CN102955944A - Human body local feature extracting method for human body detection - Google Patents
Human body local feature extracting method for human body detection Download PDFInfo
- Publication number
- CN102955944A CN102955944A CN2011102501699A CN201110250169A CN102955944A CN 102955944 A CN102955944 A CN 102955944A CN 2011102501699 A CN2011102501699 A CN 2011102501699A CN 201110250169 A CN201110250169 A CN 201110250169A CN 102955944 A CN102955944 A CN 102955944A
- Authority
- CN
- China
- Prior art keywords
- pixel
- textural characteristics
- gray
- human body
- image
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a human body local feature extracting method for human body detection. The method comprises the following steps of: designating a rectangular detection window in a designated area in a picture to be processed from a training set and/or testing set; performing texture feature extraction on all pixel points in the rectangular detection window according to the set texture feature extraction method to acquire a texture feature value of each pixel in the rectangular detection window; counting the texture feature values to acquire texture feature histogram serving as the human body local feature related to the distribution condition of the texture feature values of all pixel points in the rectangular detecting window; and applying the human body local feature into a training or a test of a human body detection system. By the method for extracting human body local feature, the calculation amount is reduced, the human body local attributes can be described well, and the method can be conveniently applied to the human body detection to meet real-time requirement.
Description
Technical field
The present invention relates to a kind of body local feature extracting method for human detection, belong to field of computer technology, particularly belong to technical field of computer vision.
Background technology
Along with the development of computer technology, computer vision is widely used in a lot of fields.Human detection based on computer vision is academia and important research topic of industry member.For the people, by observing piece image, judge in the middle of the image whether the human body appearance is arranged, be an easily thing, if but allow computing machine finish this task, remain at present a theory and technology difficult problem that solves not yet fully.
Present human body detecting method based on computer vision has a variety of, and wherein the detection method based on the body local feature is a very important developing direction, and aspect this, the local feature that how effectively to extract human body is a step of most critical.Although the method that has had some human body local features to extract is invented, all there are some technical defectives in these methods.Therefore finding a kind of body local feature extracting method that effectively is used for human detection still is technical field of computer vision technical barrier in the urgent need to address.
Summary of the invention
In view of this, the objective of the invention is to invent a kind of extracting method of body local feature, in order to be applied in the actual needs based on the human detection of computer vision.
In order to achieve the above object, the present invention proposes a kind of body local feature extracting method for human detection, described method comprises following operation steps:
(1) from training set and/or test set in the selected piece image to be processed, at hough transform window of regional assignment of appointment;
(2) for all pixels in the described hough transform window, the texture characteristic extracting method according to setting carries out texture feature extraction, obtains the textural characteristics value of each pixel in the described hough transform window;
(3) the textural characteristics value of all pixels in the described hough transform window is added up, acquisition is about the textural characteristics histogram of all pixel textural characteristics value distribution situations in the described hough transform window, this textural characteristics histogram is used among the training or test of detecting system of human body namely as the body local feature.
