CN101604395A - A kind of morphologic method for classifying invariant moment of similar objects of multi-scale mathematical that utilizes - Google Patents
A kind of morphologic method for classifying invariant moment of similar objects of multi-scale mathematical that utilizes Download PDFInfo
- Publication number
- CN101604395A CN101604395A CNA2009100893373A CN200910089337A CN101604395A CN 101604395 A CN101604395 A CN 101604395A CN A2009100893373 A CNA2009100893373 A CN A2009100893373A CN 200910089337 A CN200910089337 A CN 200910089337A CN 101604395 A CN101604395 A CN 101604395A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- invariant
- msubsup
- scale
- 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
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000000877 morphologic effect Effects 0.000 title claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 12
- 238000003909 pattern recognition Methods 0.000 abstract description 3
- 238000000605 extraction Methods 0.000 abstract description 2
- 238000007630 basic procedure Methods 0.000 abstract 1
- 230000015572 biosynthetic process Effects 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 19
- 238000004364 calculation method Methods 0.000 description 3
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 240000005373 Panax quinquefolius Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001351 cycling effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
A kind of morphologic method for classifying invariant moment of similar objects of multi-scale mathematical that utilizes, its basic procedure is a minutia of at first utilizing multi-scale mathematical morphology extraction object training image, and the multiple dimensioned minutia of object is carried out the real different parts of the various merging formation similar objects of inhomogeneity on a plurality of yardsticks.Then, the Invariant Moment Method of choosing any one kind of them is calculated the value of the invariant moments of these real different parts.Subsequently, by a kind of estimate the value of the invariant moments of real different parts is evaluated and tested selection real different parts and the corresponding invariant moments thereof that can active zone divide similar object automatically.At last, the invariant moments of the real different parts of utilization selection and correspondence thereof is to similar sorting objects.Realize effective classification of similar object.The present invention can be widely used in all kinds of Target Recognition and categorizing system, has vast market prospect and using value in Digital Image Processing and area of pattern recognition.
Description
(I) technical field
The invention relates to a method for classifying invariant moments of similar objects by utilizing multi-scale mathematical morphology, and belongs to the technical field of digital image processing and pattern recognition. The method mainly relates to mathematical morphology, invariant moment and target recognition technology, and can be widely applied to target recognition and target classification.
(II) background of the invention
The invariant moment is an important mathematical tool for target identification and classification, and the good invariant moment classification capability can effectively improve the effectiveness of target identification and classification. In order to achieve better recognition and classification effects, various invariant moments are proposed, including: hu moment, Z moment, wavelet moment, etc. While these invariant moment methods work well in distinguishing between different classes of objects, they work poorly when the shapes of the different classes of objects are very similar.
Hu first proposed seven invariant moments with translation, rotation, and scaling invariance in 1961, where the lower-order moments are mainly used to describe the overall characteristics of an object and the higher-order moments are mainly used to describe the detailed characteristics of an object. Subsequently, the Hu moment has been widely used in a variety of applications, but it also exhibits various disadvantages in various applications, such as the inability to efficiently identify different types of objects having very similar shapes. The Z moment is an effective orthogonal moment, has good orthogonality, and can be effectively applied to image reconstruction. The orthogonal characteristic of the Z moment enables the object recognition capability of the Z moment to be better than that of the Hu moment and other common orthogonal moments, and the invariant features of the image can be better described. However, since the Z moment is also a constant moment feature calculated from the overall shape of the object, the Z moment also cannot effectively distinguish different types of objects with very similar shapes. The wavelet transform can decompose the image on different scales, and can extract the details of the image on different scales. Sheng et al construct a series of wavelet invariant moments through Z moments with wavelet transform tools. The wavelet invariant moment can construct any-order invariant moment capable of reflecting the characteristics of different scales of the object, and the wavelet transformation can extract the characteristics of the object on different scales, so that the wavelet invariant moment can effectively distinguish similar objects. But the effect of wavelet invariant moment is inhibited to some extent when distinguishing different kinds of objects with similar appearances.
In fact, even if the two types of objects are very similar, the two types of objects may have large differences in some details. Therefore, if these detail differences can be efficiently extracted and used to calculate invariant moments, the classification capability of these invariant moment methods will be greatly improved. Therefore, the key for effectively distinguishing the objects with similar shapes is to find a method capable of effectively extracting real different detail features in the similar objects, and then utilize the features to calculate invariant moment, so that the classification capability of the invariant moment is improved, and the purpose of distinguishing the similar objects is achieved. Mathematical morphology is widely used in image processing, and can effectively smooth details of an image, that is, the mathematical morphology can extract the details in the image. More importantly, the multi-scale mathematical morphology can extract details of the image on different scales, which is more beneficial to the utilization of the image details.
