A global automated detection method for circular markers using close-range photogrammetry

Abstract

This invention discloses a global automated detection method for circular markers in close-range photogrammetry. First, the Canny algorithm is used to extract initial candidate edge information containing the positions of all circular markers. Then, based on the region growing algorithm, morphological methods, and ellipse / circle conditions, a large number of non-circular and pseudo-circular marker edge information are further removed. Finally, a two-stage Hough transform detection method is used to obtain all circular markers. Experimental results show that this invention can effectively and automatically detect all circular markers in the entire close-range photogrammetric image.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of close-range photogrammetry, and particularly relates to a design of a global automatic detection method for close-range photogrammetry circular mark points. BACKGROUND

[0002] Close-range photogrammetry technology refers to photogrammetry using a stereoscopic pair of images taken of a target object with a distance of not more than 300 m. It is widely used in various practical engineering fields such as civil engineering due to its non-contact measurement means and other advantages. In various practical engineering applications, circular mark points are usually laid out and detected to achieve monitoring of various target objects. However, in the existing close-range photogrammetry technology, the monitoring of target objects is usually carried out by manually selecting local areas for detection to form monitoring of all target objects in the entire image area, and there is a lack of a method that can effectively globally and automatically detect all circular mark points in the image. SUMMARY

[0003] The purpose of the present application is to solve the problem of the lack of a method that can effectively globally and automatically detect all circular mark points in the image in the existing close-range photogrammetry technology. A global automatic detection method for close-range photogrammetry circular mark points is proposed.

[0004] The technical solution of the present application is as follows: A global automatic detection method for close-range photogrammetry circular mark points, comprising the following steps:

[0005] S1. Extracting initial edge information containing the positions of all circular mark points from a close-range photogrammetry image using a Canny algorithm.

[0006] S2. Removing edge information of non-circular mark points in the initial edge information to obtain first edge information by taking the size of the circular mark point region, the major and minor radii of the ellipse, and the eccentricity of the ellipse as preliminary constraint conditions.

[0007] S3. Determining whether each region in the first edge information is closed and connected by judging the gray scale information of the eight-neighborhood of each pixel point in the region using a region growing algorithm, and further removing edge information of non-circular mark points in the first edge information to obtain second edge information.

[0008] S4. Removing edge information of pseudo-circular mark points in the second edge information using a morphological method and the circularity condition of all circular mark points to obtain third edge information.

[0009] S5. Obtaining all circular mark points from the third edge information using a Hough transform based on a two-stage algorithm.

[0010] Further, step S1 comprises the following sub-steps:

[0011] S11, Gaussian filter the close-range photographic image f(i,j) using a two-dimensional Gaussian smoothing filter function G(i,j):

[0012]

[0013] where (i,j) represents the coordinate point of the close-range photographic image, and σ represents the standard deviation of the two-dimensional Gaussian smoothing filter function.

[0014] S12, calculate the gradient amplitude L(i,j) and the gradient direction θ(i,j) of the Gaussian filtered image:

[0015]

[0016]

[0017] where G x (i,j) and G y (i,j) represent the horizontal direction gradient and the vertical direction gradient at the point (i,j), respectively, and p(i,j) represents the pixel value at the point (i,j).

[0018] S13, sequentially determine whether the gradient amplitude of each pixel point in the Gaussian filtered image is greater than the gradient amplitude of the adjacent pixel point in the gradient direction of the pixel point, if yes, determine that the pixel point is an edge, otherwise determine that the pixel point is a non-edge.

[0019] S14, remove the false edges of the image using a double threshold method to obtain initial edge information containing all the positions of the circular marker points.

[0020] Further, the preliminary constraint condition in step S2 is specifically:

[0021]

[0022] where area represents the area size of the circular marker point, Maj represents the major radius of the ellipse, and e represents the eccentricity of the ellipse.

