Corner point detection method and apparatus

Abstract

The present application provides a corner point detection method and apparatus. The corner point detection method comprises the following steps: receiving a picture to be processed, generating a corner point set, performing edge detection on said picture and obtaining an edge map, and extracting M curves from the edge map; for each curve, performing the following processing: filling gaps along the curve, determining a plurality of T-junction points on the curve, adding the plurality of T-junction points into a corner point set S, smoothing the curve, acquiring the curvature values of all points in the curve, identifying a plurality of corner points on the basis of the curvature values of all the points, and adding the plurality of corner points into the corner point set, wherein the corner points in the corner point set are corner points detected from said picture. In this way, a corner point detection method is implemented.

Description

Corner point detection method and device

[0001] This application is based on the Chinese patent application with application number 202410272676.X and application date of March 11, 2024, and claims the priority of the Chinese patent application. The entire content of the above patent application is hereby introduced into this application as a reference. Technical Field

[0002] The present application relates to a corner point detection method and device. Background Art

[0003] Corner detection, also known as feature point detection, is a method used in computer vision systems to extract image features. A corner point is typically defined as the intersection of two edges, or more strictly, as a point within a local neighborhood with boundaries between two distinct regions and orientations. These points typically exhibit significant changes in image gradients in two or more directions, and are sufficiently repeatable and significant.

[0004] Corner detection plays a very important role in computer vision because it helps us extract key features of images for various tasks such as motion detection, image matching, video tracking, 3D modeling, and object recognition. Through corner detection, we can effectively reduce the amount of information data, improve computing speed, and facilitate reliable image matching, making real-time processing possible.

[0005] Therefore, how to perform corner detection becomes a problem that needs to be solved urgently.

[0006] The reference to any prior art in the specification is not an acknowledgement or suggestion that the prior art forms part of the common general knowledge in any jurisdiction, or that it could reasonably be expected that the person skilled in the art would understand, consider relevant and / or combine with other prior art. Summary of the Invention

[0007] The purpose of this application is to provide a corner detection method and device.

[0008] To achieve one of the above-mentioned objectives, an embodiment of the present invention provides a corner detection method, comprising the following steps: receiving an image to be processed I, generating a corner point set , perform edge detection on the image to be processed I and obtain an edge map, and extract M curves from the edge map 、 ,..., , M is a natural number; the initial value of j is 1, and the first process is continued until j>M. The first process includes: filling the curve The gap on the curve Determine several T-shaped intersections and add them to the corner point set S. Smoothing is performed, and the smoothed curve Represented as a point set , get the point set The curvature values ​​of all points in , and based on the curvature values ​​of all points to identify several corner points, and add several corner points to the corner point set S, the value of j increases by 1; is a natural number; the corner points in the corner point set S are the corner points detected from the image to be processed I.

[0009] As a further improvement of the implementation mode of the present application, the "performing edge detection processing on the image to be processed I and obtaining an edge mapping map" specifically includes: using a Canny detector to perform edge detection processing on the image to be processed I and obtain an edge mapping map.

[0010] As a further improvement to the implementation of this application, the "get point set The curvature value of all points in the equation “i” specifically includes: the initial value of i is 1, and the second processing is continued until i> The second process includes: obtaining the point Support area ,point The curvature value ,in, and are all natural numbers, and the symbol Represents the Euclidean distance between two points, L is the radius of the support region, L>0, point The horizontal axis is ,point The vertical coordinate is ,point The horizontal axis is ,point The vertical coordinate is .

[0011] As a further improvement of the implementation mode of the present application, the “identifying several corner points based on the curvature values ​​of all points” specifically includes: processing the curvature values ​​of all points based on the non-minimum suppression method and threshold filtering, and identifying several corner points.

[0012] As a further improvement of the implementation manner of the present application, L=5.

