CN102622607A - Remote sensing image classification method based on multi-feature fusion - Google Patents
Remote sensing image classification method based on multi-feature fusion Download PDFInfo
- Publication number
- CN102622607A CN102622607A CN2012100430640A CN201210043064A CN102622607A CN 102622607 A CN102622607 A CN 102622607A CN 2012100430640 A CN2012100430640 A CN 2012100430640A CN 201210043064 A CN201210043064 A CN 201210043064A CN 102622607 A CN102622607 A CN 102622607A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- word
- characteristic
- bag
- feature fusion
- 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 70
- 230000004927 fusion Effects 0.000 title claims abstract description 24
- 230000000007 visual effect Effects 0.000 claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000012706 support-vector machine Methods 0.000 claims abstract description 16
- 238000012360 testing method Methods 0.000 claims abstract description 8
- 239000004744 fabric Substances 0.000 claims description 12
- 239000000284 extract Substances 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 7
- 238000013139 quantization Methods 0.000 claims description 4
- 238000001308 synthesis method Methods 0.000 abstract 1
- 230000002194 synthesizing effect Effects 0.000 abstract 1
- 230000006870 function Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 3
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 3
- 230000004069 differentiation Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007500 overflow downdraw method Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 241000282320 Panthera leo Species 0.000 description 1
- 102100029469 WD repeat and HMG-box DNA-binding protein 1 Human genes 0.000 description 1
- 101710097421 WD repeat and HMG-box DNA-binding protein 1 Proteins 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012887 quadratic function Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 238000013316 zoning Methods 0.000 description 1
Images
Abstract
The invention discloses a remote sensing image classification method based on multi-feature fusion, which includes the following steps: A, respectively extracting visual word bag features, color histogram features and textural features of training set remote sensing images; B, respectively using the visual word bag features, the color histogram features and the textural features of the training remote sensing images to perform support vector machine training to obtain three different support vector machine classifiers; and C, respectively extracting visual word bag features, color histogram features and textural features of unknown test samples, using corresponding support vector machine classifiers obtained in the step B to perform category forecasting to obtain three groups of category forecasting results, and synthesizing the three groups of category forecasting results in a weighting synthesis method to obtain the final classification result. The remote sensing image classification method based on multi-feature fusion further adopts an improved word bag model to perform visual word bag feature extracting. Compared with the prior art, the remote sensing image classification method based on multi-feature fusion can obtain more accurate classification result.
Description
Technical field
The present invention relates to a kind of image classification method, relate in particular to a kind of Classifying Method in Remote Sensing Image based on multi-feature fusion, belong to the image recognition technology field.
Background technology
Along with developing rapidly of remote sensing and infotech, every day, retrievable remote sensing image data amount increased just with surprising rapidity.In the face of the remotely-sensed data of magnanimity, mode how to utilize computing machine automatically to understand image according to people is categorized into the different semantics classification becomes a challenging problem.In the semantic content of computer vision research, the scene classification of image has not only comprised the general cognition of people to piece image, but also is that the other guide that further identifies in the image provides the foundation.Therefore the image scene classification just becomes the hot issue of current computer vision and multimedia messages process field.
Professional experiences that traditional remote sensing image visual interpretation method not only requires to enrich and sufficient open-air on-site inspection data, and because this recognition methods is to be based upon on the specific priori basis, it is very big therefore to discern difficulty; Efficient is very low, for example, and document [Zhang Jin's water; What spring sun-light, Pan Yaozhong, etc. based on the compound high spatial resolution remote sense data qualification research [J] of the multi-source information of SVM. the remote sensing journal. the 10th volume; The 1st phase, 2006:49-57.] in the method that adopted.High-resolution remote sensing image can provide a large amount of topographical features, demonstrate fully the inner differentiation of atural object obviously, aspect texture increases, details enrich and the edge gives prominence to.The atural object space distribution is complicated; It is heterogeneous that similar object demonstrates very big spectrum; Type of being embodied in internal variance becomes greatly, inter-class variance reduces, the spectrum of different atural objects is overlapped; Make traditional sorting technique cause more mistake branch easily, leak the branch phenomenon, thereby cause nicety of grading lower based on pixel.Said method does not utilize the characteristics of objects of high spatial resolution image preferably mainly according to spectral information.OO sorting technique has appearred in recent years, for example, document [Li Xinliang, Zhao Shuhe, Rui Yikang. object-oriented high-resolution classification of remote-sensing images research [J]. sensor information, 2007,6:58-61.] in the method that adopted.
Different according to the mode of describing image, current OO scene classification method is divided into based on global characteristics with based on two big types of the speech bag models of local feature.And early stage scene classification method is to describe scene content according to the global statistics characteristic of image, like color histogram, for example; Document [MICHAEL J. SWAIN, DANA H. BALLARD. Color Indexing [J]. International Journal of Computer Vision, 1991; 7 (1): 11-32.] method that is adopted in; Texture, for example, document [Tai Sing Lee. Image Representation Using 2D Gabor Wavelets. IEEE Transactions on pattern analysis and machine intelligence; 1996,18 (10): 1-13.] method that is adopted in.Color histogram has affine, unchangeability such as rotation, and counting yield is high.But because it has abandoned image space information, identical for two width of cloth distribution of color, content and different image lack differentiation power, so robustness is bad, and recognition efficiency is not high.In recent years, the main stream approach of scene classification is based on the speech bag model
[5](it represents the image as some topography's pieces for bag of visual words, method BOVW), and every width of cloth image slices sentence equally is expressed as the word of different proportion, and a dictionary is formed in the set of all words.The speech bag model have calculate simple, to noise, illumination and partial occlusion characteristics such as robust more.But because the remote sensing images texture information is abundant, local feature point is more, and direct employing differs based on clustering method acquisition sight word and reflects scene characteristic surely, thereby has caused overall nicety of grading also not high.And for the different target in the high-resolution remote sensing image, the performance of each characteristic is also inconsistent.Maybe be because the conspicuousness of its color characteristic like some type, so the most suitable employing color histogram characteristic is classified, and some type texture rich makes local feature apparent in view.If obviously adopting classifies no longer suitable with a kind of characteristic to different classes of target.
Summary of the invention
Technical matters to be solved by this invention is to overcome the deficiency of prior art, and a kind of Classifying Method in Remote Sensing Image based on multi-feature fusion is provided, and can obtain more accurate remote sensing image classification result.
The present invention is concrete to adopt following technical scheme to solve the problems of the technologies described above.
A kind of Classifying Method in Remote Sensing Image based on multi-feature fusion may further comprise the steps:
Steps A, extract visual word bag characteristic, color histogram characteristic and the textural characteristics of training set remote sensing images respectively;
Step B, utilize visual word bag characteristic, color histogram characteristic and the textural characteristics of training set remote sensing images to carry out the SVMs training respectively, obtain three different support vector machine classifiers;
Step C, for the test sample book of the unknown; Extract its visual word bag characteristic, color histogram characteristic and textural characteristics respectively; And utilize among the step B resulting corresponding support vector machine classifier to carry out the classification prediction; Obtain three categories and do not predict the outcome, adopt the weighted comprehensive method that three categories are not predicted the outcome then and carry out comprehensively, thereby obtain final classification results.
