SAR remote sensing image segmentation method and device based on DA-Unet model, medium and equipment

Abstract

Embodiments of the present application provide a SAR remote sensing image segmentation method and device based on a DA-Unet model, a medium and equipment. The method comprises: preprocessing a SAR remote sensing image to be segmented to obtain a preprocessed SAR remote sensing image, the SAR remote sensing image containing an object to be recognized; inputting the preprocessed SAR remote sensing image into a DA-Unet model that has been pre-trained to make the DA-Unet model output a segmentation image for the object to be recognized; wherein the DA-Unet model adds a ResNet101 model without a full connection layer as a second encoder on the basis of a first encoder that the original Unet model has, and the output shape of the output layer of the first encoder is the same as that of the output layer of the second encoder. The technical solution of the embodiments of the present application can improve the effectiveness of feature extraction of the SAR remote sensing image, thereby ensuring the accuracy of the image segmentation result.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and more specifically, to a SAR remote sensing image segmentation method, apparatus, medium, and device based on the DA-Unet model. Background Technology

[0002] SAR (Synthetic Aperture Radar) remote sensing imagery is a type of radar image acquired using synthetic aperture radar remote sensing technology. It can be used to obtain high-resolution images of the Earth's surface. SAR imagery is characterized by its high imaging quality, ability to penetrate clouds, and insensitivity to topographic changes, making it widely used in marine monitoring, forestry resource surveys, and urban planning. One important application of SAR imagery in marine monitoring is the segmentation of nearshore aquaculture farms. Nearshore aquaculture farms refer to artificially constructed facilities in nearshore areas primarily for the cultivation of aquatic animals and seaweed. By segmenting aquaculture farms using SAR imagery, rapid location and precise delineation of aquaculture areas can be achieved, providing a scientific basis for the protection and rational utilization of nearshore fishery resources. However, the characteristics of SAR imagery differ from multispectral data. Its reflectivity is related to factors such as the shape and texture of objects and lacks color information. Therefore, improving the effectiveness of feature extraction from SAR imagery and ensuring the accuracy of image segmentation results has become a pressing technical problem. Summary of the Invention

[0003] The embodiments of this application provide a SAR remote sensing image segmentation method, apparatus, medium, and device based on the DA-Unet model, which can at least improve the effectiveness of feature extraction from SAR remote sensing images to a certain extent, thereby ensuring the accuracy of image segmentation results.

[0004] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0005] According to one aspect of the embodiments of this application, a SAR remote sensing image segmentation method based on the DA-Unet model is provided, including:

[0006] The SAR remote sensing image to be segmented is preprocessed to obtain the preprocessed SAR remote sensing image, which contains the object to be identified.

[0007] The preprocessed SAR remote sensing image is input into the pre-trained DA-Unet model so that the DA-Unet model outputs a segmented image for the object to be identified.

[0008] The DA-Unet model adds a ResNet101 model with the fully connected layers removed as a second encoder to the original Unet model's first encoder. The output shape of the first encoder's output layer is the same as that of the second encoder's output layer.

[0009] According to one aspect of the embodiments of this application, a SAR remote sensing image segmentation device based on the DA-Unet model is provided, comprising:

[0010] The preprocessing module is used to preprocess the SAR remote sensing image to be segmented to obtain the preprocessed SAR remote sensing image, which contains the object to be identified.

[0011] The segmentation module is used to input the preprocessed SAR remote sensing image into the pre-trained DA-Unet model so that the DA-Unet model outputs a segmented image for the object to be identified.

[0012] The DA-Unet model adds a ResNet101 model with the fully connected layers removed as a second encoder to the original Unet model's first encoder. The output shape of the first encoder's output layer is the same as that of the second encoder's output layer.

[0013] According to one aspect of the embodiments of this application, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the SAR remote sensing image segmentation device method based on the DA-Unet model as described in the above embodiments.

[0014] According to one aspect of the embodiments of this application, an electronic device is provided, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the SAR remote sensing image segmentation device method based on the DA-Unet model as described in the above embodiments.

[0015] According to one aspect of the embodiments of this application, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the SAR remote sensing image segmentation device method based on the DA-Unet model provided in the above embodiments.

[0016] In some embodiments of this application, the technical solutions involve preprocessing the SAR remote sensing image to be segmented to obtain a preprocessed SAR remote sensing image containing the object to be identified. The preprocessed SAR remote sensing image is then input into a pre-trained DA-Unet model, causing the DA-Unet model to output a segmented image for the object to be identified. The DA-Unet model adds a ResNet101 model with the fully connected layers removed as a second encoder, based on the first encoder of the original Unet model. The output shape of the first encoder's output layer is the same as that of the second encoder's output layer. Thus, the first encoder can extract low-level features, such as the edges and textures of the object to be identified, while the second encoder can extract high-level features, such as the shape and contour of the object to be identified, thereby better describing the structure and shape of the object. The cooperation between the first and second encoders improves the effectiveness of feature extraction from the SAR remote sensing image, thereby ensuring the accuracy of the image segmentation results.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings:

[0019] Figure 1 A flowchart illustrating a SAR remote sensing image segmentation method based on the DA-Unet model according to an embodiment of this application is shown.

[0020] Figure 2 A schematic diagram of the structure of a DA-Unet model according to an embodiment of this application is shown;

[0021] Figure 3 A block diagram of a SAR remote sensing image segmentation apparatus based on the DA-Unet model according to an embodiment of this application is shown;

[0022] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0023] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.

[0024] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0025] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0026] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0027] Figure 1 A flowchart illustrating a SAR remote sensing image segmentation method based on the DA-Unet model according to an embodiment of this application is shown. This method can be applied to a terminal device or a server. It should be understood that the terminal device may include, but is not limited to, one or more of smartphones, tablets, laptops, or desktop computers; the server may be a physical server or a cloud server.

[0028] It should be noted that this application does not limit the number of terminal devices or servers. Depending on the implementation needs, there can be any number of terminal devices or servers. For example, the server can be a server cluster composed of multiple servers, etc.

[0029] like Figure 1 As shown, the SAR remote sensing image segmentation method based on the DA-Unet model includes at least steps S110 to S120, which are described in detail below (the following explanation uses the application of this method to a terminal device as an example, hereinafter referred to as the terminal):

[0030] In step S110, the SAR remote sensing image to be segmented is preprocessed to obtain the preprocessed SAR remote sensing image, which contains the object to be identified.

[0031] In this embodiment, the terminal can acquire captured SAR remote sensing images for image segmentation. It should be understood that image segmentation refers to the technique and process of dividing an image into regions with corresponding characteristics and extracting targets of interest. The acquired SAR remote sensing images contain objects to be identified, which can include any one of the following: nearshore aquaculture farms, urban roads, or human postures. In other words, this method can be used to segment images of any one of these three types of objects. It should be understood that the object to be identified can also be any other than the three mentioned above; those skilled in the art can determine this according to actual implementation needs, and no special limitations are made.

[0032] It should be noted that the terminal can obtain the SAR remote sensing image to be segmented from its own storage space, or from a third-party device, such as the Internet or a third-party server, without any special limitations. Furthermore, the SAR remote sensing image to be segmented can be a single image or a collection of multiple SAR remote sensing images.

[0033] After acquiring SAR remote sensing imagery, the terminal can preprocess the imagery to improve the accuracy of subsequent image segmentation. In one embodiment, this preprocessing includes sliding window overlap cropping and / or data augmentation.

[0034] Specifically, sliding window overlap cropping can segment the original SAR remote sensing image into multiple smaller images, which can then be used for training or recognition. It should be understood that because SAR remote sensing images have high resolution, directly using the entire image for training or recognition might result in an excessively large model and long training or recognition time. Sliding window overlap cropping reduces the amount of data, thereby improving recognition efficiency. In one example, the original SAR remote sensing image can be segmented into 512×512 smaller images, with a sliding window step size of 50 pixels. A certain overlap area is set between the smaller images to ensure the continuity of image information. This allows for full utilization of local image information, improving the model's ability to learn image details.

[0035] Next, the terminal can perform data augmentation on the segmented small images, including but not limited to at least one of spatial transformation, image blurring and sharpening, random noise addition, and random cropping. The spatial transformation includes random rotation, horizontal flipping, vertical flipping, random translation, and random scaling. These operations allow for the acquisition of a large amount of training data during model training, increasing data diversity and thus improving the model's generalization ability and robustness. Correspondingly, in the actual segmentation process, the SAR remote sensing images to be segmented, after undergoing the above preprocessing, are easier for the model to identify and the accuracy of the segmentation results is improved.

[0036] In step S120, the preprocessed SAR remote sensing image is input into the pre-trained DA-Unet model so that the DA-Unet model outputs a segmented image for the object to be identified; wherein, the DA-Unet model adds a ResNet101 model with the fully connected layers removed as a second encoder on the basis of the first encoder of the original Unet model, and the output shape of the output layer of the first encoder is the same as that of the output layer of the second encoder.

[0037] In this embodiment, such as Figure 2 As shown, this application proposes a DA-Unet (Dual-Encoder Atrous-Spatial-Pyramid-Pooling Unet) model, which has a first encoder in the original Unet model (the original encoder of the original Unet model is...). Figure 2 Based on the middle part of the model, a ResNet101 model with the fully connected layers removed was added as a second encoder (i.e., Figure 2 The left part of the image forms a dual encoder structure, and the output shape of the first encoder's output layer is the same as that of the second encoder's output layer. Figure 2 The right part of the image is the original decoder of the original Unet model. Because it consists of two encoders, the DA-Unet model is not a symmetrical encoder-decoder structure. The input and output image sizes of the DA-Unet model are equal, for example, both are 512×512.

[0038] The second encoder has a significantly increased network depth compared to the first encoder, resulting in a substantial improvement in its feature extraction capabilities. Specifically, the main components of the second encoder include four sets of n×Res-blocks. The second encoder uses a 7×7 convolution to correspond to the two 3×3 convolutional layers in the first layer of Unet, and then uses four sets of n×Res-blocks to correspond to the four convolutional and pooling modules of the Unet encoder. In one example, the first Res-block has n=3, an output feature map size of 128×128, and 256 convolutional kernels. The second Res-block has n=4, an output feature map size of 64×64, and 512 convolutional kernels. The third Res-block has n=23, an output feature map size of 32×32, and 1024 convolutional kernels. The fourth Res-block has n=3, an output feature map size of 16×16, and 2048 convolutional kernels.

[0039] Therefore, the first encoder can be used to extract low-level features, such as the edges and textures of the object to be identified, while the second encoder can extract high-level features, such as the shape and contour of the object to be identified, thus better describing the structure and shape of the object to be identified. The cooperation between the first encoder and the second encoder can improve the effectiveness of feature extraction from SAR remote sensing images, thereby ensuring the accuracy of image segmentation results.

[0040] It's worth noting that during the model improvement process, since the last fully connected layer of the original ResNet101 model is used for classification, it can be removed first, allowing the ResNet101 model to be used as a second encoder responsible for extracting high-level features. In one example, a 1×1 convolution and a 3×3 convolution can be added to the end of the ResNet101 model to change the number of channels, making the number of output channels and the feature map size consistent with the first encoder. It should be understood that the final output layer after forming the dual encoder should have the same shape as the output of the first encoder, thus enabling feature fusion between the second encoder and the feature maps output by the first encoder.

[0041] Furthermore, the first encoder performs four max pooling operations, reducing the feature map size four times, with a bottleneck at the bottom consisting of a 1024-channel, 32×32 feature map. Unlike the first encoder, the second encoder reduces the feature map size five times, outputting a 2048-channel, 16×16 feature map after removing the fully connected layers. Therefore, two 3×3 convolutional layers can be added to the first encoder, resulting in the same output layer feature map size for both encoders. Specifically, the two convolutional layers in the first encoder correspond to the initial 7×7 convolutional layers and max pooling layers in the second encoder, while the two convolutional layers in the second through fifth layers correspond to the four n×Res-blocks in the second encoder, where n is 3, 4, 23, and 3 respectively.

[0042] A 1×1 convolutional layer is placed between corresponding layers of the two encoders to change the number of channels, making the number of channels in the corresponding layers of the second encoder consistent with that of the first encoder. Furthermore, the features extracted from corresponding layers of the two encoders can be fused to provide richer semantic labels, enhance feature representation capabilities, and improve model performance.

[0043] based on Figure 1 The embodiment shown, in one embodiment of this application, involves adding an ASPP module to the bottleneck at the bottom of the original Unet model in the DA-Unet model.

[0044] In this embodiment, it should be noted that image segmentation methods based on the Unet model have become a research hotspot due to their small parameter count and good performance. However, in some complex scenarios, the traditional Unet model cannot cope with them. Therefore, in this application, the ASPP module is used to replace the convolutional module at the bottom bottleneck of the original Unet model in the improved DA-Unet model.

[0045] The ASPP (Atrous Spatial Pyramid Pooling) module expands the receptive field of view through parallel dilated convolutions, acquiring local and global contextual information at different scales to capture feature information at varying scales, thereby improving the accuracy of subsequent image segmentation. Thus, by setting the ASPP module, feature information at different scales can be obtained, improving the model's accuracy. Furthermore, its use of dilated convolution and pyramid pooling techniques avoids the problems of excessive model parameters and computational cost. Moreover, the ASPP module can adaptively extract features for images of different sizes, giving the model better versatility and robustness.

[0046] Specifically, in the original UNet model, the bottleneck at the bottom consists of a convolutional block combining downsampling and upsampling. This convolutional block can be replaced with an ASPP module. The ASPP module takes as input the feature map fused from the outputs of the first and second encoders, and outputs a feature map obtained through multiple parallel dilated convolution operations. Within the ASPP module, multiple dilated convolution operations at different scales are defined, each with a different dilation rate to capture feature information at different scales. Furthermore, a Batch Normalization layer and a ReLU activation function layer can be added after each dilated convolution operation to enhance the non-linear expressiveness of the features. The outputs of multiple dilated convolution operations are then concatenated to obtain a multi-channel feature map. This feature map contains multi-scale feature information, which can better describe the details and boundaries in the image.

[0047] Based on the foregoing embodiments, in one embodiment of this application, after the DA-Unet model outputs a segmented image of the object to be identified, the method further includes:

[0048] The segmented image is post-processed to obtain the target segmented image. The post-processing includes image connected component denoising and / or closing operations.

[0049] In this embodiment, the initial segmentation image output by the DA-Unet model is generally quite coarse. To obtain more accurate and refined results, the terminal can perform post-processing on the segmented image, thereby improving the accuracy of the image segmentation results. This post-processing includes image connected component denoising and / or closing operations.

[0050] Specifically, in the process of image connected component denoising, those skilled in the art can perform denoising based on a pre-set minimum connected component threshold, such as 50, 100, or 400. In one example, the minimum connected component threshold is 400 to ensure denoising effect. It should be noted that the above numbers are merely illustrative examples and are not intended to impose any special limitations.

[0051] Closing is an image morphological processing method consisting of two steps: dilation and erosion. Dilation expands parts of objects in an image, while erosion shrinks them. Closing effectively fills small holes within objects, connects adjacent objects, repairs broken contours, and smooths object boundaries while maintaining their area. It's important to note that a larger convolution kernel in the closing operation results in more pronounced dilation and erosion effects. A kernel that is too small will shrink internal holes but not completely remove them, while a kernel that is too large will cause previously separated parts to merge. In one example, to achieve the best results for the closing operation, this application sets the convolution kernel size to 7×7.

[0052] In one embodiment of this application, taking nearshore aquaculture farms as an example, a SAR remote sensing image dataset can be pre-constructed when training the DA-Unet model. Specifically, SAR remote sensing images of nearshore aquaculture farms acquired by satellite are first obtained, and the acquired SAR remote sensing images are then labeled with the nearshore aquaculture farms. During labeling, the outline of the nearshore aquaculture farm can be determined first, and then the area of ​​the nearshore aquaculture farm can be filled with a predetermined color, such as white. According to actual needs, corresponding category labels can be added to the labeled nearshore aquaculture farms; for example, the category label for nearshore aquaculture farms could be "pool." The labeling results are then saved.

[0053] It is worth noting that image resolution and pixel size need to be considered when performing annotation. Because SAR remote sensing images have large pixel sizes, annotation requires zooming in or out, which may lead to a decrease in the accuracy of the annotation results. Therefore, the image size can be adjusted appropriately during annotation to ensure the accuracy and quality of the results.

[0054] Next, the labeled SAR remote sensing images are preprocessed, including sliding window overlap cropping and data augmentation, to obtain a SAR remote sensing image dataset. This SAR remote sensing image dataset is then used to train the improved DA-Unet model.

[0055] The following describes an embodiment of the apparatus described in this application, which can be used to execute the SAR remote sensing image segmentation method based on the DA-Unet model described in the above embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the SAR remote sensing image segmentation method based on the DA-Unet model described above in this application.

[0056] Figure 3 A block diagram of a SAR remote sensing image segmentation apparatus based on the DA-Unet model according to an embodiment of this application is shown.

[0057] Reference Figure 3 As shown, a SAR remote sensing image segmentation apparatus based on the DA-Unet model according to an embodiment of this application includes:

[0058] The preprocessing module is used to preprocess the SAR remote sensing image to be segmented to obtain the preprocessed SAR remote sensing image, which contains the object to be identified.

[0059] The segmentation module is used to input the preprocessed SAR remote sensing image into the pre-trained DA-Unet model so that the DA-Unet model outputs a segmented image for the object to be identified.

[0060] The DA-Unet model adds a ResNet101 model with the fully connected layers removed as a second encoder to the original Unet model's first encoder. The output shape of the first encoder's output layer is the same as that of the second encoder's output layer.

[0061] In one embodiment of this application, in the DA-Unet model, the ASPP module is added to the bottleneck at the bottom of the original Unet model.

[0062] In one embodiment of this application, after the DA-Unet model outputs a segmented image for the object to be identified, the segmentation module is further configured to:

[0063] The segmented image is post-processed to obtain the target segmented image. The post-processing includes image connected component denoising and / or closing operations.

[0064] In one embodiment of this application, the preprocessing includes sliding window overlap cropping and / or data enhancement, wherein the data enhancement includes at least one of spatial transformation, image blurring and sharpening, random noise addition, and random cropping.

[0065] In one embodiment of this application, the object to be identified includes any one of offshore aquaculture farms, urban roads, or human postures.

[0066] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown.

[0067] It should be noted that, Figure 4 The computer system of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0068] like Figure 4 As shown, the computer system includes a Central Processing Unit (CPU) 401, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 402 or programs loaded from storage portion 408 into Random Access Memory (RAM) 403, such as performing the methods described in the above embodiments. The RAM 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An Input / Output (I / O) interface 405 is also connected to the bus 404.

[0069] The following components are connected to I / O interface 405: an input section 406 including a keyboard, mouse, etc.; an output section 407 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to I / O interface 405 as needed. A removable medium 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 410 as needed so that computer programs read from it can be installed into storage section 408 as needed.

[0070] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit (CPU) 401, it performs various functions defined in the system of this application.

[0071] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. The transmitted data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0072] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0073] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0074] In another aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the methods described in the above embodiments.

[0075] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0076] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0077] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0078] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A SAR remote sensing image segmentation method based on a DA-Unet model, characterized in that, The method comprises: preprocessing a SAR remote sensing image to be segmented to obtain a preprocessed SAR remote sensing image, the SAR remote sensing image containing an object to be recognized; inputting the preprocessed SAR remote sensing image into a DA-Unet model trained in advance to make the DA-Unet model output a segmentation image for the object to be recognized; wherein the DA-Unet model adds a ResNet101 model without a full connection layer to a second encoder on the basis of a first encoder of an original Unet model, the output shape of an output layer of the first encoder is the same as that of an output layer of the second encoder, and features extracted by corresponding layers of the two encoders are fused to provide richer semantic labels and enhance feature expression capability.

2. The method of claim 1, wherein, In the DA-Unet model, an ASPP module is added to a bottleneck at the bottom of the original Unet model.

3. The method according to any one of claims 1-2, characterized in that, The preprocessing comprises sliding window overlap cropping and / or data enhancement, and the data enhancement comprises at least one of spatial transformation, image blurring and sharpening, random noise addition, and random cropping.

4. The method according to any one of claims 1-2, characterized in that, After the DA-Unet model outputs the segmentation image for the object to be recognized, the method further comprises: postprocessing the segmentation image to obtain a target segmentation image, and the postprocessing comprises image connected domain denoising and / or closing operation.

5. The method as claimed in claim 1, wherein, The object to be recognized comprises any one of a near-shore aquaculture farm, an urban road, or a human body posture.

6. A SAR remote sensing image segmentation device based on a DA-Unet model, characterized in that, The method comprises: a preprocessing module configured to preprocess a SAR remote sensing image to be segmented to obtain a preprocessed SAR remote sensing image, the SAR remote sensing image containing an object to be recognized; a segmentation module configured to input the preprocessed SAR remote sensing image into a DA-Unet model trained in advance to make the DA-Unet model output a segmentation image for the object to be recognized; wherein the DA-Unet model adds a ResNet101 model without a full connection layer to a second encoder on the basis of a first encoder of an original Unet model, the output shape of an output layer of the first encoder is the same as that of an output layer of the second encoder, and features extracted by corresponding layers of the two encoders are fused to provide richer semantic labels and enhance feature expression capability.

7. The apparatus of claim 6, wherein, In the DA-Unet model, an ASPP module is added to a bottleneck at the bottom of the original Unet model.

8. The apparatus of any one of claims 6-7, wherein, After the DA-Unet model outputs the segmentation image for the object to be recognized, the segmentation module is further configured to: postprocess the segmentation image to obtain a target segmentation image, and the postprocessing comprises image connected domain denoising and / or closing operation.

9. A computer readable medium having stored thereon a computer program, characterized in that The computer program is executed by a processor to implement the DA-Unet model-based SAR remote sensing image segmentation method of any one of claims 1 to 5.

10. An electronic device, comprising: The method comprises: one or more processors; A storage device is configured to store one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the DA-Unet model-based SAR remote sensing image segmentation method according to any one of claims 1 to 5.

Images