WO2018072172A1 - 图像中的形状识别方法、装置、设备及计算机存储介质 - Google Patents

图像中的形状识别方法、装置、设备及计算机存储介质 Download PDF

Info

Publication number
WO2018072172A1
WO2018072172A1 PCT/CN2016/102699 CN2016102699W WO2018072172A1 WO 2018072172 A1 WO2018072172 A1 WO 2018072172A1 CN 2016102699 W CN2016102699 W CN 2016102699W WO 2018072172 A1 WO2018072172 A1 WO 2018072172A1
Authority
WO
WIPO (PCT)
Prior art keywords
point
image
shape
function
hough transform
Prior art date
Application number
PCT/CN2016/102699
Other languages
English (en)
French (fr)
Inventor
郭涛
Original Assignee
深圳配天智能技术研究院有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 深圳配天智能技术研究院有限公司 filed Critical 深圳配天智能技术研究院有限公司
Priority to PCT/CN2016/102699 priority Critical patent/WO2018072172A1/zh
Priority to CN201680026894.9A priority patent/CN107710229B/zh
Publication of WO2018072172A1 publication Critical patent/WO2018072172A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/48—Extraction of image or video features by mapping characteristic values of the pattern into a parameter space, e.g. Hough transformation

Definitions

  • the present invention relates to the field of image processing, and in particular, to a shape recognition method, apparatus, device and computer storage medium in an image.
  • Image processing and recognition is an important field of artificial intelligence. It belongs to advanced computer vision technology and has been widely used in map and terrain matching, fingerprint and seal recognition, historical text and image document restoration, various industrial and detection robots. And so on.
  • the vision system based on image recognition algorithm in the device is a key technology in the industry, which provides visual capabilities for the device, and identifies a specific shape in the process.
  • the workpiece is operated differently for workpieces of different shapes, and the workpiece generally has key features of geometric shapes such as straight lines and circles. Therefore, the visual system generally recognizes the entire workpiece by recognizing the geometric shapes.
  • the recognition of geometric shape usually adopts the Hough transform algorithm.
  • this method has certain defects.
  • the invention mainly solves the problem that the recognition accuracy of the image recognition method based on Hough transform in the prior art is poor.
  • the present invention provides a shape recognition method in an image
  • the recognition method includes performing identification of a shape to be recognized, and selecting a shape to be recognized from the candidate shapes, specifically comprising the steps of: acquiring an original image, wherein the original image Include a plurality of foreground pixels; based on the Hough transform formula corresponding to the shape to be identified, perform Hough transform on the original image to obtain a function image, and the coordinate values of each point in the function image respectively represent function parameters of the candidate shape on the original image
  • the gray value represents the number of foreground pixel points on the candidate shape respectively; calculating a stepwise mode of each point in the function image to obtain a gradient image, wherein the gray value of each point in the gradient image is proportional to each point in the function image a stepwise mode; determining a local maximum value of the gray value of each point in the gradient image to obtain a local maximum point corresponding to the local maximum; and based on the Hough transform formula corresponding to the shape to be recognized, performing the local maximum
  • the step of performing Hough transform on the original image based on the Hough transform formula corresponding to the shape to be recognized includes: confirming, according to the Hough transform formula, a candidate shape corresponding to each point coordinate in the function image on the original image; The number of foreground pixels; the gray value of each point on the function image is set according to the number of foreground pixels.
  • the identifying method comprises: recognizing a straight line; and confirming, according to the Hough transform formula, the candidate shape corresponding to each point coordinate in the function image on the original image comprises: establishing a Cartesian coordinate system on the original image; and comparing with the Cartesian coordinate system
  • the number of steps includes: counting the number N of foreground pixel points on the candidate line; and setting the gray value of each point on the function image according to the number of the foreground pixel points includes: setting according to the number N of foreground pixel points The gray value of the point in the function image corresponding to the coordinate values ⁇ , ⁇ .
  • the step of calculating a gradient of each point in the function image comprises: respectively calculating a change amount of the gray value of the coordinate value of each point in the function image, and obtaining one of each point in the function image. Step degree; modulo operation is performed on a step of each point to obtain a gradient image.
  • the identification method includes the recognition of straight lines and circles.
  • the present invention also provides a shape recognition device in an image
  • the identification device is configured to perform recognition of a shape to be recognized, and includes: an acquisition module, configured to acquire an original image, wherein the original image includes a plurality of foreground pixels a transform module is used to perform a Hough transform on the original image based on the Hough transform formula corresponding to the shape to be recognized, to obtain a function image, and the coordinate values of each point in the function image respectively represent function parameters of the candidate shape on the original image.
  • the gray value represents the number of foreground pixel points on the candidate shape, respectively;
  • the calculation module is used to calculate a stepwise modulus of each point in the function image, and a gradient image is obtained, wherein the gray value of each point in the gradient image is proportional to a stepwise mode of each point in the function image;
  • the detecting module is used to determine a local maximum value of the gray value of each point in the gradient image, and obtain a local maximum point corresponding to the local maximum value;
  • the transform module is further used as a basis and
  • the Hough transform formula corresponding to the shape to be identified is inversely transformed from the local maximum point to obtain the local maximum point.
  • the transform module further includes: a shape confirming unit configured to confirm a candidate shape corresponding to each point coordinate in the function image according to the Hough transform formula on the original image; and a quantity statistical unit used as a number of foreground pixel points on the statistical candidate shape
  • the gradation setting unit is configured to set the gradation value of each point on the function image according to the number of foreground pixel points.
  • the identification device is used for recognizing a straight line
  • the shape confirming unit is further used to establish a Cartesian coordinate system on the original image
  • the number statistic unit is further used to statistically obtain the number N of the foreground pixel points on the candidate line
  • the gradation setting unit Further, it is used to set the gradation value of the point corresponding to the coordinate value ⁇ , ⁇ in the function image in accordance with the number N of foreground pixel points.
  • the calculation module further includes: a gradient calculation unit configured to separately calculate a change amount of the gray value of each point in the function image when the coordinate value changes, and obtain a step degree of each point in the function image; A modulo operation is performed on a step of each point to obtain a gradient image.
  • a gradient calculation unit configured to separately calculate a change amount of the gray value of each point in the function image when the coordinate value changes, and obtain a step degree of each point in the function image
  • a modulo operation is performed on a step of each point to obtain a gradient image.
  • the identification device performs linear and circular recognition.
  • the present invention provides a shape recognition device in an image, comprising a processor and a memory, the processor configured to perform the steps of: acquiring an original image, wherein the original image includes a plurality of foreground pixels; based on the shape to be recognized Corresponding Hough transform formula, Hough transform is performed on the original image to obtain a function image.
  • the coordinate values of each point in the function image respectively represent the function parameters of the candidate shape on the original image, and the gray values respectively represent the foreground pixels on the candidate shape.
  • the local maximum value of the gray value of the point obtains the local maximum point corresponding to the local maximum value; based on the Hough transform formula corresponding to the shape to be recognized, the Hough inverse transform is performed on the local maximum point to obtain the candidate corresponding to the local maximum point.
  • the processor performs a Hough transform on the original image based on the Hough transform formula corresponding to the shape to be recognized, and includes: determining, according to the Hough transform formula, a candidate shape corresponding to each point coordinate in the function image on the original image; The number of foreground pixels on the candidate shape; the gray values of the points on the function image are set according to the number of foreground pixels.
  • the identification device is configured to perform line recognition
  • the step of the number of foreground pixels on the shape includes: statistically obtaining the number N of foreground pixel points on the candidate line; and the processor performing the step of setting the gray value of each point on the function image according to the number of foreground pixel points includes: The gradation value of the point corresponding to the coordinate value ⁇ , ⁇ in the function image is set according to the number N of foreground pixel points.
  • the processor performs a stepwise mode of calculating points in the function image, and the step of obtaining the gradient image includes: respectively calculating a change amount of the gray value of each point in the function image when the coordinate value changes, and obtaining each of the function images A step of the point; a grading operation is performed on a step of each point to obtain a gradient image.
  • the identification device is used for recognizing straight lines and circles.
  • the present invention provides a computer storage medium storing an executable program, the program comprising the steps of: acquiring an original image, wherein the original image includes a plurality of foreground pixels; and corresponding to the shape to be recognized Hough transform formula, Hough transform is performed on the original image to obtain a function image.
  • the coordinate values of each point in the function image respectively represent the function parameters of the candidate shape on the original image, and the gray values respectively represent the foreground pixel points on the candidate shape.
  • Quantity calculating a stepwise modulus of each point in the function image to obtain a gradient image, wherein the gray value of each point in the gradient image is proportional to a stepwise modulus of each point in the function image; determining the points in the gradient image The local maximum value of the gray value is obtained, and the local maximum point corresponding to the local maximum value is obtained; based on the Hough transform formula corresponding to the shape to be recognized, the Hough transform is performed on the local maximum point to obtain the candidate shape corresponding to the local maximum point, And as the recognition result of the shape to be identified.