Preferably, said extraction visual word bag is represented, is meant and adopts improved speech bag model, extracts the visual word bag of weighting and representes; Specifically may further comprise the steps:
The local feature of each image is put into a file with the local feature of all images in step 1, the extraction training set, forms the local feature point set;
Step 4, add up the addition summation of the word frequency of each sight word in the visual word bag characteristic of each type scene all images, thus obtain one type of scene overall visual speech bag represent;
Step 5, from the statistics of step 4, take out the bigger top n sight word of word frequency during promptly the overall visual speech bag of type scene is represented, N for preset greater than 1 integer; Word frequency to this N sight word is carried out normalization, and representes the weighted value that each sight word is corresponding through the inverse of its word frequency, and its value is respectively w
i, 0<i<n+1, i are integer; These weighted values are carried out normalization, make its weighted value between 0 to 1; Then each weighted value is added 1 computing, i.e. w
i=w
i+ 1;
The weighted value w of step 6, top n word that the word frequency during the total pronouns, general term for nouns, numerals and measure words bag of class of classification under the remote sensing images represented is bigger
iMultiply by weighted value w
iSight word word frequency in the pairing image vision speech bag characteristic so just obtains the visual word bag characteristic of the weighting of this width of cloth image.
Preferably, color histogram described in the steps A is characterized as and adopts 512 color histograms that bin quantizes.
Preferably, said textural characteristics is 5 yardsticks, the Gabor textural characteristics of 6 directions.
Preferably, when visual word bag characteristic and color histogram characteristic are carried out the SVMs training, adopt histogram intersection nuclear, when textural characteristics is carried out the SVMs training, adopt RBF as kernel function as kernel function.
Preferably; Adopt the weighted comprehensive method that three group categories results are carried out comprehensively described in the step C; Specifically according to following method: calculate test pattern visual word bag characteristic; The classification of color histogram characteristic and the pairing support vector machine classifier of textural characteristics predicts the outcome, and comprises classification prediction label and classification prediction probability value; Get first of three support vector machine classifiers and predict the outcome, and corresponding three classification prediction probabilities are sorted, obtain maximum classification prediction probability value, inferior big classification prediction probability value and minimum classification prediction probability value; If inferior big classification prediction probability value corresponding class prediction label is identical with the pairing classification prediction of minimum classification prediction probability value label; And their classification prediction probability value sum is then predicted the final classification results of label as many Feature Fusion with the classification of its representative greater than maximum classification prediction probability value; Otherwise, with the final classification results of the pairing classification prediction of maximum classification prediction probability value label as many Feature Fusion.
Compare prior art, the present invention has following beneficial effect:
One, the present invention has higher nicety of grading;
Two, the present invention adopts the character representation method of improved speech bag model as image, can when the local feature of image is unconspicuous, strengthen the character representation of image, thereby improves the nicety of grading of image;
Three, the present invention fully utilizes the various features of image, comprises local feature, and therefore color characteristic and textural characteristics can obtain higher nicety of grading in the unconspicuous situation of the single characteristic of image, make that the image range of handling is bigger, and precision is higher.
Description of drawings
Fig. 1 is the local feature synoptic diagram of remote sensing images, and wherein Fig. 1 (a) is an original image, and Fig. 1 (b) is the local feature synoptic diagram of Fig. 1 (a);
Fig. 2 is the visual word bag characteristic of remote sensing images;
Fig. 3 representes for total pronouns, general term for nouns, numerals and measure words bag of Agriculture class in the remote sensing images training set that is adopted in this paper experiment;
Fig. 4 is the color histogram characteristic synoptic diagram of remote sensing images, and wherein Fig. 4 (a) is an original image, and Fig. 4 (b) is the R histogram of component of Fig. 4 (a), and Fig. 4 (c) is the G histogram of component of Fig. 4 (a), and Fig. 4 (d) is the B component synoptic diagram of Fig. 4 (a);
Fig. 5 is the degree of accuracy of each type of BOVW generating algorithm after improving and original BOVW and total average classification degree of accuracy;
Fig. 6 is precision and total average nicety of grading of each type of single characteristic and many Feature Fusion method;
Fig. 7 is the weighted comprehensive method, the precision of majority vote method and each type of method in full accord and total average classification degree of accuracy contrast.
Embodiment
Below in conjunction with accompanying drawing technical scheme of the present invention is elaborated:
Consider notable feature that every type of image has and inconsistent; And the wrong situation of dividing of different characteristic also is not quite similar; The present invention proposes the Classifying Method in Remote Sensing Image of many Feature Fusion; Utilize visual word bag characteristic, color histogram characteristic and textural characteristics to classify respectively, and vote, finally obtain the classification results after many Feature Fusion according to the classification results of every kind of characteristic gained; The present invention further improves the visual word bag model, carries out weighting through the remarkable word feature information to different scene classes, thereby has improved the similarity degree of single image and whole class.Apparent in view to some individual differences, but semantic similar image has been obtained good classification accuracy rate.
The inventive method is specifically according to following steps:
Steps A, extract visual word bag characteristic, color histogram characteristic and the textural characteristics of remote sensing images training set respectively, specific as follows:
The local feature of each image is put into a file with the local feature of all images in steps A 1, the extraction remote sensing images training set, forms the local feature point set;
The local feature of image be according to the definition of certain conspicuousness criterion with its other image model of neighborhood phase region, be associated with the change of one or more image attributes usually, have various ways such as point, limit, image block.Can overcome visual angle and illumination variation, background obscure, block and type in the difficulty represented of vision object such as difference; Carry out the identification of target through these local characteristics; Make and adopt the Target Recognition of local feature technology under complex background that application is more and more widely arranged, and performance is preferably arranged.As have SIFT (Scale Invariant Feature Transform) the feature set feature detection of various unchangeability and be described in one; Conversion such as rotation, yardstick and brightness are remained unchanged, affined transformation, visual angle change and illumination variation are also had certain stability.The present invention preferably adopts the local feature of SIFT characteristic as image.
SIFT is characterized as prior art; But the detailed content list of references [DAVID G.LOWE. Distinctive Image Features from Scale-Invariant Keypoints [J]. International Journal of Computer Vision; 2004; 60 (2): 91-100.], it is following that it comprises step:
The building method of SIFT characteristic comprises the detection of key point and structure two parts of descriptor.
1) with gaussian kernel function continuous filtering and the down-sampling of input picture, forms the gaussian pyramid image through different scale.Two gaussian image of adjacent yardstick are subtracted each other again and obtain DoG (Difference of Gaussians), promptly the pyramid multiscale space is represented.Then the point of each point on the pyramid multiscale space with adjacent yardstick and adjacent position compared one by one, the local extremum position that obtains is residing position of key point and corresponding yardstick.
Through the three-dimensional quadratic function of match key point is carried out further accurately location, remove low contrast key point and unsettled skirt response point, to strengthen coupling stability and noise resisting ability.
2) before structure SIFT descriptor, to give a principal direction, make descriptor have rotational invariance for each key point.