[0013] The embodiment of the present application also provides a corner point detection device, comprising the following modules: an image receiving module for receiving an image to be processed I, generating a corner point set , perform edge detection on the image to be processed I and obtain an edge map, and extract M curves from the edge map 、 ,..., , M is a natural number; the detection module, the initial value of j is 1, and the first processing is continued until j>M. The first processing includes: filling the curve The gap on the curve Determine several T-shaped intersections and add them to the corner point set S. Smoothing is performed, and the smoothed curve Represented as a point set , get the point set The curvature values ​​of all points in , and based on the curvature values ​​of all points to identify several corner points, and add several corner points to the corner point set S, the value of j increases by 1; is a natural number; the corner points in the corner point set S are the corner points detected from the image to be processed I.

[0014] As a further improvement of the implementation mode of the present application, the image receiving module is further used to: use a Canny detector to detect edges of the image to be processed I and obtain an edge mapping image.

[0015] As a further improvement of the embodiment of the present application, the detection module is further configured to: the initial value of i is 1, and the second processing is continued until i> The second process includes: obtaining the point Support area ,point The curvature value ,in, and are all natural numbers, and the symbol Represents the Euclidean distance between two points, L is the radius of the support region, L>0, point The horizontal axis is ,point The vertical coordinate is ,point The horizontal axis is ,point The vertical coordinate is .

[0016] As a further improvement of the implementation manner of the present application, the detection module is further used to: process the curvature values ​​of all points based on a non-minimum suppression method and threshold filtering, and identify a number of corner points.

[0017] As a further improvement of the implementation manner of the present application, L=5.

[0018] Compared with the prior art, the technical effect of the present application is as follows: the embodiment of the present application provides a corner point detection method and apparatus thereof, the corner point detection method comprising the following steps: receiving a to-be-processed image, generating a corner point set, performing edge detection processing on the to-be-processed image and obtaining an edge map, and extracting M curves from the edge map; performing the following processing on each curve to fill gaps on the curve, determining a number of T-shaped intersections on the curve, and adding the T-shaped intersections to a corner point set S; smoothing the curve, obtaining curvature values ​​of all points in the curve, and identifying a number of corner points based on the curvature values ​​of all points, and adding the corner points to a corner point set; the corner points in the corner point set are the corner points detected from the to-be-processed image. Thus, a corner point detection method is implemented.

[0019] As used herein, the term "comprise" and variations of the term, such as "comprises," "comprised," "comprising," "including," and "containing" do not exclude other features, components, elements, or steps unless the context clearly requires otherwise. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG1 , FIG2A , FIG2B , FIG3A and FIG3B are schematic diagrams of a corner detection method in an embodiment of the present application;

[0021] FIG4 is a comparison diagram of several experiments in the examples of the present application;

[0022] Figures 5 and 6 are comparison diagrams of the results of several experiments in the examples of the present application;

[0023] FIG7 is a flowchart of a corner point detection method in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The following description and accompanying drawings sufficiently illustrate the specific embodiments herein to enable those skilled in the art to practice them. Portions and features of some embodiments may be included in or substituted for portions and features of other embodiments. The scope of the embodiments herein includes the entire scope of the claims, including all available equivalents thereof. Herein, the terms "first," "second," and the like are used solely to distinguish one element from another and do not require or imply any actual relationship or order between these elements. In practice, the first element can also be referred to as the second element, and vice versa. Furthermore, the terms "comprise," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a structure, device, or apparatus comprising a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such structure, device, or apparatus. Without further limitation, an element defined by the phrase "comprising a..." does not preclude the presence of other identical elements in the structure, device, or apparatus comprising the element. The various embodiments herein are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Similar or identical parts between the various embodiments can be referenced to each other.

[0025] The terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like in this document indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing this document and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on this application. In the description of this document, unless otherwise specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, they can be mechanical or electrical connections, or they can be internal connections between two elements. They can be directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0026] The first embodiment of the present application provides a corner detection method, as shown in FIG7 , including the following steps:

[0027] Step 701: Receive the image to be processed I and generate a set of corner points , perform edge detection on the image to be processed I and obtain an edge map, and extract M curves from the edge map 、 ,..., , M is a natural number. The main purpose of edge detection is to identify points in a digital image where brightness changes significantly. These points typically form the outline of an object. Therefore, edge detection can be understood as the process of finding the outline of an object in an image. An edge is defined as the boundary of an area in an image where grayscale changes dramatically. This grayscale change can be reflected by the gradient of the image's grayscale distribution. Therefore, edge detection typically uses local image differentiation techniques to obtain an edge detection operator. This operator is constructed for a small neighborhood of pixels in the original image to detect edges.

