CN108985191A - A kind of contour extraction method based on mobile device gesture identification - Google Patents
A kind of contour extraction method based on mobile device gesture identification Download PDFInfo
- Publication number
- CN108985191A CN108985191A CN201810689696.1A CN201810689696A CN108985191A CN 108985191 A CN108985191 A CN 108985191A CN 201810689696 A CN201810689696 A CN 201810689696A CN 108985191 A CN108985191 A CN 108985191A
- Authority
- CN
- China
- Prior art keywords
- point
- pixel
- neighborhood
- contour extraction
- extraction method
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/107—Static hand or arm
- G06V40/113—Recognition of static hand signs
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of contour extraction methods based on mobile device gesture identification, the contour extraction method carries out track-while-scan to the eight neighborhood of edge pixel, edge pixel point is searched for, and records corresponding Position Number, edge pixel point is together in series using number order, combination forms profile, only needing simple condition to carry out cyclic search can be completed, and algorithm model is simple, and calculation amount is small, with there is very high real-time on the mobile apparatus, postpone low.
Description
Technical field
The present invention relates to the technical field of gesture identification, specifically a kind of contours extract based on mobile device gesture identification
Method.
Background technique
It include mainly three Hand Gesture Segmentation, feature extraction and gesture identification steps in Gesture Recognition, and in feature
In extraction, the extraction of profile is precondition.Contours extract is typically all the first derivative extreme value or second dervative for utilizing image
Zero crossing information extract edge, edge pixel is combined to form profile.Traditional contour extraction method is all based on PC system
System, algorithm model is complicated, computationally intensive, with there is obvious delay on the mobile apparatus.
Summary of the invention
The present invention in view of the above problems, provides a kind of contours extract side based on mobile device gesture identification
Method, algorithm model is simple, computationally intensive, and being used in the not high mobile device of hardware condition has lower delay.
A kind of contour extraction method based on mobile device gesture identification of the present invention, comprising the following steps:
1) it searches first edge pixel point in order on the image, is recorded as profile point S0Into point sequence of an outline table,
The eight neighborhood model centered on the pixel is established, the pixel number of eight neighborhood is edited in a fixed order, is opened from number d
The pixel of beginning search eight neighborhood；
2) eight field pixels are successively searched for according to the number order of eight neighborhood pixel, if searching out new edge pixel
Point,
Then record profile point SnInto point sequence of an outline table, and record profile point SnNumber d ' in eight neighborhood；
3) by profile point SnIt is set as current outline point, is established with same eight neighborhood coded sequence with current outline point and is
The eight neighborhood model of the heart resets search and is numbered d=(d+3) mod8+1, continues to execute step B；
4) until the edge pixel point searched is starting point, i.e. Sn=S0, then terminate to search for.
As a further improvement, in the step A search order be by image from left to right, from top to bottom.
As a further improvement, the number order is that the nine grids upper right angle point centered on current outline point starts
Numbered counter-clockwise.
As a further improvement, the d=5.
As a further improvement, further include the extraction of edge pixel point: one one-dimensional Gaussian mask G of creation is used for figure
As I progress convolution algorithm, the gradient M of each pixel on image I is obtained, gradient value is greater than high threshold values T on imagehPixel
Point label is point, connects edge pixel point and gradient value is greater than low valve valve T1Pixel mark as picture
Vegetarian refreshments.
As a further improvement, the convolution algorithm the following steps are included:
A, the first derivative of Gaussian function is created on the ranks direction of X and Y as one-dimensional mask, obtains GxAnd Gy；
B, convolution algorithm is done to image I using one-dimensional Gaussian mask G along row and along column respectively, obtains X-component image IxAnd Y
Component image Iy；
C, G is utilizedxTo IxIt carries out convolution algorithm and obtains Ix', also with GyTo IyIt carries out convolution algorithm and obtains Iy′；
D, the intensity of skirt response, i.e. gradient are calculated in conjunction with X-component and Y-component:
As a further improvement, the ratio of the high threshold values and low valve valve is 2:1.
As a further improvement, the ratio of the high threshold values and low valve valve is 3:1.
It is searched the invention has the following advantages: the contour extraction method carries out tracking to the eight neighborhood of edge pixel
Rope searches for edge pixel point, and records corresponding Position Number, and edge pixel point is together in series using number order, combines
Form profile, it is only necessary to which simple condition, which carries out cyclic search, can be completed, and algorithm model is simple, and calculation amount is small, is used in movement
There is very high real-time in equipment, postpones low.
