CN104123561B - Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model - Google Patents

Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model Download PDF

Info

Publication number
CN104123561B
CN104123561B CN201410325747.4A CN201410325747A CN104123561B CN 104123561 B CN104123561 B CN 104123561B CN 201410325747 A CN201410325747 A CN 201410325747A CN 104123561 B CN104123561 B CN 104123561B
Authority
CN
China
Prior art keywords
pixel
remote sensing
digital image
fuzzy
classification
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.)
Active
Application number
CN201410325747.4A
Other languages
Chinese (zh)
Other versions
CN104123561A (en
Inventor
张华�
郑南山
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.)
China University of Mining and Technology CUMT
Original Assignee
China University of Mining and Technology CUMT
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 China University of Mining and Technology CUMT filed Critical China University of Mining and Technology CUMT
Priority to CN201410325747.4A priority Critical patent/CN104123561B/en
Publication of CN104123561A publication Critical patent/CN104123561A/en
Application granted granted Critical
Publication of CN104123561B publication Critical patent/CN104123561B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Image Analysis (AREA)

Abstract

A kind of fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model, automatic suitable for remote sensing image, medical image and other images are split with being used during classification.Step is:Determine the pixel number of Remote Sensing Digital Image, and image is clustered using standard FCM models and is initialized, ask spatial attraction and space constraint penalty factor between other pixels in each pixel and its neighborhood window successively afterwards, finally obtain fuzzy factorBy by fuzzy factorAddition standard FCM models circulate so as to obtain new cluster object function and seek fuzzy matrixAnd cluster centreUntil cluster centre do not continue to change or computing reach maximum iteration, then utilize the fuzzy membership matrix U={ u finally tried to achieveki}c×N, the classification described in each pixel is determined to each pixel point progress category label of remote sensing image using maximization subordinated-degree matrix criterion, classification of remote-sensing images thematic map is formed, so as to fulfill the automatic classification of Remote Sensing Digital Image.Its method is simple, the degree of automation is high, is influenced small, classification of image segmentation accuracy height by picture noise.

