CN108492306A - A kind of X-type Angular Point Extracting Method based on image outline - Google Patents

Abstract

The present invention proposes a kind of X-type Angular Point Extracting Method based on image outline, it is related to technical field of vision detection, the present invention traverses Harris angle points successively by building window on certain profile, when it is 7 to traverse angle point number, it is X-type angle point to take the point nearest apart from 7 angular coordinate mean values, and traversal next profile completes the extraction of all profile X-type angle points when extraction angle point number is 4；The present invention can filter the pseudo- angle point in image, the angle point that can be accurately detected in image, and accuracy of detection reaches 0.09 pixel；By the periphery background for setting angle point so that characteristic point is not influenced mark point in by environment, and angle point is detected in complicated scene；It is precisely extracted in conjunction with Harris and the advantages of locations of contours, sub-pixel detection is carried out to the angle point of laying, improves the coordinate precision of angle point；The threshold value that image preprocessing is adjusted by dynamic, reduces the influence of illumination in different scenes, improves the success rate of Corner Detection.

Description

A Method of Extracting X-shaped Corner Points Based on Image Contour

technical field

The invention relates to the technical field of visual detection, in particular to an X-shaped corner point extraction method based on image contours.

Background technique

With the continuous development of science and technology, cameras have also developed rapidly, and there are more and more types and functions of cameras, and people have also enjoyed various conveniences brought by the development of science and technology. For example, people can obtain the environmental information in the parking lot by setting the camera on the vehicle and collecting images through the camera.

The current cameras need to calibrate the parameters to obtain more accurate parameters, and the calculation of parameters is directly affected by the accuracy of the corner points, which requires higher accuracy of corner point extraction. The traditional corner point extraction technology has a corner point detection method based on the segment test, but the image scene is complex, and the template size of the segment test and the gray difference threshold are difficult to choose, which leads to the detection of X-shaped corner points. Cause missed inspection, multi-inspection, and more false corners.

Contents of the invention

Aiming at the deficiencies of the prior art, the present invention proposes an X-shaped corner point extraction method based on the image contour, which can accurately identify the X-shaped corner points in the image, and provides guarantee for precise camera calibration.

A method for extracting X-shaped corners based on image contours, comprising the following steps:

Step 1, obtain the image to be processed, and use the Harris corner point extraction method to extract all Harris corner points in the image;

Step 2, set the threshold value, perform binarization processing on the image, obtain a black and white image, and perform corrosion and open operation processing on the black and white image;

Step 3, extract all contours in the black-and-white image, and traverse all Harris corners on a certain contour in turn by building a window;

Step 4. Determine whether the number of Harris corner points obtained by traversal is 7, if so, take the point closest to the mean value of the coordinates of the 7 corner points as the X-shaped corner point, and traverse the next contour, otherwise, traverse the next contour;

Step 5. After traversing all contours, judge whether the number of extracted X-shaped corner points is 4, if so, complete the extraction of X-shaped corner points of all contours, otherwise, dynamically adjust the threshold and window set by the binarization process side length until the number of extracted X-shaped corner points is 4, that is, all four marked X-shaped corner points are detected, and the detection is completed.

Extract all outlines in the black-and-white image described in step 3, and traverse all Harris corner points sequentially on a certain outline by building a window;

Specifically:

Take any point on the contour, take this point as the center point, set the side length, and build a window;

Taking the center point as the starting point, traverse the pixel points in the window along the outline;

If it is traversed that there is a Harris corner falling into the window, record it, and delete the Harris corner in the image, and continue to traverse the pixels in the window along the contour until returning to the starting point.

Otherwise as described in step 4, the next contour is traversed, including restoring the deleted Harris corners.

Advantage of the present invention:

The present invention proposes an X-shaped corner point extraction method based on the image outline, which can filter out false corner points in the image, and can accurately detect the corner points in the image, and the detection accuracy can reach 0.09 pixels; by setting the corner point The surrounding background (black and white color), so that the marker points are basically not affected by the feature points in the environment, and corner points can be detected in complex scenes; combined with the advantages of Harris' precise extraction and contour positioning, it is possible to sub-pixel the laid corner points Level detection improves the coordinate accuracy of corner points; by dynamically adjusting the threshold of image preprocessing, the influence of illumination in different scenes is weakened, and the success rate of corner point detection is improved.

