Training sample set processing method, electronic device and storage medium

By storing the basic information of the remote sensing image sample set separately from the sample original data, and using multi-threaded processing technology, the problem of low training efficiency of model with large training samples is solved, and the efficiency of sample set learning and intelligent recognition is improved.

CN118711019BActive Publication Date: 2024-11-05CHINA SURVEY SURVEYING & MAPPING TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411178320.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-05
Estimated Expiration
2044-08-27

AI Technical Summary

Technical Problem

In the case of large training samples, the prior art is difficult to effectively improve the efficiency of model training, resulting in problems such as difficulty in searching and slow learning response.

Method used

By storing the basic information of the remote sensing image sample set and the sample original data into different types of databases, the retrieval performance of the sample original data is improved by using object detection, change detection and semantic segmentation threads in parallel processing.

Benefits of technology

The sample original data retrieval performance during sample set learning training is improved, and the learning model performance based on this sample set and the recognition efficiency of intelligent recognition or classification software are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118711019B_ABST
    Figure CN118711019B_ABST
Patent Text Reader

Abstract

The present application discloses a method for processing a training sample set, an electronic device and a storage medium, the method comprising: obtaining a remote sensing image sample set; determining first sample set basic information of the remote sensing image sample set and sample original data of the remote sensing image sample set according to the remote sensing image sample set, wherein the first sample set basic information at least includes identification information of the remote sensing image sample; storing the first sample set information of the remote sensing image sample set in a first database; and storing the sample original data of the remote sensing image sample set in a second database, wherein the first database and the second database differ in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type; thus, by storing the sample set basic information and the sample original data separately, the sample original data retrieval performance can be improved, and the sample original data can be retrieved more quickly.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method for processing a training sample set, an electronic device, and a storage medium. Background Art

[0002] With the research and progress of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields. Machine learning is the core of artificial intelligence and the fundamental way for computers to have intelligence. Through machine learning, neural network models that meet application requirements can be trained. However, the training sample set is the learning object of machine learning, which determines the prediction accuracy and performance of the neural network model.

[0003] In the prior art, a large number of training samples are introduced to improve the accuracy of the model. However, the large amount of training samples also has problems such as difficult search and slow learning response. Therefore, how to ensure the efficiency of model training when the amount of training sample data is large has become a technical problem that needs to be solved urgently.

[0004] The content of this background technology description is only for facilitating understanding of the relevant technology in this field and is not regarded as an admission of the prior art. Summary of the invention

[0005] Therefore, the embodiments of the present invention intend to provide a method for processing a training sample set, an electronic device, and a storage medium.

[0006] In a first aspect, an embodiment of the present invention provides a method for processing a training sample set, comprising:

[0007] Obtain remote sensing image sample set;

[0008] Determine, according to the remote sensing image sample set, first sample set basic information of the remote sensing image sample set and sample original data of the remote sensing image sample set, wherein the first sample set basic information at least includes identification information of the remote sensing image sample;

[0009] The basic information of the first sample set of the remote sensing image sample set is stored in a first database, and the original sample data of the remote sensing image sample set is stored in a second database, wherein the first database is different from the second database in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type.

[0010] In the above solution, before storing the original sample data of the remote sensing image sample set in the second database, the method includes:

[0011] Parsing the remote sensing image sample set to obtain sample original data of the remote sensing image sample set and second sample set basic information of the remote sensing image sample set, wherein the second sample set basic information includes at least one of the following: sample set target information, sample set slice information, and sample set classification information;

[0012] The method further comprises:

[0013] Importing basic information of a second sample set of the remote sensing image sample set into the first database.

[0014] In the above solution, parsing the remote sensing image sample set to obtain sample original data of the remote sensing image sample set includes:

[0015] Using a target detection thread to perform target analysis on the remote sensing image sample set, and converting target detection attribute information into first entity information;

[0016] Using a change detection thread to perform change analysis on the remote sensing image sample set, and converting the change detection attribute information into second entity information;

[0017] Performing semantic segmentation analysis on the remote sensing image sample set by using semantic segmentation, and converting the semantic segmentation attribute information into third entity information;

[0018] The storing of the original sample data of the remote sensing image sample set into the second database comprises:

[0019] The first entity information, the second entity information and the third entity information are stored in the second database.

[0020] In the above solution, parsing the remote sensing image sample set to obtain the second sample set basic information of the remote sensing image sample set includes:

[0021] Utilizing a target detection thread to perform target analysis on the remote sensing image sample set to obtain target information of the sample set;

[0022] Using a change detection thread to perform change analysis on the remote sensing image sample set to obtain slice information of the sample set;

[0023] Using a semantic segmentation thread to perform semantic segmentation analysis on the remote sensing image sample set to obtain classification information of the sample set;

[0024] The sample set target information, the sample set slice information and the sample set classification information are all stored in the first database.

[0025] In the above scheme, the target detection thread is used to perform target analysis on the remote sensing image sample set to obtain the target information of the sample set, including:

[0026] Utilizing the target detection thread to perform target analysis on individual samples in the remote sensing image sample set, and obtaining intermediate results of target information of the sample set;

[0027] The using a change detection thread to perform change analysis on the remote sensing image sample set to obtain the sample set slice information includes:

[0028] Using the change detection thread, target analysis is performed on each sample in the remote sensing image sample set to obtain an intermediate result of the sample set slice information;

[0029] The using of the semantic segmentation thread to perform semantic segmentation analysis on each sample in the remote sensing image sample set to obtain the sample classification information includes:

[0030] Using a semantic segmentation thread, semantic analysis is performed on individual samples in the remote sensing image sample set to obtain an intermediate result of classification information of the sample set;

[0031] The storing the sample set target information, the sample set slice information and the sample set classification information into the first database includes:

[0032] The intermediate results of the sample set target information, the intermediate results of the sample set slice information and the intermediate results of the sample set classification information are stored in a third database, and after all traversals are completed, the final sample set target information, the final sample set slice information and the final sample set classification information are stored in the first database, wherein the third database is different from the first database in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type; the third database is also different from the second database.

[0033] In the above scheme, the target detection thread, the change detection thread and the semantic segmentation thread are multi-threaded and performed in parallel;

[0034] and / or,

[0035] At least one of the object detection thread, the change detection thread, and the semantic segmentation thread is executed in the background.

[0036] In the above solution, determining the first sample set basic information of the remote sensing image sample set according to the remote sensing image sample set includes:

[0037] According to the remote sensing image sample set, automatically analyzing the first sample basic information of a single sample in the remote sensing image sample set;

[0038] Presenting the first sample basic information of the single sample on the terminal interface;

[0039] detecting a modification instruction from a user based on the first sample basic information, and determining modified first sample basic information;

[0040] Based on the modified first sample basic information, first sample set basic information of the remote sensing image sample set is determined.