[0028] Step 702: The initial value of j is 1, and the first process is continued until j>M. The first process includes: filling the curve The gap on the curve Determine several T-shaped intersections and add them to the corner point set S. Smoothing is performed, and the smoothed curve Represented as a point set , get the point set The curvature values ​​of all points in the image are calculated, and several corner points are identified based on the curvature values ​​of all points. These corner points are added to the corner point set S, and the value of j is increased by 1. The corner points in the corner point set S are the corner points detected from the image to be processed I. is a natural number; here, smoothing can reduce noise.

[0029] The corner detection method in this embodiment primarily includes the following operations: edge extraction, curve tracing, curve smoothing, curvature estimation, non-minimum suppression, and curvature thresholding. First, edges are extracted from the image to be processed, I. Then, in the curve tracing operation, each edge curve is traced from the edge graph and represented by a set of connected points, p. Thus, identified gaps whose separation is smaller than a preset value are connected to eliminate the gaps.

[0030] In this embodiment, the “performing edge detection processing on the image to be processed I and obtaining an edge map” specifically includes: using a Canny detector to perform edge detection processing on the image to be processed I and obtaining an edge map.

[0031] Here, the Canny detector can achieve the following three goals: (1) low error rate, that is, all edges should be found and there should be no false edges; (2) accurate edge positioning, that is, the detected edges should be close to the real edges; (3) single edge point response, that is, for edge detection, only the result of a single point thickness is returned.

[0032] The Canny detector includes the following steps: Step 1: Smooth the image using a Gaussian filter; Step 2: Calculate the gradient magnitude and edge direction; Step 3: Non-maximum suppression; Step 4: Detect and link edges using double thresholding and connectivity analysis.

[0033] In this embodiment, the "get point set" The curvature value of all points in the equation “i” specifically includes: the initial value of i is 1, and the second processing is continued until i> The second process includes: obtaining the point Support area ,point The curvature value ,in, and are all natural numbers, and the symbol Represents the Euclidean distance between two points, L is the radius of the support region, L>0, point The horizontal axis is ,point The vertical coordinate is ,point The horizontal axis is ,point The vertical coordinate is .

[0034] In this embodiment, “identifying a number of corner points based on the curvature values ​​of all points” specifically includes: processing the curvature values ​​of all points based on a non-minimum suppression method and threshold filtering, and identifying a number of corner points.

[0035] In this embodiment, L=5.

[0036] The curve extracted from the image to be processed I can be represented as a set of points , and there are connecting lines between adjacent points, where M is a natural number and M≥2. There are M points on the curve, and the discrete curvature of each point needs to be measured. To this end, a fixed radius value L of the ROS (Region Of Support) is first selected; in the corner detection method of this embodiment, the radius L of the ROS is determined at the sub-pixel position, not the original pixel position. Specifically, the point The ROS is defined as: ,in, and are all natural numbers, and They are called the backward endpoint and forward endpoint of ROS respectively and meet the following conditions: ,symbol Represents the Euclidean distance measured between the two indicated points.

[0037] Afterwards, based on the formula To determine the point The sub-pixel position of Is a parameter to be determined. , calculate the value of μ, and get 's coordinates.

[0038] Finally, the sub-pixelated chord is obtained, i.e. , as shown in Figure 1. The length of the sub-pixelated chord is at point The discrete curvature values ​​measured at . The curvature is measured as .

