阅读说明：本技术 用于多基线干涉合成孔径雷达相位解缠的聚类校正方法、系统和介质 (Cluster correction method, system and medium for multi-baseline interferometric synthetic aperture radar phase unwrapping ) 是由 袁志辉 陈天骄 徐海胜 彭葳 陈立福 邢学敏 于 2021-09-17 设计创作，主要内容包括：本发明公开了用于多基线干涉合成孔径雷达相位解缠的聚类校正方法、系统和介质,方法为：根据多基线干涉合成孔径雷达的参数,计算每幅干涉图中每个像素对应的整周模糊数,从而得到每个像素的模糊矢量；将模糊矢量相同的像素聚为同一类；对干涉图中的像素进行类校正：若干涉图尺寸小于预设尺寸值,则对所有像素,均按照像素的扩展区域内的最大类别进行类校正；若干涉图尺寸超过预设尺寸值,则先根据像素的扩展区域的密度判断该像素是否需要校正,然后对需要校正的像素,按照像素的扩展区域内的最大类别进行类校正。本发明可以提高现有多基线干涉合成孔径雷达相位解缠的像素聚类结果的准确性和有效性,进而提高解缠相位的精度。(The invention discloses a clustering correction method, a system and a medium for multi-baseline interferometric synthetic aperture radar phase unwrapping, wherein the method comprises the following steps: calculating the fuzzy number of the whole circle corresponding to each pixel in each interference pattern according to the parameters of the multi-baseline interference synthetic aperture radar, thereby obtaining the fuzzy vector of each pixel; pixels with the same fuzzy vector are gathered into the same type; carrying out class correction on pixels in the interference pattern: if the sizes of the plurality of charts are smaller than the preset size value, performing class correction on all pixels according to the maximum class in the expansion area of the pixels; and if the sizes of the plurality of images exceed the preset size value, judging whether the pixel needs to be corrected according to the density of the expansion area of the pixel, and then performing class correction on the pixel needing to be corrected according to the maximum class in the expansion area of the pixel. The method can improve the accuracy and effectiveness of the pixel clustering result of phase unwrapping of the existing multi-baseline interferometric synthetic aperture radar, and further improve the accuracy of unwrapping phases.)

1. A clustering correction method for multi-baseline interferometric synthetic aperture radar phase unwrapping, comprising:

calculating the whole-cycle fuzzy number corresponding to each pixel in each interference pattern according to the parameters of the multi-baseline interference synthetic aperture radar; the fuzzy number of the whole week corresponding to the s-th pixel in the ith interference pattern corresponding to the ith baseline is recorded as k_i(s), then the blur vector for the s-th pixel is represented as [ k ]₁(s),k₂(s),…,k_M(s)]M is the number of baselines;

and clustering all pixels according to the fuzzy vector corresponding to each pixel: pixels with the same fuzzy vector belong to the same class;

carrying out class correction on pixels in the interference pattern: if the sizes of the plurality of charts are smaller than the preset size value, performing class correction on all pixels according to the maximum class in the expansion area of the pixels; and if the sizes of the plurality of images exceed the preset size value, judging whether the pixel needs to be corrected according to the density of the expansion area of the pixel, and then performing class correction on the pixel needing to be corrected according to the maximum class in the expansion area of the pixel.

2. The method of claim 1, wherein the extended area of pixels refers to: and a rectangular window area which is expanded to the periphery to a preset size by taking the current pixel as a center.

3. The method of claim 1, wherein the largest class within a pixel expansion area is: within the current pixel extension area, a category comprising the most pixels; the performing class correction on the pixel according to the maximum class in the pixel expansion area specifically includes: and modifying the category of the current pixel into the maximum category in the expansion area of the current pixel.

4. The method of claim 1, wherein the density of the extended area of pixels is: in the expansion area, the number of pixels is the same as the type of the current pixel; alternatively, the density of the extended area of the pixel refers to: and in the expansion area, the number of all pixels of which the difference between the intercepts of the fuzzy vectors corresponding to the current pixel is smaller than a preset intercept threshold value.

5. The method of claim 1, wherein the method for determining whether the pixel needs to be corrected according to the density of the extended area of the pixel comprises: if the density of the expansion area of the pixel is greater than a preset density threshold value, the type of the current pixel is accurate and is recorded as a core pixel, namely the pixel which does not need to be corrected; otherwise, the class error of the current pixel is marked as a non-core pixel, i.e. the pixel needing to be corrected.

6. The method according to claim 1, wherein the number of the baseline M is 2, and the relation between the blur numbers of the whole cycle corresponding to each pixel in the two interferograms is expressed as:

in the formula, k_i(s) is the number of full cycle ambiguities corresponding to the s-th pixel in the ith interferogram, B_iIs the length of the vertical baseline in the ith interferogram,is the wrapping phase of the s-th pixel in the ith interferogram, i is 1, 2.

7. Cluster correction system for multi-baseline interferometric synthetic aperture radar phase unwrapping, comprising a memory and a processor, the memory having stored therein a computer program, wherein the computer program, when executed by the processor, causes the processor to carry out the method according to any one of claims 1 to 6.

8. A readable storage medium comprising computer program instructions, characterized in that the computer program instructions, when executed by a processing terminal, cause the processing terminal to perform the method of any of claims 1 to 6.

Technical Field

The invention belongs to the technical field of interferometric synthetic aperture radars, and particularly relates to a clustering correction method, a system and a medium for phase unwrapping of a multi-baseline interferometric synthetic aperture radar.

Background

The interferometric synthetic Aperture Radar (InSAR) can acquire the elevation information of a ground target according to the absolute interference phases of two SAR images corresponding to the same scene, and can be finally used for acquiring a Digital Elevation Model (DEM)^[1][2]. However, the InSAR system can only obtain the main value of the absolute phase, i.e. the winding phase. To accurately estimate terrain height, the absolute phase must be reconstructed by Phase Unwrapping (PU)^[3][4]. To solve this problem, researchers first proposed a single baseline phase unwrapping method (SBPU)^[4]. For a conventional SBPU, there are two unknowns in an equation, which results in many solutions. To ensure solution uniqueness, the SBPU assumes that the actual phase jump between neighboring pixels is less than π, which is referred to as the phase continuity assumption or Itoh condition^[5]. In order to overcome the disadvantages of SBPU, some researchers have proposed a multi-baseline phase unwrapping algorithm (MBPU) based on Chinese Remainder Theorem (CRT) to eliminate the limitation of phase continuity hypothesis^[6]。

MBPU has been studied extensively over the past several decades. Document [6] proposes for the first time three methods using multiple interferograms MBPU with different baselines or frequencies. Then, methods based on maximum likelihood estimation are proposed in documents [7] and [8 ]. Subsequently, a new strategy based on a Bayesian framework and maximum a posteriori probability (MAP) estimation is proposed in [9] and [10 ]. In [11], a multi-channel EKF phase unwrapping framework is constructed by combining a maximum likelihood criterion with an Extended Kalman Filter (EKF), and interesting performance improvement is realized. In [12], a non-local filtering technique is introduced into the MAP-based MBPU method. In [13], a two-phase planning method (TSPA) is proposed, which migrates the SBPU frame to the MBPU.

As one of the most popular MBPU methods, the MBPU algorithm based on cluster analysis has also been widely studied. In [14], a fast method based on cluster analysis, called CA for short, is proposed. The clustering method divides all pixels into different classes according to the combined information of a plurality of interferograms, and then performs PU one by one. In [15], the fuzzy vector corresponding to each pixel is solved according to the robust CRT in a closed form, then clustering is carried out according to the fuzzy vector, and some measures are designed to improve the performance. In [16], a robust CA-based multi-baseline interferogram PU algorithm (abbreviated CANOPUS) is proposed, whose recognizable mathematical model extends from the intercept dimension to the row, column and intercept dimensions. In [17], a linear combination method is employed to increase the high blur number and improve the noise robustness of the CA method. Document [18] considers the situation that the baseline is too much or the search space is too large, provides a closed solution formula of the class fuzzy vector, and also provides a new phase filtering strategy to further improve the precision of the unwrapped phase.

But the MBPU method based on the cluster analysis still has the problem of weak robustness and is easy to generate wrong clustering results, (1) the PU method directly based on the CRT is sensitive to phase noise, and the smaller phase noise can cause larger unwrapping error, so the noise robustness is poor; (2) although the CA method improves noise robustness to a certain extent and reduces PU time, when the intercept of the class with fewer pixels is very close to that of the class with more pixels, the two classes are merged together due to the existence of phase noise, resulting in an erroneous clustering result; (3) the CANOPUS method only extends the clustering information of the pixels to be clustered from the intercept to the row, the column and the intercept value thereof, but still easily obtains wrong clustering results.

Disclosure of Invention

The invention aims to provide a clustering correction method, a system and a medium for multi-baseline interferometric synthetic aperture radar phase unwrapping, and the noise robustness of a CA-based MBPU algorithm is improved.

A cluster correction method for multi-baseline interferometric synthetic aperture radar phase unwrapping, comprising:

calculating the whole-cycle fuzzy number corresponding to each pixel in each interference pattern according to the parameters of the multi-baseline interference synthetic aperture radar; the fuzzy number of the whole week corresponding to the s-th pixel in the ith interference pattern corresponding to the ith baseline is recorded as k_i(s), then the blur vector for the s-th pixel is represented as [ k ]₁(s),k₂(s),…,k_M(s)]M is the number of baselines;

and clustering all pixels according to the fuzzy vector corresponding to each pixel: pixels with the same fuzzy vector belong to the same class;

carrying out class correction on pixels in the interference pattern: if the sizes of the plurality of charts are smaller than the preset size value, performing class correction on all pixels according to the maximum class in the expansion area of the pixels; and if the sizes of the plurality of images exceed the preset size value, judging whether the pixel needs to be corrected according to the density of the expansion area of the pixel, and then performing class correction on the pixel needing to be corrected according to the maximum class in the expansion area of the pixel.

Further, the extension area of the pixel refers to: and a rectangular window area which is expanded to the periphery to a preset size by taking the current pixel as a center.

Further, the largest class within the pixel expansion area refers to: within the current pixel extension area, a category comprising the most pixels; the performing class correction on the pixel according to the maximum class in the pixel expansion area specifically includes: and modifying the category of the current pixel into the maximum category in the expansion area of the current pixel.

Further, the density of the extended area of the pixel refers to: in the expansion area, the number of pixels is the same as the type of the current pixel; alternatively, the density of the extended area of the pixel refers to: and in the expansion area, the number of all pixels of which the difference between the intercepts of the fuzzy vectors corresponding to the current pixel is smaller than a preset intercept threshold value.

Further, the method for judging whether the pixel needs to be corrected according to the density of the expansion area of the pixel comprises the following steps: if the density of the expansion area of the pixel is greater than a preset density threshold value, the type of the current pixel is accurate and is recorded as a core pixel, namely the pixel which does not need to be corrected; otherwise, the class error of the current pixel is marked as a non-core pixel, i.e. the pixel needing to be corrected.

Further, if the number M of the base lines is 2, the relational expression between the blur numbers of the whole cycles corresponding to the pixels in the two interferograms is as follows:

in the formula, k_i(s) is the number of full cycle ambiguities corresponding to the s-th pixel in the ith interferogram, B_iIs the length of the vertical baseline in the ith interferogram,is the wrapping phase of the s-th pixel in the ith interferogram, i is 1, 2.

Cluster correction system for multi-baseline interferometric synthetic aperture radar phase unwrapping, comprising a memory and a processor, the memory having stored therein a computer program, characterized in that the computer program, when executed by the processor, causes the processor to implement the cluster correction method as defined in any one of the above.

A readable storage medium comprising computer program instructions which, when executed by a processing terminal, cause the processing terminal to perform any one of the cluster correction methods described above.

Advantageous effects

The invention carries out category correction on the pixel clustering result of phase unwrapping of the existing multi-baseline interferometric synthetic aperture radar: if the sizes of the plurality of charts are smaller than the preset size value, performing class correction on all pixels according to the maximum class in the expansion area of the pixels; and if the sizes of the plurality of images exceed the preset size value, judging whether the pixel needs to be corrected according to the density of the expansion area of the pixel, and then performing class correction on the pixel needing to be corrected according to the maximum class in the expansion area of the pixel. The method can improve the accuracy and effectiveness of the pixel clustering result of phase unwrapping of the existing multi-baseline interferometric synthetic aperture radar, and further improve the accuracy of unwrapping phases.

Drawings

FIG. 1 is a simulated interferogram corresponding to a reference EDM and using long and short baselines for a first embodiment of the present application; wherein, (a) the reference DEM (unit: m), (b) the simulated interferogram corresponding to the short base line, and (c) the simulated interferogram corresponding to the long base line.

FIG. 2 is a class distribution diagram of the first embodiment of the present application after using the existing CA method classification, CANOPUS method classification, and correcting the existing CA method classification, CANOPUS method classification using the present application; the classification distribution diagram (a) corresponding to the CA method (b) corrected by the PPCC method (c) corrected by the NPCC1 method (d) corrected by the NPCC2 method (e) corrected by the CANOPUS method (f) corrected by the PPCC method (g) corrected by the CANOPUS method (g) corrected by the NPCC1 method (h) corrected by the NPCC2 method (h).

FIG. 3 is a simulated interferogram corresponding to a reference EDM and using long and short baselines for a second embodiment of the present application; the method comprises the following steps of (a) referring to a DEM (unit: m), (b) generating a noiseless simulated interferogram corresponding to a long base line, (c) generating a noiseless simulated interferogram corresponding to a short base line, (d) generating a noisy simulated interferogram corresponding to a long base line, and (e) generating a noisy simulated interferogram corresponding to a short base line.

FIG. 4 is a class distribution diagram of the second embodiment of the present application after using the existing CA method classification, CANOPUS method classification, and correcting the existing CA method classification, CANOPUS method classification using the present application; the method comprises the following steps of (a) a class distribution diagram corresponding to a CA method, (b) a CA method class distribution diagram corrected by a PPCC method, (c) a CA method class distribution diagram corrected by an NPCC1 method, (d) a CA method class distribution diagram corrected by an NPCC2 method, (e) a class distribution diagram corresponding to a CANOPUS method, (f) a CANOPUS method class distribution diagram corrected by the PPCC method, (g) a CANOPUS method class distribution diagram corrected by an NPCC1 method, and (h) a CANOPUS method class distribution diagram corrected by an NPCC2 method.

Detailed Description

The following describes embodiments of the present invention in detail, which are developed based on the technical solutions of the present invention, and give detailed implementation manners and specific operation procedures to further explain the technical solutions of the present invention.

The invention provides a clustering correction method for multi-baseline interferometric synthetic aperture radar phase unwrapping, which comprises the following steps:

step 1, calculating the whole-cycle fuzzy number corresponding to each pixel in each interferogram according to the parameters of the multi-baseline interferometric synthetic aperture radar; the fuzzy number of the whole week corresponding to the s-th pixel in the ith interference pattern corresponding to the ith baseline is recorded as k_i(s), then the blur vector for the s-th pixel is represented as [ k ]₁(s),k₂(s),…,k_M(s)]And M is the number of baselines.

Step 2, clustering all pixels according to the fuzzy vector corresponding to each pixel: pixels with the same blur vector belong to the same class.

The following explanation is made by using the basic principle of the multi-baseline phase unwrapping algorithm based on cluster analysis:

for an InSAR system, the measured phase of the target, i.e., the wrapping phase, obtained by the system can be determined by the following equation (1):

wherein the content of the first and second substances,is the wrapping phase of the s-th pixel, i.e., the principal value of the absolute phase, ψ(s) is the deplat absolute interference phase of the s-th pixel, and k(s) is the unknown integer ambiguity number of the s-th pixel. As can be seen from (1), the purpose of phase unwrapping is to find the correct value of k(s) in order to get the correct unwrapped phase ψ(s). Because there are two unknowns in one equation, the two unknowns ψ(s) and k(s) cannot be solved directly from (1). There are two ways to solve for (1). The first approach is to use the conventional SBPU, i.e. the assumption of phase continuity, of neighboring pixelsThe difference between absolute interference phases is less than pi, which is a ill-conditioned problem. Another approach is MBPU. For simplicity, only the case of two different baselines is considered in the following discussion. For InSAR systems, the terrain elevation is related to the unwrapping phase by

Wherein h(s) is the terrain height of the s-th pixel, λ is the working wavelength of the radar, r(s) represents the slant distance between the target point corresponding to the s-th pixel and the main antenna, θ is the incident angle, B is the ground angle of the main antenna, and_iis the length of the vertical base line in the ith interferogram, psi_i(s) is the deplated absolute interference phase of the s-th pixel in the ith interferogram.

The combination of (1) and (2) can obtain:

as can be seen from (3), the slope B of the line in the case of a determined baseline₂/B₁Is constant and has an intercept of B₁、B₂、Andand (4) jointly determining. Defining a blur vector k₁(s),k₂(s)]And (3) representing the fuzzy number of the s-th pixel in the two interference graphs in the whole week, and simultaneously defining that the pixels with the same fuzzy vector belong to the same class. Pixels with the same blur vector are at k₁-k₂Having the same slope B in the plane₂/B₁And pass through the same integer point (k)₁,k₂). It follows that the lines corresponding to pixels with the same blur vector overlap each other, i.e. their intercepts are the same. This is the essence of the multi-baseline phase unwrapping algorithm based on cluster analysis, i.e. all intercept phasesThe pixels corresponding to the same straight line are all classified into the same class.

The two classical CA-based MBPU algorithms described in documents [14] and [16] can meet the PU requirements to some extent, but still have some disadvantages. The multi-baseline phase unwrapping method based on the clustering analysis often has many wrong clustering results, and the false clustering results need to be corrected. Therefore, the invention provides two kinds of correction methods aiming at the size of the interferogram, and improves the accuracy of the clustering result in the prior art through the following step 3.

Step 3, carrying out class correction on the pixels in the interference pattern: if the sizes of the plurality of charts are smaller than the preset size value, performing class correction on all pixels according to the maximum class in the expansion area of the pixels; and if the sizes of the plurality of images exceed the preset size value, judging whether the pixel needs to be corrected according to the density of the expansion area of the pixel, and then performing class correction on the pixel needing to be corrected according to the maximum class in the expansion area of the pixel.

The following two explanations are provided for two types of correction methods used for two interferograms of different sizes:

a Pixel by Pixel cluster correction (PPCC) method

The size of the window is selected according to indexes such as a coherence coefficient, and the class with the highest frequency of appearance in the window is used as the class of the central pixel of the window. And performing class correction on all the pixels one by one to determine a final class corresponding to each pixel. The method comprises the following specific steps:

firstly, a pixel is selected, then a proper rectangular window is designed by taking the pixel as the center, and the size of the rectangular window is set according to the size of the interference pattern. And counting the occurrence times of each category in each rectangular window, searching the category with the maximum occurrence times in the current window, and modifying the category of the current central pixel into the category with the maximum occurrence times in the rectangular window. This process is repeated until the class correction of all pixels in the interferogram is complete. The pseudo-code for PPCC is as follows:

defining a pixel set to be corrected as M;

1：WHILE M≠Φ；

2: selecting a pixel i in M, and selecting a proper window by taking the pixel as a center;

3: counting the occurrence times of each category in the window;

4: modifying the category of the pixel i into the category with the largest occurrence frequency in the window;

5：END WHILE

b Non-kernel-pixel-class correction (NPCC) method

The method first uses density to distinguish between kernel pixels and non-kernel pixels. The density is defined in two ways, the first defines the number of pixels in the selected window with the same type as the central pixel as the density of the central pixel, and the second defines the number of pixels in the selected window with an intercept difference smaller than a fixed value from the central pixel as the density of the central pixel. If the first definition is adopted, assuming that the selected rectangular window is an N × N matrix and N pixels are the same as the central pixel in category, the density of the central pixel can be denoted as N. Then, regarding the pixel with density larger than the set threshold as the core pixel, and regarding the pixel as the pixel with correct category; pixels with density less than a set threshold are considered as non-core pixels and are considered as pixels with error classification. Class correction is then performed on the non-kernel pixels. The class correction method using the first density definition is called NPCC1 method, and the class correction method using the second density definition is called NPCC2 method. Pseudo code for the non-kernel pixel class correction method will be given below.

Defining a pixel set to be corrected as M;

1：WHILE M≠Φ；

2: selecting a pixel i in M, and selecting a proper window by taking the pixel as a center;

3: selecting a density definition, and calculating the density n corresponding to the pixel i;

4: setting a density threshold value n according to the density distribution condition in the window_T；

5: if the density of the pixel i is larger than n_TIf yes, marking as a core pixel, otherwise, marking as a non-core pixel;

6: changing the category of the non-core pixel into the category with the largest occurrence frequency in the window according to the maximum likelihood criterion;

7：END WHILE

the first method is the PPCC method, in which the category with the largest number of occurrences in the window is used as the final category of the center pixel. The second method is the NPCC method, which uses density to distinguish between kernel pixels and non-kernel pixels, and then performs class correction on the non-kernel pixels. When the interference pattern is small in scale and serious in noise, a PPCC method is adopted, and the PU result obtained in the method has high precision. Under the conditions of large interferogram scale and low noise, an NPCC method is adopted, the PU result accuracy is ensured, and meanwhile, the calculation efficiency of the algorithm is improved. The validity of the class correction method of the present invention is verified by experimental results.

The first experiment created a simple simulation scenario that showed smooth and discontinuous regions. The real elevation values of the simulation scene are set to be 35 meters and 80 meters respectively (the specific DEM is shown in FIG. 1 (a)). The lengths of the long base line and the short base line are respectively set to be 500 meters and 300 meters, and the coherence coefficients of the corresponding interferograms are respectively 0.8 and 0.7. The real intercepts (i.e. intercepts corresponding to the noise-free case) of the simulation scenario are 1 and 1/3, respectively, so that there should be only 2 classes when clustering is performed by using the cluster analysis method under the noise-free case. The simulated noisy interferograms are shown in fig. 1(b) and (c), respectively.

Then, the false clustering result generated by the existing clustering algorithm is corrected by using the class correction method provided by the invention, thereby verifying the effectiveness of the method provided by the invention.

The class distribution map generated by the CA method is shown in FIG. 2(a), and it can be seen that there are many erroneous clustering results. The corresponding class distribution map after the PPCC method is applied to the map is shown in fig. 2 (b). The class distribution map corresponding to the NPCC1 method applied to this figure is shown in fig. 2 (c). The class distribution map corresponding to the NPCC2 method applied to this figure is shown in fig. 2 (d). As can be seen from the figure, most of the error categories are now corrected.

Since different classes may have the same blur vector in the CANOPUS method, there may be many classes. The class distribution generated by CANOPUS is shown in fig. 2(e), and it can be seen that there are many very small classes, some even a few pixels, which are clearly meaningless. The corresponding class distribution map after the PPCC method is applied to the map is shown in fig. 2 (f). The class distribution map corresponding to the NPCC1 method applied to this figure is shown in fig. 2 (g). The class distribution map corresponding to the NPCC2 method applied to this figure is shown in fig. 2 (h). The number of subclasses is now significantly reduced.

From the correction results, both of the two types of correction methods proposed herein can correct the types of errors, thereby improving the accuracy of phase unwrapping.

The performance of both types of correction methods was quantitatively analyzed below. In the CANOPUS method, since the correctness of the class to which a certain pixel belongs cannot be determined directly according to the class, the correctness of the unwrapping result will be determined by the correctness of the fuzzy vector corresponding to the pixel in the following text. The phase unwrapping Success rate (phase unwrapping) is defined as the percentage of total pixels in the interferogram from which the blur number is correctly recovered. Table 1 shows the phase unwrapping success rates corresponding to the CA method and CANOPUS method in experiment 1. From this table, it can be seen that the phase unwrapping success rate is greatly improved after the class correction.

TABLE 1 disentanglement success rate of experiment 1

As a second example, the performance of the class correction method was tested on a more realistic height profile, and fig. 3(a) is a DEM corresponding to a mountain area (Isolation Peak) of colorado, usa. The corresponding height ambiguity numbers are set to be 32.1 m and 53.5 m respectively, two noiseless interferograms obtained by simulation are shown in (b) and (c) of fig. 3, and the corresponding noised interferograms are shown in (d) and (e) of fig. 3.

FIG. 4(a) is a category distribution chart generated by the CA method. Fig. 4(b) is a category distribution map corresponding to the PPCC method corrected in fig. 4 (a). Fig. 4(c) is a category distribution diagram obtained by correcting fig. 4(a) by NPCC 1. Fig. 4(d) is a category distribution map corresponding to the NPCC2 method corrected for fig. 4 (a). FIG. 4(e) is a class distribution diagram generated by the CANOPUS method. Fig. 4(f) is a category distribution map corresponding to the PPCC method corrected in fig. 4 (e). Fig. 4(g) is a category distribution diagram obtained by correcting fig. 4(e) by NPCC1 method. Fig. 4(h) is a category distribution diagram obtained by correcting fig. 4(e) by NPCC2 method. As can be seen from the figure, most of the error categories are corrected.

Table 2 shows the phase unwrapping success rates corresponding to the CA method and CANOPUS method in experiment 2. The table shows that the phase unwrapping success rate is still greatly improved after class correction, and the method provided by the invention is also effective on real terrain.

TABLE 2 disentanglement success rate of experiment 2

Although both methods rely on maximum likelihood criteria for class correction, each has advantages. The PPCC method is used for correcting all the clustered pixels and also can be used for correcting core pixels with class errors, and the PU success rate is higher than that of the NPCC method. However, when the number of pixels is relatively large or the phase noise is relatively small, since the class correction is performed on all the pixels, the PPCC method takes more time to perform the class correction than the NPCC method. Therefore, the NPCC method is more efficient and time-saving than the PPCC method when the number of pixels is large or the influence of noise is small. And simultaneously, a better phase unwrapping effect can be achieved.

The above embodiments are preferred embodiments of the present application, and those skilled in the art can make various changes or modifications without departing from the general concept of the present application, and such changes or modifications should fall within the scope of the claims of the present application.