Description

Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model
Technology neighborhood
The present invention relates to a kind of remote sensing image automatic classification method, be particularly suitable for remote sensing image, medical image and other The automatic segmentation of image and the fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model used in classification.
Background technology
Remotely-sensed data classification is that an important technology of thematic categorical data is extracted from remotely-sensed data, is every profession and trade information Provide abundant data source.Current sorting technique is broadly divided into supervised classification and non-supervised classification, with prison Sorting technique to be superintended and directed to compare, non-supervised classification can be not required priori to extract information from remotely-sensed data, therefore, Non-supervised classification has very important status in remotely-sensed data classification.
Existing non-supervised classification mainly has ISODATA, KNN, K-means, Markov random field (MRF) and the methods of Fuzzy C-Means Clustering Algorithm (FCM).Wherein, FCM is a kind of clustering algorithm based on prototype, has letter It is single, efficiently, the features such as data are adaptable, and compared with Hard clustering method, FCM can obtain each pixel and belong to each class Other degree of membership, can retain the information of image as much as possible, be more suitable for the remote sensing for representing to have a large amount of mixed pixels Image.Standard FCM algorithms are proposed by Dune, after promoted by Bezdek, be a kind of iteration optimal method.But standard The influence between adjacent picture elements is not accounted in FCM cluster process, very big mistake can be produced in the relatively low image of segmentation signal-to-noise ratio Difference.In order to overcome standard FCM algorithms to this sensitive problem of peak of noise, FCM simulated target of many researchers in standard The space constraint of image is added in function, it is proposed that many improved FCM clustering algorithms, such as FCM_S, FCM_S1, FCM_S2, BCFCM, GG-FCM, EnFCM, FGFCM etc..These innovatory algorithms improve the performance of standard FCM to a certain extent, but this The effect of methods is influenced very serious by parameters such as window size, scale factors a bit, and the selection of these parameters has very Big uncertainty, therefore the versatility of these algorithms is up for further verification.At present, there is a kind of fuzzy local message C means clustering algorithm FLICM, the local restriction of use at the same time can integrate local spatial information and pixel characteristic, And it is not required additional parameter to be calculated.However, when picture noise is larger, the segmentation accuracy of FLICM algorithms is still So relatively low, main cause is because simply simply considering the space length and neighborhood of center pel and neighborhood pixel in FLICM Pixel is subordinate to angle value, is subordinate to angle value without consideration center pel, and the fuzzy factor computational methods proposed do not have Any physical significance.Domain of the existence edge is excessively smooth in the segmentation and classification result of FLICM, and lost substantial amounts of details letter Breath.Therefore, it is badly in need of one kind and both remains edge detail information, it is contemplated that the efficient FCM algorithm of local spatial information.
The content of the invention
Technical problem:The main object of the present invention is the shortcoming overcome in prior art, there is provided a kind of method is simple, The degree of automation is high, is influenced that small, the high Fuzzy C based on spatial attraction model of classification of image segmentation accuracy is equal by picture noise It is worth remote sensing image automatic classification method.
Technical solution:For achieving the above object, the fuzzy C-mean algorithm remote sensing shadow of the invention based on spatial attraction model As automatic classification method, comprise the following steps
Step 1:Remote Sensing Digital Image to be sorted is obtained, Remote Sensing Digital Image is obtained according to picture size and wave band number Pixel number, according to actual classification it needs to be determined that cluster the classification number c and Fuzzy Exponential m of Remote Sensing Digital Image, and utilizes mark Accurate FCM models cluster Remote Sensing Digital Image to obtain initialization fuzzy membership matrix U0={ uki}c×NAnd cluster centre V0={ vk}c, N be Remote Sensing Digital Image to be sorted pixel number, ukiRepresent in Remote Sensing Digital Image to be sorted i-th pixel Belong to the degree of membership of kth class, vkFor the central point of Remote Sensing Digital Image kth class to be sorted;
Step 2:Neighborhood window size is set, is determined remote sensing according to the neighborhood window size of setting in Remote Sensing Digital Image Each pixel passes through formula as the pixel in the neighborhood window of center pel in digitized videoMeter Calculate the spatial attraction NA in each pixel and its affiliated neighborhood window between other pixels in Remote Sensing Digital Imageij,