Disclosure of the invention
1. The purpose is as follows: in order to effectively distinguish similar objects and make up for the defects of the traditional method, the invention provides a similar object invariant moment classification method by utilizing multi-scale mathematical morphology.
2. The technical scheme is as follows: in order to achieve the purpose, in the technical scheme of the invention, the process comprises the steps of firstly extracting the detail features of the object training image by utilizing the multi-scale mathematical morphology, and carrying out various combinations on the multi-scale detail features of the object to form different parts of similar objects of different types on multiple scales. Then, an optional invariant moment method calculates the invariant moment values for these truly distinct portions. Subsequently, the invariant moment values of the really different parts are evaluated through a measure, and the really different parts of the similar object and the corresponding invariant moments can be effectively distinguished through automatic selection. And finally, classifying the similar objects by using the selected really different parts and the corresponding invariant moments. And realizing effective classification of similar objects.
The invention relates to a method for classifying invariant moments of similar objects by utilizing multi-scale mathematical morphology, which comprises the following specific steps of:
n certain types of invariant moments are set, and morphological operations of a scales are required to be calculated;
the method comprises the following steps: taking two types of objects f to be distinguished1And f2The training image of (2);
step two: if i is 1, i represents the number of the invariant moment, and the range of i is more than or equal to 1 and less than or equal to n;
step three: setting q to be 0, wherein q represents the serial number of the scale needing to be calculated;
step four: let p be q +1, p represents the sequence number of another scale related to q that needs to be calculated;
step five: for object f1Calculating f1Merged multi-scale morphological detail features from scale p to scale q:
calculating f1 pqThe ith invariant moment of (c): calculating f1MM for all corresponding training imagesi pqAnd all MMs are combinedi pqForm a vector VMi pqFor another kind of object f2Another vector VM is formed in the same wayi pq；
Step six: the measure Q of the two vectors is calculated as followsi pq：
And Q isi pqPut into the measure vector Q. Wherein: σ and m are f1And f2The variance and mean of all invariant moment values corresponding to different training images;
step seven: if p is greater than a, entering step eight; otherwise, entering the step five;
step eight: q is q +1, if q is larger than a, entering the ninth step; otherwise, entering the step four;
step nine: if i is equal to i +1, if i is greater than n, entering step ten; otherwise, entering a step three;
step ten: finding out the first r minimum values in Q and the corresponding invariant moment MMi pq；
Step eleven: using r MMi pqAnd the corresponding truly different parts classify the two types of similar objects by a neighbor method.
The method utilizes multi-scale mathematical morphology to extract detail features of an object training image, and combines the multi-scale detail features of the object to form a plurality of truly different parts of different similar objects on a plurality of scales, so that the difference of the truly different parts is greater than that of the original object in overall shape; at the moment, the invariant moment values of the different parts can be calculated by any invariant moment method, and the invariant moment values are calculated on the different parts, so that the invariant moment values of different objects are obviously different; evaluating the invariant moment values of the real different parts through a measure, and automatically selecting the real different parts which can more effectively distinguish similar objects and the corresponding invariant moments; and classifying the similar objects by using the selected really different parts and the corresponding invariant moments, so as to realize effective classification of the similar objects.
3. The advantages and the effects are as follows: the method fully utilizes the possible different parts of the similar objects on the appearance details, can automatically select the real different parts of the similar objects, and replaces the value of the invariant moment obtained by the integral calculation of the similar objects by utilizing the value of the invariant moment obtained by the calculation of the local real different parts of the similar objects, thereby achieving the purpose of effectively classifying the similar objects, being widely applied to various target identification and classification systems and having wide market prospect and application value.
(IV) description of the drawings
FIG. 1 is a schematic block diagram of a method for classification of similar body invariant moment using multi-scale mathematical morphology according to the present invention.
Fig. 2 is exemplary images "1" and "I" showing the effect of the classification of similar objects according to the present invention.
FIG. 3 is a true difference between "1" and "I" of the multi-scale mathematical morphology extraction.
