Training system and method of remote sensing recognition model, device, equipment and storage medium

Abstract

The embodiment of the invention provides a remote sensing recognition model training system, a remote sensing recognition model training method, a remote sensing recognition model training device, equipment and a storage medium. The training system comprises a data management module which is used for providing a target remote sensing image to be labeled; the data annotation module is connected with the data management module and used for acquiring the target remote sensing image from the data management module and annotating the target remote sensing image; and the model training module is connected with the data labeling module and is used for training based on the labeled target remote sensing image to obtain a remote sensing recognition model, and the remote sensing recognition model comprises a small sample recognition algorithm model. The embodiment of the invention aims at training to obtain a remote sensing recognition model which is relatively high in accuracy, relatively low in cost and relatively high in generalization ability.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to a remote sensing recognition model training system, a remote sensing recognition model training method, a remote sensing recognition model training device, a computer device and a computer-readable storage medium. Background Art

[0002] Remote sensing identification is a method of obtaining and analyzing information on the earth's surface using remote sensing technology. It can obtain data such as electromagnetic wave reflection, emission or scattering of ground objects through sensors mounted on platforms such as aircraft, satellites and drones, and then identify and monitor various natural resources and environmental phenomena.

[0003] In the financial business related to agricultural insurance underwriting and verification, remote sensing recognition models are needed to identify the scope of cultivated land and crop types, so a large amount of basic data is needed to support the training of remote sensing recognition models. However, in related technologies, the acquisition channels of such data are complex, the data quality is difficult to guarantee, and the cost of obtaining big data is high. Therefore, the accuracy of the trained remote sensing recognition models is low, the cost is high, and the generalization ability is poor. Summary of the invention

[0004] The present application provides a remote sensing recognition model training system, a remote sensing recognition model training method, a remote sensing recognition model training device, a computer device and a computer-readable storage medium, aiming to train a remote sensing recognition model with high accuracy, low cost and strong generalization ability.

[0005] To achieve the above objectives, the present application provides a remote sensing recognition model training system, comprising:

[0006] A data management module, wherein the data management module is used to provide a target remote sensing image to be annotated;

[0007] A data annotation module, connected to the data management module, for acquiring the target remote sensing image from the data management module and annotating the target remote sensing image;

[0008] The model training module is connected to the data annotation module and is used to obtain a remote sensing recognition model based on the annotated target remote sensing image training, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0009] To achieve the above purpose, the present application also provides a remote sensing recognition model training method, including:

[0010] Acquire a target remote sensing image to be annotated, and annotate the target remote sensing image;

[0011] A remote sensing recognition model is obtained by training based on the labeled target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0012] To achieve the above purpose, the present application also provides a remote sensing recognition model training device, comprising:

[0013] A labeling module, the labeling module is used to obtain a target remote sensing image to be labeled and label the target remote sensing image;

[0014] A training module is used to obtain a remote sensing recognition model based on the labeled target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0015] In addition, to achieve the above-mentioned purpose, the present application also provides a computer device, which includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the steps of the remote sensing recognition model training method provided in any one of the embodiments of the present application when executing the computer program.

[0016] In addition, to achieve the above-mentioned purpose, the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor implements the steps of the remote sensing recognition model training method described in any one of the embodiments of the present application.

[0017] The embodiments of the present application disclose a training system for a remote sensing recognition model, a training method for a remote sensing recognition model, a training device for a remote sensing recognition model, a computer device, and a computer-readable storage medium. The training system for the remote sensing recognition model includes a data management module, a data annotation module, and a model training module. The data management module is used to provide a target remote sensing image to be annotated; the data annotation module is connected to the data management module, and is used to obtain the target remote sensing image from the data management module and annotate the target remote sensing image; the model training module is connected to the data annotation module, and is used to train a remote sensing recognition model based on the annotated target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model. The present application aims to annotate the target remote sensing image to be annotated based on the training system of the remote sensing recognition model, obtain the annotated target remote sensing image, and then train the annotated target remote sensing image to obtain a remote sensing recognition model with high accuracy, low cost, and strong generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 It is a scene schematic diagram of a remote sensing recognition model training method provided in an embodiment of the present application;