  • the step of performing Hough transform on the original image based on the Hough transform formula corresponding to the shape to be recognized includes: confirming, according to the Hough transform formula, the candidate shape corresponding to each point coordinate in the function image on the original image; The number of foreground pixels on the shape; the gray values of the points on the function image are set according to the number of foreground pixels.
  • the program is used for recognizing a straight line
  • the number of steps includes: counting the number N of foreground pixel points on the candidate line; and setting the gray value of each point on the function image according to the number of foreground pixel points includes: setting a function image according to the number N of foreground pixel points The gray value of the point corresponding to the coordinate values ⁇ , ⁇ .
  • the step of calculating a gradient of each point in the function image comprises: respectively calculating a change amount of the gray value of the coordinate value of each point in the function image, and obtaining one of each point in the function image. Step degree; modulo operation is performed on a step of each point to obtain a gradient image.
  • the program is used to identify lines or circles.
  • the shape recognition method in the image of the present invention comprises the steps of: step 1 first acquiring an original image, where the original image has a plurality of foreground pixels; the second step is further different from the prior art. Based on the Hough transform formula corresponding to the shape to be identified, the original image is subjected to Hough transform to obtain a function image.
  • the coordinate value of the point in the function image represents the function parameter of the candidate shape on the original image, and the gray value represents the candidate shape.
  • step 3 The number of foreground pixels; in step 3, the modulus of each point in the function image is calculated to obtain a gradient image, and the gray value of each point in the gradient image is proportional to the modulus of each point in the function image;
  • the modulo of the step indicates the rate of change of the number of foreground pixels on the candidate shape corresponding to the point when the function parameter changes, and the rate of change is larger for the shape to be recognized, so the gradient image is determined in the fourth step.
  • the local maximum value of the gray value of each point is obtained as the local maximum point corresponding to the local maximum value; the fifth step is based on the Hough transform formula corresponding to the shape to be recognized, Section maximum point inverse Hough transform candidates corresponding to the shape of the local maximum point, and as a recognition result of the shape to be identified.
  • calculating the modulus of each step in the function image can further enhance the signal of the local maximum point and reduce the signal of other points. Therefore, when the local maximum is detected, the interference is less, correspondingly The accuracy of shape recognition in the image is higher.
  • FIG. 1 is a schematic flow chart of an embodiment of a shape recognition method according to the present invention.
  • FIG. 2 is a schematic flow chart of performing line recognition by an embodiment of the shape recognition method shown in FIG. 1;
  • FIG. 3 is a schematic diagram of an original image in the flow of performing line recognition shown in FIG. 2;
  • FIG. 4 is a schematic diagram of a function image in the flow of performing line recognition shown in FIG. 2;
  • FIG. 5 is a schematic diagram of a gradient image in the flow of performing line recognition shown in FIG. 2;
  • FIG. 5 is a schematic diagram of a gradient image in the flow of performing line recognition shown in FIG. 2;
  • FIG. 6 is a schematic diagram of the recognition result in the flow of performing line recognition shown in FIG. 2;
  • Figure 7 is a schematic structural view of an embodiment of a shape recognition device of the present invention.
  • FIG. 8 is a schematic structural view of an embodiment of a shape recognition device of the present invention.
  • FIG. 9 is a block diagram showing an embodiment of a computer storage medium of the present invention.
  • FIG. 1 is a schematic flow chart of an embodiment of a shape recognition method according to an embodiment of the present invention, wherein a shape recognition method is used for identifying a shape to be recognized in an image, and selecting a shape to be recognized from the candidate shapes, specifically including the following step:
  • S1 Acquire an original image, wherein the original image includes a plurality of foreground pixels.
  • each pixel has only two possible values of 0 or 1, and the two values respectively correspond to close and open, and the close indicates that the pixel is in the background, and the pixel is turned on to indicate the pixel. It is in the foreground, so the foreground pixels in all pixels are used to define the shape to be recognized. Generally, the foreground pixel value is 1 and appears as white; the background pixel value is 0, and is rendered black. In other embodiments, other settings may be made, for example, the foreground pixel points are black, and the background pixels are Render white.
  • All pixels in the original image have their determined value of 0 or 1, and the relative position in the original image is determined.
  • a Cartesian coordinate system (X, Y) is first established with the center of the original image as an origin, and then the coordinate point and the numerical value are used to represent the pixel point, that is, (x, y, 0) or (x, y, 1), the computer can determine the location of the pixel and whether the pixel is a foreground pixel. It can be understood that other points in the original image can also be used as the coordinate system origin when establishing the direct coordinate system.
  • the core idea of the Hough transform is to map the set of foreground pixel points constituting a certain shape in the original image to a point of the function image, and this point records the number of foreground pixel points in the foreground pixel point set, and then the number of foreground pixel points.
  • the local peak search finds the point in the function image that records the maximum number of foreground pixels in the local area, and the point corresponds to the point set with the largest number of foreground pixels, and the point set constitutes the shape to be recognized in the original image.
  • the original image needs to be Hough transformed, that is, the candidate shape is Hough transformed, and converted into a point, and all the points transformed by the candidate shape constitute a function image, in which the Hough transform
  • the formula is based on the candidate shape, that is, corresponding to the shape to be recognized; by the transformation according to the formula, the coordinates of a point in the function image can uniquely represent a certain candidate shape in the original image. That is, the coordinates in the function image are function parameters of the candidate shape in the original image; and the gray value of the point represents the number of foreground pixel points on the candidate shape.
  • the coordinates of the ⁇ , ⁇ ) space and the gray value H are defined, that is, ( ⁇ , ⁇ , H), where the gray value H represents the foreground pixel point (x, on the corresponding candidate line).
  • the number N of y, 1) that is, the points obtained by the mapping are represented by ( ⁇ , ⁇ , N), and all the points together constitute a function image.
  • each point is represented by the coordinates of the space of (x0, y0, R) and the gray value H, that is, (x0, y0, R, H), and the gray value H of the point indicates its corresponding Candidate circle on foreground pixel (x, The number N of y, 1), the points obtained by the mapping are represented by (x0, y0, R, N), and all the points together constitute a function image.
  • steps S21 and S22 are usually implemented by a computer program, and when the shape to be recognized is a straight line, the computer logic language is expressed for steps S21 and S22 as follows:
  • step S2 is realized: counting the number of foreground pixel points on the candidate shape.
  • the number N is set to the gray value of each point in the function image.
  • each point will exhibit different brightness and darkness, and the foreground pixel included in the shape to be recognized in the original image
  • the maximum number is reflected on the function image, and the corresponding point gray value is the largest, which is the brightest point on the function image.
  • S3 Calculating a stepwise mode of each point in the function image, obtaining a gradient image, wherein the gray value of each point in the gradient image is proportional to a stepwise mode of each point in the function image.
  • a gradient is a vector that represents the direction in which the scalar data changes the most, while the modulus of the gradient, the length of the vector, represents the magnitude of the scalar data change.
  • a step of each point in the function image indicates the amount of change of the gray value when the coordinate value of each point changes, and a modulus of the step indicates the magnitude of the change, and corresponding to the original image, the position of the candidate shape changes slightly.
  • the rate of change in the number of foreground pixels For the shape to be identified, when the small change occurs, the number of foreground pixels will change greatly, and for the non-identified shape in the candidate shape, when the small change occurs, the change of the number of foreground pixels will be small;
  • the gradation value of each point the gradation value of the point corresponding to the shape to be recognized can be further enhanced, and the gradation value of the point corresponding to the shape to be recognized can be further weakened.
  • This step is mainly divided into two steps of obtaining a step and modulo, as follows:
  • S31 respectively calculating a change amount of the gray value of each point in the function image when the coordinate value changes, and obtaining a step degree of each point in the function image;
  • step S2 we can see that for the function image, we get the coordinates and gray value of each point, which is a set of discrete values, not a continuous function expression. Therefore, we need to use numerical differentiation to find a ladder. degree.
  • the obtained step is T(N'( ⁇ , ⁇ ), N'( ⁇ , ⁇ ));
  • a step obtained is T(N'( ⁇ x0, y0, R), N'(x0, ⁇ y0, R), N'(x0, y0, ⁇ R)).
  • S32 Perform a modulo operation on a step of each point to obtain a gradient image.
  • the modulo obtained in S31 is modulo, that is, the vector is modulo, and the modulo value
  • is proportional to the gray value H of each point in the gradient image, that is, H k
  • S4 Determine a local maximum value of the gray value of each point in the gradient image to obtain a local maximum point corresponding to the local maximum value.
  • the shape to be recognized in the original image corresponds to a point where the gray value in the gradient image is large, so that the local maximum value of the gray value is obtained from the gradient image, and the corresponding local maximum point is obtained, and the original can be obtained.
  • the shape to be recognized in the image corresponds to a point where the gray value in the gradient image is large, so that the local maximum value of the gray value is obtained from the gradient image, and the corresponding local maximum point is obtained, and the original can be obtained.
