CN103294792A - Polarimetric SAR (synthetic aperture radar) terrain classification method based on semantic information and polarimetric decomposition - Google Patents
Polarimetric SAR (synthetic aperture radar) terrain classification method based on semantic information and polarimetric decomposition Download PDFInfo
- Publication number
- CN103294792A CN103294792A CN2013101920571A CN201310192057A CN103294792A CN 103294792 A CN103294792 A CN 103294792A CN 2013101920571 A CN2013101920571 A CN 2013101920571A CN 201310192057 A CN201310192057 A CN 201310192057A CN 103294792 A CN103294792 A CN 103294792A
- Authority
- CN
- China
- Prior art keywords
- line segment
- aggregation
- region
- semantic information
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 69
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 23
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000003709 image segmentation Methods 0.000 claims abstract description 10
- 230000002776 aggregation Effects 0.000 claims description 89
- 238000004220 aggregation Methods 0.000 claims description 89
- 230000010287 polarization Effects 0.000 claims description 41
- 230000011218 segmentation Effects 0.000 claims description 23
- 238000009826 distribution Methods 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 17
- 238000004458 analytical method Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 6
- 230000004931 aggregating effect Effects 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 9
- 238000010586 diagram Methods 0.000 description 14
- 238000004088 simulation Methods 0.000 description 12
- 239000011159 matrix material Substances 0.000 description 7
- 230000007547 defect Effects 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 230000004044 response Effects 0.000 description 4
- 241000282414 Homo sapiens Species 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 239000006002 Pepper Substances 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005530 etching Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000013179 statistical model Methods 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a polarimetric SAR (synthetic aperture radar) terrain classification method based on semantic information and polarimetric decomposition. The polarimetric SAR terrain classification method includes performing mean shift on a span image, extracting a sketch map of the span image, extracting a line segment gathering region in the sketch map by an region extracting technology based on the semantic information, merging span image mean shift over-segmentation regions based on the line segment gathering region as well as by the aid of a critical region majority vote merging strategy and based on a polarimetric feature merging strategy, acquiring an image segmentation result, and fusing the image segmentation result based on the semantic information and an H/alpha-Wishart classification result based on an MRF (markov random field) to acquire a final classification result. The semantic information, an image processing technique and a polarimetric scattering property are combined by the polarimetric SAR terrain classification method, the problem of bad region consistency of existing classification techniques based on the polarimetric decomposition to classification results with gathering-featured surface features (like forest, building groups and the like) is mainly solved, and region consistency and boundary retainability of the classification results with the gathering-featured surface features are improved.
Description
Technical Field
The invention belongs to the technical field of image processing and remote sensing, relates to terrain classification of a polarized SAR image, and particularly relates to a method for classifying the terrain of the polarized SAR based on semantic information and polarization decomposition, which can be used for the terrain classification of a low-resolution polarized SAR image containing the terrain with aggregation characteristics.
Background
Polar Synthetic Aperture Radar (POLSAR) image processing is an important subject of national defense construction and economic development, and is concerned and researched by more and more people. Compared with a common single-polarization Synthetic Aperture Radar (SAR), the polarization SAR carries out full-polarization measurement, can obtain richer ground feature information of a target, and provides an important basis for more deeply researching the scattering characteristics of the target. The polarized SAR terrain classification is one of the important tasks of polarized SAR image processing and is a precondition for polarized SAR image interpretation. The key and difficulty of polarized SAR segmentation or terrain classification is the region consistency of the same terrain and the boundary maintenance between different terrain.
There are many methods for polarising SAR terrain classification, mainly classified into three types: 1) a statistical model-based classification method; 2) a classification method based on an electromagnetic wave scattering mechanism; 3) a classification method based on image processing techniques. The method based on the statistical model mainly comprises the following steps: lee et al, provide supervised polarimetric SAR classification according to the polarimetric covariance matrix satisfying complex wishart distribution. In practical applications, however, a priori knowledge about the SAR image class is very little. There are many methods based on electromagnetic wave scattering mechanism, and in 1997, Cloude et al first proposed an H/α classification method, which obtains a ground object scattering entropy H and an angle α representing the ground object scattering mechanism by decomposition, and realizes unsupervised polarized SAR image classification. In 1999, Lee et al introduced a Wishart classifier in combination with statistical distribution on the basis of the H/α classification method, and improved the classification accuracy by performing Wishart iteration on the results of the H/α classification method. In 2004, Lee et al have proposed a classification method for preserving polarization scattering characteristics, which initially classifies 3 polarization scattering mechanism components obtained by Freeman decomposition, and performs merging and class correction by Wishart iteration, thereby achieving a better classification effect.
The method well utilizes the scattering characteristic and polarization information of the polarized SAR data for classification, but the pixel-based classification method does not consider the visual characteristic of the polarized SAR image and does not combine a computer vision method and an image processing method for classification. Therefore, the conventional method for polarising SAR terrain classification, which includes the above-mentioned method, has many disadvantages: (1) the regions of the same ground object have poor consistency, and a salt-pepper noise type classification result graph is generated; (2) a polarized SAR terrain classification method based on a traditional image processing method is used for classifying terrain with alternate light and shade gray changes, for example, the traditional classification method based on pixel point and super-pixel combination is difficult to classify the terrain into one class; (3) for complex ground features such as building groups, because the ground features themselves comprise houses, roads and the like, the scattering characteristics of the ground features are not single, the ground features with alternate light and shade have scattering characteristics, and the ground features are difficult to be well divided into a complete region. Therefore, the extraction of the bottom-level features has difficulty in well grouping such surface features together, and high-level features based on the surface feature need to be further mined for classification.
In summary, the above polarized SAR terrain classification methods have fine pixel classification, but still have some defects, especially for terrain with aggregation characteristics (such as building groups, forests, etc.), because the scattering of terrain itself is not single, the terrain has scattering characteristics of alternate light and dark terrain, the classification region has poor consistency, and the boundary is susceptible to noise, so that a pepper-salt classification result is easily generated.
Disclosure of Invention
The invention aims to overcome the defects of the existing method and provides a polarized SAR terrain classification method based on semantic information and polarization decomposition.