Wherein, G is constant, for representing to adjust contribution of the space constraint to clustering object function, generally sets G=1, ukiTable Show neighborhood window center pixel xiBelong to the degree of membership of k-th of classification, ukjRepresent j-th of pixel x in neighborhood windowjBelong to The degree of membership of k classification, RijRepresent pixel xiWith pixel xjBetween theorem in Euclid space distance;
Step 3:Utilize the sky between other pixels in Remote Sensing Digital Image in each pixel and its affiliated neighborhood window Between gravitation NAij, pass through formulaThe space constraint in Remote Sensing Digital Image between pixel is obtained to punish Penalty factor wij,
Wherein, wijRepresent j-th of pixel x in neighborhood windowjTo center pel xiWeighing factor, NiImago in expression First xiNeighborhood window in pixel, NAijFor each pixel x in the image that is calculated in step 2iWith by pixel xiAs in Remaining pixel x in the neighborhood window of imago memberjBetween spatial attraction, | xi-xj| represent pixel xiWith by pixel xiAs in Pixel x in the neighborhood window of imago memberjBetween gray value difference;
Step 4:Pass through formulaObtain fuzzy factor
Step 5:Pass through formulaBy fuzzy factorIt is added to standard FCM moulds Type, obtains cluster object function
Wherein, N is the pixel number of Remote Sensing Digital Image to be sorted, and c is the cluster number of classification Remote Sensing Digital Image, Represent that in Remote Sensing Digital Image to be sorted i-th pixel belongs to the degree of membership of kth class, m is the mould of Remote Sensing Digital Image to be sorted Paste index, xiRepresent i-th of pixel of Remote Sensing Digital Image to be sorted, vkFor the center of Remote Sensing Digital Image kth class to be sorted Point;
With reference to the initially fuzzy matrix U obtained in step 10With cluster centre V0, utilize formulaCalculating is treated The new cluster centre of classification remote sensing imageIt is calculated asRepresent currently available new cluster centre;
Recycle formulaCalculate the new subordinated-degree matrix of remote sensing image to be sortedMeter ForRepresent currently available new fuzzy matrix;At this timeTo pass through initial cluster centre V0Obtain, representIt is previous Secondary obtained cluster centre;To pass through initial fuzzy matrix U0Obtain, representThe preceding fuzzy matrix once obtained;
Step 6:Judge cluster centreWhether continue change or computing reaches maximum iteration, set iteration stopping threshold Value ε=1e-5 is a small positive number, t-th of cluster centre that Remote Sensing Digital Image to be sorted is tried to achieveIn being clustered with the t-1 The heartIt is compared, if meetingOr the condition of b > T one of both, then iteration terminate, the b is initial It is worth for 0, T 100, otherwise, setting b=b+1, with currently available fuzzy matrixAnd cluster centreSubstitute respectively initial Fuzzy matrix U0With cluster centre V0, and be set asWithBack to step 2, step 2~step 6 is repeated, until MeetOr the condition of b > T one of both;
Step 7:Utilize the fuzzy membership matrix U={ u finally obtainedki}c×N, each pixel is determined according to equation below xiGeneric, i.e., for each pixel xi, its generic ciFor degree of membership ukiMiddle that maximum classification:
Ci=argk{max(uki), k=1,2,3, c
Step 8:According to each pixel xiGeneric ciDifferent cluster classifications is assigned to different colors, forms remote sensing Image classification thematic map, so as to fulfill the automatic classification of Remote Sensing Digital Image.
Beneficial effect:The present invention is by calculating the space constraint penalty factor w in Remote Sensing Digital Image between pixelij, examine Consider the gray-scale relation of local pixel in Remote Sensing Digital Image, the comprehensive spatial context characteristic that can reflect each pixel of proposing Fuzzy factor based on spatial attraction modelFor suppressing noise.Space except considering center pel and neighborhood pixel Distance and neighborhood pixel are subordinate to angle value, it is also contemplated that center pel is subordinate to angle value, and the fuzzy factor introducedWith physics Meaning, both remains edge detail information, it is contemplated that local spatial information.Present invention uses the corresponding neighbour of center pel Spatial attraction NA in the window of domain between other pixelsijIt is adaptive to should determine that influence degree of the neighborhood pixel to center pel, without Learning experience value;Introduce the fuzzy factor based on spatial attraction modelInto object function so that this classification have to containing The processing of noisy Remote Sensing Digital Image has robustness, and image detail is handled well;Its method is simple, required parameter It is identical with the FCM of standard, without other specification, small, classification of image segmentation accuracy height is influenced by picture noise, is had extensive Practicality.
Brief description of the drawings:
Fig. 1 is the flow chart of the present invention;
Fig. 2 is the pixel neighborhood window structure signal of the present invention;
Fig. 3 is the comparison diagram of the segmentation result and the prior art on 1 remote sensing image of the embodiment of the present invention;
Fig. 4 is the comparison diagram of the segmentation result and the prior art on 2 remote sensing image of the embodiment of the present invention.
Embodiment:
