CN112270717A - Checkerboard angular point detection method and device - Google Patents
Checkerboard angular point detection method and device Download PDFInfo
- Publication number
- CN112270717A CN112270717A CN202011265345.1A CN202011265345A CN112270717A CN 112270717 A CN112270717 A CN 112270717A CN 202011265345 A CN202011265345 A CN 202011265345A CN 112270717 A CN112270717 A CN 112270717A
- Authority
- CN
- China
- Prior art keywords
- checkerboard
- corner
- energy
- candidate
- extending
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30204—Marker
- G06T2207/30208—Marker matrix
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30244—Camera pose
Abstract
A checkerboard corner detection method and device, the method includes the following steps: acquiring an image and determining candidate corner points in the image; determining the number of corner point peak values of each candidate corner point and the size and the direction of each corner point peak value within the range of 0-2 pi in the boundary direction; removing candidate angular points with the angular point peak value number smaller than 4; and for each candidate corner point with the number of the corner point peak values being more than or equal to 4, determining whether to reserve the candidate corner point according to the comparison result and the direction relation of the sizes of different corner point peak values of the candidate corner point. The scheme of the invention can reduce the calculation amount in the detection process and effectively improve the detection efficiency of the candidate corner points.
Description
Technical Field
The invention relates to the technical field of images, in particular to a checkerboard corner point detection method and device.
Background
In applications of mobile camera devices, the camera head is identified as an important and indispensable step. The calibration step mainly comprises the steps of detecting corner points (triples) based on patterns of square checkerboards or circular checkerboards, and then solving internal parameters, external parameters, lens distortion parameters and the like according to the positions of the corner points.
In the existing checkerboard corner detection method, candidate corners are determined based on a Harris equiangular point detection algorithm, and then the number of corner peak values of each candidate corner, and the size and direction of each corner peak value are determined within a range of 0-pi in the boundary direction. Then, for candidate angular points with the number of peak values being more than or equal to 2 and the included angle of the direction of the peak value of each angular point being more than 0.3/pi multiplied by 180, the angular point score is obtained, and when the angular point score is more than or equal to 0.01, the candidate angular points are reserved.
In the prior art, a large amount of computing resources are consumed to determine the corner score of each candidate corner, so that the corner detection efficiency is low, and the corner score is only used in the corner detection stage, is not used in the subsequent checkerboard generation stage, and is low in utilization rate.
Furthermore, in the prior art, in order to reduce the amount of calculation, the peak value of the corner point is analyzed only in the range of 0 to pi in the boundary direction, so that the determination of each candidate corner point is not comprehensive enough, and the analysis result based on the peak value of the corner point is not accurate enough.
Disclosure of Invention
The invention solves the technical problem of providing a checkerboard corner detection method and a checkerboard corner detection device, which can reduce the calculation amount in the detection process and effectively improve the detection efficiency of candidate corners.
In order to solve the above technical problem, an embodiment of the present invention provides a checkerboard corner detection method, including the following steps: acquiring an image and determining candidate corner points in the image; determining the number of corner point peak values of each candidate corner point and the size and the direction of each corner point peak value within the range of 0-2 pi in the boundary direction; removing candidate angular points with the angular point peak value number smaller than 4; and for each candidate corner point with the number of the corner point peak values being more than or equal to 4, determining whether to reserve the candidate corner point according to the comparison result and the direction relation of the sizes of different corner point peak values of the candidate corner point.
Optionally, determining whether to reserve the candidate corner according to the comparison result of the sizes of the different corner peak values of the candidate corner and the direction relationship includes: sorting according to the sizes of the corner point peak values of the candidate corner points; if the difference value between the peak value of the first large-angle point and the peak value of the fourth large-angle point is larger than or equal to a first threshold value, classifying the candidate angle points as bright angle points, otherwise classifying the candidate angle points as dark angle points; and determining whether to reserve the candidate corner points or not by adopting different judgment conditions for the bright corner points and the dark corner points.
Optionally, the judgment condition for the bright corner point includes: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, deleting the candidate angle point.
Optionally, the judgment condition for the bright corner point includes: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, or the difference between the second big-angle peak value and the third big-angle peak value of the candidate angle point is larger than a fourth threshold value, deleting the candidate angle point.
Optionally, the judgment condition for the dark corner point includes: and if the quotient of the fifth large-angle peak value and the fourth large-angle peak value of the candidate angle point is larger than an eighth threshold value, deleting the candidate angle point.
Optionally, the judgment condition for the dark corner point includes: and if the quotient of the fifth large-angle-point peak value and the fourth large-angle-point peak value of the candidate angular point is larger than an eighth threshold value, or the minimum approach angle of the peak value direction of the candidate angular point is smaller than a seventh threshold value, deleting the candidate angular point.
Optionally, the checkerboard corner detecting method further includes:
generating an initial checkerboard based on the reserved angular points;
extending the initial checkerboards to four directions by one line, respectively obtaining first extended checkerboards after extending in each direction, respectively calculating the checkerboard energy of the four first extended checkerboards, if the first extended checkerboard with the lowest checkerboard energy in the four first extended checkerboards is lower than the checkerboard energy of the initial checkerboard, keeping the first extended checkerboard with the lowest checkerboard energy as the first checkerboard, and taking the four directions as the directions vertical to the four sides of the initial checkerboard;
selecting two directions from the four directions, wherein the checkerboard energy of the first extended checkerboard in the selected two directions is lower than the checkerboard energy of the first extended checkerboard in the other two directions;
extending the first checkerboard for multiple times towards the selected two directions, extending the first checkerboard for one line every time, obtaining second extended checkerboards after extending in each direction, calculating the checkerboard energy of the two second extended checkerboards respectively, if the energy of the second extended checkerboard with the lowest checkerboard energy in the two second extended checkerboards is lower than that of the first checkerboard, keeping the second extended checkerboard with the lowest checkerboard energy as the basis of the next extension until the checkerboard energy of the second extended checkerboard obtained after extension in the two directions is equal, or the number of lines of the second extended checkerboard obtained after extension in any one direction of the two directions reaches a preset number of lines, so as to obtain a first repeated extended checkerboard;
and if the third extension checkerboard with the lowest checkerboard energy in the two third extension checkerboards is lower than the energy of the first extension checkerboard, the third extension checkerboard with the lowest checkerboard energy is reserved as the basis of the next extension until the checkerboard energy of the third extension checkerboard in the other two directions is equal, or the number of rows of the third extension checkerboard in any one of the two directions reaches a preset number of rows after extension, so as to obtain the second extension checkerboard.