[0023] Further, step S3 includes the following sub-steps:

[0024] S31, according to the closed feature of the circle, automatically select the upper left corner pixel point in each region in the first edge information as an initial seed point, take the surrounding eight neighborhoods as the search space, and count the number of the same gray scale.

[0025] S32, loop through the number of the same gray scale in the surrounding eight neighborhoods of each other edge pixel point in the region until all the edge points in the region are counted.

[0026] S33, retain all the regions that satisfy the closed connected judgment condition to obtain the second edge information.

[0027] Further, the closed communication judgment condition in step S33 is that the number of the same gray scale in the eight-neighborhood around the edge pixel point is greater than or equal to 3.

[0028] Further, step S4 comprises the following sub-steps:

[0029] S41, the morphological reconstruction method based on geodesic dilation is used to perform hole filling processing on the second edge information:

[0030]

[0031] wherein represents the geodesic dilation with the scale n of the mark image Y in the mask image X, which is obtained by the unit scale geodesic dilation with the scale 1 is obtained by iteration n times, and p represents a point in the mask image X, d X (p, Y) represents the geodesic distance from the point p to the mark image Y, represents the iterative calculation, represents the dilation operation, and B represents a 3*3 4-connected matrix or an 8-connected matrix.

[0032] S42, the roundness of each circular filling region in the second edge information after the hole filling processing is calculated:

[0033]

[0034] wherein m represents the roundness of the circular filling region, S represents the area of the circular filling region, and P represents the circumference of the circular filling region.

[0035] S43, whether the roundness of each circular filling region is greater than or equal to a preset roundness threshold value is judged in sequence, if yes, the edge information of the circular filling region is retained, otherwise the circular filling region is determined as a pseudo-circular mark point, and the edge information thereof is removed to obtain third edge information.

[0036] Further, step S5 comprises the following sub-steps:

[0037] S51, the center feature of the circle is identified according to the gradient of each edge point on the boundary of the circular mark point in the third edge information, and the candidate center parameter is identified through local peak value detection.

[0038] S52, according to the candidate center parameter, a radius histogram is constructed by using a circular constraint condition equation, the peak value of the radius histogram is recorded as the center of the corresponding circular mark point, and the corresponding edge point is recorded as the edge of the circular mark point, so that all the circular mark points are obtained.

[0039] Further, the circular constraint condition equation in step S52 is specifically:

[0040] (x-x a ) 2 +(y-x b ) 2 = r 2

[0041] where (x, y) represents the position coordinates of the edge point, (x a , y b ) represents the position coordinates of the center point, and r represents the radius of the constraint circle.

[0042] The present application has the beneficial effects that: the present application firstly adopts the Canny algorithm to extract initial candidate edge information containing all the positions of the circular mark points, then further removes the edge information of a large number of non-circular mark points and pseudo-circular mark points based on the region growing algorithm, the morphological method and the ellipse / circular condition, and finally acquires all the circular mark points by using the two-stage algorithm Hough transform detection method, so that the present application can effectively and globally automatically detect all the circular mark points in the whole image of the close-range photogrammetry. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 Fig. 1 is a flow chart of a global automatic detection method for close-range photogrammetry circular mark points provided by an embodiment of the present application.

[0044] Figure 2 Fig. 2 is an image diagram of a three-layer frame structure photographed by a high-speed video camera provided by an embodiment of the present application.

[0045] Figure 3 Fig. 3 is a schematic diagram of initial edge information provided by an embodiment of the present application.

[0046] Figure 4 Fig. 4 is a schematic diagram of first edge information provided by an embodiment of the present application.

[0047] Figure 5 Fig. 5 is a schematic diagram of second edge information provided by an embodiment of the present application.

[0048] Figure 6 Fig. 6 is a schematic diagram of third edge information provided by an embodiment of the present application.

[0049] Figure 7 Fig. 7 is a schematic diagram of all the circular mark points of a global image provided by an embodiment of the present application. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments illustrated and described herein are merely exemplary and are not intended to limit the scope of the present application, which is defined by the appended claims.