For each characteristic area, at first gradient-norm value of all pixels
and gradient direction
in the zoning.
Wherein L representes that coordinate position is (x, the gray-scale value of y) locating in the image to be classified.
Be the center then with the unique point; The neighborhood of getting the 16*16 pixel is as sample window; Again this neighborhood is divided into the 4*4 sub regions equably; Calculate the gradient orientation histogram of each subregion, then 8 direction histograms in the subregion are sorted according to the position in order, obtained the 128 dimensional feature descriptors of 4*4*8 at last.
Image shown in Fig. 1 (a) obtains local feature shown in Fig. 1 (b) as original input picture after the SIFT feature extraction;
Steps A 2, the improved visual word bag model of employing extract the visual word bag characteristic of weighting;
1) utilize vector quantization method (like the K average) that local feature point set is carried out cluster.Cluster centre is sight word, and all cluster centres form the visual vocabulary storehouse.The K means clustering algorithm is based on cluster performance index minimization principle, and is simple, efficient, is usually used in generating in the speech bag model vision dictionary;
2) local feature of each image is distributed to word nearest with it in the visual vocabulary storehouse, add up the visual word bag characteristic that the corresponding word frequency of each word forms image then.Piece image just can be represented with the vector of a multidimensional like this.Fig. 2 has shown the visual word bag characteristic of image;
3) add up the addition sum total of the word word frequency of each word in the visual word bag characteristic of each type scene all images, thus obtain one type of scene total pronouns, general term for nouns, numerals and measure words bag represent;
4) from a last step statistics, take out the bigger top n word of word frequency during promptly total pronouns, general term for nouns, numerals and measure words bag of type scene is represented, greater than 1 integer, here the N value is 20 to N for preset.Word frequency to this N word is carried out normalization, and representes the weighted value that word is corresponding through the inverse of its word frequency, and its value is respectively wi, (0 <i < N+1, i is an integer).These weighted values are carried out normalization, make its weighted value between 0 to 1.Then each weighted value is added 1 computing, i.e. (wi=wi+1).Weighted value just is quantized between 1 and 2 like this, eliminates excessive weight adverse effect with this.Fig. 3 has shown that total pronouns, general term for nouns, numerals and measure words bag of Agriculture class is represented in the remote sensing images training set.
The weighted value wi of the top n word that 5) word frequency during the total pronouns, general term for nouns, numerals and measure words bag of class of classification under the remote sensing images is represented is bigger multiply by the word word frequency in the pairing image vision speech of the weighted value wi bag characteristic.The visual word bag characteristic of the weighting of this width of cloth image that so just obtains.。
The color histogram characteristic of steps A 3, extraction remote sensing images training set;
Histogram is the color characteristic that quilt extensively adopts in many image indexing systems, has the feature extraction advantage easy with similar calculating.Color histogram is a prior art, but concrete list of references [MICHAEL J. SWAIN, DANA H. BALLARD. Color Indexing [J]. International Journal of Computer Vision, 1991,7 (1): 11-32.].
This paper preferably adopts the color space based on HLS, and with respect to rgb color space, it more meets the visual identity characteristic of human eye.Method for distilling is following:
For a pixel in the image, according to its H, L, the value of S component is quantized to respectively among 8 bin, is combined into 8*8*8=512 bin by these 8 bin then, and adds up total number of pixels, carries out normalization at last.
This process embodiment is original input picture with image shown in Fig. 4 (a), and resulting histogram synoptic diagram is respectively shown in Fig. 4 (b), Fig. 4 (c), Fig. 4 (d) after color histogram extracts;
The textural characteristics of steps A 4, extraction remote sensing images training set;
Be proved to be a kind of method of analyzing remote sensing images efficiently through using the Gabor wave filter to obtain image texture characteristic, therefore, this paper preferably adopts the Gabor wave filter to obtain image texture characteristic.The Gabor wave filter is a kind of prior art; But concrete reference literature [Tai Sing Lee. Image Representation Using 2D Gabor Wavelets. IEEE Transactions on pattern analysis and machine intelligence; 1996; 18 (10): 1-13.], method for distilling is following:
The two-dimensional Gabor function can be expressed as:
Wherein:
Adopt 5 yardsticks in this embodiment; The Gabor wave filter of 6 directions extracts the proper vector of image, i.e. v=5, u=6; Totally 30 Gabor kernel functions; The Gabor kernel function of different frequency different directions can form 30 width of cloth image textures to be represented, calculates mean value that the texture of 30 width of cloth images representes and variance as the characteristic of extracting, thereby can obtain the proper vector of one 60 dimension.The textural characteristics that just can represent entire image with the proper vector of these 60 dimensions.
Step B, utilize visual word bag characteristic, color histogram characteristic and the textural characteristics of remote sensing images training set to carry out the SVMs training respectively, obtain three different support vector sorters;
Idiographic flow is following:
1) two types of training samples of input, one type of positive example, one type of counter-example goes out training sample matrix and corresponding class sample matrix according to these sample architecture.Value in the classification sample matrix is the class categories of sample, is generally 1 and-1.
2) adopt proper optimization algorithm and kernel function, adopt histogram intersection nuclear to local feature and color histogram in this embodiment, to textural characteristics employing RBF kernel function.In the local feature kernel function, the value of parameters C is made as 8 o'clock optimums, and in the color histogram kernel function, the value of parameters C is made as 32 o'clock optimums.In the textural characteristics kernel function, the value of parameters C is made as 128, and the value of parameter g is made as 0.00390625 o'clock optimum.Solve vectorial a, this vector magnitude equates that with number of training wherein each ai is corresponding with each training sample, and non-vanishing item correspondence is exactly support vector.
3) obtain final categorised decision function according to the vectorial a that solves.
Step C, extract visual word bag characteristic, color histogram characteristic and the textural characteristics of remote sensing images test set respectively; And utilize among the step B resulting corresponding support vector sorter to carry out the classification prediction; Obtaining three categories does not predict the outcome; Adopt the weighted comprehensive method that three categories are not predicted the outcome then and carry out comprehensively, thereby obtain final remote sensing image classification result.
Idiographic flow is following:
1) with the local feature of image to be detected, the prediction probability value that color histogram characteristic and textural characteristics carry out behind the svm classifier is designated as P1, P2, P3 respectively.
2) judge the notable feature of image to be detected, and ask its corresponding most probable value: calculate P1, P2, the maximal value max among the P3, and to obtain pairing type of label of this max be Label1.Write down second largest value and the 3rd value greatly among P1, P2, the P3 simultaneously, corresponding respectively time notable feature and notable feature least, a corresponding class label is designated as Label2 and Label3.
3) because time notable feature the and least combination of notable feature is better than notable feature possibly; The size of voting results that therefore will compare the two and notable feature probability: if Label2 is identical with Label3; And corresponding probability sum is greater than max, and then final classification results is Label2 or Label3.Otherwise Label1 is as final classification results in output.