[0041] In the above solution, storing the basic information of the first sample set in the first database includes:

[0042] determining whether sample set information associated with the first sample set information exists in the first database;

[0043] If not, importing the first sample set information into the first database;

[0044] or,

[0045] If it exists, the original sample set information is updated to the first sample set information.

[0046] In a second aspect, an embodiment of the present invention provides an electronic device, comprising: a processor and a memory storing a computer program, wherein the processor is configured to execute any method for processing a training sample set of an embodiment of the present invention when running the computer program.

[0047] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method for processing a training sample set of any embodiment of the present invention is implemented.

[0048] In the processing method of the training sample set used in the embodiment of the present invention, a remote sensing image sample set is obtained; according to the remote sensing image sample set, the first sample set basic information of the remote sensing image sample set and the sample original data of the remote sensing image sample set are determined, wherein the first sample set basic information at least includes the identification information of the remote sensing image sample set; the first sample set basic information of the remote sensing image sample set is stored in a first database, and the sample original data of the remote sensing image sample set is stored in a second database, wherein the first database and the second database are different in at least one of the following: data storage mode, data query and analysis mode, data expansion mode, and data structure and type. Based on this, the processing method of the training sample set provided in the embodiment of the present invention can be applied to models such as recognition or classification of remote sensing images to recognize or classify remote sensing images. By storing the basic information of the sample set and the sample original data separately, the sample original data retrieval performance can be improved in the sample set learning and training process, and the sample original data can be retrieved more quickly, thereby facilitating the performance of the learning model based on the sample set, and improving the recognition efficiency of the intelligent recognition or classification software based on the training sample set.

[0049] Other optional features and technical effects of the embodiments of the present invention are partially described below, and partially can be understood by reading this document. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The elements shown are not limited to the proportions shown in the accompanying drawings. The same or similar reference numerals in the accompanying drawings represent the same or similar elements, wherein:

[0051] Figure 1 A schematic diagram showing a flow chart of a method for processing a training sample set provided in accordance with an embodiment of the present invention;

[0052] Figure 2 Another schematic diagram of a process of processing a training sample set according to an embodiment of the present invention is shown;

[0053] Figure 3 A schematic diagram of a sample creation interface provided according to an embodiment of the present invention is shown;

[0054] Figure 4 It shows a schematic diagram of a multi-threaded real-time calculation process of basic information of a second sample set provided by an embodiment of the present invention;

[0055] Figure 5 A schematic diagram showing a flow chart of a method for processing a training sample set provided according to a specific embodiment of the present invention;

[0056] Figure 6 Another schematic diagram of a process of processing a training sample set according to a specific embodiment of the present invention is shown;

[0057] Figure 7 An exemplary structural diagram of a processing device for a training sample set according to an embodiment of the present invention is shown;

[0058] Figure 8 An exemplary structural diagram of an electronic device capable of implementing the method according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific implementation methods and drawings. Here, the exemplary implementation methods of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0060] As used herein, the term "including" and its variations mean open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "based at least in part on". The terms "an example embodiment" and "an embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0061] It should be noted that in the field of remote sensing, the amount of sample data is generally very large. If the sample set and the original sample data are managed offline, a lot of manpower and material costs are required. However, effective and large amounts of data support are required for model training. If the sample data cannot be effectively provided to the model, it will inevitably bring huge challenges to the high accuracy and high performance of the model. Based on this, it is necessary to find a sample set storage method that can efficiently manage or efficiently search for the desired sample data, so as to lay a solid foundation for the subsequent extraction and learning based on the sample set to build a high-accuracy and high-performance neural network model, and then provide a strong guarantee for the intelligent recognition or classification of remote sensing images in the remote sensing field.

[0062] Figure 1 FIG. 4 is a flow chart showing a method for processing a training sample set according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0063] Step 101: Obtain a remote sensing image sample set.

[0064] It should be noted that the embodiments of the present invention are applied to electronic devices. The electronic devices here include: mobile devices or fixed devices. Among them, mobile devices include: mobile phones, tablet computers, wearable devices; fixed devices may include: desktop computers, smart home devices, etc. Of course, the electronic device can also be a server or a server cluster, etc. In short, any electronic device that has a processor and can process remote sensing images and identify the image information of remote sensing images, including but not limited to the content information and label information of remote sensing images, is the electronic device described in the embodiments of the present invention.

[0065] Here, the remote sensing image sample set is a sample set composed of multiple remote sensing images for training a certain learning model or neural network model (hereinafter referred to as model). The remote sensing images here can be color images or black and white images. The remote sensing images here can be images taken by satellites or images taken by different electronic devices. In some embodiments, step 101, i.e., the step of obtaining a remote sensing image sample set, can include real-time satellite images obtained from a satellite database connected to the electronic device, and can also include images taken by other electronic devices obtained from the network or on a big data platform.

[0066] It is understandable that the acquired remote sensing images usually include not only the image content itself, but also image information such as the shooting location of the remote sensing image, the number of the remote sensing image, and the equipment used to shoot the remote sensing image.

[0067] Step 102: determining first sample set basic information of the remote sensing image sample set and sample original data of the remote sensing image sample set according to the remote sensing image sample set, wherein the first sample set basic information at least includes identification information of the sample;

[0068] The identification information of the remote sensing image sample refers to information used to identify different remote sensing images, for example, it may be the name or number of the remote sensing image sample.

[0069] It should be noted that the corresponding remote sensing image sample can be located through the first sample basic information of the remote sensing image sample, and multiple first sample basic information in the same remote sensing image sample set constitute the first sample set basic information.

[0070] Step 103: storing basic information of a first sample set of the remote sensing image sample set in a first database, and storing original data of the remote sensing image sample set in a second database, wherein the first database and the second database differ in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type.

[0071] In some embodiments, the first database may be an open source relational database management system that supports ACID transactions, complex queries, and extended functions, such as PostgreSQL; the second database may be an open source distributed search and analysis engine built on Lucene, which is mainly used to process large rule data sets and is fast, reliable, and flexible, such as Elasticsearch.