Specific embodiment
A kind of contour extraction method based on mobile device gesture identification of the present embodiment, using eight neighborhood algorithm to edge picture
Vegetarian refreshments carries out forming relevant point sequence of an outline table according to search.The extraction Canny of edge pixel point in this implementation
Method of the operator as edge detection creates an one-dimensional Gaussian mask G and is used to carry out convolution algorithm, convolution algorithm to image I
Steps are as follows:
Specific step is as follows:
A, on the ranks direction of X and Y create Gaussian function first derivative as one-dimensional mask, obtain and；
B, convolution algorithm is done to image I using one-dimensional Gaussian mask G along row and along column respectively, obtains X-component image and Y points
Spirogram picture；
C, it obtains using to progress convolution algorithm, is obtained also with to progress convolution algorithm；
D, the intensity of skirt response, i.e. gradient are calculated in conjunction with X-component and Y-component.
The basic thought of Canny Operators Algorithm is: each edge pixel is associated with a direction, in an edge picture
The size of gradient on element should be bigger than the gradient magnitude in the pixel of both sides of edges.Traditional Canny Operators Algorithm is all
Be using final step be non-maximum suppression step, wherein the pixel of non local maximum can be all removed, last image is still
Include so grayscale value, therefore also need to carry out threshold operation to judge which pixel as edge pixel, which pixel is not edge
Pixel.In the embodiment, Canny operator finally obtains the gradient M of each pixel on image I, and gradient value is greater than on image
It is point that the pixel of high threshold values, which marks, connects edge pixel point and gradient value is marked greater than the pixel of low valve valve
It is denoted as edge pixel point.Low valve valve is connected for edge, and high threshold values is used to control the initial segment at edge, the high-low threshold value ratio of recommendation
Between 2:1 to 3:1.
After edge pixel point is found out, only scattered point, there is no connections to form profile.In flat image, appoint
It anticipates a pixel, there are eight pixels to be attached thereto, must there is the profile point being attached thereto in the eight neighborhood of profile point.
The eight neighborhood of edge pixel is numbered, number order is that the nine grids upper right angle point centered on current outline point starts
Numbered counter-clockwise, number table are as follows:
| 3 | 2 | 1 |
| 4 | Sn | 8 |
| 5 | 6 | 7 |
Edge pixel point in search eight neighborhood is recorded in point sequence of an outline table, and combination forms relevant profile, specifically
Steps are as follows:
1) on the image by from left to right, sequential search from top to bottom is recorded as profile point to first edge pixel point
S0Into point sequence of an outline table, the eight neighborhood model centered on the pixel is established, edits eight neighborhood in a fixed order
Pixel number, searches for the pixel of eight neighborhood since number d；
2) eight field pixels are successively searched for according to the number order of eight neighborhood pixel, if searching out new edge pixel
Point then records profile point SnInto point sequence of an outline table, and record profile point SnNumber d ' in eight neighborhood；
3) by profile point SnIt is set as current outline point, is established with same eight neighborhood coded sequence with current outline point and is
The eight neighborhood model of the heart resets search and is numbered d=(d+3) mod8+1, continues to execute step B；
4) until the edge pixel point searched is starting point, i.e. Sn=S0, then terminate to search for.
In specific embodiment, d can take 5, search for being the pixel that number is 5 since eight neighborhood.By upper
After the search for stating step, point sequence of an outline table is obtained, the obtained new profile point in point sequence of an outline table is all the above wheel
Based on exterior feature point, profile point forms required profile according to the permutation and combination of number.
Contours extract is carried out for the image that the mobile phone of different resolution is shot in embodiment, required timetable is as follows:
| Resolution ratio | 640x480 | 1280x720 | 1088x1088 | 1280x960 | 1440x1080 | 1920x1080 |
| Extraction time (ms) | 105 | 587 | 938 | 1876 | 2560 | 3362 |
As can be seen from the table, for the mobile phone lower for resolution ratio, the contours extract time is fewer, and real-time is higher.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in of the invention
Within protection scope.
