CN110728285A - Rapid corner detection method based on dynamic frequency conversion - Google Patents
Rapid corner detection method based on dynamic frequency conversion Download PDFInfo
- Publication number
- CN110728285A CN110728285A CN201910783617.8A CN201910783617A CN110728285A CN 110728285 A CN110728285 A CN 110728285A CN 201910783617 A CN201910783617 A CN 201910783617A CN 110728285 A CN110728285 A CN 110728285A
- Authority
- CN
- China
- Prior art keywords
- gradient
- image
- mean square
- square error
- angular point
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a dynamic frequency conversion-based rapid corner detection method, which comprises the following steps: dividing an image into a plurality of blocks, calculating the mean square error of each block, and judging that the image is a flat area or an angular point rich area according to the mean square error; when the image is judged to be an angular point rich region, calculating a gradient kernel according to the mean square error; calculating the absolute gradient of the current pixel according to the gradient kernel obtained by calculation, and judging whether the current pixel is an angular point or not according to the absolute gradient; and integrating the corner points of all the blocks, wherein the integrated corner points are all the corner points of the current image. The invention realizes the rapid detection of the angular point and can effectively improve the performance of the angular point detection of the image.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a dynamic frequency conversion-based rapid corner detection method.
Background
The corner feature of an image is an important feature in image processing, and there is a generally accepted view that: the point where the brightness of the two-dimensional image changes drastically or the point where the curvature is maximum on the edge curve of the image may be referred to as the corner point. The corner points reserve important features in the image graph, can effectively reduce the redundancy of information, enables the content of the information to be high, can effectively improve the image calculation speed, is beneficial to the analysis and processing of the image, and enables the real-time processing to be possible. The corner detection plays an important role in the computer vision fields of three-dimensional scene reconstruction, motion estimation, target tracking, target identification, optical flow calculation, image registration and matching and the like.
Currently, the classical algorithms for corner detection include MIC method, SUSAN method, Harris method, and the like. Currently most used are Harris and SUSAN.
(1) The Plessey operator of Harris, the algorithm has the advantages of simple operation and implementation, is widely used in the application of corner detection at present, and has the following defects: the precision of detection positioning is not particularly ideal, especially on the detection of some special corner points, gradient information or corner points of a large obtuse angle fuzzy type are easily lost, and on the calculation time, the precision is not very ideal.
(2) Later, Smith firstly puts forward a concept of USAN, designs an angular point detection method of SUSAN based on the concept, and has the biggest advantages that the method is very simple, has an integral characteristic, is good in noise resistance, and is not influenced by the type of the angular point; the method has the disadvantages that due to the fact that a large number of fuzzy edges exist in an actual image, false response is easy to generate or real corner points are easy to lose, the detection rate of the actual image is generally inferior to that of a Harris algorithm, and in addition, the integration process is time-consuming.
(3) The MIC algorithm should be quickly required by mirosslavtrajkovic. The algorithm is probably the fastest corner detection method in the current gray-scale image processing. Unfortunately, this method, while simple, is prone to false responses, especially at straight or fuzzy edges; the detection level is generally quite sensitive to noise. However, the idea of rapidity is well popularized and can be accelerated when being incorporated into other algorithms.
In summary, although these classical algorithms have many advantages, they have the common disadvantages of long execution time, low level of accuracy in positioning the corner points, and poor performance in detecting, so it is necessary to design a method capable of accurately positioning the corner point information and having short execution speed and time, especially having high practical value in computer vision field for corner point time information and spatial information application.
Disclosure of Invention
The object of the present invention is to solve at least one of the technical drawbacks mentioned.
Therefore, the invention aims to provide a rapid corner detection method based on dynamic frequency conversion.
In order to achieve the above object, an embodiment of the present invention provides a fast corner point detection method based on dynamic frequency conversion, including the following steps:
step S1, dividing the image into a plurality of blocks, calculating the mean square error of each block, and judging the image to be a flat area or an angular point rich area according to the mean square error;
step S2, when the image is judged to be an angular point rich area, calculating a gradient kernel according to the mean square error;
step S3, calculating the absolute gradient of the current pixel according to the gradient kernel obtained by calculation, and judging whether the current pixel is an angular point or not according to the absolute gradient;
and step S4, integrating the corner points of all the blocks, wherein the integrated corner points are all the corner points of the current image.
Further, in step S1, the average value of the image is counted in each block, the variance of each pixel is calculated, the mean square error of the current block is calculated, the calculated mean square error is programmed to be between 0 and 1, a mean square error coefficient is generated, when the mean square error coefficient is smaller than a first preset threshold, the image is determined to be a flat region, otherwise, the image is determined to be a region with rich corners.
Further, in the step S2, when the mean square error is greater than a second preset threshold, a first gradient template is used to calculate a gradient kernel; and when the mean square error is less than or equal to a second preset threshold value, calculating a gradient kernel by adopting a second gradient template.
Further, the first gradient template Gx is:
the second gradient template Gy is:
wherein a is a mean square error coefficient.