FIG. 4 is a comparison of the classification effect of the present invention and the classification effect of the Hu moments. Wherein (a) is the classification effect of two invariant moments selected by the present invention, and (b) is the classification effect of the Hu moment.
Fig. 5 is a comparison of the classification effect of the present invention and the classification effect of wavelet invariant moments. Wherein (a) is the classification effect of two invariant moments selected by the present invention, and (b) is the classification effect of wavelet invariant moments.
The symbols in the figures are as follows: m1 and M2 in FIG. 4(a) represent the first and second order moments of the Hu moment, respectively; MIM in FIG. 4(b)1 51And MIM2 51The two invariant moments with better performance are obtained on the basis of the first order moment and the second order moment of the Hu moment by utilizing the method, and the first order moment and the second order moment from the fifth scale to the first scale are respectively represented; II F in FIG. 5(a)2，6，0 waveletII and F2，4，0 waveletII is two small moments that can effectively distinguish similar objects; MIM in FIG. 5(b)3 41And MIM4 41Two better invariant moments are obtained on the basis of the first, second, third and fourth orders of the Hu moment by using the inventionThird and fourth moments on the fourth to first scale, respectively.
(V) detailed description of the preferred embodiments
The prerequisites for implementing the invention are: a training image of the object to be classified must be provided.
The conditions for classification by the present invention are: it is necessary to obtain an image of the object to be classified.
For better understanding of the technical solutions of the present invention, the following further describes embodiments of the present invention with reference to the accompanying drawings.
The invention relates to a similar body invariant moment classification method by utilizing multi-scale mathematical morphology, the working flow of which is shown in figure 1, and the detailed implementation steps of each part are as follows:
the method comprises the following steps: taking two types of objects f to be distinguished1And f2The training image of (2);
step two: and (i is more than or equal to 1 and less than or equal to n), and taking the ith invariant moment. The Hu moment has poor effect when similar objects are classified, and in order to show the effectiveness of the Hu moment, the Hu moment is selected as the invariant moment used in the method. Of course, Z moment, wavelet moment, etc. may be used. The general basic definition of the 7 Hu moments is as follows:
M1＝η20+η02，
M5＝(η30-3η12)(η30+η12)[(η30+η12)2-3(η21+η03)2]+(3η21-η03)(η21+η03)[3(η30+η12)2-(η21+η03)2]，
M6＝(η20-η02)[(η30+η12)2-(η21+η03)2]+4η11(η30+η12)(η21+η03)，
M7＝(3η21-η30)(η30+η12)[(η30+η12)2-3(η21+η03)2]+(3η12-η30)(η21+η03)[3(η03+η12)2-(η21+η03)2]。
wherein,
<math>
<mrow>
<msub>
<mi>η</mi>
<mi>pq</mi>
</msub>
<mo>=</mo>
<mfrac>
<msub>
<mi>μ</mi>
<mi>pq</mi>
</msub>
<msub>
<mi>μ</mi>
<mn>00</mn>
</msub>
</mfrac>
<mo>,</mo>
</mrow>
</math>
u and v are the length and width of the image.
Since the present invention can effectively extract truly different parts of similar objects by details of similar objects, n can be selected as a low-order invariant moment with a small calculation amount, such as the first two-order invariant moment of Hu moment (M1, M2, where n is 2) or the first four-order invariant moment (M1, M2, M3, M4, where n is 4).
Step three: let q be 0, start the loop for q.
Step four: let p be q +1, start the cycle for p.
Step five: for object f1F is calculated by1The multi-scale morphological feature of (1).
The multi-scale feature of the image on the p-th scale is
The detail features of the merged image from scale p to scale q are as follows:
wherein,
○，●，
and
mathematical morphological open, close, swell and corrode operator signs, respectively.
<math>
<mrow>
<mi>qB</mi>
<mo>=</mo>
<mi>B</mi>
<mo>⊕</mo>
<mi>B</mi>
<mo>⊕</mo>
<mo>.</mo>
<mo>.</mo>
<mo>.</mo>
<mo>⊕</mo>
<mi>B</mi>
<mo>,</mo>
</mrow>
</math>
And the pair B expands q times.
Computing merged multiscale features f1 pqThe ith invariant moment of (c): calculating the value MM of the ith invariant moment corresponding to all the training images of the objecti pqAnd all MMs are combinedi pqForm a vector VMi pq。
For another kind of object f2Form another vector VMi pq。
Step six: the measure Q of the two vectors is calculated as followsi pq：
Wherein, σ and m are the variance and mean of the invariant moment values corresponding to different training images; and Q isi pqPut into the measure vector Q. From Qi pqAs can be seen from the definition of (2), Qi pqThe smaller the torque MMi pqThe classification capability of the method is stronger, and the corresponding real different parts can reflect the difference between two different objects.
