Remote sensing algorithm evaluation method and device with rapid deployment and automatic optimization functions

Abstract

The invention provides a remote sensing algorithm evaluation method and device with rapid deployment and automatic optimization functions, and the method comprises the steps: constructing a plurality of virtual mirror image environments according to a device operation environment and a dependency item on which a plurality of to-be-evaluated remote sensing algorithms depend; constructing a standardized interface; respectively deploying a corresponding remote sensing algorithm to be evaluated in each virtual mirror image environment; obtaining an evaluation scheme; creating a batch scheduling task according to the evaluation scheme; the batch scheduling task is used for acquiring remote sensing data conforming to a to-be-evaluated remote sensing algorithm according to the evaluation scheme and calculating evaluation indexes; and comparing the evaluation indexes to realize remote sensing algorithm optimization. According to the method, a virtual mirror image environment based on a USB flash disk and a cloud server can be constructed, and a unified and normalized data interface is formulated, so that rapid deployment of a remote sensing algorithm which is not limited by time, space and environment is realized. According to the method, the multi-source remote sensing algorithms of the same product can be compared, analyzed and checked, so that automatic optimization of different remote sensing algorithms is realized.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a remote sensing algorithm evaluation method and device with rapid deployment and automatic optimization functions. Background Art

[0002] With the rapid development of satellite platforms and payloads, satellite remote sensing observation capabilities have been continuously improved, and satellite remote sensing active and passive combination, spatiotemporal spectrum multi-resolution nesting, and three-dimensional comprehensive observation of thermal dynamic fields have been realized, forming an unprecedented amount of remote sensing data. Based on the electromagnetic wave radiation transmission theory, by establishing parameters and remote sensing observation scattering models from rich remote sensing observation data to quantitatively extract or invert remote sensing products with clear physical meanings, it can greatly promote the widespread application of remote sensing data in disaster prevention and mitigation, resources and environment, agricultural meteorology, global climate change and other fields related to national economic and social development.

[0003] Satellite sensor performance, geometric and radiometric calibration errors, data preprocessing level, quantitative remote sensing inversion model, remote sensing algorithm deployment and evaluation, product inspection uncertainty, etc. are a series of important factors that affect the quality of remote sensing products. Improving the quality of remote sensing products is a complex device engineering. With the continuous deepening of radiation transmission mechanism research, artificial intelligence and other computer technologies, the research and development of remote sensing product algorithms has shown a flourishing trend, and a variety of remote sensing algorithms have emerged for the same product.

[0004] Compared with the satellite platform and payload at the front end and the remote sensing industry application at the back end, the development and production of remote sensing products at the middle end, which play a bridging role, need to be further improved. Among them, the deployment and evaluation of remote sensing algorithms are important links in the development and production of remote sensing products. At present, there are two problems in the deployment and evaluation of remote sensing algorithms. On the one hand, the various remote sensing algorithms provided by algorithm developers do not have a unified and standardized data interface and programming language, which is not conducive to algorithm transplantation, scheduling and evaluation. Algorithm developers have dispersed a lot of financial resources and energy to carry out full-process algorithm deployment and evaluation, including data processing, batch product production and inspection. On the other hand, many remote sensing algorithms are only tested in local areas and have not been fully evaluated in large-scale areas. The same product faces the problem of how to compare and select multiple source algorithms. These factors restrict the uncertainty of large-scale regional production capacity and product quality of remote sensing products. Therefore, it is urgent to propose a remote sensing algorithm evaluation method and device with rapid deployment and automatic optimization functions. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides a remote sensing algorithm evaluation method and device with rapid deployment and automatic optimization functions, which can quickly deploy and automatically optimize the remote sensing algorithm.

[0006] In the first aspect, a remote sensing algorithm evaluation method with rapid deployment and automatic optimization functions includes:

[0007] Constructing multiple virtual mirror environments according to the device operating environments and dependencies on which multiple remote sensing algorithms to be evaluated depend;

[0008] Build standardized interfaces;

[0009] Deploy the corresponding remote sensing algorithm to be evaluated in each virtual mirror environment;

[0010] Get the assessment plan;

[0011] Create batch scheduling tasks according to the evaluation plan; the batch scheduling tasks are used to obtain remote sensing data that meets the remote sensing algorithm to be evaluated according to the evaluation plan and calculate the evaluation indicators;

[0012] Compare evaluation indicators to optimize remote sensing algorithms.