  • the range of the local maximum value needs to be defined, that is, the threshold value is set.
  • the gray value exceeds the threshold value, it is regarded as the local maximum value, and the corresponding point is the local maximum. point.
  • step S3 a stepwise modulus is obtained for the points in the function image, and a gradient image is obtained.
  • the gray value of the point corresponding to the shape to be recognized is further enhanced compared to the function image, and the candidate is The gray value of the point corresponding to the shape to be recognized in the shape is further weakened compared to the function image, that is, in the gradient image, the signal of the local maximum point is enhanced, and the signals of other points are weakened. It is more conducive to the search of the local maximum in this step.
  • the local maximum value is searched for the function image after step S2
  • the following problems are likely to occur: one is the candidate shape near the shape to be recognized, and the foreground pixels passing through it are also more, which is likely to be caused by corresponding to the function image.
  • the misjudgment of the local maximum point affects the recognition accuracy of the shape to be identified.
  • the range of the local maximum is not easy to define. When the definition is large, it is easy to cause adjacent shapes to be recognized to be ignored. When the definition is small, it is easy to cause a plurality of shapes to be recognized to be recognized in the vicinity of a shape to be recognized, which also causes an identification error.
  • step S3 When step S3 is completed, and the local maximum value is searched again in step S4, the corresponding advantages are as follows: First, the accuracy of the local maximum point judgment can be increased by the enhancement of the local maximum point signal and the attenuation of other point signals. Second, it is possible to set a smaller local maximum range, so that adjacent adjacent shapes to be identified are recognized, and since the signals of other points have been weakened, the definition of a smaller local maximum range is not easy to cause Misjudgment of the shape to be identified.
  • step S4 the local maximum point is further subjected to Hough transform to obtain a candidate shape corresponding to the original image, and is taken as the recognition result of the shape to be recognized.
  • the inverse Hof transform in this step is the reverse of the Hough transform in step S2.
  • the straight line and the circle or the curve in the original image can be identified.
  • the straight line and the circle appear simultaneously in the original image, they can be separately identified or simultaneously recognized.
  • the Hough transform is performed on the original image based on the two Hough transform formulas of the straight line and the circle in step S2 to obtain two function images; two gradient images are obtained in step S3;
  • the gradient image performs a local maximum point search; in step S5, the to-be-identified line corresponding to the local maximum point and the circle to be identified are simultaneously determined; the above steps can be used to identify the computing power and storage capacity requirements of the device at the same time. Bigger. Therefore, it is possible to select successive identification or simultaneous identification according to the capabilities of the device.
  • FIG. 2 is a schematic flow chart of the line recognition method according to the embodiment of the shape recognition method shown in FIG. 1 .
  • the line recognizing method of this embodiment includes the steps of:
  • the original image is shown in Figure 3. This method requires recognition of the straight line in the original image.
  • the foreground pixel representing the line appears black and the background pixel appears white.
  • a Cartesian coordinate system is created, and the pixel points on the original image can be represented by coordinates (x, y) and a value of 0 or 1.
  • S105 Set a gray value of a point corresponding to the coordinate value ⁇ , ⁇ in the function image according to the number of foreground pixel points.
  • FIG. 4 is a schematic diagram, and does not reflect the difference in brightness between dark and gray values.
  • Steps S102 to S105 are Hough transform processes, which are similar to the above-mentioned step S2, and detailed processes in the steps are not described again.
  • S106 Calculate the amount of change of the gradation value when the coordinate value of each point in the function image changes, and obtain a step degree of each point in the function image.
  • S107 Perform a modulo operation on a step of each point to obtain a gradient image, wherein the gray value of each point in the gradient image is proportional to a stepwise modulus of each point in the function image.
  • FIG. 5 is also only a schematic diagram, and does not reflect the difference between light and dark when the gray values are different.
  • Steps S106 to S107 are similar to the above step S3, and detailed processes are not described again.
  • S108 Determine a local maximum value of the gray value of each point in the gradient image, and obtain a local maximum point corresponding to the local maximum value.
  • the local maximum points A, B, C, D, E, F, G are confirmed in the gradient image FIG.
  • the straight lines a, b, c, d, e, f, g corresponding to the local maximum points A, B, C, D, E, F, G are restored to the original image, and the obtained recognition result is shown in FIG. 6.
  • the present method can identify adjacent straight lines a and f, b and g, and d and e.
  • FIG. 7 is a schematic structural view of an embodiment of the shape recognition device of the present invention.
  • the shape recognition apparatus 100 in this embodiment includes an acquisition module 11, a transformation module 12, a calculation module 13, and a detection module 14.
  • the shape recognizing device 100 of the present embodiment is capable of realizing all the steps in the aforementioned shape recognizing method.
  • the obtaining module 11 is configured to acquire an original image, wherein the original image includes foreground pixel points for defining a shape to be recognized, and the obtaining module 11 is capable of determining a coordinate position of each pixel point in the original image.
  • the transform unit 12 functions to perform a Hough transform on the original image based on the Hough transform formula corresponding to the shape to be recognized, to obtain a function image, wherein one candidate image on the original image corresponds to a point in the function image, and each of the function images
  • the coordinate values of the points represent the function parameters of the candidate shapes on the original image, respectively, and the gray values represent the number of foreground pixel points on the candidate shape, respectively.
  • the transform module 12 further includes a shape confirming unit 121, a quantity counting unit 122, and a gradation setting unit 123 for implementing the Hough transform.
  • the shape confirming unit 121 functions as a candidate shape corresponding to the coordinates of each point in the function image in the function image according to the Hough transform formula
  • the number statistic unit 122 serves as the number of foreground pixel points on the statistical candidate shape
  • the gradation setting unit 123 is used to set the gray value of each point on the function image according to the number of foreground pixel points.
  • the transformation module 12 needs to connect the acquisition module 11 to obtain the position information of each pixel in the original image, and then perform the Hough transform with the position information to obtain a function image.
  • the calculation module 13 obtains a gradient model of each point for the function image obtained by the transformation module 12 to obtain a gradient image.
  • the module for finding a step is divided into two steps, so the calculation module 13 further includes a gradient calculation unit 131 and a modulo calculation unit 132.
  • the gradient calculating unit 131 is used to calculate the amount of change of the gradation value when the coordinate values of the points in the function image are respectively changed, and obtain a step degree of each point in the function image.
  • the modulo calculation unit 132 functions as a modulo operation on a step of each point to obtain a gradient image.
  • the detecting module 14 detects the gradient image obtained by the calculating module 13 and obtains a local maximum value of the gray value of each point to obtain a local maximum point corresponding to the local maximum value.
  • the transform module 12 performs a Hough transform on the local maximum point obtained by the detecting module 14 to obtain a candidate shape corresponding thereto, and obtains the recognition result of the shape to be recognized.
  • the shape recognition device 100 of the present embodiment can recognize straight lines and circles, or curves at the same time.
  • FIG. 8 is a schematic structural diagram of an embodiment of a shape recognition device according to the present invention.
  • the shape recognition device 200 of the present embodiment includes a memory 21 and a processor 22.
  • the shape recognizing device 100 described above is a device for realizing a shape recognizing method from the viewpoint of a software architecture.
  • the shape recognizing device 200 of the present embodiment describes a device that realizes the shape recognizing method from the viewpoint of hardware.
  • the processor 22 performs processing calculation to identify the shape to be recognized in the image, and the memory 21 is used to save the data when the processor 22 performs processing calculation.
  • the processor 22 is configured to perform the following steps:
  • Hough transform is performed on the original image to obtain a function image.
  • the coordinate values of each point in the function image respectively represent the function parameters of the candidate shape on the original image, and the gray values are respectively Represents the number of foreground pixel points on the candidate shape.
  • the Hough transform is performed on the local maximum point to obtain the candidate shape corresponding to the local maximum point, and is used as the recognition result of the shape to be recognized.
  • the steps performed by the processor 22 are similar to the steps S1-S5 of the shape recognition method shown in FIG. 1, and details are not described herein. That is, the calculation processing procedure of the processor 22 in the present embodiment can implement the shape recognition method shown in FIG. 1, and can recognize lines and circles.
  • FIG. 9 is a schematic structural diagram of an embodiment of the computer storage medium of the present invention.
  • the executable program stored in the computer storage medium 300 when executed by the computer processor, includes the following steps.
  • Hough transform is performed on the original image to obtain a function image.
  • the coordinate values of each point in the function image respectively represent the function parameters of the candidate shape on the original image, and the gray values are respectively Represents the number of foreground pixel points on the candidate shape.
  • the Hough transform is performed on the local maximum point to obtain the candidate shape corresponding to the local maximum point, and is used as the recognition result of the shape to be recognized.
  • the program stored in the computer storage medium of the present embodiment can implement the shape recognition method shown in FIG. 1 when executed, and details are not described herein.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)