In order to verify effect of the present invention, adopt first large-scale remote Sensing Image Retrieval common test data set in the world, have 21 types, every type comprises 100 width of cloth images.Be the QuickBird high-resolution remote sensing image, size is 256*256.These images are downloaded from United States Geological Survey (USGS), comprise the scene in a plurality of states of the U.S.: farmland, aircraft, infield, beach, buildings; Shrubbery, jungle, forest, highway, golf course; The harbour, crossroad, medium jungle, prefabricated house district, viaduct; The parking lot, river, runway, sparse residential district, storage tank and tennis court.These images have reflected the different representations of every type of scene, have good representativeness.
From each type, take out 80 width of cloth figure as training sample, 20 width of cloth figure are as test sample book.The improved BOVW speech bag generating algorithm that adopts the present invention to propose compares with common BOVW algorithm, and the result is as shown in Figure 5.As can be seen from Figure 5, improved BOVW speech bag generating algorithm has 4% raising (78.33% vs 74.52%) approximately than the population mean nicety of grading of original BOVW speech bag generating algorithm.
From each type, choose 80 width of cloth images, utilize SVMs to train, choose 20 width of cloth images as test sample book as training sample.The parameter that each characteristic is used is provided with as follows: in the speech bag model, code book (codebook) number is 1000; In the color histogram characteristic, vectorial dimension is 512; In the Gabor texture, scale parameter is 5, and direction number is 6, and vectorial dimension is 60.Svm classifier device with training is tested.Use single tagsort result's probability statistics as shown in Figure 6.Use the probability statistics of the many Feature Fusion method gained that proposes in this paper algorithm as shown in Figure 6.Can find out through Fig. 6, except BASE, BEACH, GOLF, RIV, S RE class, the nicety of grading of other types all maximum nicety of grading than single characteristic be high, and this explain that to each characteristic image of its wrong branch is not identical, but intersection is arranged.Adopt the method for many Feature Fusion to obtain the most believable classification results according to the notable feature self-adaptation of image.And BASE, BEACH, GOLF, RIV, S RE class owing to the notable feature of individual image has been divided its mistake, thereby causes final classification error, has reduced total nicety of grading.But these reductions are to be based upon on the obvious lifting basis of other niceties of grading of most of types, so can tolerate.And see from the population mean nicety of grading of image, reached 89.0476%, to exceed 10 percentage points than the mxm. 78.33% of single tagsort precision.
At document [Peng Zhenglin, hair XianCheng, Liu Wenyi; He Meixiang. based on the classification of remote-sensing images method research [J] of Combination of Multiple Classifiers. land resources remote sensing. 2011; 89 (2): 19-25.] and document [Sun Hao, Du Peijun, Zhao Weichang. based on the high-resolution remote sensing image Target Recognition [J] of Combination of Multiple Classifiers. geographical and geography information science. 2009; 25 (1): 32-35.] in; The author proposes several frequently seen Combination of Multiple Classifiers method and is used for remote sensing image classification, votes method, maximum probability classification method and fuzzy integral diagnostic method etc.; This paper selects the weighted comprehensive method, and three kinds of Combination of Multiple Classifiers methods of majority vote method and method in full accord and this paper algorithm compare.The result is as shown in Figure 7.As can beappreciated from fig. 7, weighted comprehensive method nicety of grading is 88.08%, and is approaching with the result of the inventive method, but owing to its weight to the Gabor characteristic has reduced, so precision slightly reduces.And the overall nicety of grading of at first definite each sorter of this method requirement, can increase calculated amount.Majority vote method and method in full accord are starkly lower than this paper method, have only 81.4286% and 52.8571% respectively.
The present invention has been owing to adopted improved speech bag model, and fully utilized the various features of image, so calculated amount is bigger, but along with being showing improvement or progress day by day of science and technology, the price of hardware is more and more lower, and bigger calculated amount can remedy through the lifting of hardware.And at ask for something than higher application, whether effectively higher nicety of grading is to confirm system key factor, so the present invention has bigger using value.
Claims (9)
1. a Classifying Method in Remote Sensing Image based on multi-feature fusion is characterized in that, this method may further comprise the steps:
Steps A, extract visual word bag characteristic, color histogram characteristic and the textural characteristics of training set remote sensing images respectively;
Step B, utilize visual word bag characteristic, color histogram characteristic and the textural characteristics of training set remote sensing images to carry out the SVMs training respectively, obtain three different support vector machine classifiers;
Step C, for the test sample book of the unknown; Extract its visual word bag characteristic, color histogram characteristic and textural characteristics respectively; And utilize among the step B resulting corresponding support vector machine classifier to carry out the classification prediction; Obtain three categories and do not predict the outcome, adopt the weighted comprehensive method that three categories are not predicted the outcome then and carry out comprehensively, thereby obtain final classification results.
2. Classifying Method in Remote Sensing Image based on multi-feature fusion according to claim 1 is characterized in that, said extraction visual word bag is represented, is meant and adopts improved speech bag model, extracts the visual word bag of weighting and representes; Specifically may further comprise the steps:
The local feature of each image is put into a file with the local feature of all images in step 1, the extraction training set, forms the local feature point set;
Step 2, utilize vector quantization method that local feature point set is carried out cluster, each cluster centre is represented a sight word, and all cluster centres form sight word storehouses;
Step 3, the local feature of each image is distributed to sight word nearest with it in the visual vocabulary storehouse, add up the corresponding word frequency of each sight word then, form the visual word bag characteristic of image;
Step 4, add up the addition summation of the word frequency of each sight word in the visual word bag characteristic of each type scene all images, thus obtain one type of scene overall visual speech bag represent;
Step 5, from the statistics of step 4, take out the bigger top n sight word of word frequency during promptly the overall visual speech bag of type scene is represented, N for preset greater than 1 integer; Word frequency to this N sight word is carried out normalization, and representes the weighted value that each sight word is corresponding through the inverse of its word frequency, and its value is respectively w
i, 0<i<n+1, i are integer; These weighted values are carried out normalization, make its weighted value between 0 to 1; Then each weighted value is added 1 computing, i.e. w
i=w
i+ 1;
The weighted value w of step 6, top n word that the word frequency during the total pronouns, general term for nouns, numerals and measure words bag of class of classification under the remote sensing images represented is bigger
iMultiply by weighted value w
iSight word word frequency in the pairing image vision speech bag characteristic so just obtains the visual word bag characteristic of the weighting of this width of cloth image.
3. like the said Classifying Method in Remote Sensing Image based on multi-feature fusion of claim 2, it is characterized in that said local feature is the SIFT characteristic.
4. like the said Classifying Method in Remote Sensing Image based on multi-feature fusion of claim 2, it is characterized in that the value of said N is 20.
5. like the said Classifying Method in Remote Sensing Image based on multi-feature fusion of claim 2, it is characterized in that said vector quantization method is that K all refers to clustering method.
6. Classifying Method in Remote Sensing Image based on multi-feature fusion according to claim 1 is characterized in that, color histogram described in the steps A is characterized as and adopts 512 color histograms that bin quantizes.
7. Classifying Method in Remote Sensing Image based on multi-feature fusion according to claim 1 is characterized in that, said textural characteristics is 5 yardsticks, the Gabor textural characteristics of 6 directions.