[0020] Figure 2 It is a structural schematic diagram of a remote sensing recognition model training system provided in an embodiment of the present application;

[0021] Figure 3 It is a flowchart of a training method for a remote sensing recognition model provided in an embodiment of the present application;

[0022] Figure 4 is a schematic block diagram of a remote sensing recognition model training device provided in an embodiment of the present application;

[0023] Figure 5 It is a schematic block diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0025] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor do they have to be executed in the order described. For example, some operations / steps can also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions. In addition, although the functional modules are divided in the device schematic, in some cases, the module division can be different from that in the device schematic.

[0026] The term "and / or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0027] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0028] like Figure 1 As shown, the training method of the remote sensing recognition model provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. The application environment includes a terminal device 110 and a server 120, wherein the terminal device 110 can communicate with the server 120 through a network. Specifically, the server 120 can obtain the target remote sensing image to be annotated and annotate the target remote sensing image; and train the remote sensing recognition model based on the annotated target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model, and then send the trained remote sensing recognition model to the terminal device 110. Among them, the server 120 can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms. The terminal device 110 can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal and the server can be directly or indirectly connected by wired or wireless communication, and this application is not limited here.

[0029] See also Figure 2 , Figure 2 Schematic diagram of a remote sensing recognition model training system provided in an embodiment of the present application. Figure 2 As shown, the training system of the remote sensing recognition model includes a data management module 11, a data annotation module 12 and a model training module 13. The data management module 11 is used to provide a target remote sensing image to be annotated; the data annotation module 12 is connected to the data management module 11, and is used to obtain the target remote sensing image from the data management module 11 and annotate the target remote sensing image; the model training module 13 is connected to the data annotation module, and is used to train a remote sensing recognition model based on the annotated target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0030] It can be understood that the data management module 11 can collect, store and manage target remote sensing images, thereby ensuring that the target remote sensing images provided for subsequent processing are complete, accurate and suitable for annotation and training; the data annotation module 12 can annotate the target remote sensing images by manual annotation or automatic annotation; the model training module 13 can learn image features from the annotated target remote sensing images, and then train a remote sensing recognition model.

[0031] It should be noted that the small sample recognition algorithm model can obtain recognition results under limited annotated data conditions. Unlike traditional deep learning models that usually require a large amount of annotated data for training, the small sample recognition algorithm model can extract effective features under small sample conditions by using methods such as prior knowledge and feature transfer, thereby improving the recognition ability of the remote sensing recognition model.

[0032] The above method can provide a target remote sensing image through the data management module 11, annotate the target remote sensing image through the data annotation module 12, and finally perform model training based on the annotated target remote sensing image through the model training module 13 to obtain a remote sensing recognition model that can be effectively recognized under a small sample condition. In this way, the efficiency and accuracy of the remote sensing recognition model can be improved under the condition of limited data resources.

[0033] Optionally, the data management module 11 is also connected to a system model library and a user model library, wherein the system model library includes some cloud crop data and the user model library includes some local crop data; the data management module 11 is used to provide target remote sensing images based on the system model library and / or the user model library.

[0034] For example, the system model library includes a large amount of cloud-based crop data from different geographical locations, different time points, and covering multiple crop types and growth stages. By connecting to the system model library, the data management module 11 can access a large number of target remote sensing images, and the target remote sensing images can be regularly updated and maintained to ensure the timeliness and accuracy of the target remote sensing images. The user model library can be used to store and manage local crop data. By connecting to the user model library, the data management module 11 can access user-specific target remote sensing image data, which has high pertinence and relevance, and can better meet the specific needs of users.

[0035] In the above method, the data management module 11 can flexibly obtain target remote sensing image data from the system model library and the user model library, which not only expands the data source and enhances the diversity and applicability of the data, but also can meet the specific needs of different users and further improve the training effect and practicality of the remote sensing recognition model.