The embodiment of the present invention is further described below in conjunction with the accompanying drawings:
As shown in Figure 1, the fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model of the present invention, bag It is as follows to include step:
Step 1:Remote Sensing Digital Image to be sorted (P, Q, S) is obtained, wherein P is the line number of image, and Q is the row of image Number, the wave band number of S images, according to picture size and wave band number, 2 dimension matrix (S, P are converted to by Remote Sensing Digital Image (P, Q, S) × Q) form, i.e., the pixel number and wave band number of Remote Sensing Digital Image to be sorted are only considered, according to actual classification it needs to be determined that sense Cluster the classification number c and Fuzzy Exponential m of digitized video (S, P × Q);Using standard FCM models to Remote Sensing Digital Image (S, P × Q) clustered to obtain initialization fuzzy membership matrix U0={ uki}c×NWith cluster centre V0={ vk}c, N is to be sorted distant Feel the pixel number of digitized video, ukiRepresent that in Remote Sensing Digital Image to be sorted i-th pixel belongs to the degree of membership of kth class, vk For the central point of Remote Sensing Digital Image kth class to be sorted;
Step 2:Neighborhood window size is set, is illustrated in figure 2 a window size as 5 window, the numeral 1 in figure, 2,4 ... represent the contiguous range of center pel x, and Remote Sensing Digital Image (S, P × Q) is determined according to the neighborhood window size of setting In, each pixel in Remote Sensing Digital Image as the pixel in the neighborhood window of center pel, is passed through into formulaCalculate in Remote Sensing Digital Image in each pixel and affiliated neighborhood window between other pixels Spatial attraction NAij,
Wherein, G is constant, for representing to adjust contribution of the space constraint to clustering object function, generally sets G=1, ukiTable Show neighborhood window center pixel xiBelong to the degree of membership of k-th of classification, ukjRepresent j-th of pixel x in neighborhood windowjBelong to The degree of membership of k classification, RijRepresent pixel xiWith pixel xjBetween theorem in Euclid space distance;
Step 3:Utilize the space between other pixels in Remote Sensing Digital Image in each pixel and affiliated neighborhood window Gravitation NAij, pass through formulaObtain the space constraint punishment between pixel in Remote Sensing Digital Image Factor wij,
Wherein, wijRepresent j-th of pixel x in neighborhood windowjTo center pel xiWeighing factor, NiImago in expression First xiNeighborhood window in pixel, NAijFor each pixel x in the image that is calculated in step 2iWith by pixel xiAs in Remaining pixel x in the neighborhood window of imago memberjBetween spatial attraction, | xi-xj| represent pixel xiWith by pixel xiAs in Pixel x in the neighborhood window of imago memberjBetween gray value difference;
Step 4:Pass through formulaObtain fuzzy factor
Step 5:Pass through formulaBy fuzzy factorIt is added to standard FCM moulds Type, obtains adding fuzzy factorCluster object function
Wherein, N is the pixel number of Remote Sensing Digital Image to be sorted, both N=P × Q;C is classification Remote Sensing Digital Image Cluster number,Represent that in Remote Sensing Digital Image to be sorted i-th pixel belongs to the degree of membership of kth class, m is remote sensing number to be sorted The Fuzzy Exponential of word image, xiRepresent i-th of pixel of Remote Sensing Digital Image to be sorted, vkFor Remote Sensing Digital Image kth to be sorted The central point of class;
With reference to the initially fuzzy matrix U obtained in step 10With cluster centre V0, utilize formulaCalculate and treat point The new cluster centre of class remote sensing imageIt is calculated asRepresent currently available new cluster centre;
Recycle formulaCalculate the new subordinated-degree matrix of remote sensing image to be sortedMeter ForRepresent currently available new fuzzy matrix;At this timeTo pass through initial cluster centre V0Obtain, representIt is previous Secondary obtained cluster centre;To pass through initial fuzzy matrix U0Obtain, representThe preceding fuzzy matrix once obtained;
Step 6:Judge cluster centreWhether continue change or computing reaches maximum iteration, set operation times Iteration stopping threshold epsilon=1e-5 is a small positive number, t-th of cluster centre that Remote Sensing Digital Image to be sorted is tried to achieveWith t- 1 cluster centreIt is compared, if meetingOr the condition of b > T one of both, then iteration terminate, The b initial values are 0, T 100, otherwise, b=b+1 are set, with currently available fuzzy matrixAnd cluster centreRespectively Substitute initially fuzzy matrix U0With cluster centre V0, and be set asWithBack to step 2, step 2~step is repeated Rapid 6, until meetingOr the condition of b > T one of both;
Step 7:Utilize the fuzzy membership matrix U={ u finally obtainedki}c×N, each pixel is determined according to equation below xiGeneric, i.e., for each pixel xi, its generic ciFor degree of membership ukiMiddle that maximum classification:
Ci=argk{max(uki), k=1,2,3, c
Step 8:According to each pixel xiGeneric ciDifferent cluster classifications is assigned to different colors, forms remote sensing Image classification thematic map, so as to fulfill the automatic classification of Remote Sensing Digital Image.