8. Classifying Method in Remote Sensing Image based on multi-feature fusion according to claim 1; It is characterized in that; When visual word bag characteristic and color histogram characteristic are carried out the SVMs training; Adopt histogram intersection nuclear as kernel function, when textural characteristics is carried out the SVMs training, adopt RBF as kernel function.
9. Classifying Method in Remote Sensing Image based on multi-feature fusion according to claim 1; It is characterized in that; Adopt the weighted comprehensive method that three group categories results are carried out comprehensively described in the step C; Specifically according to following method: calculate test pattern visual word bag characteristic, the classification of color histogram characteristic and the pairing support vector machine classifier of textural characteristics predicts the outcome, and comprises classification prediction label and classification prediction probability value; Get first of three support vector machine classifiers and predict the outcome, and corresponding three classification prediction probabilities are sorted, obtain maximum classification prediction probability value, inferior big classification prediction probability value and minimum classification prediction probability value; If inferior big classification prediction probability value corresponding class prediction label is identical with the pairing classification prediction of minimum classification prediction probability value label; And their classification prediction probability value sum is then predicted the final classification results of label as many Feature Fusion with the classification of its representative greater than maximum classification prediction probability value; Otherwise, with the final classification results of the pairing classification prediction of maximum classification prediction probability value label as many Feature Fusion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210043064 CN102622607B (en) | 2012-02-24 | 2012-02-24 | Remote sensing image classification method based on multi-feature fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210043064 CN102622607B (en) | 2012-02-24 | 2012-02-24 | Remote sensing image classification method based on multi-feature fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102622607A true CN102622607A (en) | 2012-08-01 |
| CN102622607B CN102622607B (en) | 2013-09-25 |
Family
ID=46562517
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201210043064 Expired - Fee Related CN102622607B (en) | 2012-02-24 | 2012-02-24 | Remote sensing image classification method based on multi-feature fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102622607B (en) |
Cited By (57)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102867194A (en) * | 2012-09-14 | 2013-01-09 | 中国科学院对地观测与数字地球科学中心 | Method and device for detecting remote-sensing inversion accuracy |
| CN103049764A (en) * | 2012-12-13 | 2013-04-17 | 中国科学院上海微系统与信息技术研究所 | Low-altitude aircraft target identification method |
| CN103489005A (en) * | 2013-09-30 | 2014-01-01 | 河海大学 | High-resolution remote sensing image classifying method based on fusion of multiple classifiers |
| CN103489004A (en) * | 2013-09-30 | 2014-01-01 | 华南理工大学 | Method for achieving large category image identification of deep study network |
| CN103679676A (en) * | 2013-12-02 | 2014-03-26 | 西北工业大学 | Quick unordered image stitching method based on multi-level word bag clustering |
| CN103902605A (en) * | 2012-12-28 | 2014-07-02 | 重庆凯泽科技有限公司 | Compromise feature quantification method |
| CN103985130A (en) * | 2014-05-27 | 2014-08-13 | 华东理工大学 | Image significance analysis method for complex texture images |
| CN104036239A (en) * | 2014-05-29 | 2014-09-10 | 西安电子科技大学 | Fast high-resolution SAR (synthetic aperture radar) image ship detection method based on feature fusion and clustering |
| CN104134076A (en) * | 2014-07-10 | 2014-11-05 | 杭州电子科技大学 | SAR image target recognition method based on CS and SVM decision fusion |
| CN104680173A (en) * | 2015-01-26 | 2015-06-03 | 河海大学 | Scene classification method for remote sensing images |
| CN104751170A (en) * | 2015-03-05 | 2015-07-01 | 同济大学 | Method of estimating adaptability of heterologous radar image based on learning monitoring strategy |
| CN104778475A (en) * | 2015-03-30 | 2015-07-15 | 南京邮电大学 | Image classification method based on maximum frequent visual word of annular region |
| CN104966090A (en) * | 2015-07-21 | 2015-10-07 | 公安部第三研究所 | Visual word generation and evaluation system and method for realizing image comprehension |
| CN104992180A (en) * | 2015-06-26 | 2015-10-21 | 武汉大学 | Multi-feature fusion car logo recognition method and system for traffic tollgates |
| CN105005786A (en) * | 2015-06-19 | 2015-10-28 | 南京航空航天大学 | Texture image classification method based on BoF and multi-feature fusion |
| CN105005789A (en) * | 2015-07-01 | 2015-10-28 | 北京理工大学 | Vision lexicon based remote sensing image terrain classification method |
| CN105069463A (en) * | 2015-07-17 | 2015-11-18 | 重庆交通大学 | Object-oriented multiple scale mountainous city land coverage information obtaining method |
| CN105118048A (en) * | 2015-07-17 | 2015-12-02 | 北京旷视科技有限公司 | Method and device for identifying copying certificate image |
| CN105740824A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | Densely populated area recognition method with feature interaction resistance |
| CN105894030A (en) * | 2016-04-01 | 2016-08-24 | 河海大学 | High-resolution remote sensing image scene classification method based on layered multi-characteristic fusion |
| CN105930497A (en) * | 2016-05-06 | 2016-09-07 | 浙江工业大学 | Image edge and line feature based three-dimensional model retrieval method |
| CN106021330A (en) * | 2016-05-06 | 2016-10-12 | 浙江工业大学 | A three-dimensional model retrieval method used for mixed contour line views |
| CN106383888A (en) * | 2016-09-22 | 2017-02-08 | 深圳市唯特视科技有限公司 | Method for positioning and navigation by use of picture retrieval |
| CN106529583A (en) * | 2016-11-01 | 2017-03-22 | 哈尔滨工程大学 | Bag-of-visual-word-model-based indoor scene cognitive method |
| CN106570514A (en) * | 2016-05-08 | 2017-04-19 | 扬州西岐自动化科技有限公司 | Automobile wheel hub classification method based on word bag model and support vector machine |
| CN106874921A (en) * | 2015-12-11 | 2017-06-20 | 清华大学 | Image classification method and device |
| CN107067007A (en) * | 2016-12-22 | 2017-08-18 | 河海大学 | A kind of multiple features fusion crop straw burning fire detection method based on image characteristics extraction |
| CN107092871A (en) * | 2017-04-06 | 2017-08-25 | 重庆市地理信息中心 | Remote sensing image building detection method based on multiple dimensioned multiple features fusion |
| CN107292339A (en) * | 2017-06-16 | 2017-10-24 | 重庆大学 | The unmanned plane low altitude remote sensing image high score Geomorphological Classification method of feature based fusion |