[0013] Furthermore, the method for constructing the virtual image environment specifically includes:

[0014] Install Docker Engine;

[0015] Add the current user to the Docker group;

[0016] Create a Dockerfile, configure the operating environment of the Docker container, use the orchestration tool to combine multiple Docker containers according to the operating environment and dependencies of the devices that the remote sensing algorithms to be evaluated depend on, and package them into a virtual image environment;

[0017] Stores the virtual image environment.

[0018] Furthermore, the normalized interface is used to unify the specifications: input data and path, output data and path, dataset naming, and remote sensing algorithm scheduling parameters.

[0019] Furthermore, the evaluation plan includes evaluation reference data, time and space scope, time and space matching plan and evaluation indicators.

[0020] Furthermore, batch scheduling tasks are specifically used for:

[0021] Acquisition of remote sensing data;

[0022] Standardize remote sensing data to obtain standard data;

[0023] The standard data are screened within the time and space range using the time and space matching scheme to obtain remote sensing data that meets the remote sensing algorithm to be evaluated;

[0024] The evaluation indicators are calculated using remote sensing data that conforms to the remote sensing algorithm to be evaluated.

[0025] In the second aspect, a remote sensing algorithm evaluation device with rapid deployment and automatic optimization functions includes:

[0026] An image building unit: used for building multiple virtual image environments according to the device operating environments and dependencies on which multiple remote sensing algorithms to be evaluated depend;

[0027] Interface construction unit: used to build standardized interfaces;

[0028] Deployment unit: used to deploy the corresponding remote sensing algorithm to be evaluated in each virtual mirror environment;

[0029] Evaluation unit: used to obtain the evaluation plan; create batch scheduling tasks according to the evaluation plan; the batch scheduling tasks are used to obtain remote sensing data that meets the remote sensing algorithm to be evaluated according to the evaluation plan, and calculate the evaluation indicators; compare the evaluation indicators to achieve the optimization of the remote sensing algorithm.

[0030] Furthermore, the image building unit is specifically used for:

[0031] Install Docker Engine;

[0032] Add the current user to the Docker group;

[0033] Create a Dockerfile, configure the operating environment of the Docker container, use the orchestration tool to combine multiple Docker containers according to the operating environment and dependencies of the devices that the remote sensing algorithms to be evaluated depend on, and package them into a virtual image environment;

[0034] Stores the virtual image environment.

[0035] Furthermore, the normalized interface is used to unify the specifications: input data and path, output data and path, dataset naming, and remote sensing algorithm scheduling parameters.

[0036] Furthermore, the evaluation plan includes evaluation reference data, time and space scope, time and space matching plan and evaluation indicators.

[0037] Furthermore, batch scheduling tasks are specifically used for:

[0038] Acquisition of remote sensing data;

[0039] Standardize remote sensing data to obtain standard data;

[0040] The standard data are screened within the time and space range using the time and space matching scheme to obtain remote sensing data that meets the remote sensing algorithm to be evaluated;

[0041] The evaluation indicators are calculated using remote sensing data that conforms to the remote sensing algorithm to be evaluated.

[0042] It can be seen from the above technical solutions that the remote sensing algorithm evaluation method and device with rapid deployment and automatic optimization functions provided by the present invention can build a virtual mirror environment based on a USB flash drive and a cloud server, and formulate a unified and standardized data interface to achieve rapid deployment of remote sensing algorithms without time, space, and environmental restrictions. The method can also perform comparative analysis and inspection on multi-source remote sensing algorithms of the same product to achieve automatic optimization of different remote sensing algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0044] Figure 1 A flowchart of a remote sensing algorithm evaluation method provided in an embodiment.

[0045] Figure 2 A flowchart of a method for constructing a virtual image environment provided in an embodiment.

[0046] Figure 3 A flowchart of a batch scheduling task execution method provided in an embodiment.

[0047] Figure 4 A flowchart of multi-source remote sensing algorithm evaluation for a cloud detection product provided in an embodiment.

[0048] Figure 5 A module block diagram of a remote sensing algorithm evaluation device provided in an embodiment. DETAILED DESCRIPTION

[0049] The embodiments of the technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and the protection scope of the present invention cannot be limited thereto. It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should be the common meanings understood by those skilled in the art to which the present invention belongs.