Further, in the step S3, an absolute gradient is calculated using the following formula,
|G|＝|Gx|+|Gy|。
further, in step S3, the determining whether the current pixel is an angular point according to the absolute gradient includes: and when the absolute gradient is the third preset threshold, judging that the current pixel is an angular point.
According to the rapid angular point detection method based on dynamic frequency conversion, the flat area and the angular point rich area of the image are detected through scanning the image, angular point detection is not carried out on the flat area, and angular points are searched in the angular point rich area by using a convolution technology, so that the angular point is rapidly detected. The method can effectively improve the performance of detecting the corner points of the image, and has great significance for systems with high real-time requirements.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a flowchart of a fast corner detection method based on dynamic frequency conversion according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an image divided into 16x16 blocks by a block size of 32x32, according to an embodiment of the invention;
fig. 3 is a schematic diagram of a block image in fig. 2.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
As shown in fig. 1, the method for fast corner detection based on dynamic frequency conversion according to the embodiment of the present invention includes the following steps:
step S1, the image is divided into a plurality of blocks, the mean square error of each block is calculated, and the image is determined to be a flat region or a corner rich region according to the mean square error.
Specifically, referring to fig. 2 and 3, an image is divided into a plurality of blocks (blocks) with the size of 32 × 32, an average value of the image is counted in each block, then the variance of each pixel is calculated, finally, the mean square error of the current block is calculated, the calculated mean square error is planned to be between 0 and 1, a mean square error coefficient is generated, when the mean square error coefficient is smaller than a first preset threshold, it is indicated that the fluctuation of the pixel value of the block of the current image is smaller, the image is determined to be a flat area, otherwise, the image is a corner-rich area (edge-rich area). In an embodiment of the present invention, the first preset threshold may be 0.2. It should be noted that the above threshold is only for illustrative purposes, and the user may select other values to set as needed.
And step S2, when the image is judged to be the corner rich area, calculating a gradient kernel according to the mean square error.
When the mean square error is larger than a second preset threshold value, which indicates that the corner information of the current image block is rich, a more sensitive gradient template is needed to be used for detecting the corner, a first gradient template is adopted to calculate a gradient kernel,
wherein the first gradient template Gx is:
and when the mean square error is less than or equal to a second preset threshold, calculating a gradient kernel by using a gradient module with low sensitivity and a second gradient template.
Wherein the second gradient template Gy is:
where a is the mean square error coefficient calculated in step S1.
It should be noted that, if the current image block is determined as a flat region of the corner point, the corner point is not detected
In an embodiment of the present invention, the second preset threshold may be 0.5. It should be noted that the above threshold is only for illustrative purposes, and the user may select other values to set as needed.
And step S3, calculating the absolute gradient of the current pixel according to the calculated gradient kernel, and judging whether the current pixel is an angular point according to the absolute gradient.
Specifically, the absolute gradient is calculated using the following formula,
|G|＝|Gx|+|Gy|
and when the absolute gradient is larger than a third preset threshold, judging that the current pixel is an angular point, otherwise, judging that the current pixel is not an angular point.
In an embodiment of the present invention, the third preset threshold may be 2.0. It should be noted that the above threshold is only for illustrative purposes, and the user may select other values to set as needed.
And step S4, integrating the corner points of all the blocks, wherein the integrated corner points are all the corner points of the current image.
According to the rapid angular point detection method based on dynamic frequency conversion, the flat area and the angular point rich area of the image are detected through scanning the image, angular point detection is not carried out on the flat area, and angular points are searched in the angular point rich area by using a convolution technology, so that the angular point is rapidly detected. The method can effectively improve the performance of detecting the corner points of the image, and has great significance for systems with high real-time requirements.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made in the above embodiments by those of ordinary skill in the art without departing from the principle and spirit of the present invention. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (6)
1. A fast angular point detection method based on dynamic frequency conversion is characterized by comprising the following steps:
step S1, dividing the image into a plurality of blocks, calculating the mean square error of each block, and judging the image to be a flat area or an angular point rich area according to the mean square error;
step S2, when the image is judged to be an angular point rich area, calculating a gradient kernel according to the mean square error;
step S3, calculating the absolute gradient of the current pixel according to the gradient kernel obtained by calculation, and judging whether the current pixel is an angular point or not according to the absolute gradient;
and step S4, integrating the corner points of all the blocks, wherein the integrated corner points are all the corner points of the current image.
2. The method as claimed in claim 1, wherein in step S1, the average value of the image is counted in each block, then the variance of each pixel is calculated, and finally the mean square error of the current block is calculated, the calculated mean square error is planned to be between 0 and 1, a mean square error coefficient is generated, when the mean square error coefficient is smaller than a first preset threshold, the image is determined to be a flat region, otherwise, the image is determined to be a rich-corner region.
3. The method for fast corner detection based on dynamic frequency conversion according to claim 1, wherein in step S2, when the mean square error is greater than a second preset threshold, a first gradient template is used to calculate a gradient kernel; and when the mean square error is less than or equal to a second preset threshold value, calculating a gradient kernel by adopting a second gradient template.
4. The dynamic frequency conversion-based fast corner detection method according to claim 3, wherein the first gradient template Gx is:
the second gradient template Gy is:
wherein a is a mean square error coefficient.