The invention relates to a polarized SAR terrain classification method based on semantic information and polarization decomposition, which carries out unsupervised terrain classification aiming at a low-resolution polarized SAR image acquired in advance, and the classification process comprises the following steps:
And 2, carrying out semantic information analysis on the line segments in the Sketch Map, and endowing semantic information, namely two-side aggregation, one-side aggregation and isolated line segments, to the line segments according to the statistical distribution of the line segment aggregation characteristics.
And 3, in the Sketch Map, extracting a plurality of disjoint aggregation line segment sets by adopting a line segment set solving algorithm according to semantic information given to the line segments, and obtaining a line segment aggregation region R by adopting a region extraction method for each aggregation line segment set.
And 4, carrying out region merging on the over-segmentation result: performing a critical area mode voting merging strategy on the over-segmentation result graph of the span graph obtained in the step 1 and the over-segmentation area corresponding to the line segment gathering area R; extracting over-segmentation areas where the isolated line segments are located, and adopting a non-merging strategy; and for other regions, namely the residual regions, a region merging strategy based on polarization characteristics is adopted to obtain a polarized SAR image segmentation result based on semantic information.
And 5, carrying out H/alpha-Wishare classification on the polarized SAR data by utilizing polarization decomposition, and carrying out neighborhood optimization on the H/alpha-Wishare classification result by using a Markov Random Field (MRF).
And 6, fusing the polarized SAR image segmentation result based on the semantic information and the H/alpha-Wishart classification result based on the MRF through a mode voting (majauthority vote) strategy to obtain a final classification result of the ground feature classification of the polarized SAR image to be classified.
The key technology for realizing the invention is as follows: aiming at the problem of poor consistency of the classified areas of ground objects (such as building groups, forests and the like) with aggregation, the analysis shows that the low-resolution polarization SAR images generally comprise farmlands, urban areas, forests, mountains, bridges and the like, the structural line segments of the building groups are aggregated and distributed in a spherical shape according to the prior knowledge of human beings, the structural line segments of the bridges are distributed in a linear shape and the like, the cognition is used as the prior knowledge, the semantic information contained in the line segments is analyzed, and the semantic information of the line segments is endowed. By analyzing the semantic information of the line segments, a line segment gathering area can be extracted, the line segment gathering area corresponds to ground objects such as building groups, forests and the like in the image, a consistent area of the ground objects is obtained by extracting the line segment gathering area, the over-segmented image can be divided into a line segment gathering area, an isolated line segment area and a wireless segment area according to the semantic information analysis of the line segments, the line segment gathering area corresponds to the ground objects such as the building groups and the like, the isolated line segment area corresponds to line targets and the like, and the wireless segment area generally corresponds to the ground objects such as oceans, farmlands and the like. The problem of the classification region uniformity of ground objects with aggregation is poor is solved.
Compared with the prior art, the invention has the following advantages:
1. from the aspect of semantic information analysis, the Sketch Map of the span graph is obtained by using the Primal Sketch sparse representation model, the semantic information contained in the line segment is analyzed according to the Sketch Map, an area division technology based on line segment semantic information analysis is provided, and the line segment aggregation area is effectively extracted on the Sketch Map. These line segment concentration areas correspond to terrain such as urban areas, forests, etc. in the polarized SAR image. The ground objects are often classified into a plurality of types due to the gray level change between light and shade, the defect is overcome well, and the region consistency of line segment gathering region classification is effectively improved.
2. In the aspect of image processing technology, when the mean value drift over-segmentation areas are combined, different combining strategies are adopted for different types of ground feature areas, the area combination is more targeted, the different types of areas can be well combined, and the segmentation result based on the semantic information is obtained.
3. From polarization decomposition, the invention uses H/alpha-Wisharp classification, uses MRF to perform neighborhood optimization to obtain a pixel-level classification result, finally fuses segmentation and classification results, uses the segmentation region to guide the region consistency of classification, and simultaneously helps the further combination of the segmentation regions, and the segmentation and classification interaction obtains a better classification result. The method combines the image processing technology and the electromagnetic wave scattering mechanism-based technology, integrates semantic information and polarization information, combines the semantic information, the image processing technology and the polarization scattering characteristic, and improves the region consistency and the boundary retentivity of the polarized SAR terrain classification result.
Drawings
FIG. 1 is a flow chart of the present invention for classifying the terrain of polarized SAR data;
FIG. 2 is a span plot of fully polarized San Francisco data for the NASA/JPL AIRSAR L band used in the present invention;
FIG. 3 is a graph of the over-segmentation results from the mean shift in the present invention;
FIG. 4 is a Sketch of a ridge obtained by the present invention, namely Sketch Map
FIG. 5 is a block diagram of a semantic information tree for segments in the present invention;
FIG. 6 is a sketch of a ridge assigned semantic information obtained using the present invention;
FIG. 7 is a schematic diagram of a line segment aggregation region extraction process in the present invention;
FIG. 8 is a diagram of line segment aggregation area extraction results based on semantic information analysis in the present invention;
FIG. 9 is a diagram illustrating the merging results of over-segmented regions corresponding to line segment aggregation regions in the present invention;
FIG. 10 is a graph of the result of image segmentation based on semantic information in the present invention;
FIG. 11 is a schematic diagram of the segmentation and classification result fusion process of the present invention;
FIG. 12 is a span plot of fully polarized San Francisco data for the NASA/JPL AIRSAR L band used in the present invention;
FIG. 13 is a diagram of the result of MRF-based H/α -Wishart classification in the present invention;
fig. 14 is a classification result diagram of the present invention.
Detailed Description
Example 1
The invention relates to a polarized SAR terrain classification method based on semantic information and polarization decomposition, which carries out unsupervised terrain classification aiming at a low-resolution polarized SAR image acquired in advance and refers to figure 1, and the classification process of the invention comprises the following steps:
Processing polarized SAR data to obtain a covariance matrix, obtaining amplitude values of three channels according to three values of diagonal elements of the covariance matrix, and fusing the amplitude values of the three channels to obtain a span image of a polarized SAR image. The first operation performed on the span map is to use mean shift to obtain an over-segmentation result map of the span map, as shown in fig. 3.