[0036] Optionally, the data labeling module 12 includes an automatic labeling mode and an autonomous labeling mode; wherein the automatic labeling mode is used to label the target crop data through a preset labeling model, and the autonomous labeling model is used to label the target remote sensing image based on a target labeling model determined by a user.

[0037] For example, the automatic annotation mode can use a pre-trained annotation model to automatically identify and mark the features in the target crop data. This method not only reduces the workload and time of manual annotation, improves the annotation efficiency and consistency, but also achieves higher annotation accuracy.

[0038] It should be noted that the present application does not limit the above-mentioned annotation model. For example, the annotation model may include but is not limited to the U-Net model and the Mask R-CNN model. Among them, the U-Net model is a convolutional neural network for image segmentation, which has a symmetrical encoder-decoder structure, that is, a "U"-shaped architecture. The encoder part gradually extracts features and reduces the spatial resolution, while the decoder part gradually restores the spatial resolution and generates a segmented image. The skip connections between the encoder and the decoder combine high-resolution features with upsampled features, which helps to restore image details. This is especially important for fine boundaries and small targets. U-Net performs well in medical images (such as CT and MRI scans) and agricultural remote sensing images (such as crop area segmentation in satellite images), and can accurately segment target areas such as lesions, crop plots, etc.

[0039] Mask R-CNN is a region-based convolutional neural network suitable for object detection and segmentation tasks. Mask R-CNN can not only detect objects in an image, but also generate a high-quality segmentation mask for each object, that is, it can identify multiple objects in the same image and generate precise boundaries for each object. Mask R-CNN performs well in various image segmentation tasks, especially for tasks that require fine segmentation. For example, in agriculture, it can be used to accurately mark the contours and regions of crops, which helps to identify and analyze different types of crops, disease spots, etc.

[0040] The above two models have their own advantages in different labeling tasks. U-Net is suitable for labeling tasks that require fine boundaries and global context information, while Mask R-CNN is suitable for complex scenarios that require simultaneous target detection and labeling.

[0041] For example, the autonomous annotation mode can select or adjust the annotation model according to the user's specific needs and preferences, and then annotate the target remote sensing image, so that the user can adjust the annotation strategy according to different application scenarios or crop types to improve the pertinence and accuracy of the annotation. For example, the user can select a specific annotation tool, annotation format, or fine-tune the annotation model to meet specific needs.

[0042] The above method can annotate the target remote sensing image and then be used to train the remote sensing recognition model.

[0043] Optionally, the training system further includes a model recognition module, which is connected to the model training module 13 and is used to obtain a remote sensing recognition model, and recognize the remote sensing image input by the user through the remote sensing recognition model to obtain corresponding crop information.

[0044] For example, a trained remote sensing recognition model can be obtained from the model training module 13. Thus, when the user inputs the remote sensing images to be identified into the model recognition module, the model recognition module can use the trained recognition model to process and analyze these images and generate corresponding recognition results, including crop information such as crop type, distribution, and health status, for use in various application scenarios such as agricultural management and crop monitoring.

[0045] On the basis of the above embodiment, the model recognition module is further used to respond to the user's instruction to customize the model, obtain the customized model, and recognize the remote sensing image input by the user through the customized model to obtain crop information.

[0046] For example, the remote sensing recognition model training system may also support a user-defined model determined according to the user's own needs, so that the remote sensing image to be recognized is recognized through the customized model to obtain crop information.

