Deep learning-based phase unwrapping method and apparatus, device, and medium

Abstract

Embodiments of the present application provide a deep learning-based phase unwrapping method and apparatus, a device, and a medium. The method comprises: acquiring a wrapped phase map; and inputting the wrapped phase map into a pre-trained deep learning network model and performing phase unwrapping, such that the deep learning network model outputs, on the basis of a non-linear mapping relationship between wrapped phases and absolute phases, an absolute phase map corresponding to the wrapped phase map, wherein the deep learning network model adopts a U-Net framework and comprises an encoder, a decoder, a deep-supervision feature fusion module, and an atrous spatial pooling pyramid module, and the encoder and the decoder comprise Inception modules. On this basis, even in high-noise environments and in the presence of phase discontinuity, the embodiments of the present application enable a trained deep learning network model to predict a high-quality absolute phase map in just one step, thereby achieving accurate phase unwrapping.

Description

Phase unwrapping method and device based on deep learning, equipment and medium TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of phase unwrapping, in particular to a phase unwrapping method and device based on deep learning, equipment and medium. BACKGROUND

[0002] Phase unwrapping is an important signal processing problem in many applications, and the purpose is to recover the original phase from the wrapped phase. Traditional phase unwrapping algorithms are not robust to noise, and cannot complete accurate phase unwrapping when the noise is high and the phase is discontinuous. SUMMARY

[0003] Embodiments of the present application provide a phase unwrapping method and device based on deep learning, equipment and medium, which uses a U-Net framework of a deep learning network for phase unwrapping. By establishing a nonlinear mapping between the wrapped phase and the absolute phase, in a high noise environment and a phase discontinuity situation, a trained deep learning network model can predict a high-quality absolute phase map in one step, thereby completing accurate phase unwrapping.

[0004] In a first aspect, embodiments of the present application provide a phase unwrapping method based on deep learning, comprising:

[0005] obtaining a wrapped phase map;

[0006] inputting the wrapped phase map into a pre-trained deep learning network model for phase unwrapping, so that the deep network learning model outputs an absolute phase map corresponding to the wrapped phase map according to the nonlinear mapping relationship between the wrapped phase and the absolute phase, wherein the deep learning network model adopts a U-Net framework, and the deep learning network model includes an encoder, a decoder, a deep supervision feature fusion module and a dilated spatial convolution pyramid module, and the encoder and the decoder include an Inception module.

[0007] In some embodiments, the training method of the deep learning network model comprises:

[0008] obtaining a data set, the data set including a noise-wrapped phase map and a true absolute phase map;

[0009] inputting the noise-wrapped phase map into a deep learning network model to output an absolute phase map corresponding to the noise-wrapped phase map;

[0010] calculating a loss function according to the true absolute phase map and the absolute phase map corresponding to the noise-wrapped phase map;

[0011] The loss function performs parameter optimization to obtain a trained deep learning network model.

[0012] In some embodiments, the data set is obtained, including:

[0013] The data set is generated by random matrix amplification.

[0014] In some embodiments, the data set is generated by random matrix amplification, including:

[0015] A first matrix within a preset range is randomly generated;

[0016] The first matrix is amplified to 128*128 by bicubic interpolation to obtain a real absolute phase map;

[0017] The real absolute phase map is normalized to obtain a first absolute phase;

[0018] Gaussian noise is added to the first absolute phase to obtain a second absolute phase;

[0019] The second absolute phase is wrapped to [-π, π] to obtain a noise-wrapped phase map;

[0020] The data set is generated according to the noise-wrapped phase map and the real absolute phase map.

[0021] In some embodiments, the encoder includes four down-sampling operations, each down-sampling operation including an Inception module, the Inception module using multiple different branches to process a feature map in parallel, and using a residual structure to splice the different branches; the 1x1 convolution in the encoder is used to reduce the dimension of the channel number, and a batch normalization layer is used for regularization after each convolution layer; the encoder also uses a skip connection to link the feature mapping in the encoding process and the decoding process.

[0022] In some embodiments, the bottleneck layer of the deep learning network model uses the empty space convolution pyramid module to perform multi-scale extraction on deep features.

[0023] In some embodiments, the decoder includes four up-sampling operations, the feature map after up-sampling is input into the Inception module, and then a 3x3 convolution and a ReLU activation are performed, outputting an absolute phase map, the absolute phase maps output by each convolution layer in the decoding process are spliced in the last layer of the decoding process, and finally a 1x1 convolution is used to output the final target absolute phase map.

[0024] In a second aspect, the embodiments of the present application further provide a deep learning-based phase unwrapping device, the device comprising:

[0025] an acquisition module configured to acquire a wrapped phase map;

[0026] a processing module configured to input the wrapped phase map into a pre-trained deep learning network model to perform phase unwrapping, so that the deep network learning model outputs an absolute phase map corresponding to the wrapped phase map according to a nonlinear mapping relationship between the wrapped phase and the absolute phase, wherein the deep learning network model adopts a U-Net framework, and the deep learning network model comprises an encoder, a decoder, a deep supervision feature fusion module, and a dilated spatial convolution pyramid module, and the encoder and the decoder comprise Inception modules.

[0027] In a third aspect, the embodiments of the present application further provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the deep learning-based phase unwrapping method according to the first aspect when executing the computer program.

[0028] In a fourth aspect, the embodiments of the present application further provide a computer-readable storage medium storing computer-executable instructions for executing the deep learning-based phase unwrapping method according to the first aspect.

[0029] According to the deep learning-based phase unwrapping method and device, the electronic device, and the computer-readable storage medium provided by the embodiments of the present application, the deep learning-based phase unwrapping method comprises: acquiring a wrapped phase map; inputting the wrapped phase map into a pre-trained deep learning network model to perform phase unwrapping, so that the deep network learning model outputs an absolute phase map corresponding to the wrapped phase map according to a nonlinear mapping relationship between the wrapped phase and the absolute phase, wherein the deep learning network model adopts a U-Net framework, and the deep learning network model comprises an encoder, a decoder, a deep supervision feature fusion module, and a dilated spatial convolution pyramid module, and the encoder and the decoder comprise Inception modules. Based on this, the embodiments of the present application use the deep learning network U-Net framework to perform phase unwrapping, and by establishing a nonlinear mapping relationship between the wrapped phase and the absolute phase, a high-quality absolute phase map can be predicted by using the trained deep learning network model in one step in a high-noise environment and in a phase discontinuity situation, so as to complete accurate phase unwrapping. BRIEF DESCRIPTION OF DRAWINGS

[0030] FIG. 1A is a flowchart of a deep learning-based phase unwrapping method according to an embodiment of the present application;

[0031] FIG. 1B is a flowchart of a method for training a deep learning network model according to an embodiment of the present application;

[0032] FIG. 2 is a display diagram of a partial training set according to an embodiment of the present application;

[0033] FIG. 3 is a structural diagram of a deep learning network model according to an embodiment of the present application;

[0034] FIG. 4 is a structural diagram of an Inception module according to an embodiment of the present application;

[0035] FIG. 5 is a structural diagram of a cavity spatial convolution pooling pyramid module according to an embodiment of the present application;

[0036] FIG. 6 is a comparison diagram in a noiseless phase wrapping situation according to an embodiment of the present application;

[0037] FIG. 7 is a comparison diagram in a phase wrapping situation with a standard deviation of 0.1 according to an embodiment of the present application;

[0038] FIG. 8 is a comparison diagram in a phase wrapping situation with a standard deviation of 0.2 according to an embodiment of the present application;

[0039] FIG. 9 is a comparison diagram in a discontinuous phase wrapping situation according to an embodiment of the present application;

[0040] FIG. 10 is a comparison diagram in a real phase wrapping situation according to an embodiment of the present application;

[0041] FIG. 11 is a schematic diagram of a deep learning-based phase unwrapping device according to an embodiment of the present application;

[0042] FIG. 12 is a schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0043] To make the objectives, technical solutions and advantages of the present application clearer, further detailed description will be given below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.

[0044] It should be noted that although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a manner different from the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification and claims and the following drawings are used to distinguish similar objects and do not necessarily describe a specific order or sequence.

[0045] In the embodiments of the present application, the words "further", "exemplary" or "optionally" are used to indicate examples, illustrations or descriptions, and should not be interpreted as being more preferred or more advantageous than other embodiments or design schemes. The use of "further", "exemplary" or "optionally" is intended to present the relevant concept in a specific manner.

[0046] Firstly, several terms involved in the present application are analyzed:

[0047] Phase unwrapping: Due to the periodic nature of phase, when the phase change of a signal exceeds 2π (i.e. 360 degrees), phase jump phenomenon occurs, which leads to the loss or error of phase information. Phase unwrapping is a technique to solve this problem, which maps the phase value back to its original range through mathematical methods, thereby restoring the continuity of the phase.

[0048] U-Net: is a kind of deep convolutional neural network architecture, named for its unique U-shaped design. It consists of two parts: a contracting encoding path (downsampling) and a symmetric expanding path (upsampling). In the encoding path, the network gradually reduces the spatial dimensions of the image while increasing the number of feature channels, which helps to extract more abstract features. In the decoding path, the network gradually restores the spatial dimensions of the image while reducing the number of feature channels, which helps to restore the details of the image. A key feature of U-Net is the skip connection, which directly connects the output of different layers in the encoder to the corresponding layers in the decoder. This design allows the network to retain more location information and details during upsampling, which is crucial for accurate image reconstruction.

[0049] In order to more conveniently describe the working principle of the embodiments of the present application in the following, the introduction of the related technical scene is given first.

[0050] Traditional phase unwrapping methods, such as path tracking method and minimum norm method, the core is to avoid the negative impact of invalid points in wrapped phase image as much as possible. The pros and cons of the unwinding stage of the path tracking method depend on the optimal path selection, but when there is high noise or complex structure in the wrapped phase, it is difficult to determine the appropriate integration path. The minimum norm method has anti-noise ability, but because of its large amount of calculation, this method is not suitable for real-time measurement. Therefore, the traditional phase unwrapping method is only suitable for normal cases where noise is not too serious, phase is continuous and not mixed. It should be pointed out that if a pixel is misclassified, the error of the pixel will be at least an integer multiple of 2π, which is unacceptable in practical applications.

[0051] Based on this, the application provides a deep learning-based phase unwrapping method and device, equipment and medium. Wherein, the deep learning-based phase unwrapping method comprises: obtaining a wrapped phase map; inputting the wrapped phase map into a pre-trained deep learning network model for phase unwrapping, so that the deep network learning model outputs an absolute phase map corresponding to the wrapped phase map according to the nonlinear mapping relationship between the wrapped phase and the absolute phase, wherein the deep learning network model adopts a U-Net framework, and the deep learning network model comprises an encoder, a decoder, a deep supervision feature fusion module and a hollow spatial convolution pyramid module, and the encoder and the decoder comprise an Inception module. Based on this, the embodiment of the application uses the deep learning network U-Net framework for phase unwrapping, establishes the nonlinear mapping between the wrapped phase and the absolute phase, and uses the trained deep learning network model to predict a high-quality absolute phase map in one step under the condition of high noise environment and phase discontinuity, so as to complete accurate phase unwrapping.

[0052] The embodiments of the application will be further described below with reference to the drawings.

[0053] As shown in FIG. 1A, FIG. 1A is a flowchart of a deep learning-based phase unwrapping method provided by an embodiment of the application, which can comprise but is not limited to steps S101-S102.

[0054] Step S101: obtaining a wrapped phase map;

[0055] Step S102: inputting the wrapped phase map into a pre-trained deep learning network model for phase unwrapping, so that the deep network learning model outputs an absolute phase map corresponding to the wrapped phase map according to the nonlinear mapping relationship between the wrapped phase and the absolute phase, wherein the deep learning network model adopts a U-Net framework, and the deep learning network model comprises an encoder, a decoder, a deep supervision feature fusion module and a hollow spatial convolution pyramid module, and the encoder and the decoder comprise an Inception module.

[0056] The wrapped phase image is obtained, and the wrapped phase image is input into a pre-trained deep learning network model for phase unwrapping. The pre-trained deep learning network model can output an absolute phase image corresponding to the wrapped phase image according to a nonlinear mapping relationship between the wrapped phase and the absolute phase. The deep learning network model adopts a U-Net framework, and the deep learning network model includes an encoder, a decoder, a deep supervision feature fusion module and a dilated spatial convolution pyramid module, forming a U-shaped structure. This improved U-Net network structure can effectively extract and fuse multi-scale features, which helps to retain edge and detail information in the image. The encoder and the decoder include an Inception module, which is a module that can increase the depth and width of the model and effectively reduce the number of parameters.

[0057] It can be understood that the application uses the U-Net framework of deep learning network for phase unwrapping. By establishing a nonlinear mapping between the wrapped phase and the absolute phase, the trained deep learning network model can predict a high-quality absolute phase image in one step in a high-noise environment and in a phase discontinuity situation, thereby completing accurate phase unwrapping.

[0058] It can be understood that the application applies deep learning technology for phase unwrapping, which not only improves the performance of the deep network learning model in terms of noise interference resistance, generalization ability and robustness, but also successfully realizes phase unwrapping with high accuracy.

[0059] As shown in FIG. 1B, the training method of the deep learning network model includes but is not limited to steps S201 to S204.

[0060] Step S201: Obtain a data set, the data set including a noise-wrapped phase image and a true absolute phase image;

[0061] Step S202: input the noise-wrapped phase image into the deep learning network model to output an absolute phase image corresponding to the noise-wrapped phase image;

[0062] Step S203: calculate a loss function according to the true absolute phase image and the absolute phase image corresponding to the noise-wrapped phase image;

[0063] Step S204: parameter optimization is performed on the loss function to obtain a trained deep learning network model.

[0064] It can be understood that, for data set generation, unlike traditional phase unwrapping algorithms, the deep learning-based phase unwrapping algorithm needs to be trained with a large amount of data, but it is difficult to collect a large amount of wrapped phase data and matched unwrapped phase data in actual measurement. Therefore, the application generates a data set by random matrix amplification. Part of the training set in the data set is shown in FIG. 2.

[0065] First, a matrix A ranging from 2*2 to 8*8 is randomly generated, and then it is enlarged to 128*128 by bicubic interpolation to obtain the absolute phase. The absolute phase Φ(x, y) obtained by bicubic interpolation of the matrix A is expressed as:

[0066] where A(i, j) is the value of the matrix A at point (i, j), and the weight w is calculated according to the weight function of bicubic interpolation. Then, the absolute phase Φ(x, y) is normalized to 10 to 40, where 50% of the values are distributed between 10 and 30, 20% of the values are distributed between 30 and 35, and 30% of the values are distributed between 35 and 40. After normalization, Gaussian noise with a standard deviation of 0.01 to 0.2 is added to the absolute phase. The absolute phase Φ(x, y) is wrapped into [-π, π] to obtain the wrapped phase Ψ(x, y), which is expressed as:

[0067] It can be understood that the U-Net algorithm is characterized by a downsampling path called an encoder (Encoder) and an upsampling path called a decoder (Decoder), which are connected to each other through skip connections (Skip Connections). This structure enables the U-Net to capture multi-scale information and perform fine image reconstruction in the decoder. At this time, in network training, the objective function to be solved can be expressed as:

[0068] As shown in FIG. 3, the deep learning network model proposed in the present application has an encoder (Encoder) and a decoder (Decoder) as the basic structure, and uses an Inception module as the basic module in the deep learning network model, as shown in FIG. 4. The deep learning network model includes a five-level Inception encoder, a five-level Inception decoder, a deep supervision feature fusion module, and an atrous spatial pyramid pooling (ASPP) module, forming a U-shaped structure, wherein the structure of the ASPP module is shown in FIG. 5. This improved U-Net network structure can effectively extract and fuse multi-scale features, which helps to preserve edge and detail information in the image.

[0069] The encoder in the deep learning network model contains four down-sampling operations, each of which contains an Inception module, which can increase the depth and width of the model and effectively reduce the number of parameters. It uses multiple different branches to process a picture in parallel, and finally uses a residual structure to splice these different branches to ensure that the network still retains better performance in the deep layer. Among them, the role of the 1x1 convolution kernel is to reduce the dimension of the channel number, which is used to reduce the parameter amount of the model. After each convolution layer, a batch normalization (Batch Normalization) layer is used for regularization to avoid possible gradient disappearance and gradient explosion problems. In addition, the skip connection (Skip Connections) method is used to link the feature maps in the encoding process and the decoding process. The atrous spatial pyramid pooling (ASPP) module is used at the bottleneck layer of the network to extract multi-scale features from the deep layer. The decoder contains four up-sampling steps, and after the feature map is input into the Inception module after up-sampling, a 3x3 convolution and a ReLU activation are performed to output an absolute phase image. In the last layer of the decoding process, the absolute phase images output by each layer are spliced, and finally a 1x1 convolution is performed to output the final absolute phase image.

[0070] In order to evaluate the performance of the deep learning network model in multiple scales, the following tests are conducted on the deep learning network model, and all test results are evaluated comprehensively using the RMSE, PSNR and SSIM indicators.

[0071] In FIG. 6, (a) represents a noise-free wrapped phase image, i.e., the input of the deep learning network model, (b) represents the absolute phase image corresponding to (a), (c) and (d) represent the output of the deep learning network model and the error map, respectively, and (e) and (f) represent the absolute phase image and the error map of the output of the U-Net, respectively.

[0072] In FIG. 7, (a) represents a wrapped phase image with a noise standard deviation of 0.1, i.e., the input of the deep learning network model, (b) represents the absolute phase image corresponding to (a), (c) and (d) represent the output of the deep learning network model and the error map, respectively, and (e) and (f) represent the absolute phase image and the error map of the output of the U-Net, respectively.

[0073] In FIG. 8, (a) represents a wrapped phase image with a noise standard deviation of 0.2, i.e., the input of the deep learning network model, (b) represents the absolute phase image corresponding to (a), (c) and (d) represent the output of the deep learning network model and the error map, respectively, and (e) and (f) represent the absolute phase image and the error map of the output of the U-Net, respectively.

[0074] In FIG. 9, (a) represents a discontinuous wrapped phase map, i.e. the input of the deep learning network model, (b) represents the absolute phase map corresponding to (a), (c), (d) represent the output of the deep learning network model and the error map respectively, (e), (f) represent the absolute phase map and the error map of the output of the U-Net respectively.

[0075] In FIG. 10, (a) represents the real wrapped phase map, i.e. the input of the deep learning network model, (b) represents the absolute phase map corresponding to (a), (c), (d) represent the output of the deep learning network model and the error map respectively, (e), (f) represent the absolute phase map and the error map of the output of the U-Net respectively.

[0076] It should be noted that, generally speaking, when the unwrapping error is less than 5%, it can be considered that the unwrapping is correct. Therefore, the error map is generated by calculating the relative error between the model output and the true value (gt), and in the error map, any pixel point with a relative error exceeding 5% is marked as an error point.

[0077] From the data in Tables 1, 2, 3, 4, 5, it can be seen that the deep learning network model proposed in the present application is superior to the U-Net in various evaluation indicators. The experimental results fully show the advantages of the deep learning network model proposed in the present application in terms of model noise resistance, model generalization and model robustness.

[0078] Table 1 Average PSNR (dB), RMSE, SSIM of model unwrapping of noiseless phase wrapped map

[0079] Table 2 Average PSNR (dB), RMSE, SSIM of model unwrapping of noise phase wrapped map with a standard deviation of 0.1

[0080] Table 3 Average PSNR (dB), RMSE, SSIM of model unwrapping of noise phase wrapped map with a standard deviation of 0.2

[0081] Table 4 Average PSNR (dB), RMSE, SSIM of model unwrapping of discontinuous phase wrapped map

[0082] Table 5 Average PSNR (dB), RMSE, SSIM of model unwrapping of real phase wrapped map

[0083] Based on this, the application proposes a deep learning-based phase unwrapping method. Using the trained model, the absolute phase image can be solved in one step. Compared with other types of deep learning methods, the deep learning network model exhibits superiority in robustness in dealing with noise and in phase unwrapping performance in dealing with complex situations such as phase discontinuity. This means that even in scenes with large noise or dramatic phase changes, the deep learning network model can maintain high unwrapping accuracy and stability.

[0084] It should be noted that in view of the limitations of the current change detection technology, the application innovatively introduces a non-change area information suppression module, which takes into account the influence of background context information, and the cross-attention mechanism therein focuses more on the relevance between regions. The application has significant innovation in the field of change detection technology, providing a more reliable and efficient solution for modern building monitoring and illegal building inspection.

[0085] In addition, as shown in FIG. 11, one embodiment of the application also discloses a deep learning-based phase unwrapping device, which comprises:

[0086] The acquisition module 110 is configured to acquire a wrapped phase image.

[0087] The processing module 120 is configured to input the wrapped phase image into a pre-trained deep learning network model for phase unwrapping, so that the deep learning network model outputs an absolute phase image corresponding to the wrapped phase image according to the nonlinear mapping relationship between the wrapped phase and the absolute phase. The deep learning network model adopts a U-Net framework, and the deep learning network model comprises an encoder, a decoder, a deep supervision feature fusion module, and a dilated spatial convolution pyramid module. The encoder and the decoder comprise Inception modules.

[0088] The deep learning-based phase unwrapping device of the embodiment of the application is used to execute the deep learning-based phase unwrapping method in the above embodiment, and the specific processing process is the same as that of the deep learning-based phase unwrapping method in the above embodiment, which will not be repeated here.

[0089] In addition, as shown in FIG. 12, one embodiment of the application also discloses an electronic device, comprising: at least one processor 210; at least one memory 220 for storing at least one program; and when the at least one program is executed by the at least one processor 210, the deep learning-based phase unwrapping method in any of the above embodiments is implemented.

[0090] In addition, one embodiment of the present application also discloses a computer readable storage medium, wherein computer executable instructions are stored, and the computer executable instructions are used for executing the deep learning based phase unwrapping method in any of the foregoing embodiments.

[0091] The system architecture and application scenarios described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. It can be known by those skilled in the art that, with the evolution of system architecture and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0092] It can be understood by those skilled in the art that all or some steps in the method disclosed above, and the function modules / units in the system and the device can be implemented as software, firmware, hardware and appropriate combinations thereof.

[0093] In a hardware implementation, the division between the function modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component can have multiple functions, or one function or step can be performed by several physical components in cooperation. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software can be distributed on a computer readable medium, which can include computer storage media (or non-transitory media) and communication media (or transitory media). As is well known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media typically includes computer readable instructions, data structures, program modules or other data in modulated data signals such as carrier waves or other transport mechanisms, and can include any information delivery medium.

[0094] As used in this description, the terms "component," "module," "system," and the like are intended to refer to a computer-related entity, either hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, or a computer. By way of illustration, both an application running on a computing device and the computing device can be a component. One or more components can reside within a process or thread of execution and a component can be localized, partially or wholly, in one computer or distributed between two or more computers. Also, these components can execute from various computer-readable media having various data structures stored thereon. The components can communicate by way of local or remote processes such as in accordance with a signal having one or more data packets (e.g., data from one component interacting with another component in a local system, distributed system, or across a network such as the Internet with other systems via the signal).

Claims

1. A deep learning-based phase unwrapping method, comprising: obtaining a wrapped phase map; inputting the wrapped phase map into a pre-trained deep learning network model for phase unwrapping, so that the deep network learning model outputs an absolute phase map corresponding to the wrapped phase map according to a nonlinear mapping relationship between the wrapped phase and the absolute phase, wherein the deep learning network model adopts a U-Net framework, the deep learning network model comprises an encoder, a decoder, a deep supervision feature fusion module and a dilated spatial convolution pyramid module, and the encoder and the decoder comprise Inception modules.

2. The method of claim 1, wherein, The training method of the deep learning network model comprises: obtaining a data set, the data set comprising a noise-wrapped phase map and a real absolute phase map; inputting the noise-wrapped phase map into the deep learning network model to output an absolute phase map corresponding to the noise-wrapped phase map; calculating a loss function according to the real absolute phase map and the absolute phase map corresponding to the noise-wrapped phase map; the loss function is optimized to obtain a trained deep learning network model.

3. The method of claim 2, wherein, The data set is obtained by: generating the data set by random matrix amplification.

4. The method of claim 3, wherein, The data set is generated by random matrix amplification, comprising: randomly generating a first matrix within a preset range; amplifying the first matrix to 128*128 by bicubic interpolation to obtain a real absolute phase map; normalizing the real absolute phase map to obtain a first absolute phase; adding Gaussian noise to the first absolute phase to obtain a second absolute phase; wrapping the second absolute phase to [-π, π] to obtain the noise-wrapped phase map; generating the data set according to the noise-wrapped phase map and the real absolute phase map.

5. The method of claim 1, wherein, The encoder comprises four downsampling operations, each downsampling operation comprising an Inception module, the Inception module using multiple different branches to process a feature map in parallel and using a residual structure to splice the different branches; the 1x1 convolution in the encoder is used to reduce the dimension of the channel, and a batch normalization layer is used for regularization after each convolution layer; the encoder also uses a skip connection to link the feature mapping in the encoding process and the decoding process.

6. The method of claim 1, wherein, The bottleneck layer of the deep learning network model uses the dilated spatial convolution pyramid module to perform multi-scale extraction on deep-level features.

7. The method of claim 1, wherein, The decoder comprises four upsampling operations, the feature map being input into the Inception module after upsampling, and then being input into a 3x3 convolution and a ReLU activation to output an absolute phase map, the absolute phase maps output by the convolution layers in the decoding process being spliced in the last layer, and finally outputting a final target absolute phase map through a 1x1 convolution.

8. A deep learning based phase unwrapping apparatus, characterized by, The apparatus comprises: an obtaining module configured to obtain a wrapped phase map; The processing module is configured to input the wrapped phase map into a pre-trained deep learning network model to perform phase unwrapping, so that the deep learning network model outputs an absolute phase map corresponding to the wrapped phase map according to a nonlinear mapping relationship between the wrapped phase and the absolute phase. The deep learning network model adopts a U-Net framework, and includes an encoder, a decoder, a deep supervision feature fusion module, and a dilated spatial convolution pyramid module. The encoder and the decoder include Inception modules.

9. An electronic device comprising: The memory, the processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the deep learning-based phase unwrapping method according to any one of claims 1 to 7. 10.A computer readable storage medium storing computer executable instructions for performing the deep learning-based phase unwrapping method according to any one of claims 1 to 7.

Images