Described step (2) specifically comprises following operation steps:
(21) with the pixel p[i of a textural characteristics to be extracted in the described hough transform window, j] centered by, p[i, j] be illustrated in the pixel of the capable j row of i of described processing image, find out following two groups of orderly pixels set, first group of orderly pixel set be p[i, j+r], p[i-r, j+r], p[i-r, j], p[i-r, j-r], p[i, j-r], p[i+r, j-r], p[i+r, j], p[i+r, j+r] }, second group of orderly pixel set be p[i, j+R], p[i-R, j+R], p[i-R, j], p[i-R, j-R], p[i, j-R], p[i+R, j-R], p[i+R, j], p[i+R, j+R] }, wherein r and R are natural numbers, and R is greater than r;
(22) from described this width of cloth image to be processed, obtain the gray-scale value of each pixel in first group of orderly pixel set, then compare with the gray-scale value of the pixel of described textural characteristics to be extracted respectively, if the gray-scale value of i pixel was not less than the gray-scale value of the pixel of described textural characteristics to be extracted during first group of orderly pixel gathered, then allow vi get 1, otherwise allow vi get 0, i is from 1 to 8 natural number, obtains like this one by 1 and 08 string of binary characters: v1v2v3v4v5v6v7v8 that form;
(23) from described this width of cloth image to be processed, obtain the gray-scale value of each pixel in second group of orderly pixel set, then compare with the gray-scale value of the pixel of described textural characteristics to be extracted respectively, if the gray-scale value of i pixel was not less than the gray-scale value of the pixel of described textural characteristics to be extracted during second group of orderly pixel gathered, then allow Vi get 1, otherwise allow Vi get 0, i is from 1 to 8 natural number, obtains like this one by 1 and 08 string of binary characters: V1V2V3V4V5V6V7V8 that form;
(24) step (22) and resulting 8 the string of binary characters v1v2v3v4v5v6v7v8 of step (23) and V1V2V3V4V5V6V7V8 are carried out the step-by-step xor operation, obtain one by 1 and 08 string of binary characters: t1t2t3t4t5t6t7t8 that form;
(25) resulting these 8 the string of binary characters t1t2t3t4t5t6t7t8 of step (24) are carried out 7 circumference shift left operation or 7 circumference dextroposition operation, obtain 78 new strings of binary characters, add those original 8 string of binary characters t1t2t3t4t5t6t7t8, obtain altogether 88 strings of binary characters, these 88 strings of binary characters are converted to 8 decimal numbers, and that minimum decimal number of value is the textural characteristics of the pixel of described textural characteristics to be extracted.
Parameters R and parameter r will set as required suitable proportionate relationship for or set suitable difference relation in the described step (21).
If image to be processed is not gray level image, carry out again subsequent operation after can being converted into gray level image to original image first, perhaps carry out subsequent operation with one of original image suitable component image as gray level image.
Beneficial effect of the present invention is: the extracting method calculated amount of body local feature of the present invention is little, can be described the body local attribute well, can be applied to easily human detection, requirement of real time.
Description of drawings
Fig. 1 is the schematic flow sheet of a kind of body local feature extracting method for human detection of the present invention.
Fig. 2 is the synoptic diagram of two groups of orderly pixel set in one embodiment of the invention.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in further detail below in conjunction with accompanying drawing.
Referring to Fig. 1, introduce a kind of body local feature extracting method for human detection that the present invention proposes, described method comprises following operation steps:
(1) from training set and/or test set in the selected piece image to be processed, at hough transform window of regional assignment of appointment;
(2) for all pixels in the described hough transform window, the texture characteristic extracting method according to setting carries out texture feature extraction, obtains the textural characteristics value of each pixel in the described hough transform window;
(3) the textural characteristics value of all pixels in the described hough transform window is added up, acquisition is about the textural characteristics histogram of all pixel textural characteristics value distribution situations in the described hough transform window, this textural characteristics histogram is used among the training or test of detecting system of human body namely as the body local feature.
Described step (2) specifically comprises following operation steps:
(21) with the pixel p[i of a textural characteristics to be extracted in the described hough transform window, j] centered by, p[i, j] be illustrated in the pixel of the capable j row of i of described processing image, find out following two groups of orderly pixels set, first group of orderly pixel set be p[i, j+r], p[i-r, j+r], p[i-r, j], p[i-r, j-r], p[i, j-r], p[i+r, j-r], p[i+r, j], p[i+r, j+r] }, second group of orderly pixel set be p[i, j+R], p[i-R, j+R], p[i-R, j], p[i-R, j-R], p[i, j-R], p[i+R, j-R], p[i+R, j], p[i+R, j+R] }, wherein r and R are natural numbers, and R is greater than r;
For example, referring to Fig. 2, establish r=2, R=4, then first group of orderly pixel set be exactly p[i, j+2], p[i-2, j+2], p[i-2, j] and, p[i-2, j-2], p[i, j-2], p[i+2, j-2] and, p[i+2, j], p[i+2, j+2] }, second group of orderly pixel set be exactly p[i, j+4], p[i-4, j+4], p[i-4, j], p[i-4, j-4] and, p[i, j-4], p[i+4, j-4], p[i+4, j] and, p[i+4, j+4].