In order to solve the above technical problem, an embodiment of the present invention provides a checkerboard corner detecting device, including: the candidate determining module is suitable for acquiring an image and determining candidate corner points in the image; the corner peak value determining module is suitable for determining the number of corner peak values of each candidate corner and the size and the direction of each corner peak value within the range of 0-2 pi in the boundary direction; a removing module, adapted to remove candidate corner points whose number of corner point peaks is less than 4; and the determining module is suitable for determining whether to reserve the candidate corner points or not according to the comparison result and the direction relation of the sizes of different corner point peak values of the candidate corner points for each candidate corner point with the number of the corner point peak values being more than or equal to 4.
Optionally, the determining module includes: a sorting submodule adapted to sort according to the magnitude of the corner peak of the candidate corner; the classification sub-module is suitable for classifying the candidate corner points into bright corner points when the difference value between the peak value of the first large corner point and the peak value of the fourth large corner point is larger than or equal to a first threshold value, and classifying the candidate corner points into dark corner points if the difference value is not larger than the first threshold value; and the determining submodule is suitable for determining whether to reserve the candidate corner points or not by adopting different judging conditions for the bright corner points and the dark corner points.
Optionally, the determining conditions of the determining sub-module for the bright corner point include: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, deleting the candidate angle point.
Optionally, the determining conditions of the determining sub-module for the bright corner point include: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, or the difference between the second big-angle peak value and the third big-angle peak value of the candidate angle point is larger than a fourth threshold value, deleting the candidate angle point.
Optionally, the determining sub-module determines a dark corner point according to a determination condition that: and if the quotient of the fifth large-angle peak value and the fourth large-angle peak value of the candidate angle point is larger than an eighth threshold value, deleting the candidate angle point.
Optionally, the determining sub-module determines a dark corner point according to a determination condition that: and if the quotient of the fifth large-angle-point peak value and the fourth large-angle-point peak value of the candidate angular point is larger than an eighth threshold value, or the minimum approach angle of the peak value direction of the candidate angular point is smaller than a seventh threshold value, deleting the candidate angular point.
Optionally, the checker corner detection apparatus further includes: the generating module is suitable for generating an initial checkerboard based on the reserved angular points; a first extension module, adapted to extend the initial checkerboard in four directions by one line, obtain first extension checkerboards after extending in each direction, calculate checkerboard energies of the four first extension checkerboards, respectively, if a first extension checkerboard with the lowest checkerboard energy among the four first extension checkerboards is lower than the checkerboard energy of the initial checkerboard, retain the first extension checkerboard with the lowest checkerboard energy as the first checkerboard, where the four directions are directions perpendicular to four sides of the initial checkerboard; a direction selection module adapted to select two directions among the four directions, a first extended checkerboard energy in the selected two directions being lower than checkerboard energy of the first extended checkerboard in the remaining two directions; a second extending module, adapted to extend the first checkerboard multiple times in the selected two directions, each time extending for one line, obtaining a second extended checkerboard after extending in each direction, respectively calculating checkerboard energy of the two second extended checkerboards, if the energy of the second extended checkerboard with the lowest checkerboard energy is lower than the energy of the first checkerboard in the two second extended checkerboards, retaining the second extended checkerboard with the lowest checkerboard energy as a basis for next extending until the energy of the second extended checkerboard obtained after extending in the two directions is equal, or the number of lines of the second extended checkerboard obtained after extending in any one of the two directions reaches a preset number of lines, so as to obtain a first repeated extended checkerboard; and a third extending module, adapted to extend the first repeated extending checkerboard to the other two directions for multiple times, each time extending for one line, obtaining a third extending checkerboard after extending in each direction, and calculating the checkerboard energy of the two third extending checkerboards, if the energy of the third extending checkerboard with the lowest checkerboard energy in the two third extending checkerboards is lower than the energy of the first repeated extending checkerboard, keeping the third extending checkerboard with the lowest checkerboard energy as a basis for the next extending until the checkerboard energy of the third extending checkerboard obtained after extending in the other two directions is equal, or the number of lines of the third extending checkerboard obtained after extending in any one of the two directions reaches a preset number of lines, so as to obtain a second repeated extending checkerboard.
Compared with the prior art, the technical scheme of the embodiment of the invention has the following beneficial effects:
in the embodiment of the invention, an image is obtained, and candidate corner points in the image are determined; determining the number of corner point peak values of each candidate corner point and the size and the direction of each corner point peak value within the range of 0-2 pi in the boundary direction; removing candidate angular points with the angular point peak value number smaller than 4; and for each candidate corner point with the number of the corner point peak values being more than or equal to 4, determining whether to reserve the candidate corner point according to the comparison result and the direction relation of the sizes of different corner point peak values of the candidate corner point. By adopting the scheme, the corner peak value is determined in a larger boundary range (namely 0-2 pi) of each candidate corner, and compared with the prior art that the corner peak value is analyzed only in the range of 0-pi in the boundary direction, all the corner peak values of each candidate corner can be determined more comprehensively, so that a more accurate result can be obtained in the subsequent analysis of the corner peak values; compared with the prior art that the corner score is obtained for the candidate corner, the reserved candidate corner is determined based on the corner score, the calculation amount in the detection process can be reduced, and the detection efficiency of the candidate corner is effectively improved.