The second operation is to adopt an edge-ridge detection sparse coding method to extract the Sketch Map, and the extraction steps comprise:
first, a first order gaussian filter and a second order gaussian filter of N scales and M directions are constructed to form a filter bank. Wherein N takes the value of 3 and M takes the value of 18. As shown in fig. 2, the span image is convolved with the filter bank to obtain the joint response of each pixel, the maximum value of the joint response is extracted as the edge/ridge strength of the pixel, and the direction of the maximum response filter is taken as the local direction of the pixel. Carrying out non-maximum inhibition processing on the edge/ridge intensity map to obtain a suggested sketch
According to the suggested sketch
The position of the maximum joint response in the drawing, and the drawing of the suggestion
Connecting the points communicated with the position into a line segment to generate an edge/ridge original model Ssk，0；
Secondly, adding a new line segment in the edge model, evaluating the coding length gain delta L of the image, if delta L is less than epsilon, epsilon is a threshold value, taking the value as 10, refusing to accept the line segment, otherwise, accepting, searching, and suggesting a sketch
And taking a dividing line of the tail end of the new line segment and other pixels within the average fitting error as a next new suggested line segment, if the new suggested line segment exists, recalculating the image coding length gain delta L after adding the new suggested line segment, if the delta L is less than epsilon, refusing to accept the new suggested line segment, otherwise, accepting the new suggested line segment, iteratively adding the new line segment until no new suggested line segment exists, namely obtaining a ridge Sketch, and obtaining the Sketch Map as shown in FIG. 4.
And 2, analyzing semantic information of the line segments in the Sketch Map, and endowing the line segments with semantic information, namely two-side aggregation, one-side aggregation and isolated line segments according to the statistical distribution of the line segment aggregation characteristics.
2.1 for the low resolution polarized SAR image, for the ground object with the aggregation characteristic, taking a building group as an example, the line segment is formed by a bright building and a dark ground, and such a structure appears repeatedly, so that the building group is formed, the corresponding sketch line segment is generally dense in distribution, and the line segment direction is mostly approximately horizontal and vertical. For forest ground features, sketch line segments are also densely distributed, but the line segment directions are disorderly. For a bridge, its sketch segments are distributed in a manifold, etc. Therefore, the distribution structure of the line segments contains certain semantic information, and the line segments mainly correspond to three types of feature information according to different distributions of sketch line segments corresponding to different feature types: line objects, spherically aggregate distributed features, and boundaries between different features.
2.2 the distance between two line segments is defined as the Euclidean distance of the middle points of the line segments, and the aggregation degree of the line segments is represented by the average distance of the adjacent line segments K; and according to the statistical distribution of the aggregation of the line segments, the line segments are endowed with semantic information: aggregating line segments and isolated line segments; two-sided aggregation and one-sided aggregation can be classified according to the topology of the aggregation line segment.
And 2.3, representing the semantic information of the line segments in a tree structure according to the statistical distribution of the aggregative property of the line segments, as shown in fig. 5, namely a schematic diagram of the tree structure of the semantic information of the line segments. The two sides are gathered to correspond to ground objects such as forests, building groups and the like; the single side is gathered and corresponds to the boundary with ground objects such as forests or building groups on one side; the isolated line segment corresponds to a manifold feature such as a line target, a bridge, or the like, or a boundary of two different features. Fig. 6 shows a Sketch Map assigned semantic information to line segments, where gray line segments are aggregated line segments and black line segments are isolated line segments.
According to the statistical distribution of the line segment aggregation characteristics, semantic information is given to the line segments, the given semantic information comprises two-side aggregation, one-side aggregation and isolated line segments, and the semantic information analysis of the line segments is the premise of extracting line segment aggregation areas and provides basis for the subsequent ground feature area division.
And 3, in the Sketch Map, extracting a plurality of disjoint aggregation line segment sets by adopting a line segment set solving algorithm according to semantic information given to the line segments, and obtaining a line segment aggregation region R by adopting a region extraction method for each aggregation line segment set.
3.1 symbols define: set of sketch line segments is S; spatial constraint threshold δ1(ii) a Line segment growth threshold delta2(ii) a Satisfying the space constraint line segment set U; aggregating collections of line segments
Segment aggregation region R ═ { R ═ R1，r2，…，rm}；
3.2 firstly, adopting a line segment set solving algorithm, wherein the algorithm is similar to a region growing method, but the invention grows by taking line segments as elements to obtain an aggregated line segment set, which is beneficial to extracting line segment aggregated regions, and the method comprises the following specific steps:
3.2.1 first obtain sketch set S, according to the line segment aggregation property of the line segment aggregation areas such as forest, building group, etc., making statistics on the k neighbor of each line segment, calculating the average distance of the k neighbor of each line segment, finding out whether the image line segment has aggregation property from the histogram statistics of the average distance of the k neighbor, if so, indicating that the ground object exists, and according to the histogram statistics, obtaining space constraint threshold value delta1And threshold delta for line segment growth2。
3.2.2 initial setting of TiIs an empty set; obtaining an initial seed line segment according to a threshold value of the seed line segment
Randomly selecting seed line segments
Growth is carried out when，
The criterion for growth is if a certain neighbor of a line segment is present
Satisfies the line segment growth threshold delta2Grow into a collection of aggregated line segments
Traverse its k neighbors until there are no line segments that can grow, assuming this time
For T at this timeiThe segments which are not traversed in the process are sequentially used as seed segments to grow, and iterative growth is carried out until all the grown segments can not grow again, and an aggregation segment set T is obtained at the momenti。
3.2.3 if the line segments in the initial seed line segment set U do not grow, selecting one line segment as the seed line segment to continue growing, and iteratively growing until all the initial seed line segments are grown. Finally, a plurality of disjoint line segment sets T are obtainedk。
3.3 for each aggregation line segment set, adopting a region extraction method: on the basis of the line segment set, the area of the gathering line segment set is obtained by extracting the area with the circular primitive.
3.3.1 circular cell construction: taking a line segment growth threshold delta2The disks are constructed for the radius of a circle. The circle is adopted to maintain the smooth characteristic of the zone boundary, and the radius is delta2In order to ensure that the maximum gap between the line segments is filled. Since the line segment spacing should be similar in the same line segment gathering region, the growth threshold δ2Represents the maximum line segment spacing of the growing line segment set, and therefore, is taken to be δ here2As the radius of the disc.
