Polarization SAR classification method, device and equipment based on complete polarization information input

Abstract

The invention relates to a polarimetric SAR classification method, device and equipment based on complete polarimetric information input, and the method comprises the steps: carrying out the non-local mean filtering of to-be-classified complete polarimetric SAR data, carrying out the polarization classification through employing a reflection symmetric decomposition method, and obtaining a plurality of polarimetric features representing the back scattering of ground features, and performing normalization processing on different types of polarization features by using a nonlinear normalization method to obtain complete polarization information, and performing accurate ground feature classification by using the trained convolutional neural network according to the complete polarization information. According to the method, the ground features are classified by using complete polarization information, and meanwhile, a nonlinear normalization method is adopted for polarization characteristics of different ground features, so that the classification accuracy is improved.

Description

Technical Field

[0001] The present application relates to the technical field of polarimetric SAR image classification, and in particular to a polarimetric SAR classification method, device and equipment based on complete polarimetric information input. Background Art

[0002] Polarimetric synthetic aperture radar (PolSAR) can obtain complete polarimetric scattering characteristics of ground objects all day and all weather, and has been applied to a variety of remote sensing scenarios in recent years. PolSAR actively obtains polarimetric information of surface scattering, and contains more parameters describing the electromagnetic scattering characteristics of ground objects compared to traditional single-polarization SAR. How to obtain complete polarimetric scattering parameters that can reflect the backscattering of ground objects from PolSAR images, how to fully and reasonably explore these polarimetric features and apply them to the currently widely used deep learning algorithms, is crucial for the task of classification based on polarimetric SAR data.

[0003] At present, PolSAR classification methods are mainly divided into three categories. 1. Based on the features decomposed by polarization, the PolSAR image is polarized and decomposed, and the scattering features of the target objects are extracted for direct classification. Common target polarization decomposition methods include Freeman decomposition, Cloude-Potier decomposition, Huynen decomposition, etc. 2. Classification is performed based on the statistical distribution characteristics of PolSAR data. Commonly used algorithms include Wishart classification and other algorithms. 3. Classification methods using deep learning. With the rapid development of deep learning methods, relevant scholars have introduced various deep learning methods into PolSAR image classification. Because the deep learning model contains multiple convolutional layers, it can well extract the high-order features of the objects in the image and achieve better classification results. Although a large number of scholars have adopted deep learning methods to classify PolSAR images and achieved good results, there are still defects. For example, some algorithms simply stack and combine the features decomposed by polarization indiscriminately and input them into the network, but ignore the defects of the polarization decomposition method itself. When normalizing the polarization features, some methods only perform simple normalization without considering the data distribution characteristics, and also use linear normalization for nonlinear PolSAR data. When using CNN to process PolSAR images, some methods do not consider all the scattering information and various polarization scattering features of the images, and only use a single form of incomplete polarization data input scheme as the network input. These problems will lead to inaccurate classification. Summary of the invention

[0004] Based on this, it is necessary to provide a polarization SAR classification method, device and equipment based on complete polarization information input, which can use complete polarization information for accurate ground object classification in order to address the above technical problems.

[0005] A polarimetric SAR classification method based on complete polarimetric information input, the method comprising: Acquiring fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; Preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, performing polarization classification using a reflection symmetric decomposition method to obtain multiple polarization features characterizing backscattering of ground objects, and performing normalization processing on different types of polarization features using a nonlinear normalization method to obtain the complete polarization information; The complete polarization information is input into a trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

[0006] In one embodiment, the fully polarized SAR data is GF-3 L1A level data.

[0007] In one embodiment, the multiple polarization features characterizing backscattering of ground objects include 15 real polarization features and 6 complex polarization features.

[0008] In one embodiment, the total polarization power Span , the three main diagonal elements of the polarization coherence matrix, and the power value of the volume scattering component , surface scattering power value , secondary scattering power value , the total power value of the second component of reflection symmetry decomposition , the total power value of the third component of reflection symmetry decomposition , twice the orientation angle , twice the helix angle , the power proportion of spherical scattering in the second component of reflection symmetry decomposition x , the power proportion of spherical scattering in the third component of reflection symmetry decomposition y , reflection symmetry decomposition of the second component Phase of an element a , reflection symmetry decomposition of the third component Phase of an element b ; The six complex polarization features include the real part and the imaginary part of the non-main diagonal elements of the polarization coherence matrix.