Further, in the embodiment of the present invention, the candidate corner points are classified into bright corner points and dark corner points according to the magnitude sorting of the peak values of the candidate corner points, and then different determination conditions are adopted for the bright corner points and the dark corner points to determine whether to retain the candidate corner points. Compared with the prior art that the same screening condition is adopted for the bright corner points and the dark corner points, the peak values of the bright corner points are larger than the peak values of the dark corner points under the condition that other backgrounds are the same, so that the effective corner points at the dark and the bright positions are easy to be deleted by mistake when the invalid corner points at the bright positions are deleted.
Furthermore, by adopting the scheme of the embodiment of the invention, in the growth process of the checkerboards, only the checkerboards are extended towards two directions by extending one line, and then the checkerboard energy extended towards two directions is calculated, so that the checkerboards are extended towards one direction in the two directions.
Drawings
FIG. 1 is a flow chart of a method for detecting a checkerboard corner in an embodiment of the present invention;
fig. 2 is a schematic diagram of a candidate corner point in the embodiment of the present invention;
FIG. 3 is a flowchart of one specific implementation of step S14 in FIG. 1;
FIG. 4 is a partial flow diagram of another checkerboard corner detection method in accordance with an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a checkerboard corner detection apparatus in an embodiment of the present invention;
FIG. 6 is a block diagram illustrating one embodiment of determination module 54 of FIG. 5;
fig. 7 is a schematic partial structure diagram of another checker corner detection apparatus according to an embodiment of the present invention.
Detailed Description
In the existing checkerboard corner detection method, a large amount of computing resources are consumed to determine the corner score of each candidate corner, and whether the candidate corner is reserved is determined based on the corner score, so that the existing corner detection efficiency is low.
The inventor of the present invention finds, through research, that the corner score of each candidate corner is not an essential parameter, and can determine the remaining candidate corner by comparing the sizes of the different corner peak values of the candidate corner and analyzing the direction relationship of the different corner peak values of the candidate corner.
In the embodiment of the invention, an image is obtained, and candidate corner points in the image are determined; determining the number of corner point peak values of each candidate corner point and the size and the direction of each corner point peak value within the range of 0-2 pi in the boundary direction; removing candidate angular points with the angular point peak value number smaller than 4; and for each candidate corner point with the number of the corner point peak values being more than or equal to 4, determining whether to reserve the candidate corner point according to the comparison result and the direction relation of the sizes of different corner point peak values of the candidate corner point. By adopting the scheme, the corner peak value is determined in the larger boundary of each candidate corner, and compared with the prior art that the corner peak value is analyzed only in the range of 0-pi in the boundary direction, all the corner peak values of each candidate corner can be determined more comprehensively so as to obtain more accurate results in the subsequent analysis of the corner peak values; compared with the prior art that the corner score is obtained for the candidate corner, the reserved candidate corner is determined based on the corner score, the calculation amount can be reduced, and the detection efficiency of the candidate corner is effectively improved.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
Referring to fig. 1, fig. 1 is a flowchart of a checkerboard corner detection method in an embodiment of the present invention. The checkerboard corner detection method may include steps S11 through S14.
Step S11: acquiring an image and determining candidate corner points in the image;
step S12: determining the number of corner point peak values of each candidate corner point and the size and the direction of each corner point peak value within the range of 0-2 pi in the boundary direction;
step S13: removing candidate angular points with the angular point peak value number smaller than 4;
step S14: and for each candidate corner point with the number of the corner point peak values being more than or equal to 4, determining whether to reserve the candidate corner point according to the comparison result and the direction relation of the sizes of different corner point peak values of the candidate corner point.
In a specific implementation of step S11, an existing corner detection method may be used to determine candidate corners in the acquired image. Specifically, a Harris corner detection algorithm, a Susan corner detection algorithm, or the like may be employed.
In a specific implementation of the embodiment of the present invention, Corner detection is performed based on a single-scale picture and two-direction Corner models (Corner models), which are an unrotated model and a 45 ° rotated model, respectively.
Specifically, each corner model may have four different filters, which are denoted as { a, B, C, D }, and the four filters are respectively convolved with the single-scale picture, and the obtained response label is denoted as { (a, B, C, D) } or
Wherein i is used to indicate which of the two orientations is the corner model, e.g. different numbers may be used to indicate different corner models; x is used to indicate which of the four different filters the filter is, e.g. different numbers may be used to indicate different filters. It should be noted that in the embodiment of the present invention, although only two-directional corner models are used, in other specific applications, more models may be included, for example, models with different dimensions may be included.
Further, the corner similarity map C is obtained by the following formula:
wherein μ is an average of response labels obtained by four different filters, c is a corner similarity map, and candidate corners can be solved according to c through non-maximization fitting.
It should be noted that, with the solution of the embodiment of the present invention, a specific implementation manner for determining candidate corner points in an image is not limited.
Fig. 2 is a schematic diagram of a candidate corner point in the embodiment of the present invention. As shown in fig. 2, the obtained candidate angular points include an effective angular point and an ineffective angular point, the angular point of the checkerboard for camera calibration should be a critical point at the boundary of the black and white grids in the checkerboard, and is adjacent to the four black and white grids, for example, the candidate angular point a shown in fig. 2 is an effective angular point; the invalid corner points are candidate corner points at other positions than the valid corner points, such as candidate corner points b located on the boundary lines of the black and white grids, candidate corner points c located at the boundary points of the black grids and the white areas at the edges of the chessboard, candidate corner points d located on the black borders of the chessboard, and candidate corner points e beyond the range of the chessboard.