Abstract

一种图像中的形状识别方法和装置，其中识别方法包括进行待识别形状的识别，从候选形状中选出待识别形状，具体包括步骤：获取原始图像（S1）；基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像（S2）；计算函数图像中各点的一阶梯度的模，得到梯度图像（S3）；确定梯度图像中各点的灰度值的局部最大值，得到局部最大值对应的局部最大点（S4）；基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点对应的候选形状，并将此作为待识别形状的识别结果（S5）。该识别方法对于图像中形状的识别有较高的精确度。

Description

图像中的形状识别方法、装置、设备及计算机存储介质
【技术领域】
本发明涉及图像处理领域，尤其是涉及一种图像中的形状识别方法、装置、设备及计算机存储介质。
【背景技术】
图像处理和识别是人工智能的重要领域，属于高级的计算机视觉技术，现已被广泛应用在地图与地形的匹配、指纹和印章的识别、历史文字及图片文档的修复、各种工业及探测机器人等各个领域。
在现代半导体行业中，各技术都已经向智能化转换，其中设备中基于图像识别算法的视觉系统是行业中的关键技术，其为设备提供了视觉能力，在工艺过程中识别出具有特定形状的工件，以针对不同形状的工件进行不同的操作，而工件一般具有直线、圆等几何形状的关键特征，因此视觉系统一般是通过对这些几何形状的识别实现对工件整体的识别。
在现有的图像识别技术中，几何形状的识别通常采用霍夫变换的算法，然而该方法具有一定的缺陷，一是待识别几何形状附近的候选形状容易造成干扰；二是算法中采用局部峰值进行搜索时，局部峰值的定义过大，容易造成多个相邻较近的待识别几何形状被忽略，若局部峰值的定义过小，则容易造成在一个待识别几何形状附近被识别出多个几何形状，两个方面都影响了几何形状识别的精度。
【发明内容】
本发明主要解决现有技术中基于霍夫变换的图像识别方法识别精度较差的问题。
为解决上述技术问题，本发明提供一种图像中的形状识别方法，该识别方法包括进行待识别形状的识别，从候选形状中选出待识别形状，具体包括步骤：获取原始图像，其中原始图像包括多个前景像素点；基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，函数图像中各点的坐标值分别表示原始图像上的候选形状的函数参数，灰度值分别表示候选形状上的前景像素点的数量；计算函数图像中各点的一阶梯度的模，得到梯度图像，其中梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模；确定梯度图像中各点的灰度值的局部最大值，得到局部最大值所对应的局部最大点；基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点所对应的候选形状，并作为待识别形状的识别结果。
其中，基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换的步骤包括：根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状；统计候选形状上的前景像素点的数量；根据前景像素点的数量对函数图像上各点的灰度值进行设置。
其中，识别方法包括进行直线的识别；根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状的步骤包括：在原始图像上建立直角坐标系；以相对于直角坐标系的坐标轴的倾斜角度θ以及到直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取原始图像中的多个候选直线；统计候选形状上的前景像素点的数量的步骤包括：统计得到候选直线上的前景像素点的数量N；根据所述前景像素点的数量对函数图像上各点的灰度值进行设置的步骤包括：根据前景像素点的数量N设置函数图像中对应于坐标值θ，ρ的点的灰度值。
其中，计算函数图像中各点的一阶梯度的模，得到梯度图像的步骤包括：分别计算函数图像中各点的坐标值变化时其灰度值的变化量，得到函数图像中各点的一阶梯度；对各点的一阶梯度进行取模运算，进而得到梯度图像。
其中，识别方法包括进行直线和圆形的识别。
为解决上述技术问题，本发明还提供一种图像中的形状识别装置，该识别装置用于进行待识别形状的识别，包括：获取模块，用作获取原始图像，其中原始图像包括多个前景像素点；变换模块，用作基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，函数图像中各点的坐标值分别表示原始图像上的候选形状的函数参数，灰度值分别表示候选形状上的前景像素点的数量；计算模块，用作计算函数图像中各点的一阶梯度的模，得到梯度图像，其中梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模；检测模块，用作确定梯度图像中各点的灰度值的局部最大值，得到局部最大值所对应的局部最大点；变换模块进一步用作基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点所对应的候选形状，并作为待识别形状的识别结果。
其中，变换模块进一步包括：形状确认单元，用作根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状；数量统计单元，用作统计候选形状上的前景像素点的数量；灰度设置单元，用作根据前景像素点的数量对函数图像上各点的灰度值进行设置。
其中，识别装置用于进行直线的识别，形状确认单元进一步用作在原始图像上建立直角坐标系；并以相对于直角坐标系的坐标轴的倾斜角度θ以及到直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取原始图像中的多个候选直线；数量统计单元进一步用作统计得到候选直线上的所述前景像素点的数量N；灰度设置单元进一步用作根据前景像素点的数量N设置函数图像中对应于坐标值θ，ρ的点的灰度值。
其中，计算模块进一步包括：梯度计算单元，用作分别计算函数图像中各点的坐标值变化时其灰度值的变化量，得到函数图像中各点的一阶梯度；求模计算单元，用作对各点的一阶梯度进行取模运算，进而得到梯度图像。
其中，识别装置进行直线和圆形的识别。
为解决上述技术问题，本发明提供一种图像中的形状识别设备，包括处理器和存储器，处理器用于执行以下步骤：获取原始图像，其中原始图像包括多个前景像素点；基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，函数图像中各点的坐标值分别表示原始图像上的候选形状的函数参数，灰度值分别表示候选形状上的前景像素点的数量；计算函数图像中各点的一阶梯度的模，得到梯度图像，其中梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模；确定梯度图像中各点的灰度值的局部最大值，得到局部最大值所对应的局部最大点；基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点所对应的候选形状，并作为待识别形状的识别结果。
其中，处理器执行基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换的步骤包括：根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状；统计候选形状上的前景像素点的数量；根据前景像素点的数量对函数图像上各点的灰度值进行设置。
其中，识别设备用于进行直线的识别，处理器执行根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状的步骤包括：在原始图像上建立直角坐标系；以相对于直角坐标系的坐标轴的倾斜角度θ以及到直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取原始图像中的多个候选直线；处理器执行统计候选形状上的前景像素点的数量的步骤包括：统计得到候选直线上的前景像素点的数量N；处理器执行根据前景像素点的数量对函数图像上各点的灰度值进行设置的步骤包括：根据前景像素点的数量N设置函数图像中对应于坐标值θ，ρ的点的灰度值。
其中，处理器执行计算函数图像中各点的一阶梯度的模，得到梯度图像的步骤包括：分别计算函数图像中各点的坐标值变化时其灰度值的变化量，得到函数图像中各点的一阶梯度；对各点的一阶梯度进行取模运算，进而得到梯度图像。
其中，识别设备用于进行直线和圆形的识别。
为解决上述技术问题，本发明提供一种计算机存储介质，其存储有可执行程序，程序执行时包括如下步骤：获取原始图像，其中原始图像包括多个前景像素点；基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，函数图像中各点的坐标值分别表示原始图像上的候选形状的函数参数，灰度值分别表示候选形状上的前景像素点的数量；计算函数图像中各点的一阶梯度的模，得到梯度图像，其中梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模；确定梯度图像中各点的灰度值的局部最大值，得到局部最大值所对应的局部最大点；基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点所对应的候选形状，并作为待识别形状的识别结果。
其中，基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换的步骤包括：根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的所述候选形状；统计候选形状上的前景像素点的数量；根据前景像素点的数量对函数图像上各点的灰度值进行设置。