[0051] The embodiment of the present application provides a global automatic detection method for near-photogrammetry circular mark points, as shown in the following steps S1-S5: Figure 1

[0052] S1, Canny algorithm is used to extract initial edge information containing all the positions of the circular mark points from the near-photogrammetry image.

[0053] In the embodiment of the present application, black and white circular mark points are used as the detection target in near-photogrammetry, the image taken by the near-photogrammetry camera containing the circular mark points is input, and the Canny algorithm is used for edge detection to obtain all the edge information of the whole image.

[0054] Step S1 includes the following sub-steps S11-S14:

[0055] S11, a two-dimensional Gaussian smoothing filter function G(i,j) is used to perform Gaussian filtering on the near-photogrammetry image f(i,j):

[0056]

[0057] Where (i,j) represents the coordinate point of the near-photogrammetry image, and sigma represents the standard deviation of the two-dimensional Gaussian smoothing filter function. After Gaussian filtering, the interference of noise can be eliminated, and the image can be effectively smoothed.

[0058] S12, the gradient amplitude L(i,j) and the gradient direction theta(i,j) of the Gaussian filtered image are calculated:

[0059]

[0060]

[0061] Where G x (i,j) and G y (i,j) represent the horizontal direction gradient and the vertical direction gradient at the point (i,j), respectively, and p(i,j) represents the pixel value at the point (i,j).

[0062] S13, whether the gradient amplitude of each pixel point in the Gaussian filtered image is greater than the gradient amplitude of the adjacent pixel point in the gradient direction of the pixel point is judged in sequence, if yes, the pixel point is determined as an edge, otherwise, the pixel point is determined as a non-edge, and thus the non-maximum suppression is performed on the Gaussian filtered image.

[0063] S14, after the non-maximum suppression, the image still has some pseudo-edges, therefore, in the embodiment of the present application, a double-threshold method is used to remove the pseudo-edges of the image to obtain the initial edge information containing all the positions of the circular mark points.

[0064] ​S2, taking the area size of the circular mark point, the long and short radii of the ellipse and the eccentricity of the ellipse as preliminary constraint conditions, removing the edge information of the non-circular mark point in the initial edge information, and obtaining first edge information.

[0065] Since the circular mark point is projected on the image as an ellipse through the center, the area size, the long and short radii of the ellipse and the eccentricity of the ellipse are taken as preliminary constraint conditions, and a large amount of edge information of the non-circular mark point can be removed.

[0066] In the embodiment of the application, the preliminary constraint condition is specifically:

[0067]

[0068] Wherein, area represents the area size of the circular mark point, Maj represents the long radius of the ellipse, and e represents the eccentricity of the ellipse.

[0069] S3, adopting a region growing algorithm to judge the gray scale information of eight adjacent domains around each pixel point in each region of the first edge information to determine whether the region is closed and connected, further removing the edge information of the non-circular mark point in the first edge information, and obtaining second edge information.

[0070] Step S3 includes the following sub-steps S31-S33:

[0071] S31, according to the closed characteristics of the circle, automatically selecting the upper left corner pixel point in each region of the first edge information as an initial seed point, taking the eight adjacent domains as a search space, and counting the number of the same gray scale.

[0072] S32, circularly traversing the number of the same gray scale of the eight adjacent domains around other edge pixel points of each region until all edge points in the region are counted.

[0073] S33, retaining all regions satisfying the closed and connected judgment condition, and obtaining the second edge information.

[0074] In the embodiment of the application, the closed and connected judgment condition is that the number of the same gray scale of the eight adjacent domains around the edge pixel point is greater than or equal to 3.

[0075] S4, using a morphological method and the circularity condition of all circular mark points to remove the edge information of the pseudo-circular mark point in the second edge information, and obtaining third edge information.