| CN107341508A (en) * | 2017-06-22 | 2017-11-10 | 北京飞搜科技有限公司 | A kind of quick cuisines image identification method and system |
| CN107480620A (en) * | 2017-08-04 | 2017-12-15 | 河海大学 | Remote sensing images automatic target recognition method based on heterogeneous characteristic fusion |
| CN108009559A (en) * | 2016-11-02 | 2018-05-08 | 哈尔滨工业大学 | A kind of Hyperspectral data classification method based on empty spectrum united information |
| CN108053406A (en) * | 2017-09-11 | 2018-05-18 | 同济大学 | Ground mulching drafting method and device based on multiple-spatial resolution remote sensing image |
| CN108154107A (en) * | 2017-12-22 | 2018-06-12 | 北京航空航天大学 | A kind of method of the scene type of determining remote sensing images ownership |
| CN108197652A (en) * | 2018-01-02 | 2018-06-22 | 百度在线网络技术（北京）有限公司 | For generating the method and apparatus of information |
| CN108604303A (en) * | 2016-02-09 | 2018-09-28 | 赫尔实验室有限公司 | General image feature from bottom to top and the from top to bottom system and method for entity classification are merged for precise image/video scene classification |
| CN108629494A (en) * | 2018-04-19 | 2018-10-09 | 三峡大学 | Arid grade appraisal procedure and system |
| CN108710915A (en) * | 2018-05-22 | 2018-10-26 | 合肥工业大学 | Multiple features fusion gastroscope image processing method based on Multiple Kernel Learning |
| CN108764302A (en) * | 2018-05-08 | 2018-11-06 | 中山大学 | A kind of bill images sorting technique based on color characteristic and bag of words feature |
| CN108921205A (en) * | 2018-06-14 | 2018-11-30 | 浙江大学 | A kind of skin disease clinical image classification method based on multi-feature fusion |
| CN109325434A (en) * | 2018-09-15 | 2019-02-12 | 天津大学 | A kind of image scene classification method of the probability topic model of multiple features |
| CN109858577A (en) * | 2019-04-01 | 2019-06-07 | 盐城工学院 | Soybean appearance quality detection device and detection method |
| CN110490062A (en) * | 2019-07-11 | 2019-11-22 | 武汉科技大学 | A kind of bimanual input recognition methods based on multiple view |
| CN110827231A (en) * | 2019-09-19 | 2020-02-21 | 北京航空航天大学 | Active learning multi-source image fusion method under saliency driving |
| CN110956190A (en) * | 2018-09-27 | 2020-04-03 | 深圳云天励飞技术有限公司 | Image recognition method and device, computer device and computer readable storage medium |
| CN111027600A (en) * | 2019-11-25 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Image category prediction method and device |
| CN111160373A (en) * | 2019-12-30 | 2020-05-15 | 重庆邮电大学 | Method for extracting, detecting and classifying defect image features of variable speed drum parts |
| CN111582146A (en) * | 2020-05-06 | 2020-08-25 | 宁波大学 | High-resolution remote sensing image city function partitioning method based on multi-feature fusion |
| CN111965197A (en) * | 2020-07-23 | 2020-11-20 | 广东工业大学 | Defect classification method based on multi-feature fusion |
| CN112215066A (en) * | 2020-09-08 | 2021-01-12 | 北京农业信息技术研究中心 | Livestock face image recognition method and device |
| CN112287871A (en) * | 2020-11-12 | 2021-01-29 | 广东海洋大学 | Near-shore aquaculture area remote sensing image extraction method based on multi-feature and spectrum fusion |
| CN112329798A (en) * | 2020-11-27 | 2021-02-05 | 重庆理工大学 | Image scene classification method based on optimized visual bag-of-words model |
| CN112561935A (en) * | 2020-12-26 | 2021-03-26 | 广东工业大学 | Method, device and equipment for identifying Alzheimer's disease |
| CN112633142A (en) * | 2020-12-21 | 2021-04-09 | 广东电网有限责任公司电力科学研究院 | Power transmission line violation building identification method and related device |
| CN113567953A (en) * | 2021-07-28 | 2021-10-29 | 哈尔滨工业大学 | Full-waveform laser echo signal classification method based on SIFT visual word bag |
| CN113743523A (en) * | 2021-09-13 | 2021-12-03 | 西安建筑科技大学 | Visual multi-feature guided construction waste fine classification method |
| CN116720984A (en) * | 2023-08-11 | 2023-09-08 | 山东昊能电力建设有限公司 | Intelligent electric power auditing method based on satellite map technology |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101315663A (en) * | 2008-06-25 | 2008-12-03 | 中国人民解放军国防科学技术大学 | Nature scene image classification method based on area dormant semantic characteristic |
| CN101807256A (en) * | 2010-03-29 | 2010-08-18 | 天津大学 | Object identification detection method based on multiresolution frame |
| CN101853304A (en) * | 2010-06-08 | 2010-10-06 | 河海大学 | Remote sensing image retrieval method based on feature selection and semi-supervised learning |
-
2012
- 2012-02-24 CN CN 201210043064 patent/CN102622607B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101315663A (en) * | 2008-06-25 | 2008-12-03 | 中国人民解放军国防科学技术大学 | Nature scene image classification method based on area dormant semantic characteristic |
| CN101807256A (en) * | 2010-03-29 | 2010-08-18 | 天津大学 | Object identification detection method based on multiresolution frame |
| CN101853304A (en) * | 2010-06-08 | 2010-10-06 | 河海大学 | Remote sensing image retrieval method based on feature selection and semi-supervised learning |
Non-Patent Citations (2)
| Title |
|---|
| 王宇新等: "用于图像场景分类的空间视觉词袋模型", 《计算机科学》, vol. 38, no. 8, 31 August 2011 (2011-08-31), pages 265 - 268 * |
| 陈凯等: "单尺度词袋模型图像分类方法", 《计算机应用研究》, vol. 28, no. 10, 31 October 2011 (2011-10-31), pages 3986 - 3988 * |
Cited By (84)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102867194A (en) * | 2012-09-14 | 2013-01-09 | 中国科学院对地观测与数字地球科学中心 | Method and device for detecting remote-sensing inversion accuracy |
| CN102867194B (en) * | 2012-09-14 | 2015-08-05 | 中国科学院对地观测与数字地球科学中心 | A kind of remote sensing inversion accuracy detection method and device |
| CN103049764A (en) * | 2012-12-13 | 2013-04-17 | 中国科学院上海微系统与信息技术研究所 | Low-altitude aircraft target identification method |
| CN103902605A (en) * | 2012-12-28 | 2014-07-02 | 重庆凯泽科技有限公司 | Compromise feature quantification method |
| CN103489005B (en) * | 2013-09-30 | 2017-04-05 | 河海大学 | A kind of Classification of High Resolution Satellite Images method based on multiple Classifiers Combination |
| CN103489005A (en) * | 2013-09-30 | 2014-01-01 | 河海大学 | High-resolution remote sensing image classifying method based on fusion of multiple classifiers |
| CN103489004A (en) * | 2013-09-30 | 2014-01-01 | 华南理工大学 | Method for achieving large category image identification of deep study network |
| CN103679676A (en) * | 2013-12-02 | 2014-03-26 | 西北工业大学 | Quick unordered image stitching method based on multi-level word bag clustering |
| CN103985130A (en) * | 2014-05-27 | 2014-08-13 | 华东理工大学 | Image significance analysis method for complex texture images |
| CN103985130B (en) * | 2014-05-27 | 2017-07-11 | 华东理工大学 | A kind of saliency analysis method for complex texture image |