其中，程序用于进行直线的识别，根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状的步骤包括：在原始图像上建立直角坐标系；以相对于直角坐标系的坐标轴的倾斜角度θ以及到直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取原始图像中的多个候选直线；统计候选形状上的前景像素点的数量的步骤包括：统计得到候选直线上的前景像素点的数量N；根据前景像素点的数量对函数图像上各点的灰度值进行设置的步骤包括：根据前景像素点的数量N设置函数图像中对应于坐标值θ，ρ的点的灰度值。
其中，计算函数图像中各点的一阶梯度的模，得到梯度图像的步骤包括：分别计算函数图像中各点的坐标值变化时其灰度值的变化量，得到函数图像中各点的一阶梯度；对各点的一阶梯度进行取模运算，进而得到梯度图像。
其中，程序用于进行直线或圆形的识别。
本发明的有益效果是，区别于现有技术的情况，本发明图像中的形状识别方法包括步骤：第1步先获取原始图像，在此原始图像中具有多个前景像素点；第2步再基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，函数图像中点的坐标值表示原始图像上候选形状的函数参数，其灰度值则表示候选形状上前景像素点的数量；第3步中计算函数图像中各点一阶梯度的模，得到梯度图像，梯度图像中各点的灰度值则正比于函数图像中各点一阶梯度的模；一阶梯度的模表示函数参数变化时，该点所对应的候选形状上前景像素点数量的变化率，而对于待识别形状，其变化率是较大的，因此在第4步中确定梯度图像中各点的灰度值的局部最大值，得到局部最大值所对应的局部最大点；第5步则基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点对应的候选形状，并作为待识别形状的识别结果。在第3步中计算函数图像中各点一阶梯度的模能够更进一步的增强局部最大点的信号，降低其他点的信号，因此在进行局部最大值的检测时，受到的干扰较小，相应的图像中形状识别的准确度较高。
【附图说明】
为了更清楚地说明本发明实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图：
图1是本发明形状识别方法一实施方式的流程示意图；
图2是图1所示形状识别方法一实施方式进行直线识别的流程示意图；
图3是图2所示进行直线识别的流程中原始图像的示意图；
图4是图2所示进行直线识别的流程中函数图像的示意图；
图5是图2所示进行直线识别的流程中梯度图像的示意图；
图6是图2所示进行直线识别的流程中识别结果的示意图；
图7是本发明形状识别装置一实施方式的结构示意图；
图8是本发明形状识别设备一实施方式的结构示意图；
图9是本发明计算机存储介质一实施方式的结构示意图。
【具体实施方式】
下面将结合本发明实施例的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本发明的一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
请参阅图1，图1是本发明形状识别方法一实施方式的流程示意图，其中的形状识别方法用于对图像中的待识别形状进行识别，从候选形状中选出待识别形状，具体包括以下步骤：
S1：获取原始图像，其中原始图像包括多个前景像素点。
本步骤中，将原始图像处理为二值图像，每个像素点只有两个可能的数值0或1，且这两个数值分别对应关闭和打开，而关闭表示该像素处于背景，打开表示该像素处于前景，因此所有像素点中前景像素点用于定义待识别形状。一般来说，前景像素点数值为1，且呈现为白色；背景像素点数值为0，且呈现为黑色，其他实施例中也可相对的作其他设置，例如前景像素点呈现黑色、背景像素点呈现白色。
原始图像中所有像素点都有其确定的数值0或1，并且在原始图像中的相对位置确定。在定义原始图像中的像素点时，一般首先以原始图像的中心为原点建立直角坐标系(X,Y)，然后用坐标值及数值来表示像素点，即(x, y, 0)或(x, y, 1)，由此计算机即可判断像素点所在的位置以及该像素点是否为前景像素点。可以理解，在建立直接坐标系时，也可使用原始图像中的其他点作为坐标系原点。
S2：基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，其中原始图像上的一个候选形状对应于函数图像中的一个点，函数图像中各点的坐标值分别表示原始图像上的候选形状的函数参数，灰度值分别表示候选形状上前景像素点的数量。
霍夫变换的核心思想是将原始图像中构成某个形状的前景像素点集映射到函数图像的一个点上，且这个点记录了前景像素点集中前景像素点的数量，再通过前景像素点数量局部峰值的搜索，找到函数图像中的记录了局部最多前景像素点数量的点，而该点对应着前景像素点数量最多的点集，该点集则构成原始图像中待识别形状。
在本步骤中，在确认待识别形状时，需要对原始图像中的所有与待识别形状相似的候选形状进行识别，以从所有候选形状中找到待识别形状，确定待识别形状的位置。具体的识别过程中，首先需要对原始图像进行霍夫变换，即对候选形状进行霍夫变换，将其转化为一个点，而所有候选形状所转化成的点构成函数图像，在此霍夫变换的过程中，所依据的公式是与候选形状对应的，即与待识别形状所对应的；通过依据此公式的变换，一个点在函数图像中的坐标能够唯一表示原始图像中的某个候选形状，即点在函数图像中的坐标为原始图像中候选形状的函数参数；并且点的灰度值表示该候选形状上前景像素点的数量。
当待识别形状为直线时，霍夫变换公式为ρ=xcosθ+ysinθ；本步骤则是将原始图像中所有的候选直线分别映射到(ρ,θ)空间的点上，且每个点通过(ρ,θ)空间的坐标以及灰度值H来定义，即(ρ,θ,H)，其中灰度值H表示其对应的候选直线上前景像素点(x, y, 1)的数量N，即映射得到的点用(ρ,θ,N)表示，而所有点共同构成函数图像。
当待识别形状为圆形时，霍夫变换公式为(x-x0)2+(y-y0)2=R2；本步骤则是将原始图像中所有的候选圆形分别映射到(x0, y0, R)空间的点上，每个点通过(x0, y0, R)空间的坐标及灰度值H来表示，即(x0, y0, R,H)，点的灰度值H表示其对应的候选圆形上前景像素点(x, y, 1)的数量N，映射得到的点用(x0, y0, R,N)表示，且所有点共同构成函数图像。
本步骤中得到函数图像的具体过程如下：
S21：根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状；
S22：统计候选形状上前景像素点的数量；
S23：根据前景像素点的数量对函数图像上各点的灰度值进行设置。
通常通过计算机程序来实现上述步骤，当待识别形状为直线时，对于步骤S21和S22使用计算机逻辑语言表示如下：
for 遍历ρ
for 遍历θ
N(ρ,θ) = 0;
for 遍历原始图像中所有前置像素点
if (ρ== xcosθ + ysinθ)
N(ρ,θ) = N(ρ,θ) + 1;
end
end
end
end
以上计算机逻辑语言中，“for 遍历ρ”以及“for 遍历θ”表示首先设定函数图像中点的坐标(ρ,θ)；
“for遍历原始图像中所有前置像素点”，“if(ρ== xcosθ + ysinθ)”以及“N(ρ,θ) = N(ρ,θ) + 1;”表示通过ρ== xcosθ + ysinθ来判断原始图像中的前置像素点是否在(ρ,θ)所对应的候选形状上。对每个像素点进行判断，若在该候选形状上，则数量N累加1，完成所有前置像素点的遍历后即实现了步骤S2：统计候选形状上前景像素点的数量。
得到数量N后，再将其设置为函数图像中各点的灰度值，在对函数图像进行显示时，其中各点将呈现不同的明暗度，原始图像中待识别形状所包括的前景像素点的数量最多，反应到函数图像上，其对应的点灰度值最大，呈现出来是函数图像上最亮的点。
S3：计算函数图像中各点的一阶梯度的模，得到梯度图像，其中梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模。
梯度是一个向量，表示其标量数据变化最大的方向，而梯度的模即向量的长度则表示标量数据变化的大小。本步骤中计算得到函数图像中各点一阶梯度的模后，将此一阶梯度的模转化为各点的灰度值，得到梯度图像，并使得梯度图像中各点灰度值正比于一阶梯度的模。
函数图像中各点的一阶梯度表示各点坐标值变化时灰度值的变化量，一阶梯度的模则表示变化量的大小，对应到原始图像中，则表示候选形状的位置发生微小变化时，前景像素点数量的变化率。对于待识别形状，其发生微小变化时，前景像素点数量的变化会很大，而对于候选形状中的非待识别形状，其发生微小变化时，前景像素点数量的变化会很小；然后将此变化率作为各点的灰度值，则能够将待识别形状对应的点的灰度值进一步增强，而将非待识别形状对应的点的灰度值进一步减弱。
此步骤主要分为求一阶梯度和求模两个步骤，具体如下：
S31：分别计算函数图像中各点的坐标值变化时其灰度值的变化量，得到函数图像中各点的一阶梯度；
由步骤S2可知，对于函数图像，我们得到的是每个点的坐标及灰度值，为一组离散的数值，而并不是连续的函数表达式，因此需要用数值微分的方法来求一阶梯度。当待识别形状为直线时，求得的一阶梯度为T(N’(Δρ, θ), N’(ρ, Δθ))；当待识别形状为圆形时，求得的一阶梯度为T(N’(Δx0,y0,R), N’(x0,Δy0,R), N’(x0,y0,ΔR))。
S32：对各点的一阶梯度进行取模运算，得到梯度图像。