[0072] Taking PostgreSQL (hereinafter referred to as pg) as the first database and Elasticsearch (hereinafter referred to as ES) as the second database as an example, PostgreSQL uses tables to store data, each table includes multiple rows and columns; supports multiple data types, such as integer, floating point, date, text, array, etc.; Elasticsearch uses inverted index technology to support full-text search, and data is stored in JSON format; distributed architecture can distribute data on multiple nodes to improve the performance and scalability of search and analysis. PostgreSQL has excellent performance and scalability in relational databases. Elasticsearch performs well in search and analysis tasks and is suitable for large-scale data sets and real-time processing scenarios. Of course, in other embodiments, the first database can also be a database different from the second database. For example, the memory space of the first database is smaller than the memory space of the second database, and no limitation is made here.

[0073] In this way, by storing the basic information of the first sample set and the original data of the remote sensing image sample set separately in two different databases, the retrieval performance of the original data of the sample set during the sample set learning and training process can be improved, and the original data of the sample set can be retrieved more conveniently and quickly, which is beneficial to improving the performance of the model learned using the sample set, and improving the efficiency of the intelligent recognition or classification software based on the model.

[0074] In some embodiments, in step 103, before storing the original sample data of the remote sensing image sample set in the second database, the method further includes:

[0075] Parsing the remote sensing image sample set to obtain sample original data of the remote sensing image sample set and second sample set basic information of the remote sensing image sample set, wherein the second sample set basic information includes at least one of the following: sample set target information, sample set slice information, and sample set classification information;

[0076] The method further comprises:

[0077] Importing basic information of the second sample set of the remote sensing image sample set into the first database.

[0078] It can be understood that the target information of the sample set refers to the information obtained by detecting the target of the sample set. In the field of machine learning and computer vision, target detection of the sample set refers to identifying and locating specific target objects from a given sample set, which usually involves analyzing image or video data to find the target objects of interest and determine their location and position in the image or video. It can be understood that target detection is an important task in the field of computer vision, which aims to detect the existence of target objects from the input image or sample data and determine their location and size; in the target detection of the sample set, there is a set of images or sample data containing target objects, and our goal is to use these data to train a model so that it can accurately identify and locate the target objects in the image or video. In short, target detection of the sample set is an important task in the field of computer vision, which uses image or video data to train the model to achieve the recognition and location of target objects. By selecting appropriate technical methods and evaluation instructions, we can obtain a target detection model with excellent performance. The target information here refers to the learning target indicated by the remote sensing image sample, that is, the image target that the remote sensing image sample is expected to be detected or predicted, such as image type or other predicted target.

[0079] It is understandable that the sample set slicing information refers to the information obtained by slicing the sample set. The slicing detection of the sample set is usually not a term that directly corresponds to a specific technology or field. Slicing in programming and data processing usually refers to a part of the data extracted or segmented from a larger data set. Slicing in the sample set may be to select a specific data subset for further analysis or testing. For example, in a sample set containing sales data for many years, the data is sliced ​​according to attributes such as year, quarter or product type to study the sales of a specific time period or product category. The detection performed after the sample set is sliced ​​refers to a specific analysis or test for the data subset, including statistical analysis, training and evaluation of machine learning models, hypothesis testing, etc. The specific content of the detection depends on the research objectives and context. For example, in medical research, a subset of patients with specific disease characteristics is sliced ​​from a large number of patient samples, and then the genes of these patients are sliced ​​to find genetic variations associated with the disease. In this embodiment, the slice of the remote sensing pattern can be slice feature information that can characterize the type of remote sensing image (for example, people, mountain scenery or water).

[0080] It is understandable that the sample set classification information refers to the information obtained through semantic segmentation detection. Sample set semantic segmentation detection is also a technology in the field of computer vision. It involves classifying each pixel in an image or video, thereby segmenting the image into regions with different semantic information. Unlike target detection, semantic segmentation not only detects objects in the image, but also further assigns a category label to each pixel. Semantic segmentation classifies images at the pixel level through deep learning models (such as convolutional neural networks CNN). After training, the model can identify different objects and regions in the image and assign a corresponding category label to each pixel. Unlike target detection, which focuses on the location and category of objects in the image, semantic segmentation further classifies each pixel to achieve more refined image understanding.

[0081] It should be noted that, for remote sensing image samples, in addition to identifying the first sample basic information, the second sample information will also be identified.

[0082] It should be noted that in the second sample set information, the sample set target information, sample set slice information and sample set classification information are all technical types. In some embodiments, the second sample set information needs to be identified with the help of other neural network models or manually. Of course, in some other embodiments, the second sample set information may also include but is not limited to the industry type, image resolution, data source information and version number of the remote sensing image sample. It should be noted that the industry type refers to the industry to which the remote sensing image sample set belongs, such as agriculture, forestry, ecology, housing and construction, water conservancy or transportation; the data source information is used to characterize the data source of the remote sensing image sample, such as information from a satellite or an electronic device.

[0083] In some embodiments, the second sample set information here is generally only quantity information, such as the number of sample set targets of the sample set image and the target feature information characterizing the sample set targets, the number of sample set slices and the slice feature information that guarantees the sample set slices, or the number of sample set classifications and the classification feature information that characterizes the sample set classifications, and does not include all target features, all slice features, and all classification features of each sample. It can be understood that the remote sensing image sample A can include multiple target feature information, but one sample set target information only includes one target feature information. Similarly, the remote sensing image sample A can include multiple slice feature information, but one sample set slice information only includes one slice feature information. Similarly, the remote sensing image sample A can include multiple classification feature information, but one sample set classification information only includes one classification feature information. In other words, although the remote sensing image sample A and the remote sensing image sample B are different sample images, they can have the same target feature information, the same slice feature information, and the same classification feature information. In other words, the second sample set basic information described in this embodiment includes but is not limited to the sample set target information, the sample set slice information, and the sample set classification information, and these information are all technical classification information set to retrieve or locate the corresponding sample original data.

[0084] The embodiment of the present invention, by also storing the basic information of the second sample set in the first database, assists the basic information of the first sample set to accurately search for the corresponding remote sensing sample images, which can be beneficial to the efficiency and accuracy of learning samples, and further to the accuracy of the model.

[0085] In addition, since the sample set has a small amount of basic information and is more business-oriented while the sample original data has a large amount of information, the retrieval performance of the original data can be further improved by storing the sample set basic information (including the first sample set basic information and the second sample set basic information) separately from the original data.

[0086] In some embodiments, the method further includes: determining whether the remote sensing image sample set has been imported into a database corresponding to model training; if not, executing step 103.

[0087] It should be added that the step of determining whether the remote sensing image sample set has been imported into the database for corresponding model training includes:

[0088] Select the remote sensing image sample set to be imported on the network display interface of the electronic device;

[0089] After the selected remote sensing image sample set is detected, it is determined whether the remote sensing image sample set has been imported into the database for corresponding model training.