All algorithms in the present invention are all to program to realize under Matlab7.8, have chosen the distant of two different zones Feel data conduct verification data, the verification sample of classification is expired by having carried out stringent geometric correction to original remote sensing image For sufficient rectification error less than the requirement of 0.5 pixel, the method then interpreted by visual observation on the image of correction obtains required verification Sample.It finally make use of production precision, overall accuracy and Kappa coefficients to evaluate classification results precision, and this sent out Bright FLNAICM is contrasted with standard FCM and FLICM algorithm.
Embodiment 1, employ size as containing for 200 × 200 pixels be red, green and blue three wave band resolution ratio For 0.61 meter of QuickBird remote sensing image, this data is located in the urban area of Jiangsu Province, China Xuzhou City, and it is 2005 to obtain the time Year August, if Fig. 3 (a) and (b) are original classification image and reference data image respectively, image divide into 4 classifications:Structures, Bare area, water and vegetation, Fuzzy Exponential 2.
Fig. 3 (c)-(e) represents the classification results of FCM, FLICM and FLNAICM respectively, in Fig. 3 (c), due to spectrum The presence of similitude and picture noise, FCM do not account for spatial context information, make just with the spectral characteristic of image There are many " spiced salt " phenomenons into classification results.It can be seen that from Fig. 3 (d) and Fig. 3 (e), the classification of FLICM and FLNAICM Effect will be better than FCM, and the major part " miscellaneous point " in classification chart is removed, and forms preferable homogeney category regions.From Fig. 3 (d) it can be found that classification plot had smooth phenomenon in, the classification plot of large area, such as Fig. 3 (d) acceptances of the bid are formd Note at A, B, C, lost the details in many classification plot, and compared to FLICM methods, from Fig. 3 (e) as can be seen that FLNAICM Remain in many detailed information, such as Fig. 3 (e) at mark A, B, C, the reason is that since FLNAICM make use of center pel Spatial attraction between neighborhood pixel is as neighborhood pixel to center pel spacial influence degree, and this spatial attraction is fully examined The local space relation and gray-scale relation of center pel and neighborhood pixel pixel are considered, and can have been counted automatically according to the characteristic of pixel Calculate.FLICM is simply simply by the space length and degree of membership between center pel and neighborhood pixel pixel.Table 1 provides The classification results precision evaluation of FCM, FLICM and FLNAICM are as a result, relative to FCM and FLICM, FLNAICM gives highest Nicety of grading.
Table 1. implements verification sample number, production precision, overall accuracy and Kappa coefficients in 1
Embodiment 2
In example 2, employ size and contain red, green and blueness three wave bands point for 200 × 200 pixels Resolution is 0.61 meter of QuickBird remote sensing image, this data is located in the suburban areas of Jiangsu Province, China Xuzhou City, and acquisition time is In August, 2005, if Fig. 4 (a) and Fig. 4 (b) are original classification image and reference data image respectively, image divide into 4 classifications: Road, bare area, water and vegetation, Fuzzy Exponential 2.
Fig. 4 (c)-(e) is the classification results of FCM, FLICM and FLNAICM respectively, similar due to spectrum in Fig. 4 (c) Property and picture noise presence, the sorting technique of FCM do not account for spatial context letter merely with the spectral characteristic of image Cease, occur substantial amounts of " spiced salt " phenomenon in classification results.It can be seen that from Fig. 4 (d) and Fig. 4 (e), point of FLICM and FLNAICM Class effect will be better than FCM, and the major part " miscellaneous point " in classification chart is all removed.There is smooth phenomenon in Fig. 4 (d), permitted The loss in detail in more classification plot, such as in Fig. 4 (d) at mark A, B, C, and compared to FLICM methods, can from Fig. 4 (e) Go out, FLNAICM is remained in many detailed information, such as Fig. 4 (e) at mark A, B, C, the reason is that since FLNAICM is utilized Spatial attraction between center pel and neighborhood pixel as neighborhood pixel to center pel spacial influence degree, and this space Gravitation has taken into full account the local space relation and gray-scale relation of center pel and neighborhood pixel pixel, can be according to the characteristic of pixel It is automatic to calculate.Simply simply by the space length and degree of membership between center pel and neighborhood pixel pixel in FLICM.Table 2 provide the classification results precision evaluation of FCM, FLICM and FLNAICM as a result, compared to FCM and FLICM, FLNAICM gives Highest nicety of grading.
Table 2. implements verification sample number, production precision, overall accuracy and Kappa coefficients in 2
To sum up, the present invention proposes a kind of FLNAICM sorting techniques based on neighborhood Pixel domain gravitation, imago in utilization Gravitation estimation neighborhood pixel between member and neighborhood pixel reasonably introduces spatial context letter to the influence degree of center pel The FCM of standard is ceased to improve its nicety of grading and algorithm to the robustness containing noisy image classification.

