Python+OpenCV：图像Harris角点检测(Harris Corner Detection)

爱被打了一巴掌 2022-12-23 06:17 155阅读 0赞

## Python+OpenCV：图像Harris角点检测(Harris Corner Detection) ##

### 理论 ###

corners are regions in the image with large variation in intensity in all the directions.

It basically finds the difference in intensity for a displacement of (u,v) in all directions. This is expressed as below:

![20201122223812952.png][]

The window function is either a rectangular window or a Gaussian window which gives weights to pixels underneath.

A score, basically an equation, which determines if a window can contain a corner or not.

![20201122224003707.png][]

where

*  det(M)=λ1λ2
 *  trace(M)=λ1+λ2
 *  λ1 and λ2 are the eigenvalues of M

So the magnitudes of these eigenvalues decide whether a region is a corner, an edge, or flat.

*  When |R| is small, which happens when λ1 and λ2 are small, the region is flat.
 *  When R<0, which happens when λ1>>λ2 or vice versa, the region is edge.
 *  When R is large, which happens when λ1 and λ2 are large and λ1∼λ2, the region is a corner.

It can be represented in a nice picture as follows:

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70][]

### Harris Corner Detector in OpenCV ###

####################################################################################################
    # 图像Harris角点检测(Harris Corner Detection)
    def lmc_cv_image_harris_corner_detection():
        """
            函数功能: 图像Harris角点检测(Harris Corner Detection)。
        """
    
        stacking_images = []
        image_file_name = ['D:/99-Research/Python/Image/CheckerBoard.jpg',
                           'D:/99-Research/Python/Image/CheckerBoard01.jpg',
                           'D:/99-Research/Python/Image/CheckerBoard02.jpg',
                           'D:/99-Research/Python/Image/CheckerBoard07.jpg']
        for i in range(len(image_file_name)):
            # 读取图像
            image = lmc_cv.imread(image_file_name[i])
            image = lmc_cv.cvtColor(image, lmc_cv.COLOR_BGR2RGB)
            result_image = image.copy()
            gray_image = lmc_cv.cvtColor(image, lmc_cv.COLOR_BGR2GRAY)
    
            gray_image = np.float32(gray_image)
            corner_harris_image = lmc_cv.cornerHarris(gray_image, 2, 3, 0.04)
            # result is dilated for marking the corners, not important
            corner_harris_image = lmc_cv.dilate(corner_harris_image, None)
            # Threshold for an optimal value, it may vary depending on the image.
            result_image[corner_harris_image > 0.01 * corner_harris_image.max()] = [255, 0, 0]
    
            # stacking images side-by-side
            stacking_image = np.hstack((image, result_image))
            stacking_images.append(stacking_image)
    
        # 显示图像
        for i in range(len(stacking_images)):
            pyplot.figure('Harris Corner Detection %d' % (i+1))
            pyplot.subplot(1, 1, 1)
            pyplot.imshow(stacking_images[i], 'gray')
            pyplot.xticks([])
            pyplot.yticks([])
        pyplot.show()
    
        # 根据用户输入保存图像
        if ord("q") == (lmc_cv.waitKey(0) & 0xFF):
            # 销毁窗口
            pyplot.close('all')
        return

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 1][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 2][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 3][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 4][]

### Corner with SubPixel Accuracy ###

####################################################################################################
    # 图像亚像素Harris角点检测(Harris Corner Detection with SubPixel Accuracy)
    def lmc_cv_image_harris_corner_detection_subpixel():
        """
            函数功能: 图像亚像素Harris角点检测(Harris Corner Detection with SubPixel Accuracy)。
        """
    
        stacking_images = []
        image_file_name = ['D:/99-Research/Python/Image/CheckerBoard.jpg',
                           'D:/99-Research/Python/Image/CheckerBoard02.jpg',
                           'D:/99-Research/Python/Image/CheckerBoard04.jpg',
                           'D:/99-Research/Python/Image/CheckerBoard07.jpg']
        for i in range(len(image_file_name)):
            # 读取图像
            image = lmc_cv.imread(image_file_name[i])
            image = lmc_cv.cvtColor(image, lmc_cv.COLOR_BGR2RGB)
            result_image = image.copy()
            gray_image = lmc_cv.cvtColor(image, lmc_cv.COLOR_BGR2GRAY)
    
            # find Harris corners
            gray_image = np.float32(gray_image)
            corner_harris_image = lmc_cv.cornerHarris(gray_image, 2, 3, 0.04)
            corner_harris_image = lmc_cv.dilate(corner_harris_image, None)
            ret, corner_harris_image = lmc_cv.threshold(corner_harris_image, 0.01 * corner_harris_image.max(), 255, 0)
            corner_harris_image = np.uint8(corner_harris_image)
            # find centroids
            ret, labels, stats, centroids = lmc_cv.connectedComponentsWithStats(corner_harris_image)
            # define the criteria to stop and refine the corners
            criteria = (lmc_cv.TERM_CRITERIA_EPS + lmc_cv.TERM_CRITERIA_MAX_ITER, 100, 0.001)
            corners = lmc_cv.cornerSubPix(gray_image, np.float32(centroids), (5, 5), (-1, -1), criteria)
            # Now draw them
            res = np.hstack((centroids, corners))
            res = np.int0(res)
            result_image[res[:, 1], res[:, 0]] = [255, 0, 0]
            result_image[res[:, 3], res[:, 2]] = [0, 255, 0]
    
            # stacking images side-by-side
            stacking_image = np.hstack((image, result_image))
            stacking_images.append(stacking_image)
    
        # 显示图像
        for i in range(len(stacking_images)):
            pyplot.figure('Harris Corner Detection  with SubPixel Accuracy %d' % (i+1))
            pyplot.subplot(1, 1, 1)
            pyplot.imshow(stacking_images[i], 'gray')
            pyplot.xticks([])
            pyplot.yticks([])
        pyplot.show()
    
        # 根据用户输入保存图像
        if ord("q") == (lmc_cv.waitKey(0) & 0xFF):
            # 销毁窗口
            pyplot.close('all')
        return

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 5][]

[20201122223812952.png]: /images/20221120/29dcc29d5c704305830327df9efcf02a.png
[20201122224003707.png]: /images/20221120/8c2d58ad68b14836804dff3fb856d99d.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70]: /images/20221120/7d8fcc5f7e1648c5b8e0b18c4351a4b6.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 1]: /images/20221120/f9a275b645ad4a8297dd4d2526acf3e0.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 2]: /images/20221120/a705f7778af74d9ba9ba03ef5b2375cf.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 3]: /images/20221120/05056d559c5e449b963949a29946ac47.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 4]: https://img-blog.csdnimg.cn/20201122230627492.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5_size_16_color_FFFFFF_t_70 5]: https://img-blog.csdnimg.cn/20201122234608827.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2xpdWJpbmc4NjA5,size_16,color_FFFFFF,t_70