[0090] It should be noted that the corresponding model here refers to the model to be trained by the remote sensing image sample set, and the database of the corresponding model should be understood as a database originally divided into a sample set storing the corresponding model, and the database for training the corresponding model may be more than one. In some embodiments, the database of the corresponding model may be one or more.

[0091] In some other embodiments, determining whether the remote sensing image sample set has been imported into a database corresponding to model training includes:

[0092] According to the basic information of the first sample set of the remote sensing image sample set, it is determined that the remote sensing image sample set has been imported into the database for corresponding model training; if the basic information of the first sample set is not stored in the corresponding model database, it is determined that the remote sensing image sample set has not been imported into the database for corresponding model training.

[0093] Furthermore, in order to conveniently and quickly determine whether the remote sensing image sample set has been imported into the database for corresponding model training, in some embodiments, determining whether the corresponding model training stores the first sample set basic information based on the first sample set basic information in the remote sensing image sample set includes: determining whether the first sample set basic information is stored in the first database based on the first sample set basic information of the remote sensing image sample set.

[0094] In this way, by storing the basic information of the first sample set of the remote sensing image sample set in the first database, the speed of determining whether a newly added remote sensing image sample or remote sensing image sample set has a duplicate map can be increased.

[0095] In some embodiments, the method further includes: if it is determined that the remote sensing image sample set has been imported into the database corresponding to the model training, then feedback information indicating that the remote sensing image sample set has been imported is fed back.

[0096] Here, the feedback information can be displayed on the original web page through the display interface, and a voice prompt can be issued to remind the user that the remote sensing image sample set has been imported, eliminating the need for repeated import, thereby improving the efficiency of model training.

[0097] In some embodiments, see Figure 2 ,like Figure 2 As shown, in step 102, determining the first sample set basic information of the remote sensing image sample set according to the remote sensing image sample set includes:

[0098] Step 1021: automatically analyzing first sample set information of individual samples in the remote sensing image sample set according to the remote sensing image sample set;

[0099] Exemplarily, automatically parsing first sample basic information of individual samples in the remote sensing image sample set according to the remote sensing image sample set may include automatically identifying file names in the remote sensing image sample set according to the remote sensing image sample set.

[0100] Step 1022: presenting the first sample basic information of the single sample on the terminal interface;

[0101] Exemplarily, the interface may be a web interface during import.

[0102] Exemplarily, the interface may be a sample library management interface, which may present multiple sample images of the remote sensing image sample set and first sample basic information of the corresponding samples. In other embodiments, the sample library management interface may also present second sample basic information of the remote sensing image sample.

[0103] Step 1023: detecting a modification instruction from a user based on the basic information of the first sample, and determining the modified basic information of the first sample;

[0104] For example, see Figure 3 This interface can modify the first sample information of a single sample or a sample set, thereby manually adjusting the first sample information of the remote sensing image sample, thereby providing more accurate sample information for the corresponding training set model, thereby providing a strong guarantee for the recognition accuracy of the training model.

[0105] In some other embodiments, the method for obtaining the basic information of the second sample set further includes: detecting a modification instruction of the user based on the basic information of the second sample, and determining the modified basic information of the second sample. Figure 3 , the second sample basic information of the corresponding remote sensing image sample can be determined by modifying the technology type, industry type, resolution, slice size, version number, etc. In this way, the second sample basic information of the remote sensing image sample can be obtained more accurately, which is beneficial to the accuracy of model learning.

[0106] In some embodiments, in step 103, storing the basic information of the first sample set in the first database includes:

[0107] Determining whether there is sample set information associated with the first sample set information in the first database;

[0108] If it does not exist, importing the first sample set information into the first database;

[0109] or,

[0110] If it exists, the original first sample set information is updated to the unfirst sample set information.

[0111] In some other embodiments, storing the basic information of the second sample set in the first database includes:

[0112] Determining whether there is sample set information associated with the first sample set information in the first database;

[0113] If it does not exist, importing the second sample set information into the first database;

[0114] or,

[0115] If it exists, the original sample set information is updated to the second sample set information.

[0116] It should be added that each different basic information is configured with a corresponding code or unique identification code, which can facilitate the overall management. In this way, the first sample set information and / or the second sample set information in the first database can be globally managed, which is conducive to ensuring data consistency when matching with basic information later, thereby ensuring the accuracy of model learning.

[0117] Exemplarily, global configuration management is performed on the sample set in the first database, that is, the technical classification, industry classification, resolution, satellite image data source, etc. are configured. Each classification item has its own unique code and corresponding name, and the global configuration is rendered on the information needed through the network interface or web front end to ensure that all numbers or names are unified. In this way, through global configuration, that is, if it is changed in one place, the whole world will follow suit, avoiding the risks of missed changes and wrong changes, improving the efficiency of the later maintenance of the model, and improving the accuracy of the subsequent training of the model.

[0118] For example, the global configuration code may be as shown in the following table:

[0119]

[0120] In this way, by configuring a unified code, global modifications can be made to achieve global management of basic sample information.

[0121] In some embodiments, parsing the remote sensing image sample set to obtain the original data of the remote sensing image sample set includes:

[0122] Utilizing the target detection thread to perform target analysis on the remote sensing image sample set, and converting the target detection attribute information into the first entity information;

[0123] Using the change detection thread to perform change analysis on the remote sensing image sample set, and converting the change detection attribute information into the second entity information;

[0124] The semantic segmentation thread is used to perform semantic segmentation analysis on the remote sensing image sample set, and the semantic segmentation attribute information is converted into the third entity information;

[0125] The storing of the original sample data of the remote sensing image sample set into the second database comprises:

[0126] The first entity information, the second entity information and the third entity information are stored in a second database.

[0127] Exemplarily, a target detection thread is used to perform target analysis on a remote sensing image sample set, and target detection attribute information is converted into first entity information, including: the target detection thread performs image analysis on the remote sensing image sample set, generates corresponding DTO (Date Transfer Object) information, and converts the target detection attribute information into first Enity entity information that can be stored in a second database such as Elasticsearch.