Claims (8)
1. a kind of contour extraction method based on mobile device gesture identification, which comprises the following steps:
1) it searches first edge pixel point in order on the image, is recorded as profile point S0Into point sequence of an outline table, establishing should
Eight neighborhood model centered on pixel is edited the pixel number of eight neighborhood in a fixed order, is searched for since number d
The pixel of eight neighborhood；
2) eight field pixels are successively searched for according to the number order of eight neighborhood pixel, if searching out new edge pixel point,
Then record profile point SnInto point sequence of an outline table, and record profile point SnNumber d ' in eight neighborhood；
3) by profile point SnIt is set as current outline point, is established centered on current outline point by same eight neighborhood coded sequence
Eight neighborhood model resets search and is numbered d=(d+3) mod8+1, continues to execute step B；
4) until the edge pixel point searched is starting point, i.e. Sn=S0, then terminate to search for.
2. contour extraction method according to claim 1, which is characterized in that search order is by image in the step A
From left to right, from top to bottom.
3. contour extraction method according to claim 1, which is characterized in that the number order is to be with current outline point
The nine grids upper right angle point at center starts numbered counter-clockwise.
4. contour extraction method according to claim 1, which is characterized in that the d=5.
5. contour extraction method according to claim 1, which is characterized in that further include the extraction of edge pixel point: creation
One one-dimensional Gaussian mask G is used to carry out convolution algorithm to image I, obtains the gradient M of each pixel on image I, image
Upper gradient value is greater than high threshold values ThPixel mark as point, connect edge pixel point and gradient value be greater than low valve
Value T1Pixel mark as point.
6. contour extraction method according to claim 5, which is characterized in that the convolution algorithm the following steps are included:
A, the first derivative of Gaussian function is created on the ranks direction of X and Y as one-dimensional mask, obtains GxAnd Gy；
B, convolution algorithm is done to image I using one-dimensional Gaussian mask G along row and along column respectively, obtains X-component image IxAnd Y-component
Image Iy；
C, G is utilizedxTo IxIt carries out convolution algorithm and obtains I 'x, also with GyTo IyIt carries out convolution algorithm and obtains I 'y；
D, the intensity of skirt response, i.e. gradient are calculated in conjunction with X-component and Y-component:
7. contour extraction method according to claim 5, which is characterized in that the ratio of the high threshold values and low valve valve is 2:
1。
8. contour extraction method according to claim 5, which is characterized in that the ratio of the high threshold values and low valve valve is 3:
1。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810689696.1A CN108985191A (en) | 2018-06-28 | 2018-06-28 | A kind of contour extraction method based on mobile device gesture identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810689696.1A CN108985191A (en) | 2018-06-28 | 2018-06-28 | A kind of contour extraction method based on mobile device gesture identification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108985191A true CN108985191A (en) | 2018-12-11 |
Family
ID=64539493
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810689696.1A Pending CN108985191A (en) | 2018-06-28 | 2018-06-28 | A kind of contour extraction method based on mobile device gesture identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108985191A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111127404A (en) * | 2019-12-06 | 2020-05-08 | 广州柏视医疗科技有限公司 | Medical image contour rapid extraction method |
| CN112200738A (en) * | 2020-09-29 | 2021-01-08 | 平安科技（深圳）有限公司 | Method and device for identifying protrusion of shape and computer equipment |
| CN115098171A (en) * | 2022-05-20 | 2022-09-23 | 智绘微电子科技(南京)有限公司 | Method for optimizing picture compression processing in VBIOS |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679145A (en) * | 2013-12-06 | 2014-03-26 | 河海大学 | Automatic gesture recognition method |
| CN104680127A (en) * | 2014-12-18 | 2015-06-03 | 闻泰通讯股份有限公司 | Gesture identification method and gesture identification system |
| EP2980728A1 (en) * | 2014-08-01 | 2016-02-03 | Imersivo, S.L. | Procedure for identifying a hand gesture |
| CN106843472A (en) * | 2016-12-28 | 2017-06-13 | 歌尔科技有限公司 | Gesture identification method and device, virtual reality device, programmable device |