| CN104036239A (en) * | 2014-05-29 | 2014-09-10 | 西安电子科技大学 | Fast high-resolution SAR (synthetic aperture radar) image ship detection method based on feature fusion and clustering |
| CN104134076A (en) * | 2014-07-10 | 2014-11-05 | 杭州电子科技大学 | SAR image target recognition method based on CS and SVM decision fusion |
| CN104134076B (en) * | 2014-07-10 | 2017-06-06 | 杭州电子科技大学 | SAR image target recognition method based on CS and SVM decision level fusions |
| CN104680173A (en) * | 2015-01-26 | 2015-06-03 | 河海大学 | Scene classification method for remote sensing images |
| CN104680173B (en) * | 2015-01-26 | 2018-04-27 | 河海大学 | A kind of remote sensing images scene classification method |
| CN104751170A (en) * | 2015-03-05 | 2015-07-01 | 同济大学 | Method of estimating adaptability of heterologous radar image based on learning monitoring strategy |
| CN104778475B (en) * | 2015-03-30 | 2018-01-19 | 南京邮电大学 | A kind of image classification method based on annular region Maximum Frequent vision word |
| CN104778475A (en) * | 2015-03-30 | 2015-07-15 | 南京邮电大学 | Image classification method based on maximum frequent visual word of annular region |
| CN105005786A (en) * | 2015-06-19 | 2015-10-28 | 南京航空航天大学 | Texture image classification method based on BoF and multi-feature fusion |
| CN105005786B (en) * | 2015-06-19 | 2018-10-16 | 南京航空航天大学 | A kind of texture image classification method based on BoF and multiple features fusion |
| CN104992180A (en) * | 2015-06-26 | 2015-10-21 | 武汉大学 | Multi-feature fusion car logo recognition method and system for traffic tollgates |
| CN104992180B (en) * | 2015-06-26 | 2019-01-29 | 武汉大学 | A kind of multiple features fusion automobile logo identification method and system towards traffic block port |
| CN105005789A (en) * | 2015-07-01 | 2015-10-28 | 北京理工大学 | Vision lexicon based remote sensing image terrain classification method |
| CN105005789B (en) * | 2015-07-01 | 2018-04-03 | 北京理工大学 | A kind of remote sensing images terrain classification method of view-based access control model vocabulary |
| CN105118048A (en) * | 2015-07-17 | 2015-12-02 | 北京旷视科技有限公司 | Method and device for identifying copying certificate image |
| CN105118048B (en) * | 2015-07-17 | 2018-03-27 | 北京旷视科技有限公司 | The recognition methods of reproduction certificate picture and device |
| CN105069463A (en) * | 2015-07-17 | 2015-11-18 | 重庆交通大学 | Object-oriented multiple scale mountainous city land coverage information obtaining method |
| CN104966090A (en) * | 2015-07-21 | 2015-10-07 | 公安部第三研究所 | Visual word generation and evaluation system and method for realizing image comprehension |
| CN104966090B (en) * | 2015-07-21 | 2018-08-14 | 公安部第三研究所 | Realize the system and method that the vision word towards image understanding is generated and evaluated |
| CN106874921A (en) * | 2015-12-11 | 2017-06-20 | 清华大学 | Image classification method and device |
| CN105740824A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | Densely populated area recognition method with feature interaction resistance |
| US11423651B2 (en) * | 2016-02-09 | 2022-08-23 | Hrl Laboratories, Llc | System and method for the fusion of bottom-up whole-image features and top-down enttiy classification for accurate image/video scene classification |
| CN108604303A (en) * | 2016-02-09 | 2018-09-28 | 赫尔实验室有限公司 | General image feature from bottom to top and the from top to bottom system and method for entity classification are merged for precise image/video scene classification |
| CN105894030A (en) * | 2016-04-01 | 2016-08-24 | 河海大学 | High-resolution remote sensing image scene classification method based on layered multi-characteristic fusion |
| CN105894030B (en) * | 2016-04-01 | 2019-02-19 | 河海大学 | High-resolution remote sensing image scene classification method based on layering multiple features fusion |
| CN105930497B (en) * | 2016-05-06 | 2019-05-24 | 浙江工业大学 | Method for searching three-dimension model based on image border and linear feature |
| CN106021330A (en) * | 2016-05-06 | 2016-10-12 | 浙江工业大学 | A three-dimensional model retrieval method used for mixed contour line views |
| CN106021330B (en) * | 2016-05-06 | 2019-06-25 | 浙江工业大学 | A kind of method for searching three-dimension model for mixing under contour line view |
| CN105930497A (en) * | 2016-05-06 | 2016-09-07 | 浙江工业大学 | Image edge and line feature based three-dimensional model retrieval method |
| CN106570514A (en) * | 2016-05-08 | 2017-04-19 | 扬州西岐自动化科技有限公司 | Automobile wheel hub classification method based on word bag model and support vector machine |
| CN106383888A (en) * | 2016-09-22 | 2017-02-08 | 深圳市唯特视科技有限公司 | Method for positioning and navigation by use of picture retrieval |
| CN106529583A (en) * | 2016-11-01 | 2017-03-22 | 哈尔滨工程大学 | Bag-of-visual-word-model-based indoor scene cognitive method |
| CN108009559B (en) * | 2016-11-02 | 2021-12-24 | 哈尔滨工业大学 | Hyperspectral data classification method based on space-spectrum combined information |
| CN108009559A (en) * | 2016-11-02 | 2018-05-08 | 哈尔滨工业大学 | A kind of Hyperspectral data classification method based on empty spectrum united information |
| CN107067007A (en) * | 2016-12-22 | 2017-08-18 | 河海大学 | A kind of multiple features fusion crop straw burning fire detection method based on image characteristics extraction |
| CN107092871A (en) * | 2017-04-06 | 2017-08-25 | 重庆市地理信息中心 | Remote sensing image building detection method based on multiple dimensioned multiple features fusion |
| CN107292339A (en) * | 2017-06-16 | 2017-10-24 | 重庆大学 | The unmanned plane low altitude remote sensing image high score Geomorphological Classification method of feature based fusion |
| CN107341508A (en) * | 2017-06-22 | 2017-11-10 | 北京飞搜科技有限公司 | A kind of quick cuisines image identification method and system |
| CN107480620A (en) * | 2017-08-04 | 2017-12-15 | 河海大学 | Remote sensing images automatic target recognition method based on heterogeneous characteristic fusion |
| CN107480620B (en) * | 2017-08-04 | 2020-06-16 | 河海大学 | Remote sensing image automatic target identification method based on heterogeneous feature fusion |
| CN108053406A (en) * | 2017-09-11 | 2018-05-18 | 同济大学 | Ground mulching drafting method and device based on multiple-spatial resolution remote sensing image |
| CN108053406B (en) * | 2017-09-11 | 2021-05-11 | 同济大学 | Surface coverage mapping method and device based on multi-resolution remote sensing image |
| CN108154107A (en) * | 2017-12-22 | 2018-06-12 | 北京航空航天大学 | A kind of method of the scene type of determining remote sensing images ownership |
| CN108154107B (en) * | 2017-12-22 | 2021-09-14 | 北京航空航天大学 | Method for determining scene category to which remote sensing image belongs |
| CN108197652A (en) * | 2018-01-02 | 2018-06-22 | 百度在线网络技术（北京）有限公司 | For generating the method and apparatus of information |
| CN108629494A (en) * | 2018-04-19 | 2018-10-09 | 三峡大学 | Arid grade appraisal procedure and system |