With reference to fig. 1, in the specific implementation of step S12, in the range of 0-2 pi in the boundary direction of each candidate corner point, a peripheral pixel point of the candidate corner point, for example, a pixel point within 10 pixels from the candidate corner point, may be selected, where each pixel point has a gradient and a direction. And further dividing the boundary direction 0-2 pi into 32 parts, counting the direction of each pixel point, and adding the gradients of all the pixel points in each angle to obtain a weighted value in each angle, namely a 32-dimensional histogram. And performing Gaussian filtering on the 32-dimensional data in the histogram to obtain a smooth histogram. And further, solving a local maximum value in the smoothed histogram through non-maximization fitting so as to obtain a corner peak value of the candidate corner. It should be noted that, by using the scheme of the embodiment of the present invention, an implementation manner of specifically determining the corner peak value of each candidate corner point is not limited.
There may be a plurality of corner peak values calculated for each candidate corner, and each corner peak value has a magnitude and a direction.
In a specific implementation of step S13, if the number of corner peak values of a candidate corner point is less than 4, the candidate corner point is removed.
Specifically, since the checkerboard pattern is symmetrically distributed in the neighboring regions around the effective corner points, the corner point peak values in four directions around the effective corner points should be large and close in size. If the number of corner peak values of a candidate corner is less than 4, the candidate corner can be determined as an invalid corner, and the candidate corner is removed.
In a specific implementation of step S14, after removing the candidate corner points whose number of corner point peaks is less than 4, it is further determined whether to keep the candidate corner points.
Referring to fig. 3, determining whether to reserve the candidate corner according to the comparison result of the sizes of the different corner peak values of the candidate corner and the direction relationship may be implemented through steps S31 to S33, and each step is described in detail below.
Step S31: and sorting according to the sizes of the corner peak values of the candidate corner points.
Step S32: and if the difference value of the peak value of the first large-angle point and the peak value of the fourth large-angle point is larger than or equal to a first threshold value, classifying the candidate angular point as a bright angular point, otherwise classifying the candidate angular point as a dark angular point.
In particular implementation, as a non-limiting example, the first threshold value may be selected from a value of 0.6 to 0.9.
Step S33: and determining whether to reserve the candidate corner points or not by adopting different judgment conditions for the bright corner points and the dark corner points.
In a specific implementation of the embodiment of the present invention, the condition for determining the bright corner point may include: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, deleting the candidate angle point.
Candidate corners inside the white grid or in the white area of the edge of the chessboard in the chessboard and corners outside the chessboard, such as candidate corner e shown in fig. 2, can be deleted by the condition that the peak value of the first big angle point is smaller than the second threshold value, or the sum of the peak values of the first big angle point and the fourth big angle point is smaller than the third threshold value. Because the checkerboard patterns are symmetrically distributed in the adjacent regions around the effective corner points, corner point peak values in four directions are larger and close to each other around the effective corner points, when the first large-corner peak value of the candidate corner point is smaller, or the sum of the first large-corner peak value to the fourth large-corner peak value is smaller, each corner point peak value of the candidate corner points can be judged to be smaller, and is possibly an invalid corner point, and then the candidate corner points are deleted.
It should be noted that, when the determination is performed based on the above conditions, if the bright corner point and the dark corner point are not classified, the valid dark corner point is easily deleted by mistake. Specifically, the dark corner points have low calculated corner point peak values in each direction due to the small weight of the pixels in each direction, and when the first large-corner peak value of the effective dark corner point is smaller than the second threshold value, or the sum of the first large-corner peak value and the fourth large-corner peak value is smaller than the third threshold value, the false deletion is possible.
In a particular application, as a non-limiting example, the second threshold value may be selected from a value of 2.3 to 2.7 and the third threshold value may be selected from a value of 9 to 11.
Further, candidate corner points on the boundary line between the black and white grid in the checkerboard and candidate corner points on the boundary line between the black grid and the white area at the edge of the checkerboard, such as candidate corner point b shown in fig. 2, may be deleted by two conditions that the difference between the first and second peak of the candidate corner points is greater than a fifth threshold, or the quotient between the second and first peak of the candidate corner points and the first peak of the candidate corner points is less than a sixth threshold. Because the checkerboard patterns are symmetrically distributed in the adjacent areas around the effective corner points, corner point peak values in four directions around the effective corner points are large and close to each other in size, when the difference value between the first large-corner peak value and the second large-corner peak value of the candidate corner point is large, the candidate corner point is judged to be possibly an invalid corner point, and then the candidate corner point is deleted.
In a particular application, as a non-limiting example, the fifth threshold value may be selected from a value of 10 to 12 and the sixth threshold value may be selected from a value of 0.5 to 0.7.
In another specific implementation of the embodiment of the present invention, the determining condition for the bright corner point may include: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, or the difference between the second big-angle peak value and the third big-angle peak value of the candidate angle point is larger than a fourth threshold value, deleting the candidate angle point.
And by the condition that the difference value between the second large-corner peak value and the third large-corner peak value of the candidate corner points is greater than the fourth threshold, the candidate corner points at the boundary between the black grid and the white area of the chessboard edge, such as the candidate corner point c shown in fig. 2, can be deleted. Because the checkerboard patterns are symmetrically distributed in the adjacent regions around the effective corner points, corner point peak values in four directions are larger and close to each other around the effective corner points, when the difference value between the second large-corner peak value and the third large-corner peak value of the candidate corner point is larger, the candidate corner points can be judged to have larger corner point peak values in two directions only, and the corner point peak values in the other directions are smaller, so that the candidate corner points are likely to be invalid corner points, and then the candidate corner points are deleted.
