[300 opencv routines] 239. accurate positioning of Harris corner detection (cornersubpix)

『youcans Of OpenCV routine 300 piece - General catalogue 』

【youcans Of OpenCV routine 300 piece 】239. Harris Accurate positioning of corner detection （cornerSubPix）

An angle is a rapid change in the direction of a straight line . A corner is usually defined as the intersection of two edges , In other words, the neighborhood of the corner should have the boundaries of two different regions in different directions .

Angle is a highly effective feature . Corner detection （Corner Detection） It is widely used in motion detection 、 Image matching 、 Video tracking 、 3D reconstruction and target recognition .

6.2 OpenCV Medium Harris Angle detector

OpenCV Provided in Harris Corner detection function cv.cornerHarris().

Function description ：

cv.cornerHarris(src, blockSize, ksize, k[, dst=None, borderType=BORDER_DEFAULT]	) → dst

function cv.cornerHarris function Harris Angle detector .

For each pixel (x,y) Calculate the gradient covariance matrix $M(x,y)$. Then calculate the characteristics ：
$$dst(x,y)=det{M}^{(x,y)}-k\cdot (tr{M}^{(x,y)}{)}^2$$

Then the corner is a local maximum in the feature response image . Define when in practice R Greater than the set threshold , And the point that is the local maximum is the corner .

Parameter description ：

• src： The input image , Single channel 8 Bit image or floating point image
• dst： Output image ,Harris Detector response , Size and src identical , The format is CV_32FC1
• blockSize： neighborhood size
• ksize：Sobel Aperture parameter of the operator
• k：Harris Detector adjustment parameters , Usually take 0.04~0.06
• borderType： Boundary extension type
  • cv.BORDER_CONSTANT
  • cv.BORDER_REPLICATE
  • cv.BORDER_REFLECT
  • cv.BORDER_REFLECT_101
  • cv.BORDER_TRANSPARENT

routine 14.20：Harris Accurate positioning of corner detection （cornerSubPix）

OpenCV Provides functions cv.cornerSubPix() Used to refine corner position , The corner position detected with sub-pixel accuracy is refined .

cv.cornerSubPix(image, corners, winSize, zeroZone, criteria[, ]) → corners

function cv.cornerSubPix Used to refine corner position . The algorithm sets the centroid of the angle as the new center , Iterative calculation to obtain the sub-pixel precise position of corner points or radial saddle points .

Parameter description ：

• image： The input image , Single channel 8 Bit image or floating point image
• corners： The set of real Numbers , When entering, it is the initial coordinate of the corner , When output, it is the fine coordinate of the corner
• winSize： Half the side length of the search window , Such as winSize = Size(5,5) Indicates that the search window is 11*11
• zeroZone： Half the size of the dead zone in the middle of the search area ,(-1,-1) Indicates that there is no dead zone
• criteria： Criteria for the termination of iterative process , The number of iterations reaches the set value maxCount Or the corner displacement is less than the set threshold epsilon

matters needing attention ：

1. Input/output parameter corners The format of must be converted to real numbers , Is not an integer .
2. function cv.cornerSubPix() Used to refine corner position , It can not only be used for Harris Refine the corner detection results , It can also be used to refine the detection results of other corners .

    # 14.20 Harris  Accurate positioning of corner detection  (cornerSubPix)
    img = cv2.imread("../images/sign01.png", flags=1)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)  # (600, 540)
    print(img.shape)  # (600, 836, 3)

    #  Corner detection 
    gray = np.float32(gray)    # uint8,float32  All support 
    dst = cv2.cornerHarris(gray, 2, 3, 0.04)  # Harris  Corner detection 
    _, dst = cv2.threshold(dst, 0.01 * dst.max(), 255, 0)  #  Extract corners 
    dst = np.uint8(dst)  # (600, 836)
    #  Corner detection result image 
    imgCorner = np.copy(img)
    imgCorner[:,:,2] = cv2.bitwise_or(imgCorner[:,:,2], dst)  #  Filter corners , The red mark 

    #  Fine positioning of the detected corners 
    ret, labels, stats, centroids = cv2.connectedComponentsWithStats(dst)  #  Detect the connected area 
    criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 100, 0.001)  #  Stopping criterion  
    fineCorners = cv2.cornerSubPix(gray, np.float32(centroids), (5,5), (-1,-1), criteria)  # (144, 2)
    #  Fine positioning detection image 
    imgFineCorners = np.copy(img)
    centroids = centroids.astype(np.int)  #  The centroid of the connected region  (x,y)
    fineCorners = fineCorners.astype(np.int)  #  Finely positioned corners  (x,y)
    imgFineCorners[centroids[:, 1], centroids[:, 0]] = [0,255,0]  # Harris  Detection location , green 
    imgFineCorners[fineCorners[:, 1], fineCorners[:, 0]] = [0,0,255]  #  Fine detection position , Red 

    plt.figure(figsize=(9, 6))
    plt.subplot(131), plt.axis('off'), plt.title("Origin")
    plt.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
    plt.subplot(132), plt.axis('off'), plt.title("Harris corners")
    plt.imshow(cv2.cvtColor(imgCorner[115:195, 90:200], cv2.COLOR_BGR2RGB))
    plt.subplot(133), plt.axis('off'), plt.title("cornerSubPix")
    plt.imshow(cv2.cvtColor(imgFineCorners[115:195, 90:200], cv2.COLOR_BGR2RGB))
    plt.tight_layout()
    plt.show()

【 At the end of this section 】

Copyright notice ：
[email protected] Original works , Reprint must be marked with the original link ：(https://blog.csdn.net/youcans/article/details/125821029)
Copyright 2022 youcans, XUPT
Crated：2022-7-15

238. OpenCV Medium Harris Corner detection
239. Harris Accurate positioning of corner detection （cornerSubPix）
240. OpenCV Medium Shi-Tomas Corner detection