[0047] It should be noted that the present application does not limit the remote sensing recognition model, and the remote sensing recognition model may include but is not limited to a convolutional neural network model, a semantic segmentation model, a target detection model, and an instance segmentation model. Among them, the convolutional neural network model is the basic model of the remote sensing recognition model, which performs classification and recognition by extracting the spatial features of the image. Commonly used convolutional neural network models include: VGGNet, a deep network structure suitable for image classification; ResNet, which introduces residual connections and solves the degradation problem of deep network training. DenseNet, using dense connections, improves the efficiency of information flow and gradient propagation. The semantic segmentation model is used to classify each pixel in the image into a specific category. The target detection model is used to locate and identify multiple targets in an image. Common target detection models include Faster R-CNN, which performs target detection based on the region proposal network (RPN); YOLO (You Only Look Once), a real-time target detection model, suitable for application scenarios with high speed requirements; SSD (Single Shot MultiBox Detector), a target detection model that combines speed and accuracy. The instance segmentation model is a combination of object detection and semantic segmentation, which is used to identify and segment each object instance in the image. Common models include: Mask R-CNN, which is an extension of Faster R-CNN and adds a mask branch for instance segmentation; PANet (Path Aggregation Network), which improves the information flow of Feature Pyramid Network (FPN).

[0048] When processing remote sensing image data, the above models can be selected and applied according to different task requirements, thereby effectively improving the accuracy of recognition and analysis.

[0049] The training system of the remote sensing recognition model disclosed in the embodiment of the present application includes a data management module, a data annotation module and a model training module, wherein the data management module is used to provide a target remote sensing image to be annotated; the data annotation module is connected to the data management module, and is used to obtain the target remote sensing image from the data management module and annotate the target remote sensing image; the model training module is connected to the data annotation module, and is used to train a remote sensing recognition model based on the annotated target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model. The present application aims to train a training system based on a remote sensing recognition model to annotate the target remote sensing image to be annotated, obtain the annotated target remote sensing image, and then train a remote sensing recognition model with high accuracy, low cost and strong generalization ability based on the annotated target remote sensing image.

[0050] Please continue reading Figure 3 , Figure 3FIG. 1 is a flow chart of a training method for a remote sensing recognition model provided in an embodiment of the present application. Figure 3 As shown, the training method of the remote sensing recognition model can be implemented through steps S11 to S12.

[0051] Step S11: Acquire a target remote sensing image to be annotated, and annotate the target remote sensing image.

[0052] Step S12: obtaining a remote sensing recognition model based on the labeled target remote sensing image training, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0053] For example, a remote sensing image containing a plurality of targets to be annotated may be obtained for annotation. For example, the remote sensing image containing a plurality of targets to be annotated may be obtained by manual collection, web crawler or public data set, etc., which is not limited in the present application.

[0054] Furthermore, each target remote sensing image can be labeled to obtain a labeled target remote sensing image. The labeled target remote sensing image is then used as a label for the group of input data, and then each group of labeled target remote sensing images is input into an untrained small sample recognition algorithm model for supervised learning. When the training end condition is met, such as when the number of training times reaches a number threshold or the output accuracy of the model reaches an accuracy threshold, the training is terminated to obtain a trained remote sensing recognition model, that is, a trained small sample recognition algorithm model.

[0055] Optionally, based on the above embodiments, the remote sensing recognition model training method provided in the present application also includes: acquiring the remote sensing recognition model, and recognizing the remote sensing image input by the user through the remote sensing recognition model to obtain corresponding crop information.

[0056] When the user inputs the remote sensing image to be identified into the remote sensing recognition model, the remote sensing recognition model can process and analyze the remote sensing image and generate corresponding recognition results, including crop information such as crop type, distribution, and health status, for use in various application scenarios such as agricultural management and crop monitoring.

[0057] The above method can train a remote sensing recognition model, and then realize the recognition of remote sensing images based on the remote sensing recognition model to obtain the corresponding crop information.

[0058] See also Figure 4 , Figure 4 This is a schematic block diagram of a remote sensing recognition model training device provided in an embodiment of the present application. The remote sensing recognition model training device can be configured in a server to execute the aforementioned remote sensing recognition model training method.

[0059] like Figure 4As shown, the remote sensing recognition model training device 200 includes: a labeling module 201 and a training module 202.