[0039] Intuitively speaking, if the ,point and point All on the same straight line, sub-pixel chords The length of the sub-pixelated chord is close to 2L, otherwise the length is 0; it is understandable that the length of the sub-pixelated chord will change with the point For demonstration purposes, Figures 2A and 2B show the measured curvature of the test curve. It can be understood that the discrete curvature "A," measured at the sharpest angle, has the lowest curvature. In Figure 2A, point B is the starting point, and the curve tracing direction is indicated by an arrow; Figure 2B shows the actual measured curvature value.

[0040] The corner detection performance is plotted in terms of F-scores with respect to different values ​​of L, as shown in Figure 3A. The L set that achieves the highest F-score is used as the default setting for the ROS radius. When L=5, the F-score reaches its highest value.

[0041] After curvature estimation, corner points on the curve need to be identified as follows. First, non-minimum suppression is performed on h(i). Therefore, only those points with local minimum curvature values ​​are considered corner candidates. Then, a curvature threshold is used to determine "meaningful" corner points. Conversely, a high curvature threshold T may result in more false positives (i.e., non-corner points are identified as corners). Through some experiments, the default value of threshold T was determined to be 9.84, as shown in Figure 3B.

[0042] In our experiments, we used three image datasets, totaling 58 images, to evaluate the corner detection method in this embodiment. These images are either binary or grayscale, and are composed of a variety of image and vision task combinations. We carefully manually labeled the actual corner points in each image. To verify the robustness of the proposed corner detection method under different conditions (e.g., transformations, noise, JPEG compression, etc.), each of the 58 original images was processed by applying three types of transformations (e.g., rotation, scaling, and merging) and two types of degradation with different parameter settings. This resulted in 58 original images and 24,534 transformed (or degraded) images, which were used for performance evaluation.

[0043] The inventors used four common metrics for quantitative evaluation, including precision, recall, arithmetic mean of precision and recall (APR), and F-score to compare the pose corner detector with seven existing state-of-the-art corner detectors, including GCM

[0015] , MSCP

[0012] , CPDA

[0014] , CTAR

[0017] , SuperPoint[6], NewCurvature

[0018] , and DMC

[0020] . In fact, there is a trade-off between prediction and recall for each compared corner detection method. The parameters of the compared detectors are optimized to provide their best F-scores respectively.

[0044] First, for performance evaluation, 58 images were selected without applying any image transformation or degradation, and the results are recorded in Figure 5. It can be seen that the corner detection method of this embodiment has the best performance on almost all test images using all four evaluation metrics.

[0045] Next, images subjected to various transformations (or degradations) were tested, and the test results are reported in Figure 6, where the performance of each corner detector is measured based on the F-score. As can be seen, the corner detection method of this embodiment consistently produces the highest F-scores for most types of imposed image transformations (or degradations), demonstrating its robustness across diverse scenarios. Finally, sub-evaluation results obtained using binary and grayscale images are shown in Figure 4. In comparison, the corner detection method of this embodiment offers superior performance to existing corner detectors.

[0046] A second embodiment of the present invention provides a corner point detection device, including the following modules:

[0047] The image receiving module is used to receive the image to be processed I and generate a set of corner points , perform edge detection on the image to be processed I and obtain an edge map, and extract M curves from the edge map 、 ,..., , M is a natural number.

[0048] The detection module is used for the initial value of j to be 1, and the first processing is continued until j>M. The first processing includes: filling the curve The gap on the curve Determine several T-shaped intersections and add them to the corner point set S. Smoothing is performed, and the smoothed curve Represented as a point set , get the point set The curvature values ​​of all points in , and based on the curvature values ​​of all points to identify several corner points, and add several corner points to the corner point set S, the value of j increases by 1; is a natural number; the corner points in the corner point set S are the corner points detected from the image to be processed I.

[0049] In this embodiment, the image receiving module is further configured to: use a Canny detector to perform edge detection processing on the image to be processed I and obtain an edge map.

[0050] In this embodiment, the detection module is further configured to: the initial value of i is 1, and the second processing is continued until i> The second process includes: obtaining points Support area ,point The curvature value ,in, and are all natural numbers, and the symbol Represents the Euclidean distance between two points, L is the radius of the support region, L>0, point The horizontal axis is ,point The vertical coordinate is ,point The horizontal axis is ,point The vertical coordinate is .