[0050] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0051] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0052] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0053] Example:

[0054] A remote sensing algorithm evaluation method with rapid deployment and automatic optimization functions, see Figure 1 ,include:

[0055] Constructing multiple virtual mirror environments according to the device operating environments and dependencies on which multiple remote sensing algorithms to be evaluated depend;

[0056] Build standardized interfaces;

[0057] Deploy the corresponding remote sensing algorithm to be evaluated in each virtual mirror environment;

[0058] Get the assessment plan;

[0059] Create batch scheduling tasks according to the evaluation plan; the batch scheduling tasks are used to obtain remote sensing data that meets the remote sensing algorithm to be evaluated according to the evaluation plan and calculate the evaluation indicators;

[0060] Compare evaluation indicators to optimize remote sensing algorithms.

[0061] In this embodiment, the method first needs to prepare the basic environment of the device. The basic environment of the device includes the python language environment and the C language environment. The specific contents include: (1) configuring the python environment, downloading and running the Anaconda installation program, and completing the installation according to the prompts. (2) installing common remote sensing data processing dependencies through the pip command, mainly including netcdf4, h5py, GDAL, etc. (3) configuring the C language environment, mainly including installing compilers such as gcc, gfortran, Intel, etc.

[0062] In this embodiment, the remote sensing data collected by the method mainly includes data from various sensors of multi-source satellites during representative periods, such as collecting data from different seasons. The method can also collect other data, such as field observation data, numerical model and reanalysis data, static auxiliary data, national meteorological station observation data, land cover data, etc.

[0063] In this embodiment, the method packages the device operating environment and dependencies that the remote sensing algorithm depends on into an image based on container technology, obtains multiple virtual image environments, and realizes the portability and consistency of remote sensing algorithm deployment. Then the method constructs a standardized interface, for example, according to the reading requirements of remote sensing data and other data, to avoid problems such as inconsistent numerical types of input data. Then the method starts to deploy the remote sensing algorithm. The deployment method is: download and decompress the virtual image environment on the U disk or cloud server on a computer or virtual machine, obtain the device operating environment and dependencies that the remote sensing algorithm depends on, and then deploy the corresponding remote sensing algorithm according to the standardized interface in the virtual image environment to realize the rapid deployment and transplantation of multi-source remote sensing algorithms.

[0064] In this embodiment, after the remote sensing algorithm is quickly deployed, the method also needs to evaluate and select the remote sensing algorithm. When evaluating and selecting, the method first creates a batch scheduling task according to the evaluation plan; the batch scheduling task is used to obtain remote sensing data that meets the remote sensing algorithm to be evaluated according to the evaluation plan, and calculate the evaluation index. When designing the batch scheduling task, the time required for batch scheduling and the spatial input interface can be reserved, and a log output function can be designed in each step of the task to facilitate finding code errors. Different evaluation schemes result in different batch scheduling tasks. For example, when evaluating remote sensing algorithms such as cloud detection, the designed batch scheduling task is mainly divided into five steps: 1. It is necessary to obtain input data and preprocess the input data; 2. The preprocessed data is input into the cloud detection algorithm (such as physical threshold method, multi-phase method, texture and spatial feature detection method, elevation method and machine learning method) for calculation and output results; 3. The output cloud detection results are drawn with a quick view or thematic map; 4. The results are statistically analyzed; 5. The operation result status is stored in the database.

[0065] This remote sensing algorithm evaluation method can build a virtual mirror environment based on USB disk and cloud server, and formulate a unified and standardized data interface to achieve rapid deployment of remote sensing algorithms without time, space and environment restrictions. This method can also conduct comparative analysis and inspection of multi-source remote sensing algorithms for the same product to achieve automatic optimization of different remote sensing algorithms.

[0066] Further, in some embodiments, see Figure 2 , the construction method of the virtual mirror environment specifically includes:

[0067] Install Docker Engine;

[0068] Add the current user to the Docker group;

[0069] Create a Dockerfile, configure the operating environment of the Docker container, use the orchestration tool to combine multiple Docker containers according to the operating environment and dependencies of the devices that the remote sensing algorithms to be evaluated depend on, and package them into a virtual image environment;

[0070] Stores the virtual image environment.