| CN108629494B (en) * | 2018-04-19 | 2020-11-06 | 三峡大学 | Drought level assessment method and system |
| CN108764302B (en) * | 2018-05-08 | 2021-09-28 | 中山大学 | Bill image classification method based on color features and bag-of-words features |
| CN108764302A (en) * | 2018-05-08 | 2018-11-06 | 中山大学 | A kind of bill images sorting technique based on color characteristic and bag of words feature |
| CN108710915A (en) * | 2018-05-22 | 2018-10-26 | 合肥工业大学 | Multiple features fusion gastroscope image processing method based on Multiple Kernel Learning |
| CN108710915B (en) * | 2018-05-22 | 2020-09-18 | 合肥工业大学 | Multi-feature fusion gastroscope image processing method based on multi-kernel learning |
| CN108921205A (en) * | 2018-06-14 | 2018-11-30 | 浙江大学 | A kind of skin disease clinical image classification method based on multi-feature fusion |
| CN109325434A (en) * | 2018-09-15 | 2019-02-12 | 天津大学 | A kind of image scene classification method of the probability topic model of multiple features |
| CN110956190A (en) * | 2018-09-27 | 2020-04-03 | 深圳云天励飞技术有限公司 | Image recognition method and device, computer device and computer readable storage medium |
| CN109858577A (en) * | 2019-04-01 | 2019-06-07 | 盐城工学院 | Soybean appearance quality detection device and detection method |
| CN110490062A (en) * | 2019-07-11 | 2019-11-22 | 武汉科技大学 | A kind of bimanual input recognition methods based on multiple view |
| CN110827231A (en) * | 2019-09-19 | 2020-02-21 | 北京航空航天大学 | Active learning multi-source image fusion method under saliency driving |
| CN111027600A (en) * | 2019-11-25 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Image category prediction method and device |
| CN111160373A (en) * | 2019-12-30 | 2020-05-15 | 重庆邮电大学 | Method for extracting, detecting and classifying defect image features of variable speed drum parts |
| CN111160373B (en) * | 2019-12-30 | 2022-07-15 | 重庆邮电大学 | Method for extracting, detecting and classifying defect image features of variable speed drum part |
| CN111582146B (en) * | 2020-05-06 | 2021-04-13 | 宁波大学 | High-resolution remote sensing image city function partitioning method based on multi-feature fusion |
| CN111582146A (en) * | 2020-05-06 | 2020-08-25 | 宁波大学 | High-resolution remote sensing image city function partitioning method based on multi-feature fusion |
| CN111965197A (en) * | 2020-07-23 | 2020-11-20 | 广东工业大学 | Defect classification method based on multi-feature fusion |
| CN112215066A (en) * | 2020-09-08 | 2021-01-12 | 北京农业信息技术研究中心 | Livestock face image recognition method and device |
| CN112287871B (en) * | 2020-11-12 | 2023-01-17 | 广东海洋大学 | Near-shore aquaculture area remote sensing image extraction method based on multi-feature and spectrum fusion |
| CN112287871A (en) * | 2020-11-12 | 2021-01-29 | 广东海洋大学 | Near-shore aquaculture area remote sensing image extraction method based on multi-feature and spectrum fusion |
| CN112329798A (en) * | 2020-11-27 | 2021-02-05 | 重庆理工大学 | Image scene classification method based on optimized visual bag-of-words model |
| CN112329798B (en) * | 2020-11-27 | 2023-07-25 | 重庆理工大学 | Image scene classification method based on optimized visual word bag model |
| CN112633142A (en) * | 2020-12-21 | 2021-04-09 | 广东电网有限责任公司电力科学研究院 | Power transmission line violation building identification method and related device |
| CN112561935A (en) * | 2020-12-26 | 2021-03-26 | 广东工业大学 | Method, device and equipment for identifying Alzheimer's disease |
| CN113567953A (en) * | 2021-07-28 | 2021-10-29 | 哈尔滨工业大学 | Full-waveform laser echo signal classification method based on SIFT visual word bag |
| CN113743523A (en) * | 2021-09-13 | 2021-12-03 | 西安建筑科技大学 | Visual multi-feature guided construction waste fine classification method |
| CN116720984A (en) * | 2023-08-11 | 2023-09-08 | 山东昊能电力建设有限公司 | Intelligent electric power auditing method based on satellite map technology |
| CN116720984B (en) * | 2023-08-11 | 2023-12-01 | 山东昊能电力建设有限公司 | Intelligent electric power auditing method based on satellite map technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102622607B (en) | 2013-09-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102622607B (en) | Remote sensing image classification method based on multi-feature fusion | |
| Zhu et al. | Bag-of-visual-words scene classifier with local and global features for high spatial resolution remote sensing imagery | |
| Sirmacek et al. | Urban-area and building detection using SIFT keypoints and graph theory | |
| CN103942564B (en) | High-resolution remote sensing image scene classifying method based on unsupervised feature learning | |
| CN102496034B (en) | High-spatial resolution remote-sensing image bag-of-word classification method based on linear words | |
| CN101976258B (en) | Video semantic extraction method by combining object segmentation and feature weighing | |
| Luo et al. | Learning high-level features by fusing multi-view representation of MLS point clouds for 3D object recognition in road environments | |
| CN104680173B (en) | A kind of remote sensing images scene classification method | |
| CN105528794A (en) | Moving object detection method based on Gaussian mixture model and superpixel segmentation | |
| CN102054178A (en) | Chinese painting image identifying method based on local semantic concept | |
| CN105894030B (en) | High-resolution remote sensing image scene classification method based on layering multiple features fusion | |
| CN104268528A (en) | Method and device for detecting crowd gathered region | |
| CN109902618A (en) | A kind of sea ship recognition methods and device | |
| Jiang et al. | Focal-test-based spatial decision tree learning: A summary of results | |
| Zang et al. | Traffic lane detection using fully convolutional neural network | |
| Sun et al. | A multi-level convolution pyramid semantic fusion framework for high-resolution remote sensing image scene classification and annotation | |
| Wang et al. | Remote sensing scene classification using heterogeneous feature extraction and multi-level fusion | |
| KR102119136B1 (en) | Method for intelligent image classification | |
| CN105069459A (en) | Surface feature type extracting method for high-resolution SAR image | |
| Al-Ghrairi et al. | Classification of satellite images based on color features using remote sensing | |
| CN112966629A (en) | Remote sensing image scene classification method based on image transformation and BoF model | |
| Yan et al. | Graph-based semi-supervised classification on very high resolution remote sensing images | |
| CN106951873A (en) | A kind of Remote Sensing Target recognition methods | |
| Yang et al. | Supervised land-cover classification of TerraSAR-X imagery over urban areas using extremely randomized clustering forests | |
| Chu et al. | Scene classification based on SIFT combined with GIST |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130925 |