CN103761506A - Method for recognizing fissured tongue based on support vector machine - Google Patents

Method for recognizing fissured tongue based on support vector machine Download PDF

Info

Publication number
CN103761506A
CN103761506A CN201410000707.2A CN201410000707A CN103761506A CN 103761506 A CN103761506 A CN 103761506A CN 201410000707 A CN201410000707 A CN 201410000707A CN 103761506 A CN103761506 A CN 103761506A
Authority
CN
China
Prior art keywords
tongue
identified
span value
fissuring
lingual surface
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
Application number
CN201410000707.2A
Other languages
Chinese (zh)
Inventor
邵卿
李晓强
傅之诚
吴晶晶
沙彩霞
卞世敏
汪晶晶
李继德
姚谦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Shanghai for Science and Technology
Original Assignee
University of Shanghai for Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Shanghai for Science and Technology filed Critical University of Shanghai for Science and Technology
Priority to CN201410000707.2A priority Critical patent/CN103761506A/en
Publication of CN103761506A publication Critical patent/CN103761506A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Image Processing (AREA)

Abstract

The invention discloses a method for recognizing a fissured tongue based on a support vector machine. The method for recognizing the fissured tongue based on the support vector machine comprises the following steps that (1) a fissured tongue feature training sample set is established; (2) a fissured tongue classifier based on the support vector machine is established; (3) the fissured tongue classifier based on the support vector machine is used for conducting recognition, and whether each sample to be detected in a fissure feature sample set of the tongue to be detected belongs to a fissured tongue picture is recognized. According to the method for recognizing the fissured tongue based on the support vector machine, reorganization can be further conducted based on a detected suspected tongue fissure area. The method is the continuation of an existing computer fissured tongue detection method. By the adoption of the method for recognizing the fissured tongue based on the support vector machine, through training of a disposable sample set, recognition work can be completed, setting of any other parameters and artificial factor intervention are not needed in the recognition process, and full-automatic recognition of the fissured tongue is achieved.