[0060] The labeling module 201 is used to obtain a target remote sensing image to be labeled and label the target remote sensing image;

[0061] The training module 202 is used to obtain a remote sensing recognition model based on the labeled target remote sensing image through training, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0062] The training module 202 is further used to obtain a remote sensing recognition model based on the labeled target remote sensing image through training, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0063] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0064] The method and apparatus of the present application can be used in many general or special computing system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer terminal devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices.

[0065] Exemplarily, the above method and apparatus may be implemented in the form of a computer program. The computer program may be implemented in the form of a computer program. Figure 5 Runs on the computer device shown.

[0066] See also Figure 5 , Figure 5 1 is a schematic diagram of a computer device provided in an embodiment of the present application. The computer device may be a server.

[0067] like Figure 5 As shown, the computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory may include a volatile storage medium, a non-volatile storage medium and an internal memory.

[0068] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any training method for a remote sensing recognition model.

[0069] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0070] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any training method for the remote sensing recognition model.

[0071] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that the structure of the computer device 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.

[0072] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0073] In some embodiments, the processor is used to obtain a target remote sensing image to be labeled and to label the target remote sensing image; a remote sensing recognition model is obtained by training based on the labeled target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

[0074] In some embodiments, the processor is further configured to obtain the remote sensing recognition model, and recognize the remote sensing image input by the user through the remote sensing recognition model to obtain corresponding crop information.

[0075] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions, and when the program instructions are executed, a training method for any remote sensing recognition model provided in the embodiment of the present application is implemented.

[0076] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the computer device.

[0077] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function, and the like.

[0078] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A training system for remote sensing recognition models, It is characterized in that include: A data management module, wherein the data management module is used to provide a target remote sensing image to be annotated; A data annotation module, connected to the data management module, for acquiring the target remote sensing image from the data management module and annotating the target remote sensing image; The model training module is connected to the data annotation module and is used to obtain a remote sensing recognition model based on the annotated target remote sensing image training, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

2. The training system according to claim 1, It is characterized in that The data management module is also connected to a system model library and a user model library, wherein the system model library includes a number of cloud crop data, and the user model library includes a number of local crop data; The data management module is used to provide the target remote sensing image based on the system model library and / or the user model library.

3. The training system according to claim 1, It is characterized in that The data annotation module includes an automatic annotation mode and an autonomous annotation mode; wherein the automatic annotation mode is used to annotate the target crop data through a preset annotation model, and the autonomous annotation model is used to annotate the target remote sensing image based on a target annotation model determined by a user.

4. The training system according to claim 1, It is characterized in that The training system also includes a model recognition module, which is connected to the model training module and is used to obtain the remote sensing recognition model and recognize the remote sensing image input by the user through the remote sensing recognition model to obtain corresponding crop information.

5. The method according to claim 4, It is characterized in that The model recognition module is also used to respond to the instruction of the user-defined model, obtain the customized model, and recognize the remote sensing image input by the user through the customized model to obtain the crop information.

6. A training method for remote sensing recognition model, It is characterized in that The method comprises: Acquire a target remote sensing image to be annotated, and annotate the target remote sensing image; A remote sensing recognition model is obtained by training based on the labeled target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

7. The method according to claim 6, It is characterized in that The method further comprises: The remote sensing recognition model is obtained, and the remote sensing image input by the user is recognized by the remote sensing recognition model to obtain corresponding crop information.

8. A training device for remote sensing recognition model, It is characterized in that The training device comprises: A labeling module, the labeling module is used to obtain a target remote sensing image to be labeled and label the target remote sensing image; A training module is used to obtain a remote sensing recognition model based on the labeled target remote sensing image, wherein the remote sensing recognition model includes a small sample recognition algorithm model.

9. A computer device, It is characterized in that include: A memory and a processor; wherein the memory is connected to the processor and is used to store a program; the processor is used to implement the steps of the remote sensing recognition model training method as described in any one of claims 6 to 7 by running the program stored in the memory.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the steps of the remote sensing recognition model training method as described in any one of claims 6 to 7.

Images