[0128] Here, DTO information, i.e., data transfer object, is mainly used to transfer data between different layers of the application (such as business logic layer and data access layer, or between client and server). The main purpose is to separate the data structure from the business logic to improve the maintainability and scalability of the code. DTO can contain only the required data, thereby reducing the overhead of data transmission, which is very friendly for network communication or cross-layer data transmission. The Enity entity information here is expressed in transactions that can be objectively stored and distinguished from each other in the database. It represents an object with attributes and behaviors. Usually in the database, each individual of a class data object is called an entity. These entities are usually described by ER diagrams (entity-relationship) diagrams. The three elements of ER diagrams include: entity, attribute and relationship; for example, in the employee management system, "employee" is an entity, and the employee's "name", "age" and "position" are the attributes of the entity; entity information refers to the collection of employees of a certain age group or position. In this embodiment, the target detection attribute information is converted into the first Enity entity information that can be stored in the second database, such as Elasticsearch. It should be understood that it is a set of target feature information with the same target detection attribute information, or a set of target feature graphs with the same target detection attribute, and then these sets are converted into data information that can be stored in Elasticsearch for the model to retrieve or search.

[0129] Exemplarily, a change detection thread is used to perform change analysis on a remote sensing image sample set, and the change detection attribute information is converted into second entity information, including: the change detection thread performs image analysis on the remote sensing image sample set, generates corresponding DTO (Date Transfer Object) information, and converts the change detection attribute information into second Enity entity information that can be stored in a second database such as Elasticsearch.

[0130] Here, the change detection attribute information is converted into the second Enity entity information that can be stored in the second database such as Elasticsearch. It should be understood that it is a set of slice feature information with the same change detection attribute information, or a set of slice graphs with the same change detection attribute, and then these sets are converted into data information that can be stored in Elasticsearch for the model to retrieve or search.

[0131] Exemplarily, a semantic segmentation thread is used to perform change analysis on a remote sensing image sample set, and the semantic segmentation attribute information is converted into third entity information, including: the semantic segmentation thread performs image analysis on the remote sensing image sample set, generates corresponding DTO (Date Transfer Object) information, and converts the semantic segmentation attribute information into third Enity entity information that can be stored in a second database such as Elasticsearch.

[0132] Here, the semantic segmentation attribute information is converted into the third Entity entity information that can be stored in the second database, such as Elasticsearch. It should be understood that it is a set of classification feature information with the same semantic segmentation attribute information, or a set of classification graphs with the same semantic segmentation attribute, and then these sets are converted into data information that can be stored in Elasticsearch for the model to retrieve or search.

[0133] In this way, by parsing remote sensing image samples through different threads, the parsing efficiency of remote sensing image samples can be improved, thereby improving the training efficiency of the model.

[0134] In some other embodiments, parsing the remote sensing image sample set to obtain the second sample set basic information of the remote sensing image sample set further includes:

[0135] Utilizing a target detection thread to perform target analysis on the remote sensing image sample set to obtain target information of the sample set;

[0136] Using a change detection thread to perform change analysis on the remote sensing image sample set to obtain slice information of the sample set;

[0137] Using a semantic segmentation thread to perform semantic segmentation analysis on the remote sensing image sample set to obtain classification information of the sample set;

[0138] The sample set target information, the sample set slice information and the sample set classification information are all stored in the first database.

[0139] The target information, slice information and classification information here may respectively represent the number of target features (hereinafter referred to as the target number) and their names, the number of slice features (hereinafter referred to as the slice number) and their names, and the number of classification features (hereinafter referred to as the classification number) and their names.

[0140] By parsing out the number of targets, the number of slices and the number of classifications of the remote sensing image sample set and storing them in the first database, subsequent retrieval of the original data can be facilitated and the efficiency of retrieval of the original data can be improved.

[0141] In some embodiments, the using the target detection thread to perform target parsing on the remote sensing image sample set to obtain the target information of the sample set includes:

[0142] Utilizing the target detection thread to perform target analysis on individual samples in the remote sensing image sample set, and obtaining intermediate results of target information of the sample set;

[0143] The using a change detection thread to perform change analysis on the remote sensing image sample set to obtain the sample set slice information includes:

[0144] Using the change detection thread, target analysis is performed on individual samples in the remote sensing image sample set to obtain an intermediate result of the sample set slice information;

[0145] The using of the semantic segmentation thread to perform semantic segmentation analysis on each sample in the remote sensing image sample set to obtain the sample classification information includes:

[0146] Using a semantic segmentation thread, semantic analysis is performed on individual samples in the remote sensing image sample set to obtain an intermediate result of classification information of the sample set;

[0147] The storing the sample set target information, the sample set slice information and the sample set classification information into the first database includes:

[0148] The intermediate results of the sample set target information, the intermediate results of the sample set slice information and the intermediate results of the sample set classification information are stored in a third database, and after all traversals are completed, the final sample set target information, the final sample set slice information and the final sample set classification information are stored in the first database, wherein the third database is different from the first database in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type; the third database is also different from the second database.

[0149] The third database here can be a database temporarily used to cache data, such as a redis database. It is understandable that the Redis database is an open source, in-memory data structure storage system, which can be used as a database, cache and message broker. Because Redis is open source, it allows users to use, modify and distribute it freely; because Redis mainly stores data in memory, it can provide extremely high read and write speeds; in addition, although Redis mainly uses memory, it also provides two persistence mechanisms: RDB (snapshot) and AOF (append write file) to ensure the persistent preservation of data; Redis also supports a variety of data structures, such as strings, hashes, lists, sets, sorted sets, etc. Atomic operations: Redis supports atomic operations to ensure data consistency and security.

[0150] In this embodiment, by storing the intermediate results of the second sample set basic information in the intermediate third database during the calculation process, the acquisition of multiple second sample set basic information can be performed asynchronously and / or in parallel by multiple threads, thereby improving the computing performance and real-time performance.

[0151] Exemplarily, the target detection thread, the change detection thread and the semantic segmentation thread are multi-threaded and performed in parallel;

[0152] and / or,

[0153] At least one of the object detection thread, the change detection thread, and the semantic segmentation thread is executed in the background.

[0154] In other embodiments, asynchronously calculating the number of objects, the number of slices, and the number of categories of the remote sensing image sample set may specifically include:

[0155] The target detection thread, the change detection thread and the semantic segmentation thread are respectively used to simultaneously detect the number of targets, the number of slices and the number of categories of the remote sensing image sample set, and then the numbers are added up to obtain the final number of targets, the final number of slices and the final number of categories, which are stored in the first database, such as Postgresql.

[0156] It is important to understand that asynchronous computing can be processed in parallel. Asynchronous computing allows one or more calculations to be executed simultaneously while executing one calculation, thereby improving overall processing efficiency. Asynchronous computing does not require waiting. Unlike synchronous computing, the front-end thread in asynchronous computing does not need to wait for the current instruction to return the result from the back-end thread, and can continue to execute subsequent instructions, which has a non-blocking feature.