In a particular application, as a non-limiting example, the fourth threshold value may be selected from a value of 7 to 9.
For a detailed analysis of the determination conditions of other bright corners, please refer to the related description in the previous specific implementation of the embodiment of the present invention, which is not further described herein.
In another specific implementation of the embodiment of the present invention, the condition for determining the dark corner point may include: and if the quotient of the fifth large-angle peak value and the fourth large-angle peak value of the candidate angle point is larger than an eighth threshold value, deleting the candidate angle point.
Specifically, when the number of corner peak values of the candidate corner points is greater than or equal to 5, candidate corner points whose positions have colors close to the background color may be deleted in the dark outside the checkerboard range by the condition that the quotient of the fifth vertex peak value and the fourth vertex peak value of the candidate corner points is greater than the eighth threshold value.
Further, since the checkerboard pattern is symmetrically distributed in the neighboring region around the effective corner, the corner peak values in four directions around the effective corner are relatively large and close to each other, and if the fifth large corner peak value is relatively close to the fourth large corner peak value, it can be considered that the candidate corner has more than four corner peak values, and the candidate corner is likely to be an invalid corner, and then the candidate corner is deleted.
In a particular application, as a non-limiting example, the eighth threshold value may be selected from a value of 0.6 to 0.8.
In another specific implementation of the embodiment of the present invention, the condition for determining the dark corner point may include: and if the quotient of the fifth large-angle-point peak value and the fourth large-angle-point peak value of the candidate angular point is larger than an eighth threshold value, or the minimum approach angle of the peak value direction of the candidate angular point is smaller than a seventh threshold value, deleting the candidate angular point.
And deleting the candidate corner points detected by the pure white or pure black area by the condition that the minimum approach angle of the peak directions of the candidate corner points is smaller than a seventh threshold value. Since the checkerboard pattern is symmetrically distributed in the neighboring regions around the effective corner, the neighboring included angles between the directions of the first four peaks around the effective corner should be relatively close, for example, all close to 90 °, and if there is a case that the minimum included angle is very small, it can be considered that the candidate corner is most likely to be an invalid corner, and the candidate corner is deleted.
In a particular application, as a non-limiting example, said seventh threshold is selected from 10 ° to 50 °.
For detailed analysis of the determination conditions of other dark corner points, please refer to the related description in the previous specific implementation of the embodiment of the present invention, which is not described herein again.
It should be noted that, due to the complexity of the environment in which the checkerboard is located in practical applications, the candidate corner points that can be deleted by the above-mentioned determination condition for the bright corner points or the dark corner points are not limited to the above-mentioned listed areas.
By adopting the scheme of the embodiment of the invention, the corner peak value is determined in the larger boundary of each candidate corner, and compared with the prior art that the corner peak value is analyzed only in the range of 0-pi in the boundary direction, all the corner peak values of each candidate corner can be determined more comprehensively, so that more accurate results can be obtained in the subsequent analysis of the corner peak values; compared with the prior art that the corner score is obtained for the candidate corner, the reserved candidate corner is determined based on the corner score, the calculation amount can be reduced, and the detection efficiency of the candidate corner is effectively improved.
And further, classifying the candidate angular points into bright angular points and dark angular points according to the magnitude sequence of the angular point peak values of the candidate angular points, and determining whether to reserve the candidate angular points or not by adopting different judgment conditions for the bright angular points and the dark angular points. Compared with the prior art that the same screening condition is adopted for the bright corner points and the dark corner points, the peak values of the bright corner points are larger than the peak values of the dark corner points under the condition that other backgrounds are the same, so that the effective corner points at the dark and the bright positions are easy to be deleted by mistake when the invalid corner points at the bright positions are deleted.
The corner points obtained in the checkerboard corner point detection method shown in fig. 1 to 3 are the corner points of the image, and all the corner points of each checkerboard need to be determined in the subsequent camera calibration process, so that the checkerboard needs to be generated through the corner points of the image.
In the existing method for generating the checkerboards, an initial checkerboard is generally generated, the initial checkerboard is extended to four directions for one line, extended checkerboards are respectively obtained after the initial checkerboard is extended in each direction, the checkerboard energy of the four extended checkerboards is respectively calculated, and if the extended checkerboard with the lowest checkerboard energy in the four extended checkerboards is lower than the checkerboard energy of the initial checkerboard, the extended checkerboard with the lowest checkerboard energy is reserved as the basis of subsequent extension. And repeating the extending steps until the checkerboard energy of the checkerboard obtained after extending in four directions is equal, or the checkerboard obtained after extending reaches a preset size.
The energy of the checkerboard can be calculated by a Geiger method, for example, by the following formula:
E(x，y)＝Ecorners(y)+Estruct(x，y)；
Ecorners(y)＝-|{y|y≠0}|；
wherein E iscorners(y) a negative value representing the total number of corner points in the checkerboard; estruct(x, y) predicting the matching degree of a third angular point through two adjacent angular points; e (x, y) is the minimum energy measure for the current checkerboard.
Further, by calculating different Estruct(x, y) calculating the structural energy of each line and each column of adjacent three corner points in the checkerboard as a combination, and taking the maximum value of the structural energy calculated by different combinations as the checkerboard energy of the checkerboard.
In a specific implementation, since the calculation amount and the size of the checkerboard are in an exponential relationship, when the size of the checkerboard increases, the calculation amount will increase rapidly, so an optimized method for generating the checkerboard is urgently needed to reduce the calculation amount.
Fig. 4 is a partial flow chart of another checkerboard corner detection method in an embodiment of the present invention. The other checkerboard corner detecting method may include steps S41 to S45, which are described in detail below.
Step S41: based on the remaining corner points, an initial checkerboard is generated.
In a specific implementation, the initial checkerboard may comprise a 3 x 3 grid as a basis for subsequent extensions. In the embodiment of the present invention, the specific method for generating the initial checkerboard is not limited.