[0076] Step S4 includes the following sub-steps S41-S43:

[0077] S41, using a morphological reconstruction method based on geodesic dilation to perform hole filling processing on the second edge information:

[0078]

[0079] wherein represents a geodesic dilation of scale n of the marker image Y within the mask image X (where n > 0), which is obtained by iteratively applying a unit scale geodesic dilation of scale 1 to the marker image Y, and is obtained by iterating n times, and p represents a point in the mask image X, d X (p, Y) represents the geodesic distance of the point p to the marker image Y, represents an iterative computation, represents a dilation operation, and B represents a 3x3 4-connected matrix or an 8-connected matrix. In the embodiment of the present application, the marker image Y is the second edge information.

[0080] In the embodiment of the present application, the geodesic distance is defined as follows: assuming that a connected region is A, and a path P between two points a and b in the connected region A is completely contained in A, the geodesic distance between the two points a and b is defined as d A (a, b), at this time, the geodesic distance between the two points a and b in the connected region A is the shortest path in all paths, and if P does not exist, d A (a, b) = ∞.

[0081] In the embodiment of the present application, the hole filling effect can be realized by iterative computation of the basic geodesic dilation of the marker image Y in the mask image X.

[0082] S42, the roundness of each circular filling region in the second edge information after the hole filling processing is calculated.

[0083]

[0084] wherein m represents the roundness of the circular filling region, S represents the area of the circular filling region, and P represents the perimeter of the circular filling region.

[0085] S43, whether the roundness of each circular filling region is greater than or equal to a preset roundness threshold value is judged in turn, if yes, the edge information of the circular filling region is retained, otherwise the circular filling region is determined as a pseudo-circular marker point, and the edge information thereof is removed to obtain third edge information.

[0086] S5, all circular marker points are obtained from the third edge information by using Hough transform based on a two-stage algorithm.

[0087] Step S5 includes the following sub-steps S51-S52:

[0088] S51, the center feature of a circle is identified according to the gradient of each edge point on the boundary of the circular marker point in the third edge information, and a candidate center parameter is identified by local peak value detection.

[0089] S52. Second stage: Based on the candidate center parameters, construct a radius histogram using the circular constraint equation, mark the peak of the radius histogram as the center of the corresponding circular marker, and mark the corresponding edge points as the edge of the circular marker, and obtain all the circular markers.

[0090] In this embodiment of the invention, the circular constraint equation is specifically as follows:

[0091] (xx a ) 2 +(yx b ) 2 =r 2

[0092] Where (x, y) represents the position coordinates of the edge point, (x...y ... a ,x b ) represents the position coordinates of the center point, and r represents the radius of the constraint circle.

[0093] The following is a specific experimental example to further describe the effectiveness of the global automated detection method for circular markers provided in the embodiments of the present invention.

[0094] A high-speed close-up camera was used to capture images of the three-layer frame structure. Twenty-two circular markers were placed on the three-layer frame structure within the shooting area, such as... Figure 2 As shown; then, the method of the present invention is used for analysis and verification. Figure 3 The initial edge information containing the positions of all circular marker points is obtained by the Canny algorithm. Figure 4 This is the first edge information obtained based on preliminary constraints such as the region size of the circular marker points, the major and minor radii of the ellipse, and the eccentricity of the ellipse. Figure 5 The second edge information obtained by the region growing algorithm. Figure 6 The third edge information was obtained using morphological methods and roundness conditions. Figure 7 The results show that the method of this invention can automatically identify all circular markers on the entire image using the Hough transform algorithm based on a two-stage algorithm.