对S31中求得的一阶梯度进行取模，即对向量求模，得到模值|T|，然后将此模值|T|正比转化为梯度图像中各点的灰度值H，即H=k|T|，即梯度图像中各点通过坐标(ρ,θ)及灰度值H来定义，即(ρ,θ,k|T|)，相应的梯度图像则是由点(ρ,θ, k|T|)构成。
S4：确定梯度图像中各点的灰度值的局部最大值，得到与局部最大值对应的局部最大点。
由以上分析可知，原始图像中待识别形状对应梯度图像中的灰度值较大的点，因此从梯度图像中求出灰度值的局部最大值，得到相应的局部最大点，则能够得到原始图像中的待识别形状。
在寻找梯度图像中灰度值的局部最大值时，首先需要定义局部最大值的范围，即设定阈值，当灰度值超过该阈值时则认为其为局部最大值，对应的点为局部最大点。
在步骤S3中对函数图像中的点求一阶梯度的模，并得到梯度图像，在所得到的梯度图像中，待识别形状所对应点的灰度值相较于函数图像进一步增强，而候选形状中非待识别形状所对应点的灰度值相较于函数图像进一步减弱，即在梯度图像中，局部最大点的信号增强，且其他点的信号减弱。更加利于本步骤中局部最大值的搜索。
若是在步骤S2后就对函数图像进行局部最大值的搜索，则容易出现以下问题：一是待识别形状附近的候选形状，其通过的前景像素点也较多，对应到函数图像中则容易造成局部最大点的误判，影响到待识别形状的识别精度；二是局部最大值的范围不容易定义，当定义得较大时，则容易造成相邻较近的多个待识别形状被忽略，当定义得较小时，则容易造成某个待识别形状附近识别出多个待识别形状，也导致识别误差。
而完成步骤S3，在本步骤S4中再进行局部最大值的搜索，则相应的有以下优点：一是通过局部最大点信号的增强和其他点信号的减弱，能够增加局部最大点判断的精确度；二是可以设定较小的局部最大值范围，从而将相邻较近的待识别形状都识别出来，并且由于其他点的信号已减弱，因此较小的局部最大值范围的定义不易造成多个待识别形状的误判。
S5：基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点对应的候选形状，并作为待识别形状的识别结果。
在步骤S4中确定局部最大点后，将局部最大点再进行霍夫变换得到在原始图像中所对应的候选形状，并将其作为待识别形状的识别结果。本步骤中的霍夫反变换与步骤S2中的霍夫变换是相反的过程。
通过以上步骤S1-S5，可对原始图像中直线和圆形或曲线进行识别，当原始图像中同时出现直线和圆形时，可分别识别，也可同时识别。当然进行同时识别时，步骤S2中基于直线和圆形两个霍夫变换公式，对原始图像进行霍夫变换，得到两个函数图像；步骤S3中得到两个梯度图像；步骤S4中分别对两个梯度图像进行局部最大点的搜索；在步骤S5中同时确定局部最大点分别对应的待识别直线和待识别圆形；由以上步骤可看出同时进行识别对设备的计算能力及存储能力的要求更大。因此可根据设备的能力选择分别进行先后识别或同时识别。
以图像中直线的识别为例，具体的识别过程请参阅图2，图2是图1所示形状识别方法一实施方式进行直线识别的流程示意图。该实施例直线识别方法包括步骤：
S101：获取原始图像。
原始图像如图3所示，本方法需要对原始图像中的直线进行识别。其中表示直线的前景像素点呈现为黑色，背景像素点呈现为白色。
S102：在原始图像上建立直角坐标系。
建立直角坐标系，原始图像上像素点可通过坐标(x, y)以及数值0或1表示。
S103：以相对于直角坐标系的坐标轴的倾斜角度θ以及到直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取原始图像中的多个候选直线。
S104：统计得到候选直线上前景像素点的数量N。
以上两步骤S103和S104可通过上述步骤S23中所述的计算机程序实现。
S105：根据前景像素点的数量设置函数图像中对应于坐标值θ，ρ的点的灰度值。
在本步骤完成后，呈现出如图4所示的函数图像，需要说明的是图4为示意图，并未体现灰度值不同时的明暗区别。
步骤S102~S105为霍夫变换过程，类似于上述步骤S2，步骤中的详细过程不再赘述。
S106：分别计算函数图像中各点的坐标值变化时其灰度值的变化量，得到函数图像中各点的一阶梯度。
S107：对各点的一阶梯度进行取模运算，进而得到梯度图像，梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模。
在本步骤完成后得到图5所示的梯度图像，图5也仅为示意图，并未体现灰度值不同时的明暗区别。
步骤S106~S107类似于上述步骤S3，详细过程也不再赘述。
S108：确定梯度图像中各点的灰度值的局部最大值，得到局部最大值对应的局部最大点。
在梯度图像图5中确认局部最大点A，B，C，D，E，F，G。
S109：基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点对应的候选形状，并作为待识别形状的识别结果。
将局部最大点A，B，C，D，E，F，G对应的直线a，b，c，d，e，f，g还原到原始图像中，得到的识别结果如图6所示。由图6可知，采用本方法能够识别出相邻很近的直线a与f，b与g，以及d与e。
请再次参阅图7，图7是本发明形状识别装置一实施方式的结构示意图。本实施例中形状识别装置100包括获取模块11、变换模块12、计算模块13以及检测模块14。
本实施例的形状识别装置100能够实现前述形状识别方法中的所有步骤。其中获取模块11用作获取原始图像，原始图像中包括用于定义待识别形状的前景像素点，获取模块11能够确定原始图像中各个像素点的坐标位置。
变换单元12用作基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，其中原始图像上的一个候选图像对应于函数图像中的一个点，函数图像中各点的坐标值分别表示原始图像上候选形状的函数参数，灰度值分别表示候选形状上的前景像素点的数量。
变换模块12为实现霍夫变换，进一步包括形状确认单元121、数量统计单元122以及灰度设置单元123。其中，形状确认单元121用作根据霍夫变换公式确认函数图像中各点坐标在原始图像上对应的候选形状；数量统计单元122用作统计候选形状上的前景像素点的数量；灰度设置单元123用作根据前景像素点的数量对函数图像上各点的灰度值进行设置。
变换模块12在实现功能的过程中，需要连接获取模块11得到原始图像中各个像素点的位置信息，然后结合位置信息进行霍夫变换得到函数图像。
计算模块13则针对变换模块12得到的函数图像，求其中各点的一阶梯度的模，得到梯度图像。求一阶梯度的模分为两个步骤，因此计算模块13进一步包括梯度计算单元131和求模计算单元132。
其中梯度计算单元131用作分别计算函数图像中各点的坐标值变化时灰度值的变化量，得到函数图像中各点的一阶梯度。求模计算单元132用作对各点的一阶梯度进行取模运算，进而得到梯度图像。
检测模块14对计算模块13得到的梯度图像进行检测，求各点灰度值的局部最大值，得到局部最大值对应的局部最大点。
然后变换模块12再对检测模块14求得的局部最大点进行霍夫变换，得到其所对应的候选形状，并作为待识别形状的识别结果。本实施例形状识别装置100可同时识别直线和圆形，或曲线。
请参阅图8，图8是本发明形状识别设备一实施方式的结构示意图，本实施方式形状识别设备200包括存储器21和处理器22。上述形状识别装置100是从软件构架的角度来描述实现形状识别方法的装置。而本实施方式的形状识别设备200则是从硬件的角度来说明实现形状识别方法的设备。
其中，处理器22进行处理计算，以对图像中待识别的形状进行识别，存储器21则用于在处理器22进行处理计算时对数据进行保存。具体来说，处理器22用于执行以下步骤：
1、获取原始图像，其中原始图像包括多个前景像素点。
2、基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，函数图像中各点的坐标值分别表示原始图像上的候选形状的函数参数，灰度值分别表示候选形状上的前景像素点的数量。
3、计算函数图像中各点的一阶梯度的模，得到梯度图像，其中梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模。
4、确定梯度图像中各点的灰度值的局部最大值，得到局部最大值所对应的局部最大点。
5、基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点所对应的候选形状，并作为待识别形状的识别结果。
上述处理器22所执行的步骤与图1所示的形状识别方法的步骤S1-S5类似，具体不再赘述。即本实施方式中处理器22的计算处理过程能够实现图1中所示的形状识别方法，且能够进行直线和圆形的识别。
当上述实现形状识别方法以软件程序的方式呈现时，可将其作为一段可执行的程序存储于一计算机存储介质中，请参阅图9，图9是本发明计算机存储介质一实施方式的结构示意图。计算机存储介质300中存储的可执行程序在被计算机处理器执行时，包括以下步骤。
1、获取原始图像，其中原始图像包括多个前景像素点。
2、基于与待识别形状对应的霍夫变换公式，对原始图像进行霍夫变换，得到函数图像，函数图像中各点的坐标值分别表示原始图像上的候选形状的函数参数，灰度值分别表示候选形状上的前景像素点的数量。
3、计算函数图像中各点的一阶梯度的模，得到梯度图像，其中梯度图像中各点的灰度值正比于函数图像中各点的一阶梯度的模。
4、确定梯度图像中各点的灰度值的局部最大值，得到局部最大值所对应的局部最大点。
5、基于与待识别形状对应的霍夫变换公式，对局部最大点进行霍夫反变换得到局部最大点所对应的候选形状，并作为待识别形状的识别结果。
即本实施方式计算机存储介质所存储的程序在执行时能够实现了图1所示的形状识别方法，具体不再赘述。
以上所述仅为本发明的实施例，并非因此限制本发明的专利范围，凡是利用本发明说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本发明的专利保护范围内。