Claims (1)

1. a kind of fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model, it is characterised in that comprising following Step:
Step 1:Remote Sensing Digital Image to be sorted is obtained, the pixel of Remote Sensing Digital Image is obtained according to picture size and wave band number Number, according to actual classification it needs to be determined that cluster the classification number c and Fuzzy Exponential m of Remote Sensing Digital Image, and utilizes standard FCM models cluster Remote Sensing Digital Image to obtain initialization fuzzy membership matrix U0={ uki}c×NWith cluster centre V0= {vk}c, N be Remote Sensing Digital Image to be sorted pixel number, ukiRepresent that in Remote Sensing Digital Image to be sorted i-th pixel belongs to The degree of membership of kth class, kvFor the central point of Remote Sensing Digital Image kth class to be sorted;
Step 2:Neighborhood window size is set, is determined remote digital according to the neighborhood window size of setting in Remote Sensing Digital Image Each pixel passes through formula as the pixel in the neighborhood window of center pel in imageCalculate distant Feel the spatial attraction NA in each pixel and its affiliated neighborhood window between other pixels in digitized videoij,
Wherein, G is constant, for representing to adjust contribution of the space constraint to clustering object function, if G=1, ukiRepresent i-th Pixel xiBelong to the degree of membership of k-th of classification, ukjRepresent pixel xiNeighborhood window in j-th yuan of xjBelong to k-th classification Degree of membership, RijRepresent pixel xiWith pixel xjBetween theorem in Euclid space distance;
Step 3:Drawn using the space between other pixels in Remote Sensing Digital Image in each pixel and its affiliated neighborhood window Power NAij, pass through formulaObtain in Remote Sensing Digital Image space constraint punishment between pixel because Sub- wij,
Wherein, wijRepresent j-th of pixel x in neighborhood windowjTo center pel xiWeighing factor, NiRepresent center pel xi Neighborhood window in pixel, NAijFor each pixel x in the image that is calculated in step 2iWith by pixel xiAs middle imago Remaining pixel x in the neighborhood window of memberjBetween spatial attraction, | xi-xj| represent pixel xiWith by pixel xiAs middle imago Pixel x in the neighborhood window of memberjBetween gray value difference;
Step 4:Pass through formulaObtain fuzzy factor
Step 5:Pass through formulaBy fuzzy factorStandard FCM models are added to, Obtain cluster object function
Wherein, N be Remote Sensing Digital Image to be sorted pixel number, c be classification Remote Sensing Digital Image cluster number, uki mTable Show that in Remote Sensing Digital Image to be sorted i-th pixel belongs to the degree of membership of kth class, m is the fuzzy of Remote Sensing Digital Image to be sorted Index, xiRepresent i-th of pixel of Remote Sensing Digital Image to be sorted, vkFor the central point of Remote Sensing Digital Image kth class to be sorted;
With reference to the initially fuzzy matrix U obtained in step 10With cluster centre V0, utilize formulaCalculate to be sorted The new cluster centre of remote sensing imageIt is calculated asRepresent currently available new cluster centre;
Recycle formulaCalculate the new subordinated-degree matrix of remote sensing image to be sortedIt is calculated asRepresent currently available new fuzzy matrix;At this timeFor initial cluster centre V0Obtain, representIt is preceding once to obtain The cluster centre arrived;For initial fuzzy matrix U0Obtain, representThe preceding fuzzy matrix once obtained;
Step 6:Judge cluster centreWhether continue change or computing reaches maximum iteration, set iteration stopping threshold epsilon =1e-5 is a small positive number, t-th of cluster centre that Remote Sensing Digital Image to be sorted is tried to achieveWith the t-1 cluster centreIt is compared, if meetingOr the condition of b > T one of both, then iteration terminate, the b is initial It is worth for 0, T 100, otherwise, setting b=b+1, with currently available fuzzy matrixAnd cluster centreSubstitute respectively initial Fuzzy matrix U0With cluster centre V0, and be set asWithBack to step 2, step 2~step 6 is repeated, Until meetOr the condition of b > T one of both;
Step 7:Utilize the fuzzy membership matrix U={ u finally obtainedki}c×N, each pixel x is determined according to equation belowiInstitute Belong to classification, i.e., for each pixel xi, its generic ciFor degree of membership ukiMiddle that maximum classification;
Ci=argk{max(uki), k=1,2,3 ..., c
Step 8:According to each pixel xiGeneric ciDifferent cluster classifications is assigned to different colors, forms remote sensing image Classification thematic map, so as to fulfill the automatic classification of Remote Sensing Digital Image.
CN201410325747.4A 2014-07-10 2014-07-10 Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model Active CN104123561B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410325747.4A CN104123561B (en) 2014-07-10 2014-07-10 Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410325747.4A CN104123561B (en) 2014-07-10 2014-07-10 Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model