| WO2017113794A1 (en) * | 2015-12-31 | 2017-07-06 | 北京体基科技有限公司 | Gesture recognition method, control method and apparatus, and wrist-type device |
| CN107038424A (en) * | 2017-04-20 | 2017-08-11 | 华中师范大学 | A kind of gesture identification method |
| CN107330354A (en) * | 2017-03-20 | 2017-11-07 | 长沙理工大学 | A kind of natural gesture identification method |
| CN107610135A (en) * | 2017-09-21 | 2018-01-19 | 程仕燚 | Quick finger dividing method based on geodesic curve distance algorithm and morphological recognition |
-
2018
- 2018-06-28 CN CN201810689696.1A patent/CN108985191A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679145A (en) * | 2013-12-06 | 2014-03-26 | 河海大学 | Automatic gesture recognition method |
| EP2980728A1 (en) * | 2014-08-01 | 2016-02-03 | Imersivo, S.L. | Procedure for identifying a hand gesture |
| CN104680127A (en) * | 2014-12-18 | 2015-06-03 | 闻泰通讯股份有限公司 | Gesture identification method and gesture identification system |
| WO2017113794A1 (en) * | 2015-12-31 | 2017-07-06 | 北京体基科技有限公司 | Gesture recognition method, control method and apparatus, and wrist-type device |
| CN106843472A (en) * | 2016-12-28 | 2017-06-13 | 歌尔科技有限公司 | Gesture identification method and device, virtual reality device, programmable device |
| CN107330354A (en) * | 2017-03-20 | 2017-11-07 | 长沙理工大学 | A kind of natural gesture identification method |
| CN107038424A (en) * | 2017-04-20 | 2017-08-11 | 华中师范大学 | A kind of gesture identification method |
| CN107610135A (en) * | 2017-09-21 | 2018-01-19 | 程仕燚 | Quick finger dividing method based on geodesic curve distance algorithm and morphological recognition |
Non-Patent Citations (3)
| Title |
|---|
| 姚峰林: "《数字图像处理及在工程中的应用》", 30 April 2014 * |
| 毛晓洁: "基于Android平台的手势识别技术的研究与应用", 《中国优秀硕士学位论文全文数据库》 * |
| 邓继忠: "《数字图像处理技术》", 30 September 2005 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111127404A (en) * | 2019-12-06 | 2020-05-08 | 广州柏视医疗科技有限公司 | Medical image contour rapid extraction method |
| CN112200738A (en) * | 2020-09-29 | 2021-01-08 | 平安科技（深圳）有限公司 | Method and device for identifying protrusion of shape and computer equipment |
| CN115098171A (en) * | 2022-05-20 | 2022-09-23 | 智绘微电子科技(南京)有限公司 | Method for optimizing picture compression processing in VBIOS |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110738101B (en) | Behavior recognition method, behavior recognition device and computer-readable storage medium | |
| US10496901B2 (en) | Image recognition method | |
| CN106599028B (en) | Book content searching and matching method based on video image processing | |
| CN104517102B (en) | Student classroom notice detection method and system | |
| CN107316031A (en) | The image characteristic extracting method recognized again for pedestrian | |
| US9020195B2 (en) | Object tracking device, object tracking method, and control program | |
| CN104299008B (en) | Vehicle type classification method based on multi-feature fusion | |
| WO2016066038A1 (en) | Image body extracting method and system | |
| KR101636370B1 (en) | Image processing apparatus and method | |
| CN108985191A (en) | A kind of contour extraction method based on mobile device gesture identification | |
| KR20110032352A (en) | Image processing apparatus and method | |
| US10650234B2 (en) | Eyeball movement capturing method and device, and storage medium | |
| CN110472625B (en) | Chinese chess piece visual identification method based on Fourier descriptor | |
| CN104809736B (en) | Medical Slice Images closure bone contours computational methods based on priori | |
| CN108846385A (en) | Based on convolution-deconvolution neural network image recognition, bearing calibration and device | |
| US20230388449A1 (en) | Flying body control apparatus, flying body control method, and flying body control program | |
| CN104091325A (en) | Image ROI positioning method and device based on dimension invariant feature transformation during automobile instrument panel visual detection | |
| CN106127193A (en) | A kind of facial image recognition method | |
| US9286543B2 (en) | Characteristic point coordination system, characteristic point coordination method, and recording medium | |
| CN106022236A (en) | Action identification method based on human body contour | |
| CN105930789A (en) | Human body behavior recognition based on logarithmic Euclidean space BOW (bag of words) model | |
| KR20050063991A (en) | Image matching method and apparatus using image pyramid | |
| CN110598665B (en) | Pole number identification method based on vehicle-mounted mobile deep learning platform | |
| US9792675B1 (en) | Object recognition using morphologically-processed images | |
| CN104036494B (en) | A kind of rapid matching computation method for fruit image |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181211 |
|
| RJ01 | Rejection of invention patent application after publication |