Description

Fissured tongue recognition methods based on support vector machine
Technical field
The present invention relates to a kind of fissured tongue recognition methods based on support vector machine, belong to technical field of image processing.
Background technology
Traditional Chinese medical science basic skills has four kinds, that is: hope, hear, ask, cut, and in observation, a people's tongue picture can reflect his health status.The feature of tongue picture comprises the aspects such as crackle, indentation, tongue fur, ecchymosis, petechia clinically, they have played respectively different indicative functions in diagnosis, wherein in tongue, there is crackle, be commonly called as fissured tongue, its feature is on back, to form zanjon, the picture vein that the orientation of ditch has, some picture brain lines, so fissured tongue also has the title of fern leaf tongue and brain line tongue.General patient is if there is the sign of fissured tongue, and they are mostly accompanied by heat and contain that impairment of yin, the deficiency of blood are not moistened, the insufficiency of the spleen wet disease such as soak so.But the development of the traditional Chinese medical science now still exists sizable subjective experience judgement, most of fissured tongue diagnostic result is subject to the restriction of the artificial subjective factors such as know-how, thinking ability, diagnostic skill, lacks objective appraisal foundation and unified evaluation criterion.Therefore, industry is needed a kind of objectively fissuring of tongue recognition methods automatically badly.
At present, utilizing the recognition methods of the fissuring of tongue of computer technology first will from lingual surface image, extract may be the doubtful region of fissuring of tongue, and then relies on artificial demarcation to determine whether really to exist fissuring of tongue.For example, Liu L.L. and Zhang D., Extracting Tongue Cracks Using the Wide Line Detector, ICMB 2008, LNCS 4901, pp. 49-56, 2008, the wide thread detector algorithm of the method utilization with parameter processed and only obtained doubtful fissuring of tongue region lingual surface image, whether can not determine on given lingual surface image necessary being fissuring of tongue, also need doubtful fissuring of tongue region to do further recognition and verification operation, therefore, the method fissuring of tongue identification subjectivity is strong, repeatable poor, and cannot be applied to real clinical Evolution of Tongue Inspection of TCM.
Summary of the invention
The object of the invention is to the deficiency existing for prior art, a kind of fissured tongue recognition methods based on support vector machine is proposed, the method adopts support vector machine as sorter, the wide thread detector algorithm that to utilize parameter be r is processed lingual surface image, take parameter r with the statistic such as span, the area span in doubtful fissuring of tongue region compare as proper vector, utilize computer simulation Traditional Chinese Medicine experts automatically to identify crackle tonguing row, and then overcome strong, the repeatable poor shortcoming of traditional fissured tongue identification subjectivity.
In order to achieve the above object, the present invention adopts following technical proposals.
Fissured tongue recognition methods based on support vector machine of the present invention, the method comprises the steps:
(1), set up schistoglossia line features training sample set;
(2), build the fissured tongue sorter based on support vector machine;
(3), utilize fissured tongue sorter based on support vector machine to identify, identify each sample to be tested that fissuring of tongue feature samples to be measured concentrates and whether belong to crackle tongue picture.
Above-mentioned steps (1) is described sets up fissuring of tongue features training sample set, and its concrete steps are as follows:
(11), with Snake partitioning algorithm, the photo that lolls of face being carried out to lingual surface cuts apart, obtain some width lingual surface images, identifying all lingual surface images is divided into lingual surface image: crackle tongue picture class lingual surface image and non-crackle tongue picture class lingual surface image, adopt respectively " Sgn " mark 1 or-1 mark lingual surface image, and with the form of RGB image, deposit respectively that fissured tongue resembles Sample Storehouse, non-fissured tongue resembles Sample Storehouse in, the training sample set that crackle tongue picture class lingual surface image construction mark " Sgn " is 1, the training sample set that non-crackle tongue picture class lingual surface image construction mark " Sgn " is-1;
(12), adopt parameter to be
Figure 2014100007072100002DEST_PATH_IMAGE002
wide thread detector algorithm to training sample, concentrate the B in i width lingual surface image rgb space to divide spirogram to process, obtain the bianry image in the doubtful region of fissuring of tongue of i width lingual surface image, each bright connected region in the bianry image in the doubtful region of fissuring of tongue of i width lingual surface image is doubtful fissuring of tongue region, is designated as
Figure 2014100007072100002DEST_PATH_IMAGE004
;
(13) calculate, respectively the span value in the doubtful fissuring of tongue region in i width bianry image, area span value ratio, the standard deviation of the standard deviation of span value, the ratio of area span value, and then obtain crack vector, its concrete steps are as follows:
(131), calculate respectively the span value in the each doubtful fissuring of tongue region in i width bianry image , this span value is the maximal value in any two points distance on doubtful fissuring of tongue regional graphics profile, by span value sequence
Figure 188908DEST_PATH_IMAGE006
descending sort, therefrom extracts first span value, second largest span value, the third-largest span value, is designated as respectively:
Figure 2014100007072100002DEST_PATH_IMAGE008