Claims (20)

  1. 一种图像中的形状识别方法，其中，所述识别方法包括进行待识别形状的识别，从候选形状中选出待识别形状；
    具体包括步骤：
    获取原始图像，其中所述原始图像包括多个前景像素点；
    基于与所述待识别形状对应的霍夫变换公式，对所述原始图像进行霍夫变换，得到函数图像，所述函数图像中各点的坐标值分别表示所述原始图像上的所述候选形状的函数参数，灰度值分别表示所述候选形状上的所述前景像素点的数量；
    计算所述函数图像中各点的一阶梯度的模，得到梯度图像，其中所述梯度图像中各点的灰度值正比于所述函数图像中各点的一阶梯度的模；
    确定所述梯度图像中各点的灰度值的局部最大值，得到所述局部最大值所对应的局部最大点；
    基于与所述待识别形状对应的霍夫变换公式，对所述局部最大点进行霍夫反变换得到所述局部最大点所对应的所述候选形状，并作为所述待识别形状的识别结果。
  2. 根据权利要求1所述的识别方法，其中，所述基于与所述待识别形状对应的霍夫变换公式，对所述原始图像进行霍夫变换的步骤包括：
    根据所述霍夫变换公式确认所述函数图像中各点坐标在所述原始图像上对应的所述候选形状；
    统计所述候选形状上的所述前景像素点的数量；
    根据所述前景像素点的数量对所述函数图像上各点的灰度值进行设置。
  3. 根据权利要求2所述的识别方法，其中，所述识别方法包括进行直线的识别，
    所述根据所述霍夫变换公式确认所述函数图像中各点坐标在所述原始图像上对应的所述候选形状的步骤包括：
    在所述原始图像上建立直角坐标系；
    以相对于所述直角坐标系的坐标轴的倾斜角度θ以及到所述直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取所述原始图像中的多个候选直线；
    所述统计所述候选形状上的所述前景像素点的数量的步骤包括：
    统计得到所述候选直线上的所述前景像素点的数量N；
    所述根据所述前景像素点的数量对所述函数图像上各点的灰度值进行设置的步骤包括：
    根据所述前景像素点的数量N设置所述函数图像中对应于坐标值θ，ρ的点的灰度值。
  4. 根据权利要求1所述的识别方法，其中，所述计算所述函数图像中各点的一阶梯度的模，得到梯度图像的步骤包括：
    分别计算所述函数图像中各点的坐标值变化时其灰度值的变化量，得到所述函数图像中各点的一阶梯度；
    对所述各点的一阶梯度进行取模运算，进而得到梯度图像。
  5. 根据权利要求1所述的识别方法，其中，所述识别方法包括进行直线或圆形的识别。
  6. 一种图像中的形状识别装置，其中，所述识别装置用于进行待识别形状的识别，所述识别装置包括：
    获取模块，用作获取原始图像，其中所述原始图像包括多个前景像素点；
    变换模块，用作基于与所述待识别形状对应的霍夫变换公式，对所述原始图像进行霍夫变换，得到函数图像，所述函数图像中各点的坐标值分别表示所述原始图像上的候选形状的函数参数，灰度值分别表示所述候选形状上的所述前景像素点的数量；
    计算模块，用作计算所述函数图像中各点的一阶梯度的模，得到梯度图像，其中所述梯度图像中各点的灰度值正比于所述函数图像中各点的一阶梯度的模；
    检测模块，用作确定所述梯度图像中各点的灰度值的局部最大值，得到所述局部最大值所对应的局部最大点；
    所述变换模块进一步用作基于与所述待识别形状对应的霍夫变换公式，对所述局部最大点进行霍夫反变换得到所述局部最大点所对应的所述候选形状，并作为所述待识别形状的识别结果。
  7. 根据权利要求6所述的识别装置，其中，所述变换模块进一步包括：
    形状确认单元，用作根据所述霍夫变换公式确认所述函数图像中各点坐标在所述原始图像上对应的所述候选形状；
    数量统计单元，用作统计所述候选形状上的所述前景像素点的数量；
    灰度设置单元，用作根据所述前景像素点的数量对所述函数图像上各点的灰度值进行设置。
  8. 根据权利要求7所述的识别装置，其中，所述识别装置用于进行直线的识别，
    所述形状确认单元进一步用作在所述原始图像上建立直角坐标系；并以相对于所述直角坐标系的坐标轴的倾斜角度θ以及到所述直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取所述原始图像中的多个候选直线；
    所述数量统计单元进一步用作统计得到所述候选直线上的所述前景像素点的数量N；
    所述灰度设置单元进一步用作根据所述前景像素点的数量N设置所述函数图像中对应于坐标值θ，ρ的点的灰度值。
  9. 根据权利要求6所述的识别装置，其中，所述计算模块进一步包括：
    梯度计算单元，用作分别计算所述函数图像中各点的坐标值变化时其灰度值的变化量，得到所述函数图像中各点的一阶梯度；
    求模计算单元，用作对所述各点的一阶梯度进行取模运算，进而得到梯度图像。
  10. 根据权利要求6所述的识别装置，其中，所述识别装置用于进行直线和圆形的识别。
  11. 一种图像中的形状识别设备，其中，所述识别设备包括处理器和存储器，处理器用于执行以下步骤：
    获取原始图像，其中所述原始图像包括多个前景像素点；
    基于与所述待识别形状对应的霍夫变换公式，对所述原始图像进行霍夫变换，得到函数图像，所述函数图像中各点的坐标值分别表示所述原始图像上的所述候选形状的函数参数，灰度值分别表示所述候选形状上的所述前景像素点的数量；
    计算所述函数图像中各点的一阶梯度的模，得到梯度图像，其中所述梯度图像中各点的灰度值正比于所述函数图像中各点的一阶梯度的模；
    确定所述梯度图像中各点的灰度值的局部最大值，得到所述局部最大值所对应的局部最大点；
    基于与所述待识别形状对应的霍夫变换公式，对所述局部最大点进行霍夫反变换得到所述局部最大点所对应的所述候选形状，并作为所述待识别形状的识别结果。
  12. 根据权利要求11所述的识别设备，其中，
    所述处理器执行所述基于与所述待识别形状对应的霍夫变换公式，对所述原始图像进行霍夫变换的步骤包括：
    根据所述霍夫变换公式确认所述函数图像中各点坐标在所述原始图像上对应的所述候选形状；
    统计所述候选形状上的所述前景像素点的数量；
    根据所述前景像素点的数量对所述函数图像上各点的灰度值进行设置。
  13. 根据权利要求12所述的识别设备，其中，所述识别设备用于进行直线的识别，
    所述处理器执行所述根据所述霍夫变换公式确认所述函数图像中各点坐标在所述原始图像上对应的所述候选形状的步骤包括：
    在所述原始图像上建立直角坐标系；
    以相对于所述直角坐标系的坐标轴的倾斜角度θ以及到所述直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取所述原始图像中的多个候选直线；
    所述处理器执行所述统计所述候选形状上的所述前景像素点的数量的步骤包括：
    统计得到所述候选直线上的所述前景像素点的数量N；
    所述处理器执行所述根据所述前景像素点的数量对所述函数图像上各点的灰度值进行设置的步骤包括：
    根据所述前景像素点的数量N设置所述函数图像中对应于坐标值θ，ρ的点的灰度值。
  14. 根据权利要求11所述的识别设备，其中，
    所述处理器执行所述计算所述函数图像中各点的一阶梯度的模，得到梯度图像的步骤包括：
    分别计算所述函数图像中各点的坐标值变化时其灰度值的变化量，得到所述函数图像中各点的一阶梯度；
    对所述各点的一阶梯度进行取模运算，进而得到梯度图像。
  15. 根据权利要求11所述的识别设备，其中，所述识别设备用于进行直线和圆形的识别。
  16. 一种计算机存储介质，其中，所述计算机存储介质存储有可执行程序，所述程序执行时包括如下步骤：
    获取原始图像，其中所述原始图像包括多个前景像素点；
    基于与所述待识别形状对应的霍夫变换公式，对所述原始图像进行霍夫变换，得到函数图像，所述函数图像中各点的坐标值分别表示所述原始图像上的所述候选形状的函数参数，灰度值分别表示所述候选形状上的所述前景像素点的数量；
    计算所述函数图像中各点的一阶梯度的模，得到梯度图像，其中所述梯度图像中各点的灰度值正比于所述函数图像中各点的一阶梯度的模；
    确定所述梯度图像中各点的灰度值的局部最大值，得到所述局部最大值所对应的局部最大点；
    基于与所述待识别形状对应的霍夫变换公式，对所述局部最大点进行霍夫反变换得到所述局部最大点所对应的所述候选形状，并作为所述待识别形状的识别结果。
  17. 根据权利要求16所述的计算机存储介质，其中，所述基于与所述待识别形状对应的霍夫变换公式，对所述原始图像进行霍夫变换的步骤包括：
    根据所述霍夫变换公式确认所述函数图像中各点坐标在所述原始图像上对应的所述候选形状；
    统计所述候选形状上的所述前景像素点的数量；
    根据所述前景像素点的数量对所述函数图像上各点的灰度值进行设置。
  18. 根据权利要求17所述的计算机存储介质，其中，所述程序用于进行直线的识别，
    所述根据所述霍夫变换公式确认所述函数图像中各点坐标在所述原始图像上对应的所述候选形状的步骤包括：
    在所述原始图像上建立直角坐标系；
    以相对于所述直角坐标系的坐标轴的倾斜角度θ以及到所述直角坐标系的原点的距离ρ划分步长，并根据霍夫变换公式ρ=xcosθ+ysinθ获取所述原始图像中的多个候选直线；
    所述统计所述候选形状上的所述前景像素点的数量的步骤包括：
    统计得到所述候选直线上的所述前景像素点的数量N；
    所述根据所述前景像素点的数量对所述函数图像上各点的灰度值进行设置的步骤包括：
    根据所述前景像素点的数量N设置所述函数图像中对应于坐标值θ，ρ的点的灰度值。
  19. 根据权利要求16所述的计算机存储介质，其中，所述计算所述函数图像中各点的一阶梯度的模，得到梯度图像的步骤包括：
    分别计算所述函数图像中各点的坐标值变化时其灰度值的变化量，得到所述函数图像中各点的一阶梯度；
    对所述各点的一阶梯度进行取模运算，进而得到梯度图像。
  20. 根据权利要求16所述的计算机存储介质，其中，所述程序用于进行直线或圆形的识别。
PCT/CN2016/102699 2016-10-20 2016-10-20 图像中的形状识别方法、装置、设备及计算机存储介质 WO2018072172A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2016/102699 WO2018072172A1 (zh) 2016-10-20 2016-10-20 图像中的形状识别方法、装置、设备及计算机存储介质
CN201680026894.9A CN107710229B (zh) 2016-10-20 2016-10-20 图像中的形状识别方法、装置、设备及计算机存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2016/102699 WO2018072172A1 (zh) 2016-10-20 2016-10-20 图像中的形状识别方法、装置、设备及计算机存储介质