[0009] In one embodiment, the normalization process for different types of polarization features using a nonlinear normalization method includes: For the polarization total power, convert the polarization total power into a value in dB, and then perform normalization processing by using a linear normalization method; Using the polarization total power, nonlinear normalization processing is performed on polarization characteristics related to spherical scattering power value, dihedral scattering power value, 45° dihedral scattering power value, volume scattering power value, surface scattering power value, secondary scattering power value, second component total power value, and third component total power value; Using the linear normalization method, the twice orientation angle, twice the helix angle, the second component Phases of elements and third components The phase of the elements is normalized.

[0010] In one embodiment, during the training of the convolutional neural network, when constructing a training set: Randomly select a plurality of pixel points on the fully polarized SAR sample data; The selected pixel point is taken as the center according to the preset size to generate the corresponding slice sample; After preprocessing each of the slice samples, a complete polarization information sample is obtained, and the training set is constructed according to the complete polarization information sample.

[0011] The present application also provides a polarization SAR classification device based on complete polarization information input, the device comprising: A fully polarized SAR data acquisition module is used to acquire fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; A preprocessing module, used for preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, and then using a reflection symmetric decomposition method to perform polarization classification to obtain multiple polarization features that characterize backscattering of ground objects, and using a nonlinear normalization method to perform normalization processing on different types of polarization features to obtain the complete polarization information; The polarization SAR classification module is used to input the complete polarization information into the trained convolutional neural network to obtain the ground object classification results in the full polarization SAR data.

[0012] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented: Acquiring fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; Preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, performing polarization classification using a reflection symmetric decomposition method to obtain multiple polarization features characterizing backscattering of ground objects, and performing normalization processing on different types of polarization features using a nonlinear normalization method to obtain the complete polarization information; The complete polarization information is input into a trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

[0013] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps: Acquiring fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; Preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, performing polarization classification using a reflection symmetric decomposition method to obtain multiple polarization features characterizing backscattering of ground objects, and performing normalization processing on different types of polarization features using a nonlinear normalization method to obtain the complete polarization information; The complete polarization information is input into a trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

[0014] The polarization SAR classification method, device and equipment based on the input of complete polarization information, after non-local mean filtering of the fully polarized SAR data to be classified, adopts the reflection symmetric decomposition method to perform polarization classification, obtains multiple polarization features that characterize the backscattering of the ground objects, and then adopts a nonlinear normalization method to normalize different types of polarization features to obtain complete polarization information, and uses a trained convolutional neural network to accurately classify ground objects according to the complete polarization information. This method uses complete polarization information to classify ground objects, and at the same time combines the nonlinear normalization method for the polarization features of different ground objects to improve the accuracy of classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flow chart of a polarization SAR classification method based on complete polarization information input in one embodiment; Figure 2 It is a schematic block diagram of a process of performing ground object classification using polarimetric SAR classification based on complete polarization information input in one embodiment; Figure 3 is a schematic diagram of a structure of a double-branch network based on complex numbers and real numbers in one embodiment; Figure 4A complete full-polarization SAR image used in an experiment and the corresponding true value image; Figure 5 This is a diagram showing the distribution of training, validation, and test samples in an experiment. Figure 5 (a) Figure 5 (b) Figure 5 (c) and Figure 5 (d) Full polarization SAR images obtained at different times; Figure 6 This is a classification result diagram of AlexNet using this method in an experiment, where: Figure 6 (a) indicates T Schematic diagram of the classification results of the correlation coefficients of the main diagonal elements and non-main diagonal elements of the matrix. Figure 6 (b) indicates T Schematic diagram of the classification results of all matrix elements and polarization total power, Figure 6 (c) is a schematic diagram showing the classification results using the method proposed in this paper. Figure 6 (d) represents the truth map.

[0016] Figure 7 This is a classification result diagram of VGG16 using this method in an experiment, where: Figure 7 (a) indicates T Schematic diagram of the classification results of the correlation coefficients of the main diagonal elements and non-main diagonal elements of the matrix. Figure 7 (b) indicates T Schematic diagram of the classification results of all matrix elements and polarization total power, Figure 7 (c) is a schematic diagram showing the classification results using the method proposed in this paper. Figure 7 (d) represents the truth map.