Description of drawings

Fig. 1 is the flow chart of the X-type corner point extraction method based on the image contour of an embodiment of the present invention;

Fig. 2 is a schematic diagram of an image to be processed according to an embodiment of the present invention;

Fig. 3 is the Harris corner point detection image schematic diagram of an embodiment of the present invention;

Fig. 4 is a discontinuous schematic diagram of an X-shaped corner area according to an embodiment of the present invention;

Fig. 5 is a schematic diagram of an X-shaped corner area after erosion and opening operations according to an embodiment of the present invention;

Fig. 6 is a schematic diagram of a first-order differential template in an embodiment of the present invention, wherein, Figure (a) is a schematic diagram of a first-order differential template in the X direction, and Figure (b) is a schematic diagram of a first-order differential template in the Y direction;

Fig. 7 is a schematic diagram of the direction of the gradient in an embodiment of the present invention, wherein, Figure (a) is a schematic diagram of sector segmentation in the gradient direction, and Figure (b) is a schematic diagram of pixel selection in the gradient direction;

Fig. 8 is a schematic diagram of contour extraction according to an embodiment of the present invention;

FIG. 9 is a schematic diagram of the completion of extraction of contour X-shaped corner points according to an embodiment of the present invention.

Detailed ways

An embodiment of the present invention will be further described below in conjunction with the accompanying drawings.

In the embodiment of the present invention, the X-shaped corner point extraction method based on the image outline, as shown in Figure 1, includes the following steps:

Step 1, obtain the image to be processed, and use the Harris corner point extraction method to extract all Harris corner points in the image;

In the example of the present invention, as shown in Figure 2, it is the image to be processed, and the coordinate position of the X-type corner is determined by the Harris operator, and the specific steps are as follows:

Establish a coordinate system with any point in the image as the origin, preferably any one of the four corners in the image as the origin, wherein the horizontal direction of the image is the X axis, the vertical direction of the image is the Y axis, and the process is performed by the Harris operator Corner detection. After the coordinate system is established, the corner points in the image can be detected through the Harris operator. The specific detection process is as follows:

First, use the gradient operator to calculate the difference in the x and y directions, and then use the grayscale function of each pixel in the image to calculate the gradients I _x and I _y in the x-axis and y-axis directions of each pixel in the image. The calculation formula is as follows :

Then calculate the gradient products I _x ² , I _y ² and I _xy of each pixel in the image in two directions respectively, the calculation formula is as follows:

I _x ² = I _x *I _x (3)

I _y ² = I _y ^* I _y (4)

I _xy = I _x ^* I _y (5)

In this way, three new images can be obtained, and the attributes corresponding to each pixel point are the gradient products I _x ² , I _y ² , and I _xy of each pixel point in two directions, respectively, and then Gaussian filtering is performed on the three images respectively, Specifically, it can be:

Use the Gaussian function to generate a Gaussian kernel to filter I _x ² , I _y ² and I _xy respectively, where the formula of the Gaussian function is:

According to the formula (6) ~ formula (9), the corner point value R of each pixel can be calculated, and the calculation formula is as follows:

In the example of the present invention, select appropriate size window (for example 5*5), traverse all pixels in the image, calculate the local maximum value of the pixel corner amount in this window, if this value is greater than certain threshold value thresh (specifically in the example of the present invention Take a value of 0.05), then think that this pixel is a candidate point; sort according to the size of the corner point of the candidate point, get the first N (specific value is 200 in the example of the present invention) points of interest as the last Harris corner point ; In the example of the present invention, as shown in Figure 3, it is the image after Harris corner point detection;

Step 2, set threshold value (concrete value is 50 in the example of the present invention), carry out binarization process to image, obtain black-and-white image, corrode and open operation processing to black-and-white image;

In the embodiment of the present invention, in order to better detect the edge, the grayscale image to be detected is subjected to binarization processing; in the process of binarization processing, the image area farther from the camera has a smaller viewing angle and an excessive inclination degree. Large, resulting in discontinuity in the black area of the X-shaped corner point, which will cause the contour of the X-shaped area to be disconnected when extracting the contour, and eventually lead to the failure of contour matching and corner point detection. In order to solve this problem, erosion and opening operations are performed on the binarized image.

In the example of the present invention, as shown in FIG. 4 , it is a schematic diagram of an X-shaped corner region discontinuity. Figure 5 is a schematic diagram of the X-shaped corner area after erosion and opening operations.

Step 3. Extract all contours in the black-and-white image, and traverse all Harris corner points on a certain contour in turn by building a window; specifically: take any point on the contour, take this point as the center point, and set the side length, Build a window; starting from the center point, traverse the pixels in the window along the outline; if there is a Harris corner point that falls into the window, record it, delete the Harris corner point in the image, and continue to traverse the pixels in the window along the outline point until returning to the starting point;

In the example of the present invention, after corroding the image to be detected, the edge of the image is obtained through the canny operator, and the Gaussian filter function is first used to smooth the image, and then the first-order difference shown in Figure 6 (a) and Figure (b) is used The template enhances the edge of the image, calculates the magnitude and direction of the gradient, uses the direction of the gradient as shown in Figure 7 to perform non-maximum suppression on the calculated gradient, and retains the pixel with the largest local gradient.