[0071] In this embodiment, when building a virtual image environment, the method first uses Docker as a software container for packaging the operating environment and dependencies, and selects a suitable Docker engine installation according to the CentOs operating device. Then the current user is added to the Docker group to avoid running the Docker command as the root user of the current user. Then create a Dockerfile file, for example, create a Dockerfile file in the root directory of an application, and configure the operating environment of the Docker container, wherein the operating environment includes network settings, volume mounting, environment variables, etc. Use orchestration tools such as Docker Compose or Docker Swarm to combine multiple containers, and package the containers into images through the commit command provided by Docker. Finally, use the save command provided by Docker to form a tarball of the virtual image environment, and copy the tarball to a USB flash drive or a cloud server, so that other devices can download and deploy the virtual image environment, thereby ensuring the portability and consistency of the deployment of various remote sensing algorithms.

[0072] Furthermore, in some embodiments, the standardized interface is used to unify the specifications: input data and path, output data and path, data set naming, and remote sensing algorithm scheduling parameters.

[0073] In this embodiment, the standardized interface is used to unify the specifications: input data and path, output data and path, data set naming, and remote sensing algorithm scheduling parameters. For example, the data set naming for atmospheric quantitative products is shown in Table 1:

[0074] Table 1:

[0075]

[0076] For example, the format of remote sensing algorithm scheduling parameters can be: job name_star_instrument name_data level_station_date_time_orbit number_real-time / supplementary processing flag, where the definitions of the above fields are shown in Table 2:

[0077] Table 2:

[0078]

[0079]

[0080] For example, the dispatch control parameter specification is:

[0081] F3EGMCMPMWRXXXXXXXXR_FY3E_MWHS-_L2_ORBT_20200426_0046_12680_R;

[0082] FY3EPMOLRGDXXXXXXXXG_FY3E_MERSI_L2_GRAN_20220802_0140_05578_R.

[0083] Furthermore, in some embodiments, the evaluation scheme includes evaluation reference data, a time and space scope, a time and space matching scheme, and evaluation indicators.

[0084] In this embodiment, the evaluation scheme specifically includes: a) reference data. For example, when using star cross comparison evaluation, the reference data is EOS / MODIS cloud detection data and CALIPSO cloud detection data. Using satellite-ground authenticity comparison evaluation, the parameter data is ground station observation data. b) Time and space scope. For example, the time and space scope can cover different seasons, a certain important weather process period, low and medium latitudes, Asia, etc. c) Time and space matching scheme. The time and space matching scheme includes a time matching scheme and a space matching scheme. For example, the time matching scheme uses the nearest time and a certain time threshold for matching; the space matching scheme uses the nearest distance and a certain distance threshold for matching. d) Evaluation indicators. For example, for numerical products, the evaluation indicators include correlation coefficient, deviation, root mean square error and other indicators. For binary products, the evaluation indicators include false alarm rate, accuracy rate, etc.

[0085] Further, in some embodiments, see Figure 3 , batch scheduling tasks are specifically used for:

[0086] Acquisition of remote sensing data;

[0087] Standardize remote sensing data to obtain standard data;

[0088] The standard data are screened within the time and space range using the time and space matching scheme to obtain remote sensing data that meets the remote sensing algorithm to be evaluated;

[0089] The evaluation indicators are calculated using remote sensing data that conforms to the remote sensing algorithm to be evaluated.

[0090] In this embodiment, the batch scheduling task is specifically used for:

[0091] (1) Remote sensing data collection: According to the established evaluation plan, obtain sample data that meets the evaluation plan.

[0092] (2) Data standardization processing: Through the data standardization processing script, the sample data and reference data are parsed into the same format, and remote sensing elements, observation time, longitude and latitude and other information are extracted.

[0093] (3) Carry out the evaluation according to the evaluation plan. This mainly includes using the time-space matching plan to screen the standard data within the time and space range to obtain the screened data, that is, the remote sensing data that meets the remote sensing algorithm to be evaluated; using the screened data to calculate the evaluation indicators, and selecting the best according to the evaluation indicators.

[0094] (5) Visualization of evaluation results. For example, based on the spatiotemporal matching results and indicator calculation results, a scatter point probability density distribution diagram, a deviation distribution histogram, a spatiotemporal distribution diagram, etc. can be drawn.

[0095] (6) Output the evaluation report. Summarize the evaluation plan and evaluation results to generate an evaluation report.

[0096] Furthermore, the steps of this method are further explained below using a cloud detection product as an example.