[0017] Figure 8 is a structural block diagram of a polarization SAR classification device based on complete polarization information input in one embodiment; Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0019] In the prior art, when using neural networks to classify polarimetric SAR images, some algorithms indiscriminately decompose various polarization features, simply stack and combine them, and input them into the network without considering the defects of the polarization decomposition method itself. Some methods simply normalize the polarization features without considering the characteristics of the data distribution itself, and even apply some linear normalization methods to nonlinear PolSAR data. These preprocessing processes before inputting data into the neural network will make it difficult to converge in the subsequent neural network training, and the trained neural network has the problem of inaccurate classification. In this application, Figure 1 As shown, a polarimetric SAR classification method based on complete polarization information input is provided, which specifically includes the following steps: Step S100: acquiring fully polarized SAR data to be classified, where the fully polarized SAR data is data related to ground objects.

[0020] Step S110, preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, and then using a reflection symmetric decomposition method to perform polarization classification to obtain multiple polarization features that characterize backscattering of ground objects, and using a nonlinear normalization method to perform normalization processing on different types of polarization features to obtain complete polarization information.

[0021] Step S120, inputting the complete polarization information into the trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

[0022] In this method, based on the complete polarization decomposition algorithm, the features extracted at the input are used, and a nonlinear normalization algorithm is adopted to classify the PolSAR images through the classical convolutional neural network, which further improves the classification accuracy and obtains the input scheme of the deep learning method with complete polarization information.

[0023] In step S100, the method in the present application classifies the types of ground objects, that is, the fully polarized SAR data is a whole image obtained by detecting the ground from a high altitude, and its purpose is to classify different ground objects on the ground.

[0024] Furthermore, the fully polarized SAR data is the GF-3 L1A data, that is, the fully polarized raw data. In remote sensing data processing, the L1A data is generally processed by preliminary radiation correction, but there may still be noise and other problems, which require further filtering and other processing to improve the data quality. Therefore, in the subsequent preprocessing process, non-local mean filtering is performed after obtaining the L1A data, which is aimed at the L1A data itself, which is convenient for subsequent processing.

[0025] In step S110, the fully polarized SAR data is first processed by non-local mean filtering to eliminate possible noise problems, and then a polarization decomposition algorithm that can completely extract the polarization information of the ground objects, that is, a reflection symmetric decomposition method, is used to perform polarization feature processing on the processed data to obtain a variety of polarization features that can completely characterize the scattering information of different ground objects in the fully polarized SAR data.

[0026] In this embodiment, the multiple polarization features characterizing the backscattering of ground objects include 15 real polarization features and 6 complex polarization features, including power value polarization features, angle polarization features, phase polarization features, power ratio polarization features, and complex polarization features.

[0027] Specifically, the 15 real polarization scattering characteristics include: polarization total power Span , the three main diagonal elements of the polarization coherence matrix, and the power value of the volume scattering component , surface scattering power value , secondary scattering power value , the total power value of the second component of reflection symmetry decomposition , the total power value of the third component of reflection symmetry decomposition , twice the orientation angle , twice the helix angle , the power proportion of spherical scattering in the second component of reflection symmetry decomposition x , the power proportion of spherical scattering in the third component of reflection symmetry decomposition y , reflection symmetry decomposition of the second component Phase of an element a , reflection symmetry decomposition of the third component Phase of an element b .

[0028] Specifically, the six complex polarization features include: the real part and the imaginary part of the non-main diagonal elements of the polarization coherence matrix.

[0029] In this embodiment, in order to meet the requirements of the neural network input, the polarization characteristic parameters need to be normalized. Considering that polarization SAR data is essentially a nonlinear feature, a nonlinear normalization method needs to be adopted to increase the distinguishability of polarization characteristic parameters of different ground objects and normalize the distribution range of polarization characteristics to 0~1.

[0030] Furthermore, the nonlinear normalization method is used to normalize different types of polarization characteristics, including: for the total polarization power, the total polarization power is converted into a value in dB, and then a linear normalization method is used for normalization. The polarization characteristics related to the spherical scattering power value, the dihedral scattering power value, the 45° dihedral scattering power value, the volume scattering power value, the surface scattering power value, the secondary scattering power value, the second component total power value, and the third component total power value are normalized using the total polarization power. The linear normalization method is used to normalize the double orientation angle, the double spiral angle, the second component Phases of elements and third components The phase of the elements is normalized.

[0031] Specifically, before normalizing each polarization feature, the total polarization power generated by polarization decomposition needs to be Span To process. In the matrix, the total polarization power Span Value is The sum of the matrix diagonal values, that is: .

[0032] Next, in order to better express Span ,Will Span To convert to a quantity in dB, use the following formula: .

[0033] at this time, is a linear variable, and a linear normalization method is used for the variable. In one embodiment, a maximum and minimum value normalization method can be used, which is expressed as: .

[0034] In the above formula, represents the normalized physical quantity, represents the physical quantity to be processed, , They are respectively the maximum and minimum values ​​of the domain of the physical quantity to be processed.

[0035] Furthermore, due to , so, for The main diagonal elements in the matrix are processed using (in i,j express The row and column numbers of the matrix) Normalize the values ​​of the elements in the matrix.

[0036] In this embodiment, for the volume scattered power value , surface scattering power value , secondary scattering power value , the total power value of the second component , the total power value of the third component , these physical quantities are all less than the total polarization power Span Therefore, for the processing of these physical quantities, we divide by Span value to achieve normalization.

[0037] Furthermore, twice the orientation angle The value range is (-π / 2, π / 2]. Twice the helix angle φ , the range is [-π / 4, π / 4], the second component T 12 Phase of an element a , the third ingredient T 12 Phase of an element b , the value range of both is (-π, π]. Therefore, for the processing of these physical quantities, the above-mentioned linear normalization method is adopted, that is, the maximum and minimum value normalization method is adopted.

[0038] Furthermore, due to the proportion of spherical scattering power in the second component x , the power proportion of spherical scattering in the third component y， The value range of these two physical quantities is [0,1], so there is no need to normalize them.

[0039] For the six complex polarization features, the non-main diagonal elements of the polarization coherence matrix are real and imaginary parts, using and Normalization is achieved by the method, where Re() represents the real part of the complex polarization characteristic, Im () represents the imaginary part of the complex polarization characteristic.

[0040] like Figure 2 As shown, it is a schematic flowchart of the process of ground object classification using polarization SAR classification based on complete polarization information input.

[0041] In this embodiment, during the training of the convolutional neural network, when constructing the training data: on the full polarization SAR sample data, a plurality of pixels are randomly selected, and the selected pixels are used as the center according to a preset size to generate corresponding slice samples, and each slice sample is preprocessed to obtain a complete polarization information sample, and a training set is constructed based on the complete polarization information sample. In the training set, the ground feature category of each slice sample is also included as a true value label.

[0042] Specifically, when training the convolutional neural network, samples selected from the training set are randomly input in batches, and the convolutional neural network is used to calculate the probability that the sample belongs to all candidate label categories, compare the real object labels, and output the prediction results. According to the number of real labels predicted by the model and the number of misclassifications, the accuracy of the training samples and the verification samples is calculated, and the accuracy of the two is used to determine whether the training model is appropriately fitted. Whenever the accuracy on the verification set is improved, the weight at this time is retained, and when the training is completed, the weight that makes the verification set most accurate is stored. The model with the highest accuracy in the verification set is saved, and finally the data with the highest accuracy is obtained, and this model is used for testing subsequent data.

[0043] In one embodiment, the batch data size is 64, the initial value of the learning rate is 0.1, it decays once every 10 epochs, the decay coefficient is 0.1, the weight coefficient is 0.9, and the weight coefficient is 0.0005.

[0044] In this embodiment, in the convolutional neural network, the eigenvalue size can be obtained according to the complete polarization information through convolution. The fully connected layer and the Softmax function are used to determine the category to which the ground object belongs. Then, the result of the determination is filled into an empty matrix of the same size as the predicted image to obtain the classification result of the entire image.

[0045] In this embodiment, the convolutional neural network can adopt AlexNet convolutional neural network, VGG16 convolutional neural network and dual-branch network to process real and complex polarization features respectively.

[0046] like Figure 3 As shown, in this method, a dual-branch network structure based on complex and real numbers is also proposed to process real and complex polarization features respectively.

[0047] Specifically, the real number branch network is built based on the ResNet50 residual structure and the CBAM attention mechanism. The branch network consists of a convolutional module (ConvBlock) and three residual modules (RCBlock) that integrate the CBAM attention mechanism. ConvBlock first extracts the input real number parameter features, and the features extracted by the ConvBlock convolution block are recorded as , , and then after the first RCBlock convolution block, the feature , Similarly, the second feature after the RCBlock convolution block is, , , and finally the features extracted by the real branch network (Features extracted by the 4th RCBlock). Among them, Indicates element-wise multiplication. During the multiplication process, the channel attention mechanism broadcasts along the spatial dimension, It is the feature after the CBAM attention mechanism.

[0048] Specifically, the complex branch network includes three parallel complex blocks with different convolution kernel sizes. Each complex block includes two complex convolution layers (CCD) with convolution kernel sizes of 3, 7, and 11, a complex batch normalization layer (CBN), a complex activation function layer (CRL) layer, and a complex pooling layer (CMP). Specifically, the extracted feature map is input into the complex layers of three convolution kernels of different sizes, and then the extracted features are additively summed as the input features of the next complex block.

[0049] Finally, a fusion classification network is used to fuse the features extracted by the complex branch network with the features extracted by the real branch network, and then a fully connected layer is used to classify objects based on the fused features. Among them, the size adjustment is achieved by using convolution kernels of different sizes in the pooling layer.

[0050] In this paper, experiments are conducted on AlexNet convolutional neural network and VGG16 convolutional neural network to demonstrate the effectiveness of this method.

[0051] like Figure 4 As shown in FIG. 1 , the complete fully polarized SAR image used in the experiment and the corresponding true value image are shown in FIG.

[0052] like Figure 5 Figure 1 shows the distribution of training, validation, and test samples. Figures (a), (b), (c), and (d) are fully polarized SAR images acquired at different times. In the experiment, polarization images acquired at different times are used to highlight the universality of the proposed method.

[0053] like Figure 6 As shown in the figure, the result of classification using this method on the AlexNet convolutional neural network is shown. From this figure, it can be seen that the method proposed in this paper achieves the best overall classification effect and classification effect between ground object classes.

[0054] like Figure 7 As shown in the figure, the classification result after using this method on the VGG16 convolutional neural network is shown. From this figure, it can be seen that the method proposed in this paper achieves the best overall classification effect and classification effect between ground object classes.

[0055] In the polarization SAR classification method based on complete polarization information input, the polarization SAR classification method, device and equipment based on complete polarization information input perform non-local mean filtering on the full polarization SAR data to be classified, and then use the reflection symmetric decomposition method to perform polarization classification to obtain multiple polarization features that characterize the backscattering of the ground objects, and then use a nonlinear normalization method to perform normalization processing on different types of polarization features to obtain complete polarization information, and use the trained convolutional neural network to accurately classify the ground objects according to the complete polarization information. This method uses complete polarization information to classify ground objects, and at the same time combines the nonlinear normalization method for the polarization features of different ground objects to improve the accuracy of classification. This method is based on a reflection symmetric decomposition algorithm that can completely extract polarization features, extracts complete backscattering information of ground objects, and solves the problem of incomplete information input of deep learning networks.

[0056] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0057] In one embodiment, Figure 8 As shown, a polarization SAR classification device based on complete polarization information input is provided, comprising: a full polarization SAR data acquisition module 200, a preprocessing module 210 and a polarization SAR classification module 220, wherein: A fully polarized SAR data acquisition module 200 is used to acquire fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; A preprocessing module 210 is used to preprocess the full polarization SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the full polarization SAR data, and then using a reflection symmetric decomposition method to perform polarization classification to obtain multiple polarization features that characterize backscattering of ground objects, and using a nonlinear normalization method to perform normalization processing on different types of polarization features to obtain the complete polarization information; The polarization SAR classification module 220 is used to input the complete polarization information into the trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

[0058] For the specific definition of the polarization SAR classification device based on complete polarization information input, please refer to the definition of the polarization SAR classification method based on complete polarization information input above, which will not be repeated here. Each module in the above-mentioned polarization SAR classification device based on complete polarization information input can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0059] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a polarization SAR classification method based on the input of complete polarization information is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0060] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0061] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented: Acquiring fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; Preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, performing polarization classification using a reflection symmetric decomposition method to obtain multiple polarization features characterizing backscattering of ground objects, and performing normalization processing on different types of polarization features using a nonlinear normalization method to obtain the complete polarization information; The complete polarization information is input into a trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

[0062] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: Acquiring fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; Preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, performing polarization classification using a reflection symmetric decomposition method to obtain multiple polarization features characterizing backscattering of ground objects, and performing normalization processing on different types of polarization features using a nonlinear normalization method to obtain the complete polarization information; The complete polarization information is input into a trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

[0063] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0064] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0065] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the method of the present application shall be subject to the attached claims.

Claims

1. A polarimetric SAR classification method based on complete polarization information input, characterized in that: The method comprises: Acquiring fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; The fully polarized SAR data is preprocessed to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, and then performing polarization decomposition using a reflection symmetric decomposition method to extract multiple polarization features that characterize backscattering of ground objects, wherein the multiple polarization features include 15 real polarization features: total polarization power, three main diagonal elements of a polarization coherence matrix, a volume scattering component power value, a surface scattering power value, a secondary scattering power value, a total power value of the second component of reflection symmetric decomposition, a total power value of the third component of reflection symmetric decomposition, twice the orientation angle, twice the spiral angle, a power proportion of spherical scattering in the second component of reflection symmetric decomposition, a power proportion of spherical scattering in the third component of reflection symmetric decomposition, a phase of an element of the second component of reflection symmetric decomposition, and a phase of an element of the third component of reflection symmetric decomposition, with and, 6 complex polarization features: the real part and the imaginary part of the non-main diagonal elements of the polarization coherence matrix; a nonlinear normalization method is used to perform normalization processing on different types of polarization features to obtain the complete polarization information, wherein, for the polarization total power, after the polarization total power is converted into a value in dB units, a linear normalization method is used for normalization processing, and the polarization total power is used to perform nonlinear normalization processing on the polarization features related to the spherical scattering power value, the dihedral scattering power value, and the 45° dihedral scattering power, as well as the volume scattering power value, the surface scattering power value, the secondary scattering power value, the second component total power value, and the third component total power value, and the linear normalization method is used to normalize the twice orientation angle, twice the spiral angle, the phase of the second component element, and the phase of the third component element; The complete polarization information is input into a trained convolutional neural network to obtain the ground object classification result in the full polarization SAR data.

2. The polarimetric SAR classification method based on complete polarimetric information input according to claim 1, characterized in that: The fully polarized SAR data is GF-3 L1A level data.

3. The polarimetric SAR classification method based on complete polarization information input according to claim 2, characterized in that: When training a convolutional neural network, when constructing a training set: On the full polarization SAR sample data, multiple pixels are randomly selected; The selected pixel point is taken as the center according to the preset size to generate the corresponding slice sample; After preprocessing each of the slice samples, a complete polarization information sample is obtained, and the training set is constructed according to the complete polarization information sample.

4. A polarimetric SAR classification device based on complete polarization information input, characterized in that: The device comprises: A fully polarized SAR data acquisition module is used to acquire fully polarized SAR data to be classified, wherein the fully polarized SAR data is data related to ground objects; A preprocessing module, used for preprocessing the fully polarized SAR data to obtain complete polarization information, wherein the preprocessing includes performing non-local mean filtering on the fully polarized SAR data, and then adopting a reflection symmetric decomposition method to perform polarization grading to obtain multiple polarization features characterizing backscattering of ground objects, wherein the multiple polarization features include 15 real polarization features: total polarization power, three main diagonal elements of a polarization coherence matrix, a volume scattering component power value, a surface scattering power value, a secondary scattering power value, a total power value of the second component of reflection symmetric decomposition, a total power value of the third component of reflection symmetric decomposition, twice the orientation angle, twice the spiral angle, a power proportion of spherical scattering in the second component of reflection symmetric decomposition, a power proportion of spherical scattering in the third component of reflection symmetric decomposition, a phase of an element of the second component of reflection symmetric decomposition, a phase of an element of the third component of reflection symmetric decomposition, and 6 complex polarization features: real and imaginary parts of non-main diagonal elements of the polarization coherence matrix; a polarization feature processing module, configured to perform normalization processing on different types of polarization features by using a nonlinear normalization method to obtain the complete polarization information, wherein, for the polarization total power, after converting the polarization total power into a value in dB, a linear normalization method is used for normalization processing, and the polarization features related to the spherical scattering power value, the dihedral scattering power value, and the 45° dihedral scattering power, as well as the volume scattering power value, the surface scattering power value, the secondary scattering power value, the second component total power value, and the third component total power value are subjected to nonlinear normalization processing by using the polarization total power, and the twice orientation angle, twice spiral angle, the phase of the second component element, and the phase of the third component element are subjected to normalization processing by using a linear normalization method; The polarization SAR classification module is used to input the complete polarization information into the trained convolutional neural network to obtain the ground object classification results in the full polarization SAR data.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images