[0157] For example, see Figure 4 ,like Figure 4 As shown, the method includes:

[0158] Start multiple threads;

[0159] Initialization: The target detection thread detects the number of targets, the number of slices, and the number of categories, and stores them in redis;

[0160] Initialization: the change detection thread detects the number of targets, slices, and categories, and stores them in redis;

[0161] Initialization: The semantic segmentation thread detects the number of targets, slices, and categories, and stores them in redis;

[0162] The above three steps can be executed asynchronously and in parallel with multiple threads;

[0163] The target detection thread traverses the files in the labelText folder, one file corresponds to one slice, and updates the number of slices in redis;

[0164] The target detection thread traverses the entries in each file, each entry corresponds to a target, uniquely classifies and updates the number of targets and classifications in redis;

[0165] Confirm whether the traversal is complete. If so, execute the step "Is the file traversal under the labelText folder completed?"; otherwise, return to the step "The target detection thread traverses the entries under each file";

[0166] Whether the file traversal under the labelText folder is completed, if yes, execute 523; otherwise, return to step "confirm whether the traversal is completed";

[0167] The change detection thread traverses the files in the labelText folder, one file corresponds to one slice, and updates the number of slices in redis;

[0168] The default number of targets and categories is 1.

[0169] Update the number of slices, categories, and targets in redis to Postgresql.

[0170] The semantic segmentation thread traverses all labeled files in the target dataset, calculates the number of slices, and updates them to redis;

[0171] The number of targets is 1, the number of categories is 1, and then step 523 is executed;

[0172] Finish.

[0173] In the above embodiment, the identification, parsing and storage processes of the second sample set basic information and the original data can all be performed in an asynchronous manner. Since the sample size is relatively large, the identification, parsing and storage processes take a long time, while parsing the basic information of the sample set is relatively simple. The identification, parsing and storage processes of the sample set and the sample original data information are processed separately in an abnormal manner, that is, the basic information of the sample set is first parsed and stored in the database, and after success, it is returned to the web front-end for presentation, and then the sample original data information is parsed in an asynchronous multi-threaded manner and stored in the second database. Finally, after the original data is parsed and stored in the database, the es_flag status of the sample entry in the corresponding database is updated, thereby ensuring the timeliness of business processing.

[0174] In addition, due to the associated target detection, change detection and semantic segmentation detection of each sample set, the number of classifications and targets only needs to be calculated once. The process of importing all sample set information is completed through asynchronous multi-threading and the intermediate calculation results are stored in redis to complete the calculation of the number of slices, the number of classifications and the number of targets. Compared with the existing technology, it can improve the computing performance and real-time performance.

[0175] In order to better understand the method for processing the training sample set described in the embodiment of the present invention, the present application also provides a specific embodiment. Specifically, please refer to Figure 5 ,

[0176] like Figure 5 As shown, the processing method of the training sample set can be integrated into a sample management service system. The embodiment process of the sample management service includes:

[0177] Deploy ES and create object detection, change detection, and semantic segmentation detection threads;

[0178] Deploy the database pg and create sample set tables and statistical tables;

[0179] Deploy redis;

[0180] Deploy backend service programs;

[0181] Put the sample set into the specified directory for automatic parsing by the service program;

[0182] Deploy front-end service programs;

[0183] Log in to the system through the web access address and create a series of user operations such as sample sets.

[0184] The ES here refers to the second database mentioned above, the pg here refers to the first database mentioned above, the redis here refers to the third database mentioned above, and the sample set here refers to the remote sensing image sample set mentioned above; the sample set table here is used to record the first sample set information, and the statistical table here is used to record the basic information of the second sample set.

[0185] For the specific execution process, please refer to Figure 6 ,like Figure 6 As shown, the method includes:

[0186] Web filter the sample set to be imported;

[0187] Has it been imported? If yes, proceed to step "Prompt: This sample set has been imported". If no, proceed to step "Automatically identify the sample set file name and fill it in the web form";

[0188] Tip: This sample set has been imported;

[0189] Automatically identify the sample set file name and fill it in the web form;

[0190] Analyze the features, regions, technology types, industry types, resolutions, data sources, etc.;

[0191] Convert the above analysis into a recognizable code;

[0192] Return to the web front-end for presentation and fill in the web form, which the user can modify again;

[0193] The server-side function processes the business attributes of the sample set and stores them in the database pg;

[0194] Whether the database already exists, if yes, execute the step "update the database sample set attribute information", if no, execute the step "convert the business attributes into an entity class that can be recognized by the database and store it in the database pg";

[0195] Update database sample set attribute information;

[0196] Convert business attributes into entity classes that can be recognized by the database and store them in the database pg;

[0197] Return the datesetld of the database record to the front end;

[0198] Multithreaded traversal of sample set;

[0199] Parse the target detection sample set file and generate the corresponding DTO information;

[0200] Convert target detection attribute information into Entily entity information that can be entered into ES;

[0201] Parse the change detection sample set file and generate the corresponding DTO information;

[0202] Convert the change detection attribute information into Entily entity information that can be entered into ES;

[0203] Parse the semantic segmentation sample set file and generate corresponding DTO information;

[0204] Convert semantic segmentation attribute information into Entily entity information that can be entered into ES;

[0205] Enable multithreading, execute ES operations, and record success and failure data;

[0206] Whether the traversal has been completed; if so, execute the step "asynchronously calculate the number of sample set targets, the number of slices, the number of categories, and update pg", if not, return to the step "multi-thread traversal of the sample set";

[0207] Asynchronously calculate the number of sample set targets, slices, and categories, and update pg;

[0208] The number of targets, number of slices, and number of categories in the target detection sample set are calculated in redis;

[0209] The number of targets, slices, and classifications in the change detection sample set are calculated using rerids;

[0210] The number of targets, slices, and categories in the semantic segmentation sample set are calculated in redis;

[0211] Update the target number, slice number, and classification number of the corresponding sample set to pg;

[0212] Update the database sample set to see if ES is successful, es_flag is successful.