[0095] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for global automated detection of close-range photogrammetric circular mark points, characterized in that, The method comprises the following steps: S1, using Canny algorithm to extract initial edge information containing all circular mark point positions from the close-range photographic image; S2, taking the area size of the circular mark point, the long and short radii of the ellipse and the eccentricity of the ellipse as preliminary constraint conditions, removing the edge information of the non-circular mark points in the initial edge information to obtain first edge information; S3, using a region growing algorithm to determine whether each region in the first edge information is closed and connected by judging the gray information of the eight adjacent domains around each pixel point in the region, further removing the edge information of the non-circular mark points in the first edge information to obtain second edge information; S4, using a morphological method and the circularity condition of all the circular mark points to remove the edge information of the pseudo-circular mark points in the second edge information to obtain third edge information; S5, using a Hough transform based on a two-stage algorithm to obtain all the circular mark points from the third edge information. The step S4 comprises the following steps: S41, using a morphological reconstruction method based on geodesic dilation to perform hole filling processing on the second edge information; wherein denotes a point in the mask image X , Y denotes a geodesic dilation of the scale n 1 of the scale 1 unit scale geodesic dilation is iterated n times and , denotes a point in the mask image X , denotes the geodesic distance of the point p to the marker image Y , denotes an iterative computation, denotes a dilation operation, B denotes a 3x3 4-connected matrix or an 8-connected matrix; S42, calculating the circularity of each circular filling region in the second edge information after the hole filling processing; wherein a circularity of the circular fill area, S an area of the circular fill area, P a circumference of the circular fill area; S43, sequentially judging whether the circularity of each circular filling region is greater than or equal to a preset circularity threshold value, if yes, retaining the edge information of the circular filling region, otherwise, determining that the circular filling region is a pseudo-circular mark point and removing the edge information thereof to obtain the third edge information.

2. The globally automated detection method of close-range photogrammetry circular mark points according to claim 1, characterized in that, The step S1 comprises the following steps: S11, a two-dimensional Gaussian smoothing filter function is adopted to near-view photographic images Gaussian filtering is performed: wherein represents a coordinate point of a close-up photograph image, represents a standard deviation of a two-dimensional Gaussian smoothing filter function; S12, calculate gradient amplitude of the image after Gaussian filtering and gradient direction : wherein and respectively denote the horizontal and vertical gradient at the point , denotes the pixel value at the point . S13, sequentially judging whether the gradient amplitude of each pixel point in the image after the Gaussian filtering is greater than the gradient amplitude of the adjacent pixel points in the gradient direction of the pixel point, if yes, determining that the pixel point is an edge, otherwise, determining that the pixel point is a non-edge; S14, using a double-threshold method to remove the pseudo-edges of the image to obtain the initial edge information containing all the positions of the circular mark points.

3. The method of claim 1, wherein, The preliminary constraint conditions in the step S2 are specifically: wherein represents the area size of a circular marker point, represents the long radius of an ellipse, represents the eccentricity of an ellipse.

4. The method of claim 1, wherein, The step S3 comprises the following steps: S31, according to the closed characteristics of the circle, automatically selecting the upper left corner pixel point in each region of the first edge information as an initial seed point, taking the eight adjacent domains as a search space, and counting the number of the same gray scales; S32, cyclically traversing the number of the same gray scales of the eight adjacent domains around other edge pixel points of each region until all the edge points in the region are counted; S33, retaining all the regions satisfying the closed and connected judgment conditions to obtain the second edge information.

5. The globally automated detection method of close-range photogrammetry circular mark points according to claim 4, characterized in that, The closed and connected judgment condition in the step S33 is that the number of the same gray scales of the eight adjacent domains around the edge pixel point is greater than or equal to 3.

6. The globally automated detection method of close-range photogrammetry circular mark points according to claim 1, characterized in that, The step S5 comprises the following steps: S51, according to the gradient center identification of each edge point on the boundary of the circular mark point in the third edge information, and through local peak value detection, identifying candidate center parameters; S52, according to the candidate center parameters, using a circular constraint condition equation to construct a radius histogram, recording the peak value of the radius histogram as the center of the corresponding circular mark point, recording the corresponding edge point as the edge of the circular mark point, and obtaining all the circular mark points.

7. The globally automated detection method of close-range photogrammetry circular mark points according to claim 6, characterized in that, The circular constraint condition equation in the step S52 is specifically: wherein denotes the position coordinates of the edge point, denotes the position coordinates of the center point, r denotes the radius of the constraint circle.

Images