Step S42: and if the first extension checkerboard with the lowest checkerboard energy in the four first extension checkerboards is lower than the checkerboard energy of the initial checkerboard, the first extension checkerboard with the lowest checkerboard energy is reserved as the first checkerboard, and the four directions are the directions perpendicular to the four sides of the initial checkerboard.
Step S43: two directions are selected among the four directions, the checkerboard energy of the first extended checkerboard in the selected two directions being lower than the checkerboard energy of the first extended checkerboard in the remaining two directions.
Step S44: and if the second extension checkerboard with the lowest checkerboard energy in the two second extension checkerboards is lower than the energy of the first checkerboard, the second extension checkerboard with the lowest checkerboard energy is kept as the basis of the next extension until the checkerboard energy of the second extension checkerboard obtained after extension in the two directions is equal, or the number of rows of the second extension checkerboard obtained after extension in any one direction of the two directions reaches a preset number of rows, so as to obtain the first repeated extension checkerboard.
Step S45: and if the third extension checkerboard with the lowest checkerboard energy in the two third extension checkerboards is lower than the energy of the first extension checkerboard, the third extension checkerboard with the lowest checkerboard energy is reserved as the basis of the next extension until the checkerboard energy of the third extension checkerboard in the other two directions is equal, or the number of rows of the third extension checkerboard in any one of the two directions reaches a preset number of rows after extension, so as to obtain the second extension checkerboard.
By adopting the scheme of the embodiment of the invention, in the growth process of the checkerboards, only the checkerboards extend towards two directions by extending one line, and then the checkerboard energy after extending towards two directions is calculated, so that the checkerboards are extended towards one direction in the two directions.
Fig. 5 is a schematic structural diagram of a checkerboard corner detecting device in an embodiment of the present invention, where the checkerboard corner detecting device may include: a candidate determination module 51, a corner peak determination module 52, a removal module 53 and a determination module 54.
The candidate determining module 51 is adapted to acquire an image and determine candidate corner points in the image;
the corner peak determining module 52 is adapted to determine the number of corner peaks of each candidate corner and the size and direction of each corner peak within a range of 0-2 pi in the boundary direction;
the removing module 53 is adapted to remove the candidate corner points with the number of corner point peaks smaller than 4;
the determining module 54 is adapted to determine, for each candidate corner point with the number of corner point peak values being greater than or equal to 4, whether to retain the candidate corner point according to the comparison result and the direction relationship of the sizes of the different corner point peak values of the candidate corner point.
Further, fig. 6 shows a schematic structural diagram of a specific implementation of the determining module 54, and the determining module 54 may include an ordering sub-module 541, a classifying sub-module 542, and a determining sub-module 543.
Wherein, the sorting submodule 541 is adapted to sort according to the sizes of the corner peak values of the candidate corners;
the classification submodule 542 is adapted to classify the candidate corner point as a bright corner point when a difference between the peak value of the first large corner point and the peak value of the fourth large corner point is greater than or equal to a first threshold value, and classify the candidate corner point as a dark corner point otherwise;
the determining sub-module 543 is adapted to determine whether to reserve the candidate corner by using different determining conditions for the bright corner and the dark corner.
Still further, the determining sub-module 543 may determine the condition for the bright corner point by: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, deleting the candidate angle point.
Still further, the determining sub-module 543 may determine the condition for the bright corner point by: and if the first big-angle peak value of the candidate angle point is smaller than a second threshold value, or the sum of the first big-angle peak value and a fourth big-angle peak value of the candidate angle point is smaller than a third threshold value, or the difference between the first big-angle peak value and the second big-angle peak value of the candidate angle point is larger than a fifth threshold value, or the quotient of the second big-angle peak value and the first big-angle peak value of the candidate angle point is smaller than a sixth threshold value, or the difference between the second big-angle peak value and the third big-angle peak value of the candidate angle point is larger than a fourth threshold value, deleting the candidate angle point.
Further, the determining condition of the determining sub-module 543 for the dark corner point may include: and if the quotient of the fifth large-angle peak value and the fourth large-angle peak value of the candidate angle point is larger than an eighth threshold value, deleting the candidate angle point.
Further, the determining condition of the determining sub-module 543 for the dark corner point may include: and if the quotient of the fifth large-angle-point peak value and the fourth large-angle-point peak value of the candidate angular point is larger than an eighth threshold value, or the minimum approach angle of the peak value direction of the candidate angular point is smaller than a seventh threshold value, deleting the candidate angular point.
For more details of the checker corner detection apparatus, please refer to the related description of the checker corner detection method shown in fig. 1 to 3, and the description thereof is omitted here for brevity.
Fig. 7 is a schematic partial structure diagram of another checker corner detection apparatus according to an embodiment of the present invention. The further checker corner detection means may comprise a generating module 71, a first extending module 72, a direction selecting module 73, a second extending module 74 and a third extending module 75.