[0213] In this way, the above embodiment makes the configuration global, that is, if it is changed in one place, the whole world will be changed accordingly, avoiding the risks of missing changes, wrong changes, etc., and improving the efficiency of research and development, testing, and later maintenance. The basic information of the sample set and the original data information of the sample are stored separately. Because the basic information of the sample set is small and the business is relatively strong, it is stored through the database pg; because the amount of information of the original data of the sample is relatively large, it is stored with Elasticsearch to improve the retrieval performance of the original data of the sample. The process of identifying, parsing, and warehousing the basic information of the sample set and the original data of the sample is processed in an asynchronous decoupling manner. Because the original data information of the sample is relatively large, it is only necessary to update the es_flag status of the corresponding sample set after the processing of the original data information of the sample is completed, thereby ensuring the timeliness of business processing. Improve the real-time performance of calculation. Because the number of slices, classifications, and targets of the associated target detection, change detection, and semantic separation of each data set only needs to be calculated once, so in the process of importing the data set, the number of slices, classifications, and targets of the data set are calculated by asynchronous multi-threading and redis to store the intermediate results of calculation; compared with the prior art by scheduling task calculation, asynchronous multi-threading combined with redis improves computing performance and real-time performance.

[0214] Figure 7FIG. 4 is a block diagram showing a processing device for a training sample set according to an embodiment of the present invention. Figure 7 , the device comprises:

[0215] An acquisition module 801 is configured to acquire a remote sensing image sample set;

[0216] The determination module 802 is configured to determine first sample set basic information of the remote sensing image sample set and sample original data of the remote sensing image sample set according to the remote sensing image sample set, wherein the first sample set basic information at least includes identification information of the remote sensing image sample;

[0217] The storage module 803 is configured to store the basic information of the first sample set of the remote sensing image sample set in a first database, and store the original sample data of the remote sensing image sample set in a second database, wherein the first database is different from the second database in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type.

[0218] In an optional embodiment, the device further includes: a parsing module configured to:

[0219] Parsing the remote sensing image sample set to obtain sample original data of the remote sensing image sample set and second sample set basic information of the remote sensing image sample set, wherein the second sample set basic information includes at least one of the following: sample set target information, sample set slice information, and sample set classification information;

[0220] The device further comprises: an import module configured to: import the second sample set basic information of the remote sensing image sample set into the first database.

[0221] In an optional embodiment, the parsing module is further configured to:

[0222] Using a target detection thread to perform target analysis on the remote sensing image sample set, and converting target detection attribute information into first entity information;

[0223] Using a change detection thread to perform change analysis on the remote sensing image sample set, and converting the change detection attribute information into second entity information;

[0224] Using a semantic segmentation thread to perform semantic segmentation analysis on the remote sensing image sample set, and converting the semantic segmentation attribute information into third entity information;

[0225] The storage module 803 is further configured to:

[0226] The first entity information, the second entity information and the third entity information are stored in the second database.

[0227] In an optional embodiment, the parsing module is further configured to:

[0228] Utilizing a target detection thread to perform target analysis on the remote sensing image sample set to obtain target information of the sample set;

[0229] Using a change detection thread to perform change analysis on the remote sensing image sample set to obtain slice information of the sample set;

[0230] Using a semantic segmentation thread to perform semantic segmentation analysis on the remote sensing image sample set to obtain classification information of the sample set;

[0231] The sample set target information, the sample set slice information and the sample set classification information are all stored in the first database.

[0232] In an optional embodiment, the parsing module is further configured to:

[0233] Utilizing the target detection thread to perform target analysis on individual samples in the remote sensing image sample set, and obtaining intermediate results of target information of the sample set;

[0234] Using the change detection thread, target analysis is performed on individual samples in the remote sensing image sample set to obtain an intermediate result of the sample set slice information;

[0235] Using a semantic segmentation thread, semantic analysis is performed on individual samples in the remote sensing image sample set to obtain an intermediate result of classification information of the sample set;

[0236] The storage module 803 is further configured to:

[0237] The intermediate results of the sample set target information, the intermediate results of the sample set slice information and the intermediate results of the sample set classification information are stored in a third database, and after all traversals are completed, the final sample set target information, the final sample set slice information and the final sample set classification information are stored in the first database, wherein the third database is different from the first database in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type; the third database is also different from the second database.

[0238] In an optional embodiment, the target detection thread, the change detection thread and the semantic segmentation thread are multi-threaded and performed in parallel;

[0239] and / or,

[0240] At least one of the object detection thread, the change detection thread, and the semantic segmentation thread is executed in the background.

[0241] In an optional embodiment, the determining module 802 is further configured to:

[0242] According to the remote sensing image sample set, automatically analyzing the first sample basic information of a single sample in the remote sensing image sample set;

[0243] Presenting the first sample basic information of the single sample on the terminal interface;

[0244] detecting a modification instruction from a user based on the first sample basic information, and determining modified first sample basic information;

[0245] Based on the modified first sample basic information, first sample set basic information of the remote sensing image sample set is determined.

[0246] In an optional embodiment, the storage module 803 is further configured to:

[0247] determining whether sample set information associated with the first sample set information exists in the first database;

[0248] If not, importing the first sample set information into the first database;

[0249] or,

[0250] If it exists, the original sample set information is updated to the first sample set information.

[0251] In some embodiments, the training sample set processing device may be combined with the features of the training sample set processing method of any embodiment, and vice versa, which will not be elaborated herein.

[0252] In an embodiment of the present invention, an electronic device is provided, including: a processor and a memory storing a computer program, wherein the processor is configured to execute any training sample set processing method of the embodiment of the present invention when running the computer program.

[0253] Figure 8 A schematic diagram of an electronic device 900 that can implement a method or implement an embodiment of the present invention is shown, which may include more or fewer electronic devices than shown in the figure in some embodiments. In some embodiments, it can be implemented using a single or multiple electronic devices. In some embodiments, it can be implemented using cloud or distributed electronic devices.

[0254] like Figure 8As shown, the electronic device 900 includes a processor 901, which can perform various appropriate operations and processes according to the programs and / or data stored in the read-only memory (ROM) 902 or the programs and / or data loaded from the storage part 908 to the random access memory (RAM) 903. The processor 901 can be a multi-core processor or can include multiple processors. In some embodiments, the processor 901 can include a general main processor and one or more special coprocessors, such as a central processing unit (CPU), a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), etc. In the RAM 903, various programs and data required for the operation of the electronic device 900 are also stored. The processor 901, the ROM 902 and the RAM 903 are connected to each other via a bus 904. The input / output (I / O) interface 905 is also connected to the bus 904.

[0255] The processor and the memory are used together to execute the program stored in the memory. When the program is executed by the computer, the methods, steps or functions described in the above embodiments can be implemented.

[0256] The following components are connected to the I / O interface 905: an input section 906 including a keyboard, a mouse, a touch screen, etc.; an output section 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, a modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage section 908 as needed. Figure 8 Only some components are schematically shown in the figure, which does not mean that the computer system 900 only includes Figure 8 Components shown.