3.3.2 closing operation: the closing operation of a set A with a structuring element B, denoted A.B, is defined as
Wherein,it is shown that B performs an expansion operation on a,
b shows the etching operation on A.
This formula illustrates that the closing of a with the structuring element B is performed by expanding a with B and then eroding the result with B. Fig. 7 is a schematic diagram of a line segment aggregation area extraction process in the present invention, and in fig. 7(a), a structural element B is a circular primitive constructed above, and a set a is a set composed of line segments. Expanding the set A means that each point on a line segment in the image A moves by using the structure B, and the set of all displacements is the expanded result. The expansion operation is shown in FIG. 7(b), and the expansion result is shown in FIG. 7 (c). After expansion, an etching operation is performed, as shown in fig. 7(d), and the final closing operation result is shown in fig. 7 (e). As can be seen from the figure, the closing operation obtains the region where the line segment set A is located, eliminates the long and narrow slits, and obtains a consistent connected region. Region extraction is performed on each aggregation line segment set to obtain a line segment aggregation region R, and fig. 8 shows a result of the line segment aggregation region extraction.
And 4, carrying out region merging on the over-segmentation result: adopting a critical area mode voting merging strategy in an over-segmentation area corresponding to the line segment gathering area R; extracting over-segmentation areas where the isolated line segments are located, and adopting a non-merging strategy; and adopting a region merging strategy based on polarization characteristics for other regions to obtain a polarized SAR image segmentation result.
4.1 the over-segmentation region corresponding to the line segment aggregation region adopts a critical region mode voting merging strategy:because the region consistency of the line segment aggregation region is good, but the boundary is not accurate, and the boundary of the over-segmentation region is accurate, a critical region mode voting merging strategy is adopted for the condition that the boundary of the line segment aggregation region is not matched with the boundary of the over-segmentation region; there are two cases of overlap between the line segment aggregation region and the over-segmentation region: firstly, some over-segmentation areas are completely covered by the line segment gathering area; secondly, the edge area of the line segment aggregation area is partially overlapped with the over-segmentation area, and the edge partially overlapped area is called a critical area. For the first case, directly combining the mean shift over-segmentation areas, for the second case, according to a mode voting strategy, if the line segment aggregation areas account for more than 50% of the over-segmentation areas, combining all the over-segmentation areas into the line segment aggregation areas, otherwise, dividing the line segment aggregation areas into wireless segment areas; finally, obtaining a merged line segment gathering area in the over-segmentation graphThis ensures that these over-segmented regions, which are difficult to merge, are well merged. Fig. 9 shows the result of the line segment aggregation areas after merging, and it can be seen that such line segment aggregation areas of the building groups are well merged.
And 4.2, extracting the over-segmentation area of the isolated line segment. These regions are not merged. According to the semantic information analysis of the line segments, for the isolated line segments corresponding to the line target in the image or the boundary of two ground objects, when the regions are merged, if the regions where the isolated line segments are located are merged, the line target disappears, or two different regions are merged. Therefore, the invention does not carry out region combination on the regions where the isolated line segments are located.
4.3 for other regions, defined as wireless segment regions, a merging strategy based on polarization characteristics is adopted. Firstly, regarding each over-segmentation region obtained by mean shift as a super-pixel, counting the polarization characteristic of the super-pixel, adopting three-channel gray histogram statistics as a feature, quantizing the gray value into 16 parts for each channel, and then calculating the region histogram in the feature space. Three-way pipeThe lane had 16 × 3 ═ 48 parts. Each region can be represented by a 48-dimensional vector, such as by HistpRepresenting the normalized histogram feature of the region P.
The similarity ρ (P, Q) of the two regions P and Q is calculated according to a Bhattacharyya coefficient calculation formula, and ρ (P, Q) is defined as follows:
wherein, HistPAnd HistQNormalized histograms of R and Q, respectively. The superscript u denotes the u-th component of the histogram.
And setting a merging threshold U, merging the adjacent regions with the similarity larger than the threshold, calculating the histogram characteristics of the merged regions again, and iteratively merging until no combinable regions exist to obtain a segmentation result based on the semantic information. Fig. 10 shows a segmentation result based on semantic information.
The invention not only provides a line segment aggregation region extraction method based on semantic information to extract line segment aggregation regions on a ridge sketch, but also adopts different strategies to merge over-segmentation regions: for the line segment aggregation areas, guiding the area combination of the over-segmentation blocks by adopting a critical area mode voting strategy; adopting a non-merging strategy for the over-segmentation area where the isolated line segment is located; the rest area is a wireless section area, and a region merging strategy based on polarization information is adopted. The method combines semantic information to extract the line segment aggregation areas, adopts different merging strategies for different types of areas, and well solves the problem that the areas aggregated by the line segments are difficult to classify.
And 5, carrying out H/alpha-Wishart classification on the polarized SAR data by utilizing polarization decomposition, and carrying out neighborhood optimization on the H/alpha-Wishart classification result by using MarkovRandom Field.
5.1 obtaining initial classification result by using H/alpha-Wishart classification method
Where S is the set of pixel points. The Wishart distance adopts the distance measure based on the Wishart distribution after being corrected by Kersten and the like. l[0]Of each pixel mark
L is the total number of categories. Where L is 8.
5.2 given a set of observations O ═ TsI S belongs to S, wherein TsIs the polarization coherence matrix of the pixel s. The covariance matrix is known to obey a complex wishart distribution. Using observation samples of class i based on the initial classification resultsTo estimate the distribution parameter σ of the class and calculate the distance matrix D between classes of L × L:
wherein DijDenotes the distance between the ith and jth classes, and d denotes the euclidean distance of the average coherence matrix.
5.3 based on the MRF framework, the data item is the similarity value of each pixel point, and the smooth item is the inter-class distance. The minimum energy function is as follows:
wherein,
is the class conditional probability, N, of the observed data at pixel ssIs a neighborhood set of pixels of pixel s. Lambda [ alpha ]1Is a regularization parameter. The total energy in equation (3) is minimized by the alpha-expansion algorithm.
And 6, fusing the segmentation result based on the semantic information and the H/alpha-Wishart classification result based on MRF.
The method combines the advantages of region consistency of the segmentation result and pixel-level accuracy of the classification result to obtain a better classification result. The fusion strategy combines unsupervised segmentation and pixel-based classification results, and classification is carried out based on a priority volume strategy to obtain a final classification result of the ground feature classification of the polarized SAR image to be classified. Fig. 11 is a schematic diagram of a segmentation and classification result fusion process, which mainly includes the following steps:
6.1, segmentation: dividing to obtain consistent regions, wherein the number of the regions is larger than the number of the final categories and slightly higher than the number of the categories; FIG. 11(a) is a schematic diagram of 4 divided regions, wherein 1-4 are used to represent 4 divided regions;
6.2 classification based on pixels: the classification at the pixel level is performed based on the scattering characteristics of the image, and fig. 11(b) is a schematic diagram of the classification based on pixel points, in which three classes are represented by white, black, and gray.
6.3 fusion segmentation and classification: and adopting a majpriority vote strategy, selecting the class with the maximum number of pixels in the corresponding classification result as the class of each region in the segmentation graph, and marking the corresponding region of the final classification result graph as the class. Thus, the region consistency of the classification result is greatly improved. It should be noted that in the majpriority volume, the neighborhood of pixels is not a fixed neighborhood window, but rather, the pixels belonging to the same region are segmented. Fig. 11(c) is a schematic diagram of fusing the segmentation map and the classification result based on the pixel points, and a mode voting strategy is applied to each region in the map to obtain the classification result shown in fig. 11 (d). And obtaining a final classification result diagram of the ground feature classification of the polarized SAR image to be classified through fusion segmentation and classification results, as shown in FIG. 14.
The method utilizes the Primal Sketch sparse representation model to obtain the Sketch Map of the span image, analyzes the semantic information contained in the line segment according to the Sketch Map, provides a line segment aggregation area extraction technology based on line segment semantic information analysis, and effectively extracts the line segment aggregation area on the Sketch Map. These line segment concentration areas correspond to terrain such as urban areas, forests, etc. in the polarized SAR image. The ground objects are often classified into a plurality of types due to the gray level change between light and shade, the defect is overcome well, and the region consistency of line segment gathering region classification is effectively improved. Meanwhile, in order to keep the polarization scattering property, H/alpha-Wishart classification is carried out on the polarization SAR data, and neighborhood optimization is carried out by using MRF. The classification result based on polarization decomposition is fine, but the number of the miscellaneous points is large, so the method provided by the invention integrates the segmentation result and the H/alpha-Wisharp classification result based on MRF to obtain the ground feature classification result of the polarized SAR image to be classified. And (5) effectively fusing semantic information and polarization decomposition to obtain a final classification result.
Example 2
The method for classifying the polarized SAR terrain based on semantic information and polarization decomposition is the same as that in embodiment 1, and simulation data and images are described as follows:
1. simulation conditions
(1) Selecting full-polarization San Francisco data of a NASA/JPL AIRSAR L wave band;
(2) in a simulation experiment, a parameter N in the Primal Sketch sparse representation model takes a value of 3, a parameter M takes a value of 18, and a threshold value epsilon takes a value of 20;
(3) in a simulation experiment, the neighbor number k is 9;
(4) in simulation experiment, seed line segment threshold value delta1Taking 20; line segment growth threshold delta2Taking 12;
(5) in a simulation experiment, a region merging threshold U is 0.7;
(6) in the simulation experiment, the neighborhood window in the MRF-based H/alpha-Wishart classification was selected to be 3 x 3.
2. Simulation content and results
The invention was used to classify the terrain using the fully polarized San Francisco data from the NASA/JPLAIRSAR L band. Fig. 12 is a span chart, which is the same as fig. 2, and fig. 12, fig. 13, and fig. 14 are collectively shown for convenience of evaluation of the classification result, and fig. 14 is a classification result chart of the present invention. As can be seen from the figure, the region consistency of the classification result is good, the boundary part is accurate, a large consistent region can be obtained particularly for a building group region, the understanding of human vision to images can be better met, and for a bridge line target, the strategy of the invention can obtain a good classification result and can well separate the bridge. In conclusion, due to the addition of semantic information, the method and the device can obtain a classification result more suitable for human image understanding, and the region consistency and the edge accuracy of the ground features are improved.
Example 3
The method for classifying the polarized SAR terrain based on semantic information and polarization decomposition is the same as the embodiment 1-2, wherein the MRF-based H/alpha-Wishart classification method is the same as the step 5 in the embodiment 1, and as a comparison experiment of the invention, simulation data and results are as follows:
1. simulation conditions
(1) Selecting full-polarization San Francisco data of the NASA/JPLAIRSAR L wave band;
(2) in the simulation experiment, the neighborhood window in the MRF-based H/alpha-Wishart classification was selected to be 3 x 3.
2. Simulation content and results
Utilizing the fully polarized San Francisco data of the NASA/JPL AIRSAR L wave band, and classifying by using an H/alpha-Wishart classification method based on MRF, wherein the method is a classification method based on pixel points, FIG. 12 is a span graph, and FIG. 13 is a result of the H/alpha-Wishart classification method based on MRF. As can be seen from the figure, the method is fine in classification, but generates a pepper salt type classification result, particularly for the buildings, such as the buildings and roads, which have aggregation characteristics, the scattering types of the buildings and the roads are inconsistent, so that an inconsistent classification result is generated, but for the low-resolution polarization SAR image, the consistent building group classification result is expected to be obtained when the image understanding is carried out, so that the method has poor consistency on the classification area of the buildings with aggregation characteristics, and the boundary is also easily influenced by noise.
Compared with the result of the H/alpha-Wishart classification method based on MRF, the method of the invention has the following steps:
the invention is compared with the ground feature classification result of the H/alpha-Wishart classification based on MRF. The experimental results are as follows, fig. 12 is a span graph, fig. 13 is a graph showing the results of H/α -Wishart classification based on MRF, and fig. 14 is a graph showing the classification results of the present invention. Comparing fig. 13 and fig. 14, it can be seen that, compared with the H/α -Wishart classification based on MRF, the building group area of the invention adopts the area extraction method based on semantic information analysis, which improves the area consistency of such complex ground features, and the segmentation results are merged based on mean shift, so that the boundary is more accurate. And finally, the classification precision is improved by fusing the method with a classification method based on MarkovRandom Field and polarization information.
In conclusion, the polarized SAR terrain classification method based on semantic information and polarization decomposition is disclosed by the invention. The realization comprises the following steps: carrying out mean shift on the span graph, extracting a side ridge sketch of the span graph, and extracting a line segment aggregation area in the side ridge sketch by using an area extraction technology based on semantic information; merging the span graph mean shift over-segmentation areas by adopting a critical area mode voting merging strategy and a polarization characteristic merging strategy to obtain a segmentation result; and fusing the image segmentation result based on the semantic information and the H/alpha-Wishart classification result based on MRF to obtain a final classification result. The invention combines semantic information, an image processing technology and polarization scattering characteristics, mainly solves the problem that the existing classification technology based on polarization decomposition has poor region consistency on classification results of ground objects with aggregation characteristics, improves the region consistency and the boundary retentivity of the classification results of the ground objects with aggregation characteristics (such as forests, building groups and the like), overcomes the defect of pixel-level classification, and obtains good polarization SAR ground object classification effect.
Claims (6)
1. A polarized SAR terrain classification method based on semantic information and polarization decomposition is characterized in that: the method comprises the following steps:
step 1, inputting data of a polarized SAR image to be classified, processing the polarized SAR data to obtain amplitude values of three channels of the polarized SAR data, fusing the amplitude values of the three channels to obtain a power map, namely a span map, of the polarized SAR image, and obtaining an over-segmentation result map of the span map by using mean shift; extracting a side ridge Sketch consisting of line segments of the span graph, namely the Sketch Map, according to the prime Sketch sparse representation model;
step 2, semantic information analysis is carried out on the line segments in the Sketch Map, and semantic information, namely two-side aggregation, one-side aggregation and isolated line segments, is given to the line segments according to the statistical distribution of the line segment aggregation characteristics;
step 3, in the Sketch Map, extracting a plurality of disjoint aggregation line segment sets by adopting a line segment set solving algorithm according to semantic information given to the line segments, and obtaining a line segment aggregation region R by adopting a region extraction method for each aggregation line segment set;
and 4, carrying out region merging on the over-segmentation result: adopting a critical area mode voting merging strategy for the over-segmentation area corresponding to the line segment gathering area R; the over-segmentation areas where the isolated line segments are located are not merged; adopting a region merging strategy based on polarization characteristics for other regions to obtain a polarized SAR image segmentation result based on semantic information;
step 5, carrying out H/alpha-Wishart classification on the polarized SAR data by utilizing polarization decomposition, and carrying out neighborhood optimization on the H/alpha-Wishart classification result by using Markov Random Field;
and 6, fusing the segmentation result based on the semantic information and the H/alpha-Wishart classification result based on the MRF, adopting mode voting, selecting the class with the largest number of pixels in the corresponding classification result as the class of the region for each region in the segmentation graph, and endowing the class to the corresponding region in the final classification result graph to obtain the final classification result graph of the ground feature classification of the polarized SAR image to be classified.
2. The polarized SAR terrain classification method based on semantic information and polarization decomposition according to claim 1, characterized in that: the step 2 of analyzing the line segment semantic information, representing the semantic information and calculating to construct a semantic classification tree of the line segment comprises the following steps:
2.1 obtaining the information that the line segment mainly corresponds to three ground feature information according to different distributions of sketch line segments corresponding to different ground feature types: a line object, a spherical aggregate distribution feature and a boundary between the features;
2.2 the distance between two line segments is defined as the Euclidean distance of the middle points of the line segments, and the average distance of the adjacent line segments K represents the aggregative property of the line segments; and according to the statistical distribution of the aggregation of the line segments, the line segments are endowed with semantic information: aggregating line segments and isolated line segments; two-side aggregation and one-side aggregation can be divided according to the topological structure of the aggregation line segment;
2.3 according to the statistical distribution of the line segment aggregation, expressing the semantic information of the line segment in a tree structure, and aggregating ground objects corresponding to forests, building groups and the like on two sides; the single side is gathered and corresponds to the boundary with ground objects such as forests, building groups and the like on one side; the isolated line segment corresponds to a line target, a bridge, or the like, manifold feature, or the boundary of two features.
3. The polarized SAR image terrain classification method according to claim 2, characterized in that: in the step 3, in the Sketch Map, according to semantic information given to the line segment, extracting an aggregated line segment set by using a line segment set solving algorithm, and obtaining a line segment aggregated region R by using a region extraction method for the aggregated line segment set, the method includes:
3.1 symbol definition
Defining: set of sketch line segments is S; spatial constraint threshold δ1(ii) a Line segment growth threshold delta2(ii) a Satisfying the space constraint line segment set U; aggregating collections of line segments
Segment aggregation region R ═ { R ═ R1，r2，…，rm}；
3.2 obtaining a plurality of disjoint aggregation line segment sets T by adopting a line segment set solving algorithmk；
And 3.3, obtaining a line segment aggregation region R by adopting a region extraction method for each aggregation line segment set.
4. The polarized SAR terrain classification method based on semantic information and polarization decomposition according to claim 3, characterized in that: carrying out region merging on the over-segmentation result in the step 4 to obtain a polarized SAR image segmentation result based on semantic information; the method comprises the following steps:
4.1 adopting a critical area mode voting merging strategy for the over-segmentation area corresponding to the line segment aggregation area: there are two cases of overlap between the line segment aggregation region and the over-segmentation region: directly merging the over-segmentation areas by the whole coverage areas of the line segment aggregation areas, and adopting a critical area mode voting merging strategy for the condition that the over-segmentation areas and the boundary areas of the line segment aggregation areas are partially overlapped, wherein the partially overlapped areas at the edges are called critical areas, if the line segment aggregation areas account for more than 50% of the over-segmentation areas, the over-segmentation areas are completely merged into the line segment aggregation areas, and if not, the line segment aggregation areas are divided into wireless segment areas; finally, obtaining a merged line segment gathering area in the over-segmentation graph
4.2, extracting the over-segmentation area of the isolated line segment; the areas are not merged, and the area where the isolated line segment is located is reserved;
4.3 for other areas, defined as wireless section areas, adopting a combination strategy based on polarization characteristics; firstly, regarding each over-segmentation block with mean shift as a super-pixel, counting the polarization characteristic of the super-pixel, adopting three-channel gray level histogram statistics as a feature, quantizing the gray level into 16 parts for each channel, and then calculating a region histogram in the feature space; the total of the three channels is 16 × 3 ═ 48; each region can be represented by a 48-dimensional vector, the characteristics of the histogram are normalized, and the similarity of the two regions is calculated according to a Bhattacharyya coefficient calculation formula; setting a merging threshold U, merging adjacent regions with similarity larger than the threshold, calculating histogram features of the merged regions again, and iteratively merging until no regions which can be merged exist;
and obtaining a final region merging result, namely a polarized SAR image segmentation result based on semantic information, by the three merging strategies.
5. The polarized SAR image terrain classification method of claim 3, characterized in that: the line segment set solving process comprises the following steps:
3.2.1 first get sketch set S, according to the aggregation of the line segments in the forest, building group and other areas, making statistics on the k neighbor of each line segment, calculating the average distance of the k neighbor of each line segment, finding out whether the image line segment has aggregation from the histogram statistics of the average distance of the k neighbor, if so, indicating that such a feature exists, and according to the histogram statistics, getting space constraint threshold delta1And threshold delta for line segment growth2；
3.2.2 initial setting of TiIs an empty set; obtaining an initial seed line segment according to a threshold value of the seed line segment
Randomly selecting seed line segments
The growth is carried out, at which time,
the criterion for growth is if a certain neighbor of a line segment is present
Satisfies the line segment growth threshold delta2Grow into a collection of aggregated line segments
Traverse its k neighbors until there are no line segments that can grow, assuming this time
For T at this timeiThe segments which are not traversed in the process are sequentially used as seed segments to grow, iterative growth is carried out until all the grown segments can not be regrown, and an aggregation segment set T is obtainedi；
3.2.3 if any line segment in the initial seed line segment set U does not grow, selecting a line segment as a seed line segment to continue iterative growth until all initial seed line segments grow; finally, a plurality of disjoint aggregation line segment sets T are obtainedk。
6. The polarized SAR image terrain classification method of claim 3, characterized in that: the process of extracting the region of the aggregated line segment set comprises the following steps:
3.3.1 circular cell construction: taking a line segment growth threshold delta2Constructing a circular disc for the radius of the circle, i.e. a circular element;
3.3.2 closing operation: performing a closing operation on a set A by using a structural element B, wherein the structural element B is a constructed circular primitive, and the set A is an obtained aggregation line segment set; and obtaining the line segment aggregation region R through the closing operation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310192057.1A CN103294792B (en) | 2013-05-13 | 2013-05-13 | Based on the polarization SAR terrain classification method of semantic information and polarization decomposing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310192057.1A CN103294792B (en) | 2013-05-13 | 2013-05-13 | Based on the polarization SAR terrain classification method of semantic information and polarization decomposing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103294792A true CN103294792A (en) | 2013-09-11 |
| CN103294792B CN103294792B (en) | 2016-02-10 |
Family
ID=49095654
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310192057.1A Active CN103294792B (en) | 2013-05-13 | 2013-05-13 | Based on the polarization SAR terrain classification method of semantic information and polarization decomposing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103294792B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103675818A (en) * | 2013-12-03 | 2014-03-26 | 中国科学院深圳先进技术研究院 | Method and system for measuring polarimetric synthetic aperture radar data ground object three-dimensional distance |
| CN104123555A (en) * | 2014-02-24 | 2014-10-29 | 西安电子科技大学 | Super-pixel polarimetric SAR land feature classification method based on sparse representation |
| CN104463882A (en) * | 2014-12-15 | 2015-03-25 | 西安电子科技大学 | SAR image segmentation method based on shape completion area chart and feature coding |
| CN104809690A (en) * | 2015-04-10 | 2015-07-29 | 西北工业大学 | Intelligent fusing method for multiple rough drawings |
| CN107944370A (en) * | 2017-11-17 | 2018-04-20 | 西安电子科技大学 | Classification of Polarimetric SAR Image method based on DCCGAN models |
| CN109325529A (en) * | 2018-09-06 | 2019-02-12 | 安徽大学 | Sketch identification method and application of sketch identification method in commodity retrieval |
| CN109740475A (en) * | 2018-12-25 | 2019-05-10 | 杭州世平信息科技有限公司 | A kind of remote sensing images ground scene classification method |
| CN111062445A (en) * | 2019-12-23 | 2020-04-24 | 中国科学院自动化研究所 | Polarimetric SAR image classification method based on collaborative regularization and superpixel |
| CN112084923A (en) * | 2020-09-01 | 2020-12-15 | 西安电子科技大学 | Semantic segmentation method for remote sensing image, storage medium and computing device |
| CN112712050A (en) * | 2021-01-12 | 2021-04-27 | 西安电子科技大学 | DS evidence fusion-based polarimetric SAR image semantic change detection method |
| CN113609898A (en) * | 2021-06-23 | 2021-11-05 | 国网山东省电力公司泗水县供电公司 | Power transmission line icing monitoring method and system based on SAR image |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070071280A1 (en) * | 2001-10-15 | 2007-03-29 | Jessica Fridrich | Lossless embedding of data in digital objects |
| CN102622598A (en) * | 2012-01-13 | 2012-08-01 | 西安电子科技大学 | SAR (Synthesized Aperture Radar) image target detection method based on zone markers and grey statistics |
| CN102708167A (en) * | 2012-04-27 | 2012-10-03 | 上海交通大学 | Web-based semantic annotation system and Web-based semantic annotation method for high resolution SAR (synthetic aperture radar) image interpretation |
-
2013
- 2013-05-13 CN CN201310192057.1A patent/CN103294792B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070071280A1 (en) * | 2001-10-15 | 2007-03-29 | Jessica Fridrich | Lossless embedding of data in digital objects |
| CN102622598A (en) * | 2012-01-13 | 2012-08-01 | 西安电子科技大学 | SAR (Synthesized Aperture Radar) image target detection method based on zone markers and grey statistics |
| CN102708167A (en) * | 2012-04-27 | 2012-10-03 | 上海交通大学 | Web-based semantic annotation system and Web-based semantic annotation method for high resolution SAR (synthetic aperture radar) image interpretation |
Non-Patent Citations (1)
| Title |
|---|
| 陶旸等: "基于语义和剖面特征匹配的地形粗糙度模型评价", 《地理研究》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103675818B (en) * | 2013-12-03 | 2017-01-18 | 中国科学院深圳先进技术研究院 | Method and system for measuring polarimetric synthetic aperture radar data ground object three-dimensional distance |
| CN103675818A (en) * | 2013-12-03 | 2014-03-26 | 中国科学院深圳先进技术研究院 | Method and system for measuring polarimetric synthetic aperture radar data ground object three-dimensional distance |
| CN104123555A (en) * | 2014-02-24 | 2014-10-29 | 西安电子科技大学 | Super-pixel polarimetric SAR land feature classification method based on sparse representation |
| CN104123555B (en) * | 2014-02-24 | 2017-05-24 | 西安电子科技大学 | Super-pixel polarimetric SAR land feature classification method based on sparse representation |
| CN104463882A (en) * | 2014-12-15 | 2015-03-25 | 西安电子科技大学 | SAR image segmentation method based on shape completion area chart and feature coding |
| CN104463882B (en) * | 2014-12-15 | 2017-07-28 | 西安电子科技大学 | SAR image segmentation method based on shape completion administrative division map and feature coding |
| CN104809690A (en) * | 2015-04-10 | 2015-07-29 | 西北工业大学 | Intelligent fusing method for multiple rough drawings |
| CN104809690B (en) * | 2015-04-10 | 2017-10-27 | 西北工业大学 | A kind of intelligent method for fusing of many sketches |
| CN107944370A (en) * | 2017-11-17 | 2018-04-20 | 西安电子科技大学 | Classification of Polarimetric SAR Image method based on DCCGAN models |
| CN109325529B (en) * | 2018-09-06 | 2021-07-30 | 安徽大学 | Sketch identification method and application of sketch identification method in commodity retrieval |
| CN109325529A (en) * | 2018-09-06 | 2019-02-12 | 安徽大学 | Sketch identification method and application of sketch identification method in commodity retrieval |
| CN109740475A (en) * | 2018-12-25 | 2019-05-10 | 杭州世平信息科技有限公司 | A kind of remote sensing images ground scene classification method |
| CN111062445A (en) * | 2019-12-23 | 2020-04-24 | 中国科学院自动化研究所 | Polarimetric SAR image classification method based on collaborative regularization and superpixel |
| CN111062445B (en) * | 2019-12-23 | 2023-05-19 | 中国科学院自动化研究所 | Polarized SAR image classification method based on collaborative regularization and superpixel |
| CN112084923A (en) * | 2020-09-01 | 2020-12-15 | 西安电子科技大学 | Semantic segmentation method for remote sensing image, storage medium and computing device |
| CN112084923B (en) * | 2020-09-01 | 2023-12-22 | 西安电子科技大学 | Remote sensing image semantic segmentation method, storage medium and computing device |
| CN112712050A (en) * | 2021-01-12 | 2021-04-27 | 西安电子科技大学 | DS evidence fusion-based polarimetric SAR image semantic change detection method |
| CN112712050B (en) * | 2021-01-12 | 2023-05-16 | 西安电子科技大学 | Polarized SAR image semantic change detection method based on DS evidence fusion |
| CN113609898A (en) * | 2021-06-23 | 2021-11-05 | 国网山东省电力公司泗水县供电公司 | Power transmission line icing monitoring method and system based on SAR image |
| CN113609898B (en) * | 2021-06-23 | 2023-09-29 | 国网山东省电力公司泗水县供电公司 | SAR image-based power transmission line icing monitoring method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103294792B (en) | 2016-02-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103294792B (en) | Based on the polarization SAR terrain classification method of semantic information and polarization decomposing | |
| Kim et al. | Deep-hurricane-tracker: Tracking and forecasting extreme climate events | |
| Wang et al. | Urban impervious surface detection from remote sensing images: A review of the methods and challenges | |
| CN103578110B (en) | Multiband high-resolution remote sensing image dividing method based on gray level co-occurrence matrixes | |
| CN101976504B (en) | Multi-vehicle video tracking method based on color space information | |
| CN103632363A (en) | Object-level high-resolution remote sensing image change detection method based on multi-scale fusion | |
| CN106096655A (en) | A kind of remote sensing image airplane detection method based on convolutional neural networks | |
| CN105469393A (en) | Shallow water depth multi-temporal remote sensing image inversion method based on decision fusion | |
| Demir et al. | Integrated shoreline extraction approach with use of Rasat MS and SENTINEL-1A SAR Images | |
| Mahrooghy et al. | A machine learning framework for detecting landslides on earthen levees using spaceborne SAR imagery | |
| Zhao et al. | Change detection method of high resolution remote sensing image based on DS evidence theory feature fusion | |
| CN114612315B (en) | High-resolution image missing region reconstruction method based on multitask learning | |
| Barkhade et al. | Flood extent mapping with unmanned aerial vehicles data using deep convolutional neural network | |
| Lizarazo et al. | Segmentation of remotely sensed imagery: moving from sharp objects to fuzzy regions | |
| Nurwauziyah et al. | Satellite image classification using decision tree, SVM and k-nearest neighbor | |
| CN103854285A (en) | SAR image ground object cutting method based on random projection and improved spectral cluster | |
| Zaletnyik et al. | LIDAR waveform classification using self-organizing map | |
| CN116543300A (en) | Cloud-aerosol hierarchical classification method based on semantic segmentation | |
| Wang et al. | Damaged buildings recognition of post-earthquake high-resolution remote sensing images based on feature space and decision tree optimization | |
| Widyaningrum et al. | Skeleton-based automatic road network extraction from an orthophoto colored point cloud | |
| Budde et al. | Identification of misclassified pixels in semantic segmentation with uncertainty evaluation | |
| CN113963270A (en) | High resolution remote sensing image building detection method | |
| Kuras | Airborne hyperspectral imaging for multisensor data fusion | |
| Mercovich et al. | Utilizing the graph modularity to blind cluster multispectral satellite imagery | |
| Chen et al. | Learning Transformations between Heterogeneous SAR and Optical Images for Change Detection |
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 |