Wherein the generating module 71 is adapted to generate an initial checkerboard based on the reserved corner points;
the first extending module 72 is adapted to extend the initial checkerboard to four directions by one line, obtain first extended checkerboards after extending in each direction, calculate checkerboard energies of the four first extended checkerboards, respectively, if a first extended checkerboard with the lowest checkerboard energy among the four first extended checkerboards is lower than the checkerboard energy of the initial checkerboard, keep the first extended checkerboard with the lowest checkerboard energy as the first checkerboard, where the four directions are directions perpendicular to four sides of the initial checkerboard;
said direction selection module 73 adapted to select two directions among said four directions, a first extended checkerboard energy in the selected two directions being lower than the checkerboard energy of the first extended checkerboard in the remaining two directions;
the second extending module 74 is adapted to extend the first checkerboard for multiple times in the selected two directions, each time extend for one line, obtain a second extended checkerboard after extending in each direction, calculate checkerboard energies of the two second extended checkerboards, respectively, if the energy of the second extended checkerboard with the lowest checkerboard energy in the two second extended checkerboards is lower than the energy of the first checkerboard, keep the second extended checkerboard with the lowest checkerboard energy as a basis for the next extending until the energy of the second extended checkerboard obtained after extending in the two directions is equal, or the number of lines of the second extended checkerboard obtained after extending in any one of the two directions reaches a preset number of lines, so as to obtain a first repeated extended checkerboard;
the third extending module 75 is adapted to extend the first repeated extending checkerboard to the other two directions for multiple times, each time extend for one line, obtain a third extending checkerboard after extending in each direction, calculate checkerboard energy of two third extending checkerboards, if the energy of the third extending checkerboard with the lowest checkerboard energy in the two third extending checkerboards is lower than the energy of the first repeated extending checkerboard, keep the third extending checkerboard with the lowest checkerboard energy as a basis for next extending until the energy of the third extending checkerboard obtained after extending in the other two directions is equal, or the number of lines of the third extending checkerboard obtained after extending in any one of the two directions reaches a preset number of lines, so as to obtain a second repeated extending checkerboard.
For more details about the another checkered corner detecting apparatus, please refer to the related description about another checkered corner detecting method shown in fig. 4 and the foregoing, and details are not repeated herein.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer-readable storage medium, and the storage medium may include: ROM, RAM, magnetic or optical disks, and the like.
Although the present invention is disclosed above, the present invention is not limited thereto. Various changes and modifications may be effected therein by one skilled in the art without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (2)
1. A checkerboard corner detection method is characterized by comprising the following steps:
acquiring an image and determining candidate corner points in the image;
determining the number of corner point peak values of each candidate corner point and the size and the direction of each corner point peak value within the range of 0-2 pi in the boundary direction;
removing candidate angular points with the angular point peak value number smaller than 4;
for each candidate corner point with the number of the corner point peak values being more than or equal to 4, determining whether to reserve the candidate corner point according to the comparison result and the direction relation of the sizes of different corner point peak values of the candidate corner point;
the checkerboard corner detection method further comprises the following steps:
generating an initial checkerboard based on the reserved angular points;
extending the initial checkerboards to four directions by one line, respectively obtaining first extended checkerboards after extending in each direction, respectively calculating the checkerboard energy of the four first extended checkerboards, if the first extended checkerboard with the lowest checkerboard energy in the four first extended checkerboards is lower than the checkerboard energy of the initial checkerboard, keeping the first extended checkerboard with the lowest checkerboard energy as the first checkerboard, and taking the four directions as the directions vertical to the four sides of the initial checkerboard;
selecting two directions from the four directions, wherein the checkerboard energy of the first extended checkerboard in the selected two directions is lower than the checkerboard energy of the first extended checkerboard in the other two directions;
extending the first checkerboard for multiple times towards the selected two directions, extending the first checkerboard for one line every time, obtaining second extended checkerboards after extending in each direction, calculating the checkerboard energy of the two second extended checkerboards respectively, if the energy of the second extended checkerboard with the lowest checkerboard energy in the two second extended checkerboards is lower than that of the first checkerboard, keeping the second extended checkerboard with the lowest checkerboard energy as the basis of the next extension until the checkerboard energy of the second extended checkerboard obtained after extension in the two directions is equal, or the number of lines of the second extended checkerboard obtained after extension in any one direction of the two directions reaches a preset number of lines, so as to obtain a first repeated extended checkerboard;
and if the third extension checkerboard with the lowest checkerboard energy in the two third extension checkerboards is lower than the energy of the first extension checkerboard, the third extension checkerboard with the lowest checkerboard energy is reserved as the basis of the next extension until the checkerboard energy of the third extension checkerboard in the other two directions is equal, or the number of rows of the third extension checkerboard in any one of the two directions reaches a preset number of rows after extension, so as to obtain the second extension checkerboard.
2. A checker corner detection apparatus, comprising:
the candidate determining module is suitable for acquiring an image and determining candidate corner points in the image;
the corner peak value determining module is suitable for determining the number of corner peak values of each candidate corner and the size and the direction of each corner peak value within the range of 0-2 pi in the boundary direction;
a removing module, adapted to remove candidate corner points whose number of corner point peaks is less than 4;
the determining module is suitable for determining whether to reserve the candidate corner or not according to the comparison result and the direction relation of the sizes of different corner peak values of the candidate corner for each candidate corner with the number of the corner peak values being more than or equal to 4;
the checker corner detection device further includes:
the generating module is suitable for generating an initial checkerboard based on the reserved angular points;