Step seven: cycling p, setting p to be p +1, and if p is larger than a, entering the step eight; otherwise, entering the step five. a is the scale degree of the multi-scale morphological operation, can be selected according to practical application, and can be generally selected to be a less than 10.
Step eight: circulating q, setting q to be q +1, and if q is larger than a, entering a ninth step; otherwise, entering the step four.
Step nine: circulating i, setting i to be i +1, and if i is larger than n, entering a step ten; otherwise, entering the step three.
Step ten: selecting invariant moment with good classification capability
From Qi pqAs defined in (1), Qi pqThe smaller the value is, the better the capability of the corresponding invariant moment to distinguish the current different types of objects is, so that only the invariant moment MM corresponding to the first r minimum values in Q needs to be foundi pqAnd its corresponding truly different part fpqAnd is used for distinguishing two similar objects. r may be chosen according to the specific application, and is usually 2 for both effectiveness and robustness.
Step eleven: classifying similar objects
First of all by using fpqCalculate the true different parts of the two similar object images and then use the corresponding r MMsi pqThe invariant moments of two classes of similar objects were calculated on truly different parts, and finally the most common neighbor method was used (see: frontier, zhang chei, et al, pattern recognition, qinghua university press, beijing, 1999,chapter six.) classify the invariant moments of two types of objects to achieve the purpose of distinguishing the two types of objects.
In order to demonstrate the effect of the present invention, "1" and "I" with similar shapes were used as experimental images, and a partial image is shown in fig. 2. As can be seen in fig. 2, "1" and "I" are very similar, with only minor differences at the top. The truly different parts of "1" and "I", one of which is shown in fig. 3, are extracted using multi-scale mathematical morphology. As can be seen from fig. 3, the extracted portion of fig. 3 is indeed much different from that of fig. 2. In order to better show the effect of the invention, the Hu moment with relatively poor classification capability is selected as the invariant moment MM required to be used in the invention, the invariant moment MM is marked as MIM, and the invariant moment MM is compared with the Hu moment and the wavelet invariant moment capable of effectively distinguishing similar objects. The two low-order invariant moments of the first and second orders with better stability are used in the method proposed by the present invention (n is 2), fig. 4 is the classification effect of the present invention and the Hu moment, although the first and second orders of the Hu moment cannot distinguish between "1" and "I" (fig. (a)), while the present invention using multi-scale mathematical morphology can effectively distinguish between "1" and "I" (fig. (b)). To better demonstrate the effect of the present invention, the present invention is compared with wavelet invariant moment, and fig. 5 is a classification effect. In fig. 5(a), although the wavelet invariant moment can distinguish between "1" and "I", it can just as well distinguish. However, the present invention (n-4) (fig. 5(b)) using 4 low-order Hu moments can distinguish "1" from "I" well. This shows that the present invention can effectively distinguish similar objects and has better effect than the traditional method. In addition, any invariant moment method can be used for the invariant moment MM required by the invention instead of the Hu moment, and the better the classification capability of the invariant moment MM is, the better the capability of the invention for distinguishing similar objects is. Therefore, the method can be used for classifying similar objects, can also be widely used for improving the invariant moment classification capability when distinguishing similar objects by invariant moment, and has wide application prospect in various applications.