, ,
Figure 2014100007072100002DEST_PATH_IMAGE012
;
(132), calculate respectively first span value, second largest span value, the third-largest span value
Figure 589934DEST_PATH_IMAGE008
,
Figure 934327DEST_PATH_IMAGE010
,
Figure 166726DEST_PATH_IMAGE012
the corresponding area of corresponding bright connected region
Figure 2014100007072100002DEST_PATH_IMAGE014
,
Figure 2014100007072100002DEST_PATH_IMAGE016
,
Figure 2014100007072100002DEST_PATH_IMAGE018
, calculate according to the following formula the ratio of the span value bright connected region area corresponding to it in i width bianry image, be denoted as ;
Figure 2014100007072100002DEST_PATH_IMAGE022
，
Figure 2014100007072100002DEST_PATH_IMAGE024
(133), the standard deviation of the doubtful fissuring of tongue region span value in i width bianry image is set, be denoted as , computing method are as follows;
Figure 2014100007072100002DEST_PATH_IMAGE028
(134), the standard deviation of the ratio of the doubtful fissuring of tongue region area span value in i width bianry image is set, be denoted as , computing method are as follows;
(135), by described in above-mentioned steps (12), step (131), step (132), step (133), step (134)
Figure 2014100007072100002DEST_PATH_IMAGE034
, , ,
Figure 2014100007072100002DEST_PATH_IMAGE040
,
Figure 2014100007072100002DEST_PATH_IMAGE042
,
Figure 2014100007072100002DEST_PATH_IMAGE044
, ,
Figure 2014100007072100002DEST_PATH_IMAGE048
,
Figure DEST_PATH_IMAGE050
and the crack vector of a training sample of Sgn composition;
(14), the fissured tongue proper vector of all training samples has formed fissuring of tongue features training sample set.
The fissured tongue sorter of structure described in above-mentioned steps (2) based on support vector machine, its concrete steps are as follows:
(21) the fissuring of tongue features training sample set input support vector machine, step (1) being obtained;
(22), fissuring of tongue features training sample set is trained, obtain training pattern, by this training pattern, be built into the fissured tongue sorter based on support vector machine;
The fissured tongue sorter of utilization described in above-mentioned steps (3) based on support vector machine identified, and identifies each concentrated sample to be tested of fissuring of tongue feature samples to be measured and whether belongs to crackle tongue picture, and its concrete steps are as follows:
(31), utilize Snake partitioning algorithm that the face photo that lolls is carried out lingual surface and cut apart, obtain some lingual surface images to be identified, by some lingual surface image construction to be identified lingual surface image pattern to be identified collection;
(32), adopt parameter to be
Figure DEST_PATH_IMAGE052
wide thread detector algorithm to lingual surface image pattern to be identified, concentrate the B in i width lingual surface image rgb space to divide spirogram to process, obtain the bianry image in the doubtful region of fissuring of tongue in i width lingual surface image to be identified, each bright connected region in the bianry image in the doubtful region of fissuring of tongue of this i width lingual surface image to be identified is doubtful fissuring of tongue region, is designated as
Figure DEST_PATH_IMAGE054
;
(33) calculate, respectively the standard deviation of the ratio of standard deviation, the area span value of the span value in the doubtful fissuring of tongue region in above-mentioned i lingual surface bianry image to be identified to be identified, obtain crack vector, its concrete steps are as follows:
(331), calculate the span value of the each bright connected region in i width bianry image to be identified
Figure DEST_PATH_IMAGE056
, this span value to be identified is the maximal value in any two points distance on bright connected region graph outline, by span value sequence
Figure 623858DEST_PATH_IMAGE056
descending sort is also therefrom extracted first span value to be identified, second largest span value to be identified, the third-largest span value to be identified, is designated as respectively:
Figure DEST_PATH_IMAGE058
、
Figure DEST_PATH_IMAGE060
、
Figure DEST_PATH_IMAGE062
；
(332), calculate respectively first span value to be identified, second largest span value to be identified, the third-largest span value to be identified
Figure 512180DEST_PATH_IMAGE058
,
Figure 394685DEST_PATH_IMAGE060
,
Figure 747169DEST_PATH_IMAGE062
the corresponding area of corresponding bright connected region
Figure DEST_PATH_IMAGE064
, calculate according to the following formula the ratio of the each span value bright connected region area corresponding to it in i width bianry image to be identified, be denoted as
Figure DEST_PATH_IMAGE066
Figure DEST_PATH_IMAGE068
;
Figure DEST_PATH_IMAGE070
(333), the standard deviation of the bright connected region span value in lingual surface i width bianry image to be identified is set, be denoted as
Figure DEST_PATH_IMAGE072
;
Figure DEST_PATH_IMAGE074
(334), the standard deviation of the bright connected region area span ratio in lingual surface i width bianry image to be identified is set
Figure DEST_PATH_IMAGE076
;
Figure DEST_PATH_IMAGE078
(335), by described in above-mentioned steps (32), step (331), step (332), step (333), step (334)
Figure DEST_PATH_IMAGE080
, ,
Figure DEST_PATH_IMAGE084
,
Figure DEST_PATH_IMAGE086
,
Figure DEST_PATH_IMAGE088
, ,
Figure DEST_PATH_IMAGE092
,
Figure DEST_PATH_IMAGE094
,
Figure DEST_PATH_IMAGE096
,
Figure DEST_PATH_IMAGE098
the crack vector of a training sample to be identified of composition;
(34), the proper vector of the crackle of all samples to be identified forms crack sample set to be identified;
(35), utilize the fissured tongue sorter based on support vector machine described in step (2) to identify, identify the described concentrated sample to be tested of fissuring of tongue feature samples to be identified of step (34) and whether belong to crackle tongue picture.
The advantage having compared with the method for the fissured tongue recognition methods based on support vector machine of the present invention and Traditional Chinese Medicine visual observations crackle tongue picture is: the method adopts support vector machine as sorter, the wide thread detector algorithm that to utilize parameter be r is processed lingual surface image, fissuring of tongue proper vector take statistics such as the span value in parameter r and doubtful fissuring of tongue region, area span ratios as training sample, utilize computer simulation Traditional Chinese Medicine experts to identify crackle tongue picture, there is repeatability; The method is detecting on the basis in doubtful fissuring of tongue region, can provide further identification, it is the continuation of active computer fissuring of tongue detection method, the method is utilized disposable sample set training, can complete identification work, and in crackle tongue picture identifying without any need for other parameter arrange and artifact intervention, realized full-automatic identification fissured tongue; The method simulation traditional Chinese medical science, to the identification of crackle tongue picture, overcomes the defect that subjectivity is strong, nothing is sought unity of standard of traditional Chinese medical science visual observations crackle tongue picture.
Accompanying drawing explanation
Fig. 1 is the main-process stream block diagram of the fissured tongue recognition methods based on support vector machine of the present invention;
Fig. 2 is the described process flow diagram of asking crack vector of step in Fig. 1 (1);
Fig. 3 is the described process flow diagram of asking crack vector of step in Fig. 1 (3);
Fig. 4 adopts the fissured tongue recognition methods based on support vector machine of the present invention to identify the lingual surface image of a width fissured tongue;
Fig. 5 is the classifying quality table 1 of 100 lingual surface images to be identified in the embodiment of the present invention.
Embodiment
Below in conjunction with accompanying drawing 1, embodiments of the invention are described in further detail.
The present embodiment is implemented take technical scheme of the present invention as prerequisite:
As shown in Figure 1, the fissured tongue recognition methods based on support vector machine of the present invention, its concrete steps are as follows:
(1), set up fissuring of tongue features training sample set, as shown in Figure 2, its concrete steps are as follows:
(11), adopting Snake partitioning algorithm to carry out lingual surface to the photo that lolls of 441 width faces cuts apart, obtain 441 width lingual surface images, identify all lingual surface images and whether have crackle, all lingual surface images of having identified are divided into: crackle tongue picture class lingual surface image, with non-crackle tongue picture class lingual surface image, adopt respectively " Sgn " mark 1 or-1 mark lingual surface image, and with the form of RGB image, deposit respectively fissured tongue in and resemble Sample Storehouse, non-fissured tongue resembles Sample Storehouse, the training sample set that crackle tongue picture class lingual surface image construction mark " Sgn " is 1, the training sample set that non-crackle tongue picture class lingual surface image construction mark " Sgn " is-1,
(12), adopt parameter
Figure DEST_PATH_IMAGE100
be that 12 wide thread detector algorithm concentrates the B in i width lingual surface image rgb space to divide spirogram to process to training sample, obtain the bianry image in the doubtful region of fissuring of tongue of i width lingual surface image, each bright connected region in the bianry image in the doubtful region of fissuring of tongue of this i width lingual surface image is doubtful fissuring of tongue region, is designated as
Figure 443380DEST_PATH_IMAGE004
;
(13) calculate, respectively the standard deviation of the ratio of standard deviation, the area span value of the span value in the doubtful fissuring of tongue region in i width bianry image, and then obtain crack vector, its concrete steps are as follows:
(131), calculate respectively the span value in the each doubtful fissuring of tongue region in i width bianry image
Figure 818997DEST_PATH_IMAGE006
, this span value is the maximal value in any two points distance on doubtful fissuring of tongue regional graphics profile, by span value sequence
Figure 505193DEST_PATH_IMAGE006
descending sort, therefrom extracts first span value, second largest span value, the third-largest span value, is designated as respectively:
Figure 712184DEST_PATH_IMAGE008
,
Figure 825633DEST_PATH_IMAGE010
,
Figure 750864DEST_PATH_IMAGE012
;
(132), calculate respectively first span value, second largest span value, the third-largest span value
Figure 912855DEST_PATH_IMAGE008
,
Figure 239931DEST_PATH_IMAGE010
,
Figure 258703DEST_PATH_IMAGE012
the corresponding area of corresponding bright connected region
Figure 671230DEST_PATH_IMAGE014
, ,
Figure 818494DEST_PATH_IMAGE018
, calculate according to the following formula the ratio of the span value bright connected region area corresponding to it in i width bianry image, be denoted as ;
Figure 907990DEST_PATH_IMAGE022
，
Figure 474101DEST_PATH_IMAGE024
(134), the standard deviation of the doubtful fissuring of tongue region span value in i width bianry image is set, be denoted as
Figure 775769DEST_PATH_IMAGE026
, computing method are as follows;
Figure 339605DEST_PATH_IMAGE028
(135), the standard deviation of the ratio of the doubtful fissuring of tongue region area span value in i width bianry image is set, be denoted as
Figure 461145DEST_PATH_IMAGE030
, computing method are as follows;
Figure 830947DEST_PATH_IMAGE032
(135), by described in above-mentioned steps (12), step (131), step (132), step (133), step (134)
Figure 987122DEST_PATH_IMAGE034
, ,
Figure 330695DEST_PATH_IMAGE038
,
Figure 238608DEST_PATH_IMAGE040
,
Figure 249290DEST_PATH_IMAGE042
,
Figure 217246DEST_PATH_IMAGE044
,
Figure 251061DEST_PATH_IMAGE046
,
Figure 962665DEST_PATH_IMAGE048
, ,
Figure DEST_PATH_IMAGE102
the crack vector of a training sample of composition;
(14), the fissured tongue proper vector of all training samples has formed fissuring of tongue features training sample set;
(2), build fissured tongue sorter based on support vector machine, its concrete steps are as follows:
(21) the fissuring of tongue features training sample set input support vector machine, step (1) being obtained;
(22), fissuring of tongue features training sample set is trained, obtain training pattern, by this training pattern, be built into the fissured tongue sorter based on support vector machine;
(3), utilize fissured tongue sorter based on support vector machine to identify, identify each sample to be tested that fissuring of tongue feature samples to be measured concentrates and whether belong to crackle tongue picture, as shown in Figure 4, its concrete steps are as follows:
(31), utilize Snake partitioning algorithm that the 100 width faces photo that lolls is carried out lingual surface and cut apart, obtain 100 lingual surface images to be identified, by 100 lingual surface image construction to be identified lingual surface image pattern to be identified collection;
(32), adopt parameter to be
Figure 966710DEST_PATH_IMAGE100
wide thread detector algorithm to lingual surface image pattern to be identified, concentrate the B in i width lingual surface image rgb space to divide spirogram to process, obtain the bianry image in the doubtful region of fissuring of tongue of i width to be identified, the each bright connected region in the bianry image in this doubtful region of fissuring of tongue to be identified;
(33) calculate, respectively the standard deviation of the ratio of standard deviation, the area span value of the span value in the each doubtful fissuring of tongue region in above-mentioned i lingual surface bianry image to be identified to be identified, obtain crack vector, as shown in Figure 3, its concrete steps are as follows:
(331), calculate the span value of the each bright connected region in i width bianry image to be identified
Figure 484891DEST_PATH_IMAGE056
this span value to be identified is the maximal value in any two points distance on bright connected region graph outline, by span value sequence
Figure 186DEST_PATH_IMAGE056
descending sort is also therefrom extracted first span value to be identified, second largest span value to be identified, the third-largest span value to be identified, is designated as respectively:
Figure 719881DEST_PATH_IMAGE058
、
Figure 29639DEST_PATH_IMAGE060
、
Figure 100364DEST_PATH_IMAGE062
；
(332), calculate respectively first span value to be identified, second largest span value to be identified, the third-largest span value to be identified ,
Figure 931233DEST_PATH_IMAGE060
,
Figure 411893DEST_PATH_IMAGE062
the corresponding area of corresponding bright connected region
Figure DEST_PATH_IMAGE104
,
Figure DEST_PATH_IMAGE106
,
Figure DEST_PATH_IMAGE108
, calculate according to the following formula the ratio of the each span value bright connected region area corresponding to it in i width bianry image to be identified, be denoted as
Figure 907597DEST_PATH_IMAGE068
;
Figure DEST_PATH_IMAGE110
Figure DEST_PATH_IMAGE112
(333), the standard deviation of the bright connected region span value in lingual surface i width bianry image to be identified is set, be denoted as
Figure 436798DEST_PATH_IMAGE072
;
Figure 131085DEST_PATH_IMAGE074
(334), the standard deviation of the bright connected region area span ratio in lingual surface i width bianry image to be identified is set
Figure 48225DEST_PATH_IMAGE076
;
Figure 827962DEST_PATH_IMAGE078
(335), by described in above-mentioned steps (32), step (331), step (332), step (333), step (334)
Figure 223171DEST_PATH_IMAGE080
, ,
Figure 797689DEST_PATH_IMAGE084
,
Figure 64723DEST_PATH_IMAGE086
,
Figure 263623DEST_PATH_IMAGE088
, , ,
Figure 617878DEST_PATH_IMAGE094
,
Figure 354890DEST_PATH_IMAGE096
,
Figure 878275DEST_PATH_IMAGE098
the crack vector of a training sample to be identified of composition;
(34), the proper vector of the crackle of all samples to be identified forms crack sample set to be identified;
(35), utilize the fissured tongue sorter based on support vector machine described in step (2) to identify, identify the described concentrated sample to be tested of fissuring of tongue feature samples to be identified of step (34) and whether belong to crackle tongue picture.
For the recognition effect of verifying the fissured tongue recognition methods based on support vector machine of the present invention has carried out following experiment, in experiment, adopting wide thread detector parameter r is 12, training sample source is the photo that lolls of 441 width faces, wherein, crackle tongue picture picture is 220 width, and non-crackle tongue picture picture is 221 width.After passing through the above-mentioned fissured tongue sorter training pattern based on support vector machine of training sample training formation, clinically 200 lingual surface images to be identified have been randomly drawed, wherein 100 width are crackle tongue picture, 100 width are non-crackle tongue picture, after artificial mark, utilize training pattern of the present invention to classify, its classification results, as shown in Figure 5, in table 1, there is flawless to represent that classification results is crackle tongue picture and non-crackle tongue picture; Accuracy rate is the number of sample and the ratio of such total sample number that certain class is correctly classified; Total accuracy rate is number of samples that in two types of samples, all quilts are correctly classified and the ratio of two types of total sample number; Total accuracy rate of final fissured tongue classification is 91.5%, and effect is stood head and shoulders above others.As can be seen from Table 1, in crackle tongue picture identifying, realized full-automatic identification fissured tongue, the method simulation traditional Chinese medical science, to the identification of crackle tongue picture, overcomes the strong shortcoming of subjectivity of traditional Chinese medical science visual observations crackle tongue picture.

Claims (4)

1. the fissured tongue recognition methods based on support vector machine, is characterized in that, the method comprises the steps:
(1), set up schistoglossia line features training sample set;
(2), build the fissured tongue sorter based on support vector machine;
(3), utilize fissured tongue sorter based on support vector machine to identify, identify each sample to be tested that fissuring of tongue feature samples to be measured concentrates and whether belong to crackle tongue picture.
2. the fissured tongue recognition methods based on support vector machine according to claim 1, is characterized in that, above-mentioned steps (1) is described sets up fissuring of tongue features training sample set, and its concrete steps are as follows:
(11), with Snake partitioning algorithm, the photo that lolls of face being carried out to lingual surface cuts apart, obtain some width lingual surface images, identifying all lingual surface images is divided into lingual surface image: crackle tongue picture class lingual surface image and non-crackle tongue picture class lingual surface image, adopt respectively " Sgn " mark 1 or-1 mark lingual surface image, and with the form of RGB image, deposit respectively that fissured tongue resembles Sample Storehouse, non-fissured tongue resembles Sample Storehouse in, the training sample set that crackle tongue picture class lingual surface image construction mark " Sgn " is 1, the training sample set that non-crackle tongue picture class lingual surface image construction mark " Sgn " is-1;
(12), adopt parameter to be
Figure 5154DEST_PATH_IMAGE001
wide thread detector algorithm to training sample, concentrate the B in i width lingual surface image rgb space to divide spirogram to process, obtain the bianry image in the doubtful region of fissuring of tongue of i width lingual surface image, each bright connected region in the bianry image in the doubtful region of fissuring of tongue of i width lingual surface image is doubtful fissuring of tongue region, is designated as
Figure 2014100007072100001DEST_PATH_IMAGE002
;
(13) calculate, respectively the span value in the doubtful fissuring of tongue region in i width bianry image, area span value ratio, the standard deviation of the standard deviation of span value, the ratio of area span value, and then obtain crack vector, its concrete steps are as follows:
(131), calculate respectively the span value in the each doubtful fissuring of tongue region in i width bianry image
Figure 735344DEST_PATH_IMAGE003
, this span value is the maximal value in any two points distance on doubtful fissuring of tongue regional graphics profile, by span value sequence
Figure 515081DEST_PATH_IMAGE003
descending sort, therefrom extracts first span value, second largest span value, the third-largest span value, is designated as respectively:
Figure 2014100007072100001DEST_PATH_IMAGE004
,
Figure 972607DEST_PATH_IMAGE005
,
Figure DEST_PATH_IMAGE006
;
(132), calculate respectively first span value, second largest span value, the third-largest span value
Figure 786979DEST_PATH_IMAGE004
, ,
Figure 692454DEST_PATH_IMAGE006
the corresponding area of corresponding bright connected region
Figure 891354DEST_PATH_IMAGE007
,
Figure DEST_PATH_IMAGE008
,
Figure 622550DEST_PATH_IMAGE009
, calculate according to the following formula the ratio of the span value bright connected region area corresponding to it in i width bianry image, be denoted as
Figure 2014100007072100001DEST_PATH_IMAGE010
;
Figure 615914DEST_PATH_IMAGE011
，
Figure 2014100007072100001DEST_PATH_IMAGE012
(133), the standard deviation of the doubtful fissuring of tongue region span value in i width bianry image is set, be denoted as , computing method are as follows;
Figure 2014100007072100001DEST_PATH_IMAGE014
(134), the standard deviation of the ratio of the doubtful fissuring of tongue region area span value in i width bianry image is set, be denoted as
Figure 982621DEST_PATH_IMAGE015
, computing method are as follows;
Figure 2014100007072100001DEST_PATH_IMAGE016
(135), by described in above-mentioned steps (12), step (131), step (132), step (133), step (134)
Figure 506006DEST_PATH_IMAGE017
,
Figure DEST_PATH_IMAGE018
,
Figure 748900DEST_PATH_IMAGE019
,
Figure DEST_PATH_IMAGE020
, ,
Figure DEST_PATH_IMAGE022
,
Figure 327966DEST_PATH_IMAGE023
,
Figure DEST_PATH_IMAGE024
,
Figure 784486DEST_PATH_IMAGE025
, a training sample of Sgn composition crack vector;
(14), the fissured tongue proper vector of all training samples has formed fissuring of tongue features training sample set.
3. the fissured tongue recognition methods based on support vector machine according to claim 2, is characterized in that, the fissured tongue sorter of the described structure of above-mentioned steps (2) based on support vector machine, and its concrete steps are as follows:
(21) the fissuring of tongue features training sample set input support vector machine, step (1) being obtained;
(22), fissuring of tongue features training sample set is trained, obtain training pattern, by this training pattern, be built into the fissured tongue sorter based on support vector machine.
4. the fissured tongue recognition methods based on support vector machine according to claim 3, it is characterized in that, the fissured tongue sorter of utilization described in above-mentioned steps (3) based on support vector machine identified, identify each concentrated sample to be tested of fissuring of tongue feature samples to be measured and whether belong to crackle tongue picture, its concrete steps are as follows:
(31), utilize Snake partitioning algorithm that the face photo that lolls is carried out lingual surface and cut apart, obtain some lingual surface images to be identified, by some lingual surface image construction to be identified lingual surface image pattern to be identified collection;
(32), adopt parameter to be
Figure DEST_PATH_IMAGE026
wide thread detector algorithm to lingual surface image pattern to be identified, concentrate the B in i width lingual surface image rgb space to divide spirogram to process, obtain the bianry image in the doubtful region of fissuring of tongue in i width lingual surface image to be identified, each bright connected region in the bianry image in the doubtful region of fissuring of tongue of this i width lingual surface image to be identified is doubtful fissuring of tongue region, is designated as
Figure 181969DEST_PATH_IMAGE027
;
(33) calculate, respectively the standard deviation of the ratio of standard deviation, the area span value of the span value in the doubtful fissuring of tongue region in above-mentioned i lingual surface bianry image to be identified to be identified, obtain crack vector, its concrete steps are as follows:
(331), calculate the span value of the each bright connected region in i width bianry image to be identified
Figure DEST_PATH_IMAGE028
, this span value to be identified is the maximal value in any two points distance on bright connected region graph outline, by span value sequence
Figure 910891DEST_PATH_IMAGE028
descending sort is also therefrom extracted first span value to be identified, second largest span value to be identified, the third-largest span value to be identified, is designated as respectively: ;
(332), calculate respectively first span value to be identified, second largest span value to be identified, the third-largest span value to be identified
Figure 297802DEST_PATH_IMAGE029
the corresponding area of corresponding bright connected region
Figure DEST_PATH_IMAGE030
,
Figure 131766DEST_PATH_IMAGE031
,
Figure DEST_PATH_IMAGE032
, calculate according to the following formula the ratio of the each span value bright connected region area corresponding to it in i width bianry image to be identified, be denoted as
Figure 82405DEST_PATH_IMAGE033
Figure DEST_PATH_IMAGE034
;
Figure 777959DEST_PATH_IMAGE035
(333), the standard deviation of the bright connected region span value in lingual surface i width bianry image to be identified is set, be denoted as
Figure DEST_PATH_IMAGE036
;
Figure 130443DEST_PATH_IMAGE037
(334), the standard deviation of the bright connected region area span ratio in lingual surface i width bianry image to be identified is set
Figure DEST_PATH_IMAGE038
;
Figure 135308DEST_PATH_IMAGE039
(335), by described in above-mentioned steps (32), step (331), step (332), step (333), step (334)
Figure DEST_PATH_IMAGE040
,
Figure 386292DEST_PATH_IMAGE041
,
Figure DEST_PATH_IMAGE042
,
Figure 806909DEST_PATH_IMAGE043
,
Figure DEST_PATH_IMAGE044
,
Figure 341796DEST_PATH_IMAGE045
, ,
Figure 268295DEST_PATH_IMAGE047
,
Figure DEST_PATH_IMAGE048
, and the crack vector of a training sample to be identified of composition;
(34), the proper vector of the crackle of all samples to be identified forms crack sample set to be identified;
(35), utilize the fissured tongue sorter based on support vector machine described in step (2) to identify, identify the described concentrated sample to be tested of fissuring of tongue feature samples to be identified of step (34) and whether belong to crackle tongue picture.
CN201410000707.2A 2014-01-02 2014-01-02 Method for recognizing fissured tongue based on support vector machine Pending CN103761506A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410000707.2A CN103761506A (en) 2014-01-02 2014-01-02 Method for recognizing fissured tongue based on support vector machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410000707.2A CN103761506A (en) 2014-01-02 2014-01-02 Method for recognizing fissured tongue based on support vector machine

Publications (1)

Publication Number Publication Date
CN103761506A true CN103761506A (en) 2014-04-30

Family

ID=50528742

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410000707.2A Pending CN103761506A (en) 2014-01-02 2014-01-02 Method for recognizing fissured tongue based on support vector machine

Country Status (1)

Country Link
CN (1) CN103761506A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104537379A (en) * 2014-12-26 2015-04-22 上海大学 High-precision automatic tongue partition method
CN105787486A (en) * 2015-12-11 2016-07-20 昆明理工大学 Method for detecting girder cracks based on image processing
CN106683088A (en) * 2016-12-28 2017-05-17 西北工业大学 Road crack type identification method based on mobile phone sensor

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102426583A (en) * 2011-10-10 2012-04-25 北京工业大学 Chinese medicine tongue manifestation retrieval method based on image content analysis

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102426583A (en) * 2011-10-10 2012-04-25 北京工业大学 Chinese medicine tongue manifestation retrieval method based on image content analysis

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
LAURA LI LIU ET AL.: "Tongue line extraction", 《19TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION,2008》 *
傅之成等: "基于径向边缘检测和Snake模型的舌像分割", 《中国图象图形学报》 *
刘哲等: "基于视觉注意和支持向量机的舌体自动分割方法的探讨", 《北京中医药大学学报》 *
杨朝晖: "计算机舌诊中裂纹舌图像的诊断分类研究", 《万方数据企业知识服务平台》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104537379A (en) * 2014-12-26 2015-04-22 上海大学 High-precision automatic tongue partition method
CN105787486A (en) * 2015-12-11 2016-07-20 昆明理工大学 Method for detecting girder cracks based on image processing
CN105787486B (en) * 2015-12-11 2019-04-09 昆明理工大学 A kind of girder steel crack detecting method based on image procossing
CN106683088A (en) * 2016-12-28 2017-05-17 西北工业大学 Road crack type identification method based on mobile phone sensor
CN106683088B (en) * 2016-12-28 2019-06-21 西北工业大学 Road crack-type recognition methods based on mobile phone sensor

Similar Documents

Publication Publication Date Title
CN105224921B (en) A kind of facial image preferentially system and processing method
CN107368670A (en) Stomach cancer pathology diagnostic support system and method based on big data deep learning
CN109858540B (en) Medical image recognition system and method based on multi-mode fusion
CN105160346B (en) A kind of greasy recognition methods of curdy fur on tongue based on texture and distribution characteristics
CN108717554A (en) A kind of thyroid tumors histopathologic slide image classification method and its device
CN105893980A (en) Method and system for evaluating concentration degree of attention
CN101706445B (en) Image processing mthod for beef marbling grain grade scoring
CN107368671A (en) System and method are supported in benign gastritis pathological diagnosis based on big data deep learning
CN107369151A (en) System and method are supported in GISTs pathological diagnosis based on big data deep learning
CN106408566B (en) A kind of fetal ultrasound image quality control method and system
Hatanaka et al. Vertical cup-to-disc ratio measurement for diagnosis of glaucoma on fundus images
CN106529571A (en) Multilayer image characteristics extraction modeling, image recognition method and apparatus based on neural network
CN110378232B (en) Improved test room examinee position rapid detection method of SSD dual-network
CN104198325B (en) Stem ratio measuring method in pipe tobacco based on computer vision
CN107564000B (en) Hericium erinaceus non-destructive testing stage division based on computer vision
CN106780465A (en) Retinal images aneurysms automatic detection and recognition methods based on gradient vector analysis
CN110189383A (en) Chinese medicine tongue color coating colour quantitative analysis method based on machine learning
Maji et al. An automated method for counting and characterizing red blood cells using mathematical morphology
CN104198497A (en) Surface defect detection method based on visual saliency map and support vector machine
Aiadi et al. A new method for automatic date fruit classification
Virk et al. Cup-to-disk ratio (CDR) determination for glaucoma screening
CN106548131A (en) A kind of workmen's safety helmet real-time detection method based on pedestrian detection
CN106599880A (en) Discrimination method of the same person facing examination without monitor
CN103761506A (en) Method for recognizing fissured tongue based on support vector machine
CN111816314B (en) Chest card selection, labeling and verification method for artificial intelligent screening of pneumoconiosis

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20140430

WD01 Invention patent application deemed withdrawn after publication