a first extension module, adapted to extend the initial checkerboard in four directions by one line, obtain first extension checkerboards after extending in each direction, calculate checkerboard energies of the four first extension checkerboards, respectively, if a first extension checkerboard with the lowest checkerboard energy among the four first extension checkerboards is lower than the checkerboard energy of the initial checkerboard, retain the first extension checkerboard with the lowest checkerboard energy as the first checkerboard, where the four directions are directions perpendicular to four sides of the initial checkerboard;
a direction selection module adapted to select two directions among the four directions, a first extended checkerboard energy in the selected two directions being lower than checkerboard energy of the first extended checkerboard in the remaining two directions;
a second extending module, adapted to extend the first checkerboard multiple times in the selected two directions, each time extending for one line, obtaining a second extended checkerboard after extending in each direction, respectively calculating checkerboard energy of the two second extended checkerboards, if the energy of the second extended checkerboard with the lowest checkerboard energy is lower than the energy of the first checkerboard in the two second extended checkerboards, retaining the second extended checkerboard with the lowest checkerboard energy as a basis for next extending until the energy of the second extended checkerboard obtained after extending in the two directions is equal, or the number of lines of the second extended checkerboard obtained after extending in any one of the two directions reaches a preset number of lines, so as to obtain a first repeated extended checkerboard;
and a third extending module, adapted to extend the first repeated extending checkerboard to the other two directions for multiple times, each time extending for one line, obtaining a third extending checkerboard after extending in each direction, and calculating the checkerboard energy of the two third extending checkerboards, if the energy of the third extending checkerboard with the lowest checkerboard energy in the two third extending checkerboards is lower than the energy of the first repeated extending checkerboard, keeping the third extending checkerboard with the lowest checkerboard energy as a basis for the next extending until the checkerboard energy of the third extending checkerboard obtained after extending in the other two directions is equal, or the number of lines of the third extending checkerboard obtained after extending in any one of the two directions reaches a preset number of lines, so as to obtain a second repeated extending checkerboard.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011265345.1A CN112270717B (en) | 2017-04-01 | 2017-04-01 | Checkerboard angular point detection method and device |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011265345.1A CN112270717B (en) | 2017-04-01 | 2017-04-01 | Checkerboard angular point detection method and device |
| CN201710215927.0A CN108665500B (en) | 2017-04-01 | 2017-04-01 | Checkerboard angular point detection method and device |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710215927.0A Division CN108665500B (en) | 2017-04-01 | 2017-04-01 | Checkerboard angular point detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112270717A true CN112270717A (en) | 2021-01-26 |
| CN112270717B CN112270717B (en) | 2022-05-31 |
Family
ID=63784877
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710215927.0A Active CN108665500B (en) | 2017-04-01 | 2017-04-01 | Checkerboard angular point detection method and device |
| CN202011265345.1A Active CN112270717B (en) | 2017-04-01 | 2017-04-01 | Checkerboard angular point detection method and device |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710215927.0A Active CN108665500B (en) | 2017-04-01 | 2017-04-01 | Checkerboard angular point detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN108665500B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113610927B (en) * | 2021-08-06 | 2022-04-15 | 镁佳(北京)科技有限公司 | AVM camera parameter calibration method and device and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101477687A (en) * | 2009-01-22 | 2009-07-08 | 上海交通大学 | Checkerboard angle point detection process under complex background |
| CN104517276A (en) * | 2013-09-28 | 2015-04-15 | 沈阳新松机器人自动化股份有限公司 | Checker corner detection method |
| US20160227206A1 (en) * | 2015-02-04 | 2016-08-04 | Sony Corporation | Calibration methods for thick lens model |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102663810B (en) * | 2012-03-09 | 2014-07-16 | 北京航空航天大学 | Full-automatic modeling approach of three dimensional faces based on phase deviation scanning |
| CN104299218B (en) * | 2013-07-17 | 2017-02-22 | 南京邮电大学 | Projector calibration method based on lens distortion rule |
-
2017
- 2017-04-01 CN CN201710215927.0A patent/CN108665500B/en active Active
- 2017-04-01 CN CN202011265345.1A patent/CN112270717B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101477687A (en) * | 2009-01-22 | 2009-07-08 | 上海交通大学 | Checkerboard angle point detection process under complex background |
| CN104517276A (en) * | 2013-09-28 | 2015-04-15 | 沈阳新松机器人自动化股份有限公司 | Checker corner detection method |
| US20160227206A1 (en) * | 2015-02-04 | 2016-08-04 | Sony Corporation | Calibration methods for thick lens model |
Non-Patent Citations (1)
| Title |
|---|
| JONG-EUN HA: "《Automatic detection of calibration markers on a chessboard》", 《OPTICAL ENGINEERING》, vol. 46, no. 10, 30 October 2007 (2007-10-30) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108665500B (en) | 2020-09-29 |
| CN112270717B (en) | 2022-05-31 |
| CN108665500A (en) | 2018-10-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109978839B (en) | Method for detecting wafer low-texture defects | |
| CN108805023B (en) | Image detection method, device, computer equipment and storage medium | |
| CN108121991B (en) | Deep learning ship target detection method based on edge candidate region extraction | |
| US8805077B2 (en) | Subject region detecting apparatus | |
| CN113109368B (en) | Glass crack detection method, device, equipment and medium | |
| JP2017521779A (en) | Detection of nuclear edges using image analysis | |
| CN104637046B (en) | Image detection method and device | |
| US8923610B2 (en) | Image processing apparatus, image processing method, and computer readable medium | |
| US20120320433A1 (en) | Image processing method, image processing device and scanner | |
| CN110415208A (en) | A kind of adaptive targets detection method and its device, equipment, storage medium | |
| CN115205223B (en) | Visual inspection method and device for transparent object, computer equipment and medium | |
| CN109255792B (en) | Video image segmentation method and device, terminal equipment and storage medium | |
| CN104732510A (en) | Camera lens black spot detecting method and device | |
| JP2015204023A (en) | Subject detection device, subject detection method, and program | |
| CN111951283A (en) | Medical image identification method and system based on deep learning | |
| CN108665500B (en) | Checkerboard angular point detection method and device | |
| CN106778822B (en) | Image straight line detection method based on funnel transformation | |
| CN114693543B (en) | Image noise reduction method and device, image processing chip and image acquisition equipment | |
| CN115249024A (en) | Bar code identification method and device, storage medium and computer equipment | |
| CN110349133B (en) | Object surface defect detection method and device | |
| JP2018120642A (en) | Subject detection device, subject detection method, and program | |
| CN111476821B (en) | Target tracking method based on online learning | |
| CN111738045B (en) | Image detection method and device, electronic equipment and storage medium | |
| Singh et al. | Multi-level threshold based edge detector using logical operations | |
| JP2018106456A (en) | Image processing device and image processing method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |