Deep network model self-supervised training method and device and land utilization change detection method and device
Through self-supervised training of deep network models, the positive and negative sample pairs and multi-angle contrast loss functions are constructed using multi-angle remote sensing images, which solves the problem of neglecting manual annotation dependence and remote sensing observation mechanism in the existing technology, and improves the efficiency of land use change detection and the degree of automation of the model.
Patent Information
- Application Number
- CN202411098438.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-12
AI Technical Summary
The existing land use monitoring method based on deep learning relies on a large number of manual labeling samples, which is very cost-effective, making it difficult to adapt to the acquisition speed of remote sensing images and the change speed of land use. It ignores the multi-angle, multi-spectral, and multi-scale remote sensing observation mechanism, resulting in the failure of the model. Fully describe surface properties.
The self-supervised training method of deep network model is adopted. By obtaining multi-angle remote sensing images of two time phases, positive and negative sample pairs are constructed, multi-angle contrast loss function is constructed, and the deep network model is self-supervised and trained, making full use of label-free Data, integrated into multi-angle remote sensing observation characteristics.
It effectively alleviates the dependence on manual labeling samples, improves the efficiency of land use change detection, and promotes the personalized use of deep learning in the field of remote sensing. The model parameters are small and easy to set, and is suitable for large-scale urban land use monitoring.
Smart Images

Figure CN118968223A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of remote sensing image processing, and relates to a method and device for self-supervised training of a deep network model and detection of land use changes. Background Art
[0002] Monitoring urban land use can provide data support for urban planning, construction and management, and is an important guarantee for achieving sustainable urban development. Remote sensing technology has been widely used in land use monitoring due to its advantages such as wide coverage, fine granularity, and high revisit period. Deep learning is an important means to extract land use information from remote sensing images. However, the current land use monitoring based on deep learning relies on a large number of manually annotated samples, which has a high acquisition cost and is difficult to adapt to the speed of remote sensing image acquisition and the speed of land use change. Self-supervised learning is a new machine learning paradigm that has gradually developed in recent years. It mainly uses the characteristics and structure of the data itself to learn general feature representations from a large amount of unlabeled data for downstream tasks, thereby reducing dependence on manual annotation. Self-supervised learning has been initially developed in the field of remote sensing. However, existing algorithms usually only consider the visual features of the image, ignoring the remote sensing observation mechanisms such as multi-angle, multi-spectral, and multi-scale, making the model unable to fully describe the surface properties. Summary of the invention
[0003] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method and device for self-supervised training of a deep network model and land use change detection to improve the efficiency of land use change detection.
[0004] To achieve the above object, the present invention is implemented by adopting the following technical solutions:
[0005] In a first aspect, the present invention provides a method for self-supervised training of a deep network model, comprising:
[0006] Acquire a first time phase training image and a second time phase training image, wherein the first time phase training image and the second time phase training image include multi-angle images of the same scene;
[0007] Inputting the first time phase training image and the second time phase training image into a deep network model to obtain first time phase image features and second time phase image features;
[0008] Constructing a positive sample pair using image features of the same scene and different angles in the first time phase image features and the second time phase image features, and constructing a negative sample pair using image features of different scenes in the first time phase image features and the second time phase image features;
[0009] Construct a multi-angle contrast loss function;
[0010] The deep network model is self-supervised trained according to the positive sample pairs, negative sample pairs and multi-angle contrast loss function.
[0011] Furthermore, the first time phase training image and the second time phase training image are remote sensing images.
[0012] Furthermore, each scene in the first time phase training image and the second time phase training image includes images from three perspectives, namely, a front view, a front view, and a rear view.
[0013] Furthermore, according to the positive sample pairs, the negative sample pairs and the multi-angle contrast loss function, the deep network model is self-supervised trained, including:
[0014] grouping the first phase training images by scene;
[0015] Taking each angle image of each scene as the main perspective, the positive sample pairs and negative sample pairs formed by the second phase training image are calculated through the multi-angle contrast loss function to obtain multiple angle loss values;
[0016] Add multiple angle loss values to obtain multi-angle contrast loss values;
[0017] The deep network model is trained by self-supervision according to the multi-angle contrast loss values.
[0018] Furthermore, the first phase training image The calculation formula for the contrast loss of the front view of a scene as the main perspective is:
[0019] ,
[0020] in, represents the front view, Represents the front view, Rear view. Indicates scenes; Indicates the first phase of the training image A front view of a scene; Indicates the first The scene in The image under the observation angle; represents the number of scenes in the second phase training images; To distinguish the function, the calculation formula is:
[0021] ,
[0022] in, To image Input the image features obtained by the deep network model; is the L2 norm of the feature vector; is the training parameter of the deep network model; is the hyperparameter of the deep network model;
[0023] In the first phase training image The calculation formula of the multi-angle contrast loss value of a scene is:
[0024] .
[0025] In a second aspect, the present invention further provides a deep network model self-supervised training device, the device comprising:
[0026] A training image acquisition module, used to acquire a first time phase training image and a second time phase training image, wherein the first time phase training image and the second time phase training image include multi-angle images of the same scene;
[0027] An image feature acquisition module, used to input the first time phase training image and the second time phase training image into a deep network model to acquire first time phase image features and second time phase image features;
[0028] A sample pair construction module, configured to construct a positive sample pair from image features of the same scene and different angles in the first time phase image features and the second time phase image features, and to construct a negative sample pair from image features of different scenes in the first time phase image features and the second time phase image features;
[0029] Loss function construction module, used to construct multi-angle comparison loss function;
[0030] The deep network model self-supervised training module is used to perform self-supervised training on the deep network model based on the positive sample pairs, negative sample pairs and multi-angle contrast loss function.
[0031] In a third aspect, the present invention further provides a land use change detection method, comprising:
[0032] Obtain remote sensing images of the same scene at different phases to be detected;
[0033] Inputting the remote sensing image to be detected into a pre-established land use change detection model to obtain the land use change in the remote sensing image;
[0034] Among them, the pre-training method of the land use change detection model includes the above-mentioned deep network model self-supervised training method.
[0035] In a fourth aspect, the present invention further provides a land use change detection device, comprising:
[0036] A remote sensing image acquisition module is used to acquire remote sensing images of the same scene at different time phases to be detected;
[0037] The land use change detection module is used to input the remote sensing image to be detected into a pre-established land use change detection model to obtain the land use change in the remote sensing image.
[0038] In a fifth aspect, the present invention further provides a computer device, comprising:
[0039] Memory for storing computer programs;
[0040] A processor is used to execute the computer program to implement the steps of the above-mentioned deep network model self-supervised training method or the steps of the above-mentioned land use change detection method.
[0041] In a sixth aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the steps of the above-mentioned deep network model self-supervised training method or the steps of the above-mentioned land use change detection method.
[0042] Compared with the prior art, the present invention has the following beneficial effects:
[0043] The present invention constructs multi-angle positive sample pairs and negative sample pairs through multi-view images of two phases, inputs them into a deep network for feature extraction, and constructs a multi-angle self-supervised contrast loss function to train the deep network. After the network training, it is applied to the downstream task of land use change detection, realizing self-supervised learning and feature expression of the deep network, making full use of the large amount of unlabeled data that can be obtained, and effectively alleviating the problem of excessive reliance on manually labeled samples in the current deep learning-based land use change detection method; it incorporates multi-angle remote sensing observation characteristics, mines the unique content and structural characteristics of remote sensing data, and promotes the personalized use of deep learning in the remote sensing field; the model requires fewer parameters to be manually adjusted and is relatively easy to set, has a high degree of automation, and can be widely used in large-scale urban land use monitoring. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 A flowchart of a self-supervised training method for a deep network model provided by an embodiment of the present invention;
[0045] Figure 2 Schematic diagram of the framework of self-supervised training of a deep network model in an embodiment of the present invention;
[0046] Figure 3 A schematic diagram of the structure of a deep network model self-supervised training device provided by an embodiment of the present invention;
[0047] Figure 4A schematic diagram of a flow chart of a land use change detection method provided by an embodiment of the present invention;
[0048] Figure 5 Schematic diagram of the framework of the land use change detection model in the embodiment of the present invention;
[0049] Figure 6 A schematic diagram of the structure of a land use change detection device provided by an embodiment of the present invention;
[0050] Figure 7 An internal structure diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0051] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. The same reference numerals in the accompanying drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. The embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.
[0052] The term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.
[0053] Embodiment 1:
[0054] like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a self-supervised training method for a deep network model. Figure 1 The flowchart is a flowchart of the self-supervised training method of the deep network model. This flowchart only shows the logical sequence of the method described in this embodiment. Under the premise of no conflict, in other possible embodiments of the present invention, different Figure 1 The steps shown or described are accomplished in the order shown.
[0055] The deep network model self-supervised training method provided in this embodiment can be applied to a terminal, and can be executed by a deep network model self-supervised training device, which can be implemented by software and / or hardware, and the device can be integrated in a terminal.
[0056] See also Figure 1 The method of the embodiment of the present invention specifically includes the following steps 101 to 105. Among them:
[0057] Step 101: Acquire a first time phase training image and a second time phase training image.
[0058] The first phase and the second phase refer to images acquired at two time points. In an embodiment of the present invention, the first phase training images and the second phase training images include multi-angle remote sensing images of the same scene, and each scene image includes three angles, namely, a front view, a front view, and a rear view.
[0059] In addition, the embodiment of the present invention also performs geometric registration on the acquired first phase training image and the second phase training image. Geometric registration is used to align image data acquired at different times, different bands or different sensors in spatial position and direction so that the same feature points in the image can be accurately matched. This process is particularly important in the field of remote sensing because it involves integrating image data from different sources to ensure the accuracy and consistency of spatial information. Geometric registration of images can be achieved through remote sensing image processing software such as ENVI, ERDAS, ArcGIS, etc. The original image is then cropped into several fixed-size image blocks to meet the input requirements of the deep network model. The cropped image blocks are usually 256×256 or 512×512 in size.
[0060] Step 102: Input the first time phase training image and the second time phase training image into a deep network model to obtain first time phase image features and second time phase image features.
[0061] The embodiment of the present invention is Figure 2 The model structure shown in the figure is used as an example for self-supervised training. The model consists of two parts: an encoder and a mapping unit. The encoder is used to extract the semantic features of the input image. It can use the deep network structure commonly used in image interpretation, such as ResNet-50 based on convolutional layers and Swin Transformer based on Transformer structure. The mapping unit takes the encoder output features after global spatial pooling as input to project the features. It is usually a multi-layer perceptron (MLP) composed of fully connected layers.
[0062] Step 103: constructing a positive sample pair using the image features of the same scene and different angles in the first time phase image features and the second time phase image features, and constructing a negative sample pair using the image features of different scenes in the first time phase image features and the second time phase image features.
[0063] Step 104: Construct a multi-angle contrast loss function.
[0064] The goal of the multi-angle contrast loss function is to shorten the distance between positive samples and increase the distance between negative samples.
[0065] Step 105: Perform self-supervised training on the deep network model according to the positive sample pairs, negative sample pairs and multi-angle contrast loss function.
[0066] Step 105 specifically includes: grouping the first phase training images according to scenes; taking each angle image of each scene as the main perspective, calculating the contrast loss of the positive sample pairs and negative sample pairs composed of the positive sample pairs and the negative sample pairs formed by the second phase training images through a multi-angle contrast loss function, and obtaining multiple angle loss values; adding the multiple angle loss values to obtain a multi-angle contrast loss value; and performing self-supervised training on the deep network model according to the multi-angle contrast loss value.
[0067] The first phase training image Taking a scene as an example, the process of calculating its contrast loss includes:
[0068] First, calculate the first phase training image The contrast loss value of the front view of a scene is calculated as:
[0069] ,
[0070] in, represents the front view, Represents the front view, Rear view. Indicates scenes; Indicates the first phase of the training image A front view of a scene; Indicates the first The scene in The image under the observation angle; represents the number of scenes in the second phase training images; To distinguish the function, the calculation formula is:
[0071] ,
[0072] in, To image Input the image features obtained by the deep network model; is the L2 norm of the feature vector; is the training parameter of the deep network model; is the hyperparameter of the deep network model.
[0073] Then, the first phase training image is calculated in the same way as above. The contrast loss values of the front and back views of a scene.
[0074] Finally, the first phase training image is calculated. The sum of the contrast loss values of the front view, front view and back view of a scene is calculated as:
[0075] ,
[0076] The scene in the first phase training image Multi-angle comparison loss value.
[0077] According to the constructed loss function, the deep network model is trained using the gradient descent method until the network converges. Commonly used optimizers for network training are SGD, RMSprop, Adam, etc., which can be directly called in common deep learning frameworks such as Tensorflow and Pytorch.
[0078] The self-supervised training method of the deep network model of the present invention constructs a multi-angle self-supervised contrast loss function to train the network with the goal of shortening the distance between positive sample pairs and widening the distance between negative sample pairs. This allows the deep network to learn the remote sensing observation characteristics of the same scene from different perspectives and achieve perspective-independent deep semantic feature expression. In the application of land use change detection, it helps to suppress pseudo-changes caused by differences in observation angles between two phases. After multi-angle self-supervised pre-training, the network learns rich feature representations of land objects from a large amount of unlabeled data, and its encoder can be applied to downstream tasks of land use change detection. By using unsupervised change detection methods or fine-tuning the network with a small number of samples, a more ideal land use change detection result can be obtained.
[0079] Embodiment 2:
[0080] Based on the same inventive concept as Example 1, the embodiment of the present invention also provides a deep network model self-supervised training device for implementing the above-mentioned deep network model self-supervised training method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the detection model training device embodiment provided below can refer to the limitations of the deep network model self-supervised training method above, and will not be repeated here.
[0081] like Figure 3 As shown, an embodiment of the present invention provides a deep network model self-supervised training device, comprising:
[0082] A training image acquisition module, used to acquire a first time phase training image and a second time phase training image, wherein the first time phase training image and the second time phase training image include multi-angle images of the same scene;
[0083] An image feature acquisition module, used to input the first time phase training image and the second time phase training image into a deep network model to acquire first time phase image features and second time phase image features;
[0084] A sample pair construction module, configured to construct a positive sample pair from image features of the same scene and different angles in the first time phase image features and the second time phase image features, and to construct a negative sample pair from image features of different scenes in the first time phase image features and the second time phase image features;
[0085] Loss function construction module, used to construct multi-angle comparison loss function;
[0086] The deep network model self-supervised training module is used to perform self-supervised training on the deep network model based on the positive sample pairs, negative sample pairs and multi-angle contrast loss function.
[0087] Embodiment 3:
[0088] like Figure 4 and Figure 5 As shown, an embodiment of the present invention provides a method for detecting land use changes. Figure 4 The flowchart of the land use change detection method is shown in FIG. 1 . This flowchart only shows the logical sequence of the method described in this embodiment. In other possible embodiments of the present invention, different methods may be used without conflict. Figure 4 The steps shown or described are accomplished in the order shown.
[0089] The land use change detection method provided in this embodiment can be applied to a terminal, and can be executed by a land use change detection device, which can be implemented by software and / or hardware, and can be integrated in a terminal.
[0090] See also Figure 4 , the method of the embodiment of the present invention specifically includes:
[0091] Step 301: Acquire remote sensing images of the same scene at different time phases to be detected.
[0092] Step 302: Input the remote sensing image to be detected into a pre-established land use change detection model to obtain the land use change in the remote sensing image.
[0093] Among them, Figure 5 As shown, the pre-training method of the land use change detection model of the present invention includes the deep network model self-supervised training method in the aforementioned embodiment, and the trained detection model is applied to the downstream task of land use change detection to obtain the change detection result.
[0094] Land use change detection can be achieved using the following two methods:
[0095] (1) For the two phases of input remote sensing images, the image features of the two phases output by the encoder are directly differentiated to obtain the change intensity map :
[0096] ,
[0097] in, and are the input remote sensing images of two phases, is the trained network encoder, is the number of features extracted by the encoder. After normalization to the range of [0,1], the Otsu method is used for threshold segmentation to obtain the change detection results.
[0098] (2) Based on the encoder, an additional change detection discriminant network (such as BiT) is introduced, and a small number of land use change detection samples are used to fine-tune the overall network (i.e., based on the self-supervised pre-trained encoder, the network is further trained using the cross entropy loss function). Then, the remote sensing images of the two time phases are input into the fine-tuned network to obtain the change detection results.
[0099] Embodiment 4:
[0100] Based on the same inventive concept as in Example 3, the embodiment of the present invention further provides a land use change detection device for implementing the above-mentioned land use change detection method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the land use change detection device embodiment provided below can refer to the limitations of the land use change detection method above, and will not be repeated here.
[0101] like Figure 6 As shown, an embodiment of the present invention provides a land use change detection device, comprising:
[0102] A remote sensing image acquisition module is used to acquire remote sensing images of the same scene at different time phases to be detected;
[0103] The land use change detection module is used to input the remote sensing image to be detected into a pre-established land use change detection model to obtain the land use change in the remote sensing image.
[0104] Embodiment 5:
[0105] The embodiment of the present invention further provides a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication 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, the deep network model self-supervised training method or the land use change detection method in the aforementioned embodiment is implemented.
[0106] Those skilled in the art will understand that Figure 7 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.
[0107] Embodiment 6:
[0108] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned deep network model self-supervised training method or land use change detection method.
[0109] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products, and therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0110] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0111] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0113] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.
Claims
1. A self-supervised training method for a deep network model, characterized in that: include: Acquire a first time phase training image and a second time phase training image, wherein the first time phase training image and the second time phase training image include multi-angle images of the same scene; Inputting the first time phase training image and the second time phase training image into a deep network model to obtain first time phase image features and second time phase image features; Constructing a positive sample pair using image features of the same scene and different angles in the first time phase image features and the second time phase image features, and constructing a negative sample pair using image features of different scenes in the first time phase image features and the second time phase image features; Construct a multi-angle contrast loss function; The deep network model is self-supervised trained according to the positive sample pairs, negative sample pairs and multi-angle contrast loss function.
2. The deep network model self-supervised training method according to claim 1, characterized in that: The first phase training image and the second phase training image are remote sensing images; The method further includes: geometrically registering the first phase training image and the second phase training image.
3. The deep network model self-supervised training method according to claim 1, characterized in that: Each scene in the first phase training image and the second phase training image includes images from three perspectives, namely, a front view, a front view, and a rear view.
4. The deep network model self-supervised training method according to claim 3, characterized in that: According to the positive sample pairs, the negative sample pairs and the multi-angle contrast loss function, the deep network model is self-supervised trained, including: grouping the first phase training images by scene; Taking each angle image of each scene as the main perspective, the positive sample pairs and negative sample pairs formed by the second phase training image are calculated through the multi-angle contrast loss function to obtain multiple angle loss values; Add multiple angle loss values to obtain multi-angle contrast loss values; Based on the multi-angle contrast loss values, the deep network model is trained self-supervised.
5. The deep network model self-supervised training method according to claim 4, characterized in that: The first phase training image The calculation formula for the contrast loss of the front view of a scene as the main perspective is: , in, represents the front view, Represents the front view, Rear view. Indicates scenes; Indicates the first phase of the training image A front view of a scene; Indicates the first The scene in The image under the observation angle; represents the number of scenes in the second phase training images; To distinguish the function, the calculation formula is: , in, To image Input the image features obtained by the deep network model; is the L2 norm of the feature vector; is the training parameter of the deep network model; is the hyperparameter of the deep network model; Scenes in the first phase training images The calculation formula of the multi-angle contrast loss value is: 。 6. A self-supervised training device for a deep network model, characterized in that: include: A training image acquisition module, used to acquire a first time phase training image and a second time phase training image, wherein the first time phase training image and the second time phase training image include multi-angle images of the same scene; An image feature acquisition module, used to input the first time phase training image and the second time phase training image into a deep network model to acquire first time phase image features and second time phase image features; A sample pair construction module, configured to construct a positive sample pair from image features of the same scene and different angles in the first time phase image features and the second time phase image features, and to construct a negative sample pair from image features of different scenes in the first time phase image features and the second time phase image features; Loss function construction module, used to construct multi-angle comparison loss function; The deep network model self-supervised training module is used to perform self-supervised training on the deep network model based on the positive sample pairs, negative sample pairs and multi-angle contrast loss function.
7. A land use change detection method, characterized in that: include: Obtain remote sensing images of the same scene at different phases to be detected; Inputting the remote sensing image to be detected into a pre-established land use change detection model to obtain the land use change in the remote sensing image; Among them, the pre-training method of the land use change detection model includes the self-supervised training method of the deep network model as described in any one of claims 1 to 5.
8. A land use change detection device, characterized in that: include: A remote sensing image acquisition module is used to acquire remote sensing images of the same scene at different time phases to be detected; The land use change detection module is used to input the remote sensing image to be detected into a pre-established land use change detection model to obtain the land use change in the remote sensing image.
9. A computer device, characterized in that: include: Memory for storing computer programs; A processor, used to execute the computer program to implement the steps of the deep network model self-supervised training method described in any one of claims 1 to 5 or the steps of the land use change detection method described in claim 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the steps of the deep network model self-supervised training method described in any one of claims 1 to 5 or the steps of the land use change detection method described in claim 7.
Citation Information
Patent Citations
Mammary gland molybdenum target image segmentation method and system based on multi-view self-supervised deep learning
CN115170505A
Image classification model training method, image classification method, equipment and medium
CN115496955A
Remote sensing image building change detection method, device and equipment and storage medium
CN116310776A
Optical remote sensing image self-supervised contrast learning change detection method and device
CN117576517A
Image recognition model training method, and interaction method and device based on image recognition
CN117809084A
Cited By
Network training method and device for light variation curve data
CN121117521A