Process image for being positioned near the pixel of the textural characteristics to be extracted of position, image border, the position coordinates of some pixel is certain to exceed the scope of the image of processing in two groups of orderly pixels set recited above, at this moment can adopt the method that original image is carried out mirror image or carries out periodic extension to process.
(22) from described this width of cloth image to be processed, obtain the gray-scale value of each pixel in first group of orderly pixel set, then compare with the gray-scale value of the pixel of described textural characteristics to be extracted respectively, if the gray-scale value of i pixel was not less than the gray-scale value of the pixel of described textural characteristics to be extracted during first group of orderly pixel gathered, then allow vi get 1, otherwise allow vi get 0, i is from 1 to 8 natural number, obtains like this one by 1 and 08 string of binary characters: v1v2v3v4v5v6v7v8 that form;
For example: if the gray-scale value of each pixel sequentially is { 80,125,90,200 in first group of orderly pixel set, 30,80,230,15}, the gray-scale value of the pixel of textural characteristics to be extracted is 100, the string of binary characters v1v2v3v4v5v6v7v8 that then obtains is exactly: 01010010.
(23) from described this width of cloth image to be processed, obtain the gray-scale value of each pixel in second group of orderly pixel set, then compare with the gray-scale value of the pixel of described textural characteristics to be extracted respectively, if the gray-scale value of i pixel was not less than the gray-scale value of the pixel of described textural characteristics to be extracted during second group of orderly pixel gathered, then allow Vi get 1, otherwise allow Vi get 0, i is from 1 to 8 natural number, obtains like this one by 1 and 08 string of binary characters: V1V2V3V4V5V6V7V8 that form;
For example: if the gray-scale value of each pixel sequentially is { 110,135,80,70 in second group of orderly pixel set, 130,180,130,75}, the gray-scale value of the pixel of textural characteristics to be extracted is 100, the string of binary characters V1V2V3V4V5V6V7V8 that then obtains is exactly: 11001110.
(24) step (22) and resulting 8 the string of binary characters v1v2v3v4v5v6v7v8 of step (23) and V1V2V3V4V5V6V7V8 are carried out the step-by-step xor operation, obtain one by 1 and 08 string of binary characters: t1t2t3t4t5t6t7t8 that form;
For example: if we carry out the step-by-step xor operation to the string of binary characters v1v2v3v4v5v6v7v8 that obtains previously (namely 01010010) and V1V2V3V4V5V6V7V8 (namely 11001110), the value of the string of binary characters t1t2t3t4t5t6t7t8 that then obtains is exactly: 10011100.
(25) resulting these 8 the string of binary characters t1t2t3t4t5t6t7t8 of step (24) are carried out 7 circumference shift left operation or 7 circumference dextroposition operation, obtain 78 new strings of binary characters, add those original 8 string of binary characters t1t2t3t4t5t6t7t8, obtain altogether 88 strings of binary characters, these 88 strings of binary characters are converted to 8 decimal numbers, and that minimum decimal number of value is the textural characteristics of the pixel of described textural characteristics to be extracted.
For example: if we carry out the operation of shifting left of 7 circumference again to the string of binary characters t1t2t3t4t5t6t7t8 that obtains previously (namely 10011100), can obtain respectively: 00111001,01110010,11100100,11001001,10010011,00100111 and 01001110, add those original 8 strings of binary characters 10011100, obtain altogether 88 strings of binary characters, these 88 strings of binary characters are converted to 8 decimal numbers, respectively 57,114,228,201,147,39,78,156, minimum value is 39, and this value is the textural characteristics of the pixel of described textural characteristics to be extracted.
Parameters R and parameter r will set as required suitable proportionate relationship for or set suitable difference relation in the described step (21).A R/r=3 or R-r=3 pixel for example.
If image to be processed is not gray level image, carry out again subsequent operation after can being converted into gray level image to original image first, perhaps carry out subsequent operation with one of original image suitable component image as gray level image.
For example: if image to be processed is the RGB coloured image, carry out again relevant treatment after then can selecting this RGB coloured image is converted to gray level image, perhaps select the R component image of this coloured image to carry out subsequent operation as gray level image.
Claims (4)
1. body local feature extracting method that is used for human detection, it is characterized in that: described method comprises following operation steps:
(1) from training set and/or test set in the selected piece image to be processed, at hough transform window of regional assignment of appointment;
(2) for all pixels in the described hough transform window, the texture characteristic extracting method according to setting carries out texture feature extraction, obtains the textural characteristics value of each pixel in the described hough transform window;
(3) the textural characteristics value of all pixels in the described hough transform window is added up, acquisition is about the textural characteristics histogram of all pixel textural characteristics value distribution situations in the described hough transform window, this textural characteristics histogram is used among the training or test of detecting system of human body namely as the body local feature.
2. a kind of body local feature extracting method for human detection according to claim 1, it is characterized in that: described step (2) specifically comprises following operation steps:
(21) with the pixel p[i of a textural characteristics to be extracted in the described hough transform window, j] centered by, p[i, j] be illustrated in the pixel of the capable j row of i of described processing image, find out following two groups of orderly pixels set, first group of orderly pixel set be p[i, j+r], p[i-r, j+r], p[i-r, j], p[i-r, j-r], p[i, j-r], p[i+r, j-r], p[i+r, j], p[i+r, j+r] }, second group of orderly pixel set be p[i, j+R], p[i-R, j+R], p[i-R, j], p[i-R, j-R], p[i, j-R], p[i+R, j-R], p[i+R, j], p[i+R, j+R] }, wherein r and R are natural numbers, and R is greater than r;
(22) from described this width of cloth image to be processed, obtain the gray-scale value of each pixel in first group of orderly pixel set, then compare with the gray-scale value of the pixel of described textural characteristics to be extracted respectively, if the gray-scale value of i pixel was not less than the gray-scale value of the pixel of described textural characteristics to be extracted during first group of orderly pixel gathered, then allow vi get 1, otherwise allow vi get 0, i is from 1 to 8 natural number, obtains like this one by 1 and 08 string of binary characters: v1v2v3v4v5v6v7v8 that form;
(23) from described this width of cloth image to be processed, obtain the gray-scale value of each pixel in second group of orderly pixel set, then compare with the gray-scale value of the pixel of described textural characteristics to be extracted respectively, if the gray-scale value of i pixel was not less than the gray-scale value of the pixel of described textural characteristics to be extracted during second group of orderly pixel gathered, then allow Vi get 1, otherwise allow Vi get 0, i is from 1 to 8 natural number, obtains like this one by 1 and 08 string of binary characters: V1V2V3V4V5V6V7V8 that form;
(24) step (22) and resulting 8 the string of binary characters v1v2v3v4v5v6v7v8 of step (23) and V1V2V3V4V5V6V7V8 are carried out the step-by-step xor operation, obtain one by 1 and 08 string of binary characters: t1t2t3t4t5t6t7t8 that form;
(25) resulting these 8 the string of binary characters t1t2t3t4t5t6t7t8 of step (24) are carried out 7 circumference shift left operation or 7 circumference dextroposition operation, obtain 78 new strings of binary characters, add those original 8 string of binary characters t1t2t3t4t5t6t7t8, obtain altogether 88 strings of binary characters, these 88 strings of binary characters are converted to 8 decimal numbers, and that minimum decimal number of value is the textural characteristics of the pixel of described textural characteristics to be extracted.
3. a kind of body local feature extracting method for human detection according to claim 2 is characterized in that: parameters R and parameter r will set as required suitable proportionate relationship for or set suitable difference relation in the described step (21).