Publications (2)

Publication Number Publication Date
CN104123561A CN104123561A (en) 2014-10-29
CN104123561B true CN104123561B (en) 2018-04-13

Family

ID=51768966

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410325747.4A Active CN104123561B (en) 2014-07-10 2014-07-10 Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model

Country Status (1)

Country Link
CN (1) CN104123561B (en)

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105654453B (en) * 2014-11-10 2018-09-28 华东师范大学 A kind of FCM image partition methods of robustness
CN105825222A (en) * 2015-01-07 2016-08-03 王伟 Land automatic classification method based on multisource and multi-temporal satellite image data
CN104834942B (en) * 2015-05-22 2018-02-09 武汉大学 Remote sensing image variation detection method and system based on mask classification
CN105320753B (en) * 2015-09-30 2018-07-06 重庆大学 A kind of unbalanced data sorting technique and its system based on level gravity model
CN106408571B (en) * 2016-09-20 2018-11-02 辽宁工程技术大学 A kind of variable class remote sensing image segmentation method based on the selection of optimal fuzzy factor
CN106600605A (en) * 2016-12-14 2017-04-26 陕西科技大学 Unsupervised fast image segmentation algorithm
CN107492092A (en) * 2017-07-13 2017-12-19 青岛黄海学院 The medical image cutting method of GSA algorithms is improved based on FCM algorithm fusions
CN108830297B (en) * 2018-05-19 2021-04-27 烟台大学 Multispectral remote sensing image ground feature classification method
CN109064470B (en) * 2018-08-28 2022-02-22 河南工业大学 Image segmentation method and device based on self-adaptive fuzzy clustering
CN109697466B (en) * 2018-12-20 2022-10-25 烟台大学 Terrain classification method of adaptive interval type spatial fuzzy C-means
CN109993204A (en) * 2019-02-25 2019-07-09 重庆邮电大学 The classification of remote-sensing images algorithm of FCM cluster is selected and improved based on mRMR
CN109829519B (en) * 2019-03-22 2020-04-03 兰州交通大学 Remote sensing image classification method and system based on self-adaptive spatial information
CN110135432A (en) * 2019-05-24 2019-08-16 哈尔滨工程大学 A kind of high-spectrum remote sensing dividing method based on K-means cluster
CN110415208B (en) * 2019-06-10 2023-10-17 西安电子科技大学 Self-adaptive target detection method and device, equipment and storage medium thereof
CN111898690B (en) * 2020-08-05 2022-11-18 山东大学 Power transformer fault classification method and system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102289678A (en) * 2011-08-10 2011-12-21 武汉大学 Fuzzy supervised classification method for multiband remote sensing image based on non-equal weight distances
CN103530875A (en) * 2013-10-09 2014-01-22 哈尔滨工程大学 End member extraction data preprocessing method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060291751A1 (en) * 2004-12-16 2006-12-28 Peyman Milanfar Robust reconstruction of high resolution grayscale images from a sequence of low-resolution frames (robust gray super-resolution)