[0257] The systems, devices, modules or units described in the above embodiments may be implemented by a computer or its associated components. The computer may be, for example, a mobile terminal, a smart phone, a personal computer, a laptop computer, a vehicle-mounted human-computer interaction device, a personal digital assistant, a media player, a navigation device, a game console, a tablet computer, a wearable device, a smart TV, an Internet of Things system, a smart home, an industrial computer, a server or a combination thereof.

[0258] Although not shown, in an embodiment of the present invention, a storage medium is provided, the storage medium storing a computer program, the computer program being configured to execute any method for processing a training sample set according to an embodiment of the present invention when executed.

[0259] Storage media in embodiments of the present invention include permanent and non-permanent, removable and non-removable items that can be used to store information by any method or technology. Examples of storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0260] The methods, programs, systems, devices, etc. of the embodiments of the present invention may be executed or implemented in a single or multiple networked computers, or may be practiced in a distributed computing environment. In the embodiments of this specification, in these distributed computing environments, tasks may be performed by remote processing devices connected via a communication network.

[0261] Those skilled in the art should understand that the embodiments of the present specification may be provided as methods, systems or computer program products. Therefore, those skilled in the art may imagine that the functional modules / units or controllers and related method steps described in the above embodiments may be implemented in software, hardware or a combination of software / hardware.

[0262] Unless explicitly stated, the actions or steps of the methods, programs, and embodiments of the present invention do not have to be performed in a specific order and can still achieve the desired results. In some implementations, multitasking and parallel processing are also possible or may be advantageous.

[0263] In this article, multiple embodiments of the present invention are described, but for the sake of simplicity, the description of each embodiment is not exhaustive, and the same or similar features or parts between the embodiments may be omitted. In this article, "one embodiment", "some embodiments", "example", "specific example", or "some examples" are intended to be applicable to at least one embodiment or example of the present invention, but not all embodiments. The above terms do not necessarily mean to refer to the same embodiment or example. In the absence of contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples.

[0264] The exemplary systems and methods of the present invention have been specifically shown and described with reference to the above embodiments, which are merely examples of the best modes for implementing the present systems and methods. It will be appreciated by those skilled in the art that various changes may be made to the embodiments of the systems and methods described herein when implementing the present systems and / or methods without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A method for processing a training sample set, characterized in that: include: Obtain remote sensing image sample set; Determine, according to the remote sensing image sample set, first sample set basic information of the remote sensing image sample set and sample original data of the remote sensing image sample set, wherein the first sample set basic information at least includes identification information of the remote sensing image sample; storing basic information of a first sample set of the remote sensing image sample set in a first database, and storing original sample data of the remote sensing image sample set in a second database, wherein the first database is different from the second database in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type; Before storing the original sample data of the remote sensing image sample set in the second database, the method further comprises: Parsing the remote sensing image sample set to obtain sample original data of the remote sensing image sample set and second sample set basic information of the remote sensing image sample set, wherein the second sample set basic information includes at least one of the following: sample set target information, sample set slice information, and sample set classification information; The method further comprises: Importing basic information of a second sample set of the remote sensing image sample set into the first database; Parsing the remote sensing image sample set to obtain sample original data of the remote sensing image sample set includes: Using a target detection thread to perform target analysis on the remote sensing image sample set, and converting target detection attribute information into first entity information; Using a change detection thread to perform change analysis on the remote sensing image sample set, and converting the change detection attribute information into second entity information; Using a semantic segmentation thread to perform semantic segmentation analysis on the remote sensing image sample set, and converting the semantic segmentation attribute information into third entity information; The storing of the original sample data of the remote sensing image sample set into the second database comprises: The first entity information, the second entity information and the third entity information are stored in the second database.

2. The method according to claim 1, characterized in that: Parsing the remote sensing image sample set to obtain second sample set basic information of the remote sensing image sample set includes: Utilizing the target detection thread to perform target analysis on the remote sensing image sample set to obtain target information of the sample set; Using a change detection thread to perform change analysis on the remote sensing image sample set to obtain slice information of the sample set; Using a semantic segmentation thread to perform semantic segmentation analysis on the remote sensing image sample set to obtain classification information of the sample set; The sample set target information, the sample set slice information and the sample set classification information are all stored in the first database.

3. The method according to claim 2, characterized in that The using the target detection thread to perform target analysis on the remote sensing image sample set to obtain target information of the sample set includes: Utilizing the target detection thread to perform target analysis on individual samples in the remote sensing image sample set, and obtaining intermediate results of target information of the sample set; The using a change detection thread to perform change analysis on the remote sensing image sample set to obtain the sample set slice information includes: Using the change detection thread, target analysis is performed on individual samples in the remote sensing image sample set to obtain an intermediate result of the sample set slice information; The using of the semantic segmentation thread to perform semantic segmentation analysis on individual samples in the remote sensing image sample set to obtain classification information of the sample set includes: Using a semantic segmentation thread, semantic analysis is performed on individual samples in the remote sensing image sample set to obtain an intermediate result of classification information of the sample set; The storing the sample set target information, the sample set slice information and the sample set classification information into the first database includes: The intermediate results of the sample set target information, the intermediate results of the sample set slice information and the intermediate results of the sample set classification information are stored in a third database, and after all traversals are completed, the final sample set target information, the final sample set slice information and the final sample set classification information are stored in the first database, wherein the third database is different from the first database in at least one of the following: data storage method, data query and analysis method, data expansion method, and data structure and type; the third database is also different from the second database.

4. The method according to claim 1 or 2, characterized in that: The target detection thread, the change detection thread and the semantic segmentation thread are multi-threaded and performed in parallel; and / or, At least one of the object detection thread, the change detection thread, and the semantic segmentation thread is executed in the background.

5. The method according to claim 1, characterized in that Determining first sample set basic information of the remote sensing image sample set according to the remote sensing image sample set includes: According to the remote sensing image sample set, automatically analyzing the first sample basic information of a single sample in the remote sensing image sample set; Presenting the first sample basic information of the single sample on the terminal interface; detecting a modification instruction from a user based on the first sample basic information, and determining modified first sample basic information; Based on the modified first sample basic information, first sample set basic information of the remote sensing image sample set is determined.

6. The method according to claim 1 or 5, characterized in that: Storing the first sample set basic information in a first database includes: determining whether sample set information associated with the first sample set basic information exists in the first database; If not, importing the basic information of the first sample set into the first database; or, If it exists, the original sample set information is updated to the first sample set basic information.

7. An electronic device, characterized in that: include: A processor and a memory storing a computer program, wherein the processor is configured to implement the method according to any one of claims 1 to 6 when running the computer program.

8. A storage medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.