5. The dynamic frequency conversion-based fast corner detection method according to claim 1, wherein in said step S3, the absolute gradient is calculated by using the following formula,
|G|＝|Gx|+|Gy|。
6. the method as claimed in claim 1, wherein in step S3, said determining whether the current pixel is a corner point according to the absolute gradient includes: and when the absolute gradient is the third preset threshold, judging that the current pixel is an angular point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910783617.8A CN110728285B (en) | 2019-08-23 | 2019-08-23 | Rapid corner detection method based on dynamic frequency conversion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910783617.8A CN110728285B (en) | 2019-08-23 | 2019-08-23 | Rapid corner detection method based on dynamic frequency conversion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110728285A true CN110728285A (en) | 2020-01-24 |
| CN110728285B CN110728285B (en) | 2022-06-24 |
Family
ID=69217736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910783617.8A Active CN110728285B (en) | 2019-08-23 | 2019-08-23 | Rapid corner detection method based on dynamic frequency conversion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110728285B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS61120002A (en) * | 1984-11-16 | 1986-06-07 | Toyota Central Res & Dev Lab Inc | Method and device for detecting corner point of image |
| JP2014006682A (en) * | 2012-06-25 | 2014-01-16 | Juki Corp | Image processing apparatus |
| WO2017049994A1 (en) * | 2015-09-25 | 2017-03-30 | 深圳大学 | Hyperspectral image corner detection method and system |
| US20180174328A1 (en) * | 2016-05-12 | 2018-06-21 | Huizhou University | Turning radius-based corner detection algorithm |
| CN108537796A (en) * | 2018-03-19 | 2018-09-14 | 太原理工大学 | Adaptive H arris angular-point detection methods based on template edge |
| CN108615041A (en) * | 2018-05-09 | 2018-10-02 | 桂林电子科技大学 | A kind of angular-point detection method |
-
2019
- 2019-08-23 CN CN201910783617.8A patent/CN110728285B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS61120002A (en) * | 1984-11-16 | 1986-06-07 | Toyota Central Res & Dev Lab Inc | Method and device for detecting corner point of image |
| JP2014006682A (en) * | 2012-06-25 | 2014-01-16 | Juki Corp | Image processing apparatus |
| WO2017049994A1 (en) * | 2015-09-25 | 2017-03-30 | 深圳大学 | Hyperspectral image corner detection method and system |
| US20180174328A1 (en) * | 2016-05-12 | 2018-06-21 | Huizhou University | Turning radius-based corner detection algorithm |
| CN108537796A (en) * | 2018-03-19 | 2018-09-14 | 太原理工大学 | Adaptive H arris angular-point detection methods based on template edge |
| CN108615041A (en) * | 2018-05-09 | 2018-10-02 | 桂林电子科技大学 | A kind of angular-point detection method |
Non-Patent Citations (1)
| Title |
|---|
| 程希希 等: "基于模板匹配的快速角点检测方法", 《计算机工程》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110728285B (en) | 2022-06-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112115953B (en) | Optimized ORB algorithm based on RGB-D camera combined plane detection and random sampling coincidence algorithm | |
| WO2022002039A1 (en) | Visual positioning method and device based on visual map | |
| CN109784250B (en) | Positioning method and device of automatic guide trolley | |
| CN111899334A (en) | Visual synchronous positioning and map building method and device based on point-line characteristics | |
| CN108573471B (en) | Image processing apparatus, image processing method, and recording medium | |
| CN108257153B (en) | Target tracking method based on direction gradient statistical characteristics | |
| US10643338B2 (en) | Object detection device and object detection method | |
| CN112861870B (en) | Pointer instrument image correction method, system and storage medium | |
| CN117557820B (en) | Quantum dot optical film damage detection method and system based on machine vision | |
| CN106771329B (en) | Method for detecting running speed of unmanned aerial vehicle in deceleration process | |
| CN113256729A (en) | External parameter calibration method, device, equipment and storage medium for laser radar and camera | |
| US20190279022A1 (en) | Object recognition method and device thereof | |
| JP5371685B2 (en) | Information processing apparatus, control method, and program | |
| CN107808165B (en) | Infrared image matching method based on SUSAN corner detection | |
| CN103714550A (en) | Image registration automatic optimization algorithm based on matching of curve characteristic evaluation | |
| WO2020007323A1 (en) | Supermarket shopping cart positioning method, supermarket shopping cart positioning system, and supermarket shopping cart | |
| CN114674826A (en) | Visual detection method and detection system based on cloth | |
| CN110728285B (en) | Rapid corner detection method based on dynamic frequency conversion | |
| CN117893468A (en) | Rapid 3D hole detection method, system, computer equipment and medium | |
| CN114578341A (en) | Radar tracking method, noise elimination method, device and equipment | |
| CN110349169B (en) | Linear measuring method | |
| CN107067411B (en) | Mean-shift tracking method combined with dense features | |
| CN114049380B (en) | Target object positioning and tracking method, device, computer equipment and storage medium | |
| CN110458881B (en) | Method for positioning center of oblique-incidence laser spot without threshold | |
| CN104616302A (en) | Real-time object identification 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 |