In the example of the present invention, 4 pairs of sectors of Figure (a) in Figure 7 correspond to four combinations of Figure (b) templates in Figure 7 respectively; Utilize the gradient direction to select the label of the sector, then compare the gradient of this point according to the direction of the label The magnitude and the gradient magnitude of adjacent pixels, if the value is greater than the other two, keep the original value, otherwise set the value to 0. For example, if the gradient direction of the middle point of the template points to the sector label 3, then compare the gradient amplitudes of the middle point and the neighborhood points 4 and 8, and if the values of the middle point are greater than the values of points 4 and 8, then M remains unchanged, otherwise M =0.

In the example of the present invention, the gradient magnitude is processed by a bilinear threshold value, specifically, the upper limit value and the lower limit value are respectively selected, and if the gradient of a pixel is greater than the upper limit threshold value, then the point is considered to be an edge pixel point; if If the value is lower than the lower threshold, the point is discarded; if the value is greater than the lower threshold and smaller than the upper threshold, it is accepted only when the point is connected to a pixel higher than the upper threshold, so that it can be determined The straight lines where the connected pixels are located are the edges of the image.

In the example of the present invention, after obtaining the edge image, the image can be extracted to obtain at least one area outline of the black and white grid in the X-shaped corner area, as shown in Figure 8; after obtaining at least one X-shaped corner area outline Finally, the acquired area contours can be matched to obtain the matching area contours; specifically, each X-shaped corner area contour in the at least one X-shaped corner area contour can be traversed sequentially according to the preset window side length , judge whether there are corner points in the window domain; if there are corner points, then determine the contour of the region containing 7 initial corner points as the contour of the matching region;

In the example of the present invention, it can be seen from Fig. 3 and Fig. 4 that there are 7 corner points on the outline of each X-shaped corner region. That is, as long as the contour with 7 initial corner points is matched, the marked point in the image is found. When matching the contour, you can select an appropriate window side length (the specific value is 3 in the example of the present invention), traverse all the pixels of the contour in turn, and find whether there is a corner in the neighborhood of the pixel. In the specific implementation process, after the erosion operation, the contour of the X-shaped corner area becomes a connected one, which makes the distance between the detected contour and the real edge larger, so it is necessary to choose the window side length reasonably to detect the corner. If there are corner points, the number of corner points is accumulated, and each corner point is only counted once until all pixels are traversed.

Step 4. Determine whether the number of Harris corner points obtained by traversal is 7, and if so, take the point closest to the average value of the coordinates of the 7 corner points as the X-shaped corner point, and traverse the next contour, otherwise, restore the deleted Harris corner point , to traverse the next contour;

Step 5. After traversing all the contours, judge whether the number of extracted X-shaped corner points is 4, if so, complete the extraction of all contour X-shaped corner points, as shown in Figure 9, otherwise, dynamically adjust the binarization Process the set threshold and window side length until the number of extracted X-shaped corner points is 4, that is, all four marked X-shaped corner points are detected, and the detection is completed.

Claims (3)

1. a kind of X-type Angular Point Extracting Method based on image outline, which is characterized in that include the following steps：

Step 1 obtains pending image, is carried to Harris angle points all in image using Harris Angular Point Extracting Methods It takes；

Step 2, setting threshold value, carry out binary conversion treatment to image, obtain black white image, corroded and opened fortune to black white image Calculation is handled；

All profiles in step 3, extraction black white image traverse institute successively by way of building window on a certain profile There are Harris angle points；

Step 4 judges to traverse whether obtained Harris angle points number is 7, if so, taking apart from 7 angular coordinate mean values most Close point is X-type angle point, and traverses next profile and otherwise traverse next profile；

Step 5, after having traversed all profiles, whether the number for judging to extract the X-type angle point is 4, if so, completing all wheels The extraction of wide X-type angle point, otherwise, dynamic adjust the threshold value and the window length of side of binary conversion treatment setting, until extraction X-type angle point Number is 4, that is, 4 marked X-type angle point all detects that detection finishes.

2. the X-type Angular Point Extracting Method according to claim 1 based on image outline, which is characterized in that described in step 3 All profiles in extraction black white image traverse all angles Harris successively by way of building window on a certain profile Point；

Specially：

It is arbitrarily taken on profile a bit, the point centered on the point, the length of side is set, build window；

Using central point as starting point, the pixel along profile cycling among windows is played；

If traversing out, there are Harris angle points to fall into window, records, and deletes the Harris angle points in the picture, continues edge Pixel in profile cycling among windows, until being back to starting point.

3. the X-type Angular Point Extracting Method according to claim 1 based on image outline, which is characterized in that described in step 4 Otherwise, next profile is traversed, further includes the Harris angle points to undelete.