[0051] In this embodiment, the detection module is further configured to process the curvature values ​​of all points based on a non-minimum suppression method and threshold filtering, and identify a number of corner points.

[0052] In this embodiment, L=5.

[0053] It should be noted that although the above describes the various steps in a specific order, it does not mean that the steps must be performed in the above specific order. In fact, some of these steps can be executed concurrently or even in a different order as long as the required functions can be achieved.

[0054] The present invention may be a system, a method and / or a computer program product. The computer program product may include a readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0055] A readable storage medium may be a tangible device that holds and stores instructions used by an instruction execution device. The readable storage medium may include, for example, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination thereof.

[0056] While various embodiments of the present invention have been described above, the above descriptions are intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A corner point detection method, characterized in that: The following steps are involved: Receive the image to be processed I and generate a set of corner points , perform edge detection on the image to be processed I and obtain an edge map, and extract M curves from the edge map 、 、...、 , M is a natural number; The initial value of j is 1, and the first process is continued until j>M. The first process includes: filling the curve The gap on the curve Determine several T-shaped intersections and add them to the corner point set S. Smoothing is performed, and the smoothed curve Represented as a point set , get the point set The curvature values ​​of all points in , and based on the curvature values ​​of all points to identify several corner points, and add several corner points to the corner point set S, the value of j increases by 1; is a natural number; the corner points in the corner point set S are the corner points detected from the image to be processed I.

2. The corner detection method according to claim 1, wherein: The “performing edge detection processing on the image to be processed I and obtaining an edge map” specifically includes: Use the Canny detector to perform edge detection on the image to be processed I and obtain an edge map.

3. The corner detection method according to claim 1, wherein: The "get point set The curvature values ​​of all points in the ” specifically include: The initial value of i is 1, and the second process is continued until i> The second process includes: obtaining points Support area ,point The curvature value ,in, and are all natural numbers, and the symbol Represents the Euclidean distance between two points, L is the radius of the support region, L>0, point The horizontal axis is ,point The vertical coordinate is ,point The horizontal axis is ,point The vertical coordinate is 。 4. The corner detection method according to claim 1, wherein: The “identifying a number of corner points based on the curvature values ​​of all points” specifically includes: The curvature values ​​of all points are processed based on the non-minimum suppression method and threshold filtering, and several corner points are identified.

5. The corner detection method according to claim 3, wherein: L=5。 6. A corner point detection device, characterized in that: Includes the following modules: The image receiving module is used to receive the image to be processed I and generate a set of corner points , perform edge detection on the image to be processed I and obtain an edge map, and extract M curves from the edge map 、 、...、 , M is a natural number; The detection module is used for the initial value of j to be 1, and the first processing is continued until j>M. The first processing includes: filling the curve The gap on the curve Determine several T-shaped intersections and add them to the corner point set S. Smoothing is performed, and the smoothed curve Represented as a point set , get the point set The curvature values ​​of all points in , and based on the curvature values ​​of all points to identify several corner points, and add several corner points to the corner point set S, the value of j increases by 1; is a natural number; the corner points in the corner point set S are the corner points detected from the image to be processed I.

7. The corner detection device according to claim 6, characterized in that: The picture receiving module is also used for: Use the Canny detector to perform edge detection on the image to be processed I and obtain an edge map.

8. The corner point detection device according to claim 6, characterized in that: The detection module is further configured to: The initial value of i is 1, and the second process is continued until i> The second process includes: obtaining points Support area ,point The curvature value ,in, and are all natural numbers, and the symbol Represents the Euclidean distance between two points, L is the radius of the support region, L>0, point The horizontal axis is ,point The vertical coordinate is ,point The horizontal axis is ,point The vertical coordinate is 。 9. The corner detection device according to claim 6, characterized in that: The detection module is further configured to: The curvature values ​​of all points are processed based on the non-minimum suppression method and threshold filtering, and several corner points are identified.

10. The corner point detection device according to claim 8, characterized in that: L=5。

Images