                        [1, 1, 0, 1],                        [0, 0, 0, 1]])        # Obtain respective homogeneous points on the image plane    pts_src = (perspective_matrix @ pts_dst.T).T    # Convert homogeneous coordinates to Cartesian coordinates    pts_src_cart = np.array([[x/w, y/w] for x, y, w in pts_src])    pts_dst_cart = np.array([[x/w, y/w] for x, y, z, w in pts_dst])        # Find the 3x3 Homography Matrix for transforming image plane to floorplane    H, status = cv2.findHomography(pts_src_cart, pts_dst_cart)    return Hdef project_to_floor(image_coordinates: List[int], H: np.array) -> List[int]:    """ This method takes the Homography matrix and the 2D image Cartesian coordinates.     It returns the (x, y) Cartesian coordinates in 3D world coordinates on floorplane (z=0). """    # Adding 1 for homogeneous coordinate system    x, y, w = H @ np.array([[*image_coordinates, 1]]).T    return [x/w, y/w]# Example usagep = np.random.rand(3, 4)H = get_inverse_perspective(p)src_point = (5, 10)dst_point = project_to_floor(src_point, H)

    ????

  • ?????????

    ???????z=0???????z??????3x4???????3x3???????????

  • ????

    ???????????????????????????????4?????

  • OpenCV??

    ?OpenCV??????cv2.findHomography??????????

  • ????????????2D????????????3D????????????

    ????

  • OpenCV-Contrib?????????

    ???????????????????????????????????OpenCV???????????????????SFM????????????????????????????????

  • Python??????52?

    ????????????????Python???????????????????????????????????????????????????????????????????31????????????????????

  • OpenCV????20?

    ????????????????OpenCV????20?????????20???OpenCV????????????OpenCV???

  • ??????????????????SLAM??????????????????????????????????GAN???????????????????????????+??/??+???????????? ???? ??SLAM????????????????????????????????????????????????????????????~

    转载地址:http://yyaq.baihongyu.com/

    你可能感兴趣的文章
    Qt网络编程之实例二POST方式
    查看>>
    POJ 1765 November Rain
    查看>>
    poj 1860 Currency Exchange
    查看>>
    POJ 1961 Period
    查看>>
    POJ 2019 Cornfields (二维RMQ)
    查看>>
    poj 2057 The Lost House 贪心思想在动态规划上的应用
    查看>>
    poj 2057 树形DP,数学期望
    查看>>
    poj 2112 最优挤奶方案
    查看>>
    Qt编写自定义控件12-进度仪表盘
    查看>>
    SpringBoot主启动原理在SpringApplication类《第六课》
    查看>>
    poj 2186 Popular Cows :求能被有多少点是能被所有点到达的点 tarjan O(E)
    查看>>
    POJ 2186:Popular Cows Tarjan模板题
    查看>>
    POJ 2229 Sumsets(递推,找规律)
    查看>>
    poj 2236
    查看>>
    POJ 2243 Knight Moves
    查看>>
    POJ 2262 Goldbach's Conjecture
    查看>>
    POJ 2362 Square DFS
    查看>>
    Qt笔记——解决添加Qt Designer Form Class时“allocation of incomplete type Ui::”
    查看>>
    poj 2386 Lake Counting(BFS解法)
    查看>>
    poj 2387 最短路模板题
    查看>>