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102289678A (en) * 2011-08-10 2011-12-21 武汉大学 Fuzzy supervised classification method for multiband remote sensing image based on non-equal weight distances
CN103530875A (en) * 2013-10-09 2014-01-22 哈尔滨工程大学 End member extraction data preprocessing method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
《A Robust Fuzzy Local Information C-Means Clustering Algorithm》;Stelios Krinidis,Vassilios Chatzis;《IEEE TRANSACTIONS ON IMAGE PROCESSING》;20100531;第29卷(第5期);第1328-1337页 *
《遥感数据可靠性分类方法研究》;张华;《中国博士学位论文全文数据库(信息科技辑)》;20130615(第09期);正文第106-111页 *

Also Published As

Publication number Publication date
CN104123561A (en) 2014-10-29

Similar Documents

Publication Publication Date Title
CN104123561B (en) Fuzzy C-mean algorithm remote sensing image automatic classification method based on spatial attraction model
CN110443143B (en) Multi-branch convolutional neural network fused remote sensing image scene classification method
CN109344736B (en) Static image crowd counting method based on joint learning
CN103098090B (en) Multiparameter three-dimensional magnetic resonance imaging brain tumor partition method
CN103761726B (en) Block adaptive image partition method based on FCM
CN112396619B (en) Small particle segmentation method based on semantic segmentation and internally complex composition
CN106096561A (en) Infrared pedestrian detection method based on image block degree of depth learning characteristic
CN111639587B (en) Hyperspectral image classification method based on multi-scale spectrum space convolution neural network
CN103150731B (en) A kind of fuzzy clustering image partition method
CN109829519B (en) Remote sensing image classification method and system based on self-adaptive spatial information
CN105427309A (en) Multiscale hierarchical processing method for extracting object-oriented high-spatial resolution remote sensing information
CN106611423B (en) SAR image segmentation method based on ridge ripple filter and deconvolution structural model
CN105894490A (en) Fuzzy integration multiple classifier integration-based uterine neck cell image identification method and device
CN108960404B (en) Image-based crowd counting method and device
CN104966085A (en) Remote sensing image region-of-interest detection method based on multi-significant-feature fusion
CN105261004A (en) Mean shift and neighborhood information based fuzzy C-mean image segmentation method
CN103996193B (en) Brain MR image segmentation method combining weighted neighborhood information and biased field restoration
CN105809175A (en) Encephaledema segmentation method and system based on support vector machine algorithm
CN103996018A (en) Human-face identification method based on 4DLBP
CN106611422B (en) Stochastic gradient Bayes's SAR image segmentation method based on sketch structure
CN104102928B (en) A kind of Classifying Method in Remote Sensing Image based on texture primitive
CN107403434A (en) SAR image semantic segmentation method based on two-phase analyzing method
CN106991686A (en) A kind of level set contour tracing method based on super-pixel optical flow field
CN104252625A (en) Sample adaptive multi-feature weighted remote sensing image method
CN108537751A (en) A kind of Thyroid ultrasound image automatic segmentation method based on radial base neural net

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
CP02 Change in the address of a patent holder

Address after: 221116 Research Institute of China University of Mining and Technology,, Jiangsu

Patentee after: China University of Mining & Technology

Address before: 221116 Research Institute, China University of Mining and Technology, Xuzhou University, Jiangsu, China,

Patentee before: China University of Mining & Technology

CP02 Change in the address of a patent holder
GR01 Patent grant
GR01 Patent grant