Claims (1)
1. A method for classification of analogous volumetric invariant moments using multi-scale mathematical morphology, provided with n selected types of invariant moments and requiring a number a of scales of morphological operations to be calculated, characterized in that: the method comprises the following specific steps:
the method comprises the following steps: taking two types of objects f to be distinguished1And f2The training image of (2);
step two: if i is 1, i represents the number of the invariant moment, and the range of i is more than or equal to 1 and less than or equal to n;
step three: setting q to be 0, wherein q represents the serial number of the scale needing to be calculated;
step four: let p be q +1, p represents the sequence number of another scale related to q that needs to be calculated;
step five: for object f1Calculating f1Merged multi-scale morphological detail features from scale p to scale q:
calculating f1 pqThe ith invariant moment of (c): calculating f1MM for all corresponding training imagesi pqAnd all MMs are combinedi pqForm a vector VMi pqFor another kind of object f2Another vector VM is formed in the same wayi pq；
Step six: the measure Q of the two vectors is calculated as followsi pq：
And Q isi pqPut into a measure vector Q, where: σ and m are f1And f2The variance and mean of all invariant moment values corresponding to different training images;
step seven: if p is greater than a, entering step eight; otherwise, entering the step five;
step eight: q is q +1, if q is larger than a, entering the ninth step; otherwise, entering the step four;
step nine: if i is equal to i +1, if i is greater than n, entering step ten; otherwise, entering a step three;
step ten: finding out the first r minimum values in Q and the corresponding invariant moment MMi pq；
Step eleven: using r MMi pqAnd pairThe real different parts classify the two similar objects by a neighbor method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100893373A CN101604395B (en) | 2009-07-16 | 2009-07-16 | Method for classifying invariant moment of similar objects by multi-scale mathematical morphology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100893373A CN101604395B (en) | 2009-07-16 | 2009-07-16 | Method for classifying invariant moment of similar objects by multi-scale mathematical morphology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101604395A true CN101604395A (en) | 2009-12-16 |
| CN101604395B CN101604395B (en) | 2011-12-14 |
Family
ID=41470115
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009100893373A Expired - Fee Related CN101604395B (en) | 2009-07-16 | 2009-07-16 | Method for classifying invariant moment of similar objects by multi-scale mathematical morphology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101604395B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679664A (en) * | 2013-12-30 | 2014-03-26 | 北京航空航天大学 | Enhancing method capable of preserving image details by using alternative filter based on mathematical morphology |
-
2009
- 2009-07-16 CN CN2009100893373A patent/CN101604395B/en not_active Expired - Fee Related
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679664A (en) * | 2013-12-30 | 2014-03-26 | 北京航空航天大学 | Enhancing method capable of preserving image details by using alternative filter based on mathematical morphology |
| CN103679664B (en) * | 2013-12-30 | 2016-07-06 | 北京航空航天大学 | A kind of Enhancement Method that can retain image detail utilizing mathematical morphology alternative filter |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101604395B (en) | 2011-12-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107748873B (en) | A kind of multimodal method for tracking target merging background information | |
| CN110084249A (en) | The image significance detection method paid attention to based on pyramid feature | |
| CN112184752A (en) | Video target tracking method based on pyramid convolution | |
| CN103208097B (en) | Filtering method is worked in coordination with in the principal component analysis of the multi-direction morphosis grouping of image | |
| Smith et al. | Classification of archaeological ceramic fragments using texture and color descriptors | |
| CN105184772A (en) | Adaptive color image segmentation method based on super pixels | |
| CN101807294B (en) | Image enhancing method using multi-dimensional multi-structural element mathematical morphology | |
| Khmag et al. | Recognition system for leaf images based on its leaf contour and centroid | |
| CN100461217C (en) | Method for cutting complexity measure image grain | |
| CN104616308A (en) | Multiscale level set image segmenting method based on kernel fuzzy clustering | |
| CN104537673A (en) | Infrared image segmentation method based on multiple threshold values and self-adaptation fuzzy clustering | |
| CN102254172B (en) | Method for segmenting fingerprint image based on cellular neural network and morphology | |
| Li et al. | Designing compact Gabor filter banks for efficient texture feature extraction | |
| CN106845516A (en) | A kind of footprint image recognition methods represented based on multisample joint | |
| CN109858386A (en) | A kind of microalgae cell recognition methods based on fluorescence microscope images | |
| CN103839066A (en) | Feature extraction method from biological vision | |
| CN103246877B (en) | Based on the recognition of face novel method of image outline | |
| Singh et al. | Performance analysis of various local and global shape descriptors for image retrieval | |
| CN104268549A (en) | Anti-noise multi-scale local binary pattern characteristic representation method | |
| CN115147613A (en) | Infrared small target detection method based on multidirectional fusion | |
| Lu et al. | Extended biologically inspired model for object recognition based on oriented Gaussian–Hermite moment | |
| CN112070116B (en) | Automatic artistic drawing classification system and method based on support vector machine | |
| CN101604395B (en) | Method for classifying invariant moment of similar objects by multi-scale mathematical morphology | |
| CN111127407B (en) | Fourier transform-based style migration forged image detection device and method | |
| CN102270297B (en) | Fingerprint image enhancement method |
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: 20111214 Termination date: 20150716 |
|
| EXPY | Termination of patent right or utility model |