According to claim 1 with 2 described a kind of body local feature extracting methods for human detection, it is characterized in that: if image to be processed is not gray level image, carry out again subsequent operation after can being converted into gray level image to original image first, perhaps carry out subsequent operation with one of original image suitable component image as gray level image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110250169.9A CN102955944B (en) | 2011-08-29 | 2011-08-29 | A kind of body local feature extracting method for human detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110250169.9A CN102955944B (en) | 2011-08-29 | 2011-08-29 | A kind of body local feature extracting method for human detection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102955944A true CN102955944A (en) | 2013-03-06 |
CN102955944B CN102955944B (en) | 2015-08-19 |
Family
ID=47764729
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201110250169.9A Expired - Fee Related CN102955944B (en) | 2011-08-29 | 2011-08-29 | A kind of body local feature extracting method for human detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102955944B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103455817A (en) * | 2013-04-28 | 2013-12-18 | 南京理工大学 | Method for extracting human body features of robust time-space domain |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101739548A (en) * | 2009-02-11 | 2010-06-16 | 北京智安邦科技有限公司 | Eye detection method and system |
US20100329517A1 (en) * | 2009-06-26 | 2010-12-30 | Microsoft Corporation | Boosted face verification |
CN101996308A (en) * | 2009-08-19 | 2011-03-30 | 北京中星微电子有限公司 | Human face identification method and system and human face model training method and system |
-
2011
- 2011-08-29 CN CN201110250169.9A patent/CN102955944B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101739548A (en) * | 2009-02-11 | 2010-06-16 | 北京智安邦科技有限公司 | Eye detection method and system |
US20100329517A1 (en) * | 2009-06-26 | 2010-12-30 | Microsoft Corporation | Boosted face verification |
CN101996308A (en) * | 2009-08-19 | 2011-03-30 | 北京中星微电子有限公司 | Human face identification method and system and human face model training method and system |
Non-Patent Citations (1)
Title |
---|
江小辉: "基于改进型LBP特征的人脸识别研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》, 31 March 2010 (2010-03-31) * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103455817A (en) * | 2013-04-28 | 2013-12-18 | 南京理工大学 | Method for extracting human body features of robust time-space domain |
Also Published As
Publication number | Publication date |
---|---|
CN102955944B (en) | 2015-08-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
CN109509187B (en) | Efficient inspection algorithm for small defects in large-resolution cloth images | |
WO2022236876A1 (en) | Cellophane defect recognition method, system and apparatus, and storage medium | |
CN104361314B (en) | Based on infrared and transformer localization method and device of visual image fusion | |
CN104574353B (en) | The surface defect decision method of view-based access control model conspicuousness | |
CN108596102B (en) | RGB-D-based indoor scene object segmentation classifier construction method | |
CN106650770A (en) | Mura defect detection method based on sample learning and human visual characteristics | |
CN103729649B (en) | A kind of image rotation angle detection method and device | |
CN110232379A (en) | A kind of vehicle attitude detection method and system | |
CN108364281A (en) | A kind of ribbon edge hair defect defect inspection method based on convolutional neural networks | |
CN105354847A (en) | Fruit surface defect detection method based on adaptive segmentation of sliding comparison window | |
CN108416774A (en) | A kind of fabric types recognition methods based on fine granularity neural network | |
CN105389541B (en) | The recognition methods of fingerprint image and device | |
CN111476758A (en) | Defect detection method and device for AMO L ED display screen, computer equipment and storage medium | |
CN105572143A (en) | Method for detecting periodic defect of calendered material surface in calendering process | |
CN110020691A (en) | LCD screen defect inspection method based on the training of convolutional neural networks confrontation type | |
WO2021227289A1 (en) | Deep learning-based low-quality two-dimensional barcode detection method in complex background | |
CN116310548A (en) | Method for detecting invasive plant seeds in imported seed products | |
CN111161228B (en) | Button surface defect detection method based on transfer learning | |
CN102955944A (en) | Human body local feature extracting method for human body detection | |
CN110503627A (en) | Building tear detection method, device, storage medium and computer equipment | |
Yang et al. | Cherry recognition based on color channel transform | |
CN113901947A (en) | Intelligent identification method for tire surface flaws under small sample | |
CN112669300A (en) | Defect detection method and device, computer equipment and storage medium | |
CN110660042A (en) | Traditional Chinese medicine bag number identification algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150819 Termination date: 20160829 |
|
CF01 | Termination of patent right due to non-payment of annual fee |