Publications (1)

Publication Number Publication Date
WO2018072172A1 true WO2018072172A1 (zh) 2018-04-26

Family

ID=61168984

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/102699 WO2018072172A1 (zh) 2016-10-20 2016-10-20 图像中的形状识别方法、装置、设备及计算机存储介质

Country Status (2)

Country Link
CN (1) CN107710229B (zh)
WO (1) WO2018072172A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109447068A (zh) * 2018-10-26 2019-03-08 信雅达系统工程股份有限公司 一种从图像中分离印章并校准印章的方法
CN109242807B (zh) * 2018-11-07 2020-07-28 厦门欢乐逛科技股份有限公司 渲染参数自适应的边缘软化方法、介质及计算机设备
CN113807325B (zh) * 2021-11-17 2022-02-22 南京三叶虫创新科技有限公司 一种基于图像处理的线型识别方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060210116A1 (en) * 2005-03-18 2006-09-21 Honda Elesys Co., Ltd. Lane recognition apparatus
CN101625723A (zh) * 2009-07-02 2010-01-13 浙江省电力公司 电力线轮廓的快速图像识别方法
CN103605979A (zh) * 2013-12-03 2014-02-26 苏州大学张家港工业技术研究院 一种基于形状片段的物体识别方法及系统
CN105760812A (zh) * 2016-01-15 2016-07-13 北京工业大学 一种基于Hough变换的车道线检测方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7016539B1 (en) * 1998-07-13 2006-03-21 Cognex Corporation Method for fast, robust, multi-dimensional pattern recognition
CN102509017A (zh) * 2011-11-10 2012-06-20 浙江大学 一种用计算机预测刨花板强度的方法
DE102014109063A1 (de) * 2014-06-27 2015-12-31 Connaught Electronics Ltd. Verfahren zur Detektion eines Objekts mit einer vorbestimmten geometrischen Form in einem Umgebungsbereich eines Kraftfahrzeugs, Kamerasystem und Kraftfahrzeug

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060210116A1 (en) * 2005-03-18 2006-09-21 Honda Elesys Co., Ltd. Lane recognition apparatus
CN101625723A (zh) * 2009-07-02 2010-01-13 浙江省电力公司 电力线轮廓的快速图像识别方法
CN103605979A (zh) * 2013-12-03 2014-02-26 苏州大学张家港工业技术研究院 一种基于形状片段的物体识别方法及系统
CN105760812A (zh) * 2016-01-15 2016-07-13 北京工业大学 一种基于Hough变换的车道线检测方法

Also Published As

Publication number Publication date
CN107710229B (zh) 2021-02-26
CN107710229A (zh) 2018-02-16

Similar Documents

Publication Publication Date Title
WO2018072172A1 (zh) 图像中的形状识别方法、装置、设备及计算机存储介质
CN103424409B (zh) 一种基于dsp的视觉检测系统
WO2016070462A1 (zh) 基于方向梯度直方图的显示面板缺陷检测方法
CN109426789B (zh) 手和图像检测方法和系统、手分割方法、存储介质和设备
WO2010041836A2 (en) Method of detecting skin-colored area using variable skin color model
WO2016180246A1 (zh) 蓝宝石的激光加工方法、设备和存储介质
CN110598634B (zh) 一种基于图例库的机房草图识别方法及其装置
CN113222940B (zh) 一种基于rgb-d图像和cad模型的机器人自动抓取工件方法
CN110473184A (zh) 一种pcb板缺陷检测方法
CN107167172A (zh) 一种总线式汽车数字仪表指针功能的在线监测方法
CN111624203A (zh) 一种基于机器视觉的继电器接点齐度非接触式测量方法
CN104751141A (zh) 基于特征图像全像素灰度值的elm手势识别算法
Zhang et al. Multi-scale defect detection of printed circuit board based on feature pyramid network
WO2020235854A1 (ko) 불량 이미지 생성 장치 및 방법
CN105619741A (zh) 一种基于Tegra K1的模具智能检测方法
WO2019113968A1 (zh) 基于图像内容的投射结构光方法、深度检测方法及结构光投射装置
Ding et al. Detection and recognition method for pointer-type meter based on deep learning
CN115689990A (zh) 一种基于机器视觉的地毯灯杂光缺陷智能检测算法及设备
KR100893086B1 (ko) 조명 변화에 강인한 얼굴 검출 방법
Hu et al. A novel vision-based mold monitoring system in an environment of intense vibration
WO2018213980A1 (zh) 一种机器人的标定方法、系统及标定板
Shao et al. High-Speed and Accurate Method for the Gear Surface Integrity Detection Based on Visual Imaging
Yang et al. Automatic AFM images distortion correction based on adaptive feature recognition algorithm
Zhang et al. Design of classification grasping algorithm for robot based on color information
LU502459B1 (en) Intelligent Counting Method And Device For Number Of Defects In Continuous Target Online Detection Based On Machine Vision

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16919466

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16919466

Country of ref document: EP

Kind code of ref document: A1