[0097] Evaluation plan: Inspection period: January 1, 2022 - December 31, 2022;

[0098] Inspection area: within the global area;

[0099]

[0100] See also Figure 4 First, the obtained inspection data and inspection source data are parsed to unify the data range, time, spatial resolution, etc. The inspection data is the Fengyun-3D nebula detection data, including baseline algorithm data, integrated algorithm data, and machine learning algorithm data. There are two main types of inspection source data, including EOS / MODIS cloud detection and GEO data (MYD35_L2, MOD35_L2 and MYD03, MOD03), and CALIPSO L2 cloud detection products. Data processing mainly uses python to read the longitude and latitude, time, and cloud detection information in each data, and store each information as a numpy array for subsequent spatiotemporal matching calls.

[0101] The spatiotemporal matching method is to call cKDTree (also known as scipy.spatial.cKDTree) in Python, build a Kdtree using the test data, and then use the kdtree.query method to find the test source data closest to the test data. The obtained test data and test source data are compared, and the data with a time difference within plus or minus five minutes are selected to obtain remote sensing data that meets the remote sensing algorithm to be evaluated.

[0102] Algorithm optimization: Calculate evaluation indicators. For example, evaluation indicators include: clear sky detection accuracy (POD_clear), cloud detection accuracy (POD_cloud), clear sky misjudgment rate (FAR_clear), cloud misjudgment rate (FAR_cloud), total detection accuracy (HR), and detection index (KSS).

[0103] Finally, this method can also draw images, such as the spatial distribution map of the inspection source and the inspected cloud detection product, the true color map at daytime, the infrared window area and channel synthetic cloud map at night, and the deviation spatial distribution map. By analyzing images and indicators, the optimal remote sensing algorithm can be selected.

[0104] A remote sensing algorithm evaluation device with rapid deployment and automatic optimization functions, see Figure 5 ,include:

[0105] An image building unit: used for building multiple virtual image environments according to the device operating environments and dependencies on which multiple remote sensing algorithms to be evaluated depend;

[0106] Interface construction unit: used to build standardized interfaces;

[0107] Deployment unit: used to deploy the corresponding remote sensing algorithm to be evaluated in each virtual mirror environment;

[0108] Evaluation unit: used to obtain the evaluation plan; create batch scheduling tasks according to the evaluation plan; the batch scheduling tasks are used to obtain remote sensing data that meets the remote sensing algorithm to be evaluated according to the evaluation plan, and calculate the evaluation indicators; compare the evaluation indicators to achieve the optimization of the remote sensing algorithm.

[0109] Furthermore, in some embodiments, the mirror image construction unit is specifically used to:

[0110] Install Docker Engine;

[0111] Add the current user to the Docker group;

[0112] Create a Dockerfile, configure the operating environment of the Docker container, use the orchestration tool to combine multiple Docker containers according to the operating environment and dependencies of the devices that the remote sensing algorithms to be evaluated depend on, and package them into a virtual image environment;

[0113] Stores the virtual image environment.

[0114] Furthermore, in some embodiments, the standardized interface is used to unify the specifications: input data and path, output data and path, data set naming, and remote sensing algorithm scheduling parameters.

[0115] Furthermore, in some embodiments, the evaluation scheme includes evaluation reference data, a time and space scope, a time and space matching scheme, and evaluation indicators.

[0116] Furthermore, in some embodiments, the batch scheduling task is specifically used to:

[0117] Acquisition of remote sensing data;

[0118] Standardize remote sensing data to obtain standard data;

[0119] The standard data are screened within the time and space range using the time and space matching scheme to obtain remote sensing data that meets the remote sensing algorithm to be evaluated;

[0120] The evaluation indicators are calculated using remote sensing data that conforms to the remote sensing algorithm to be evaluated.

[0121] The device provided in the embodiment of the present invention is briefly described. For matters not mentioned in the embodiment part, reference can be made to the corresponding contents in the aforementioned embodiment.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A remote sensing algorithm evaluation method with rapid deployment and automatic optimization functions, characterized in that: include: Constructing multiple virtual mirror environments according to the device operating environments and dependencies on which multiple remote sensing algorithms to be evaluated depend; Build standardized interfaces; Deploying the corresponding remote sensing algorithm to be evaluated in each of the virtual mirror environments; Get the assessment plan; Creating a batch scheduling task according to the evaluation plan; the batch scheduling task is used to obtain remote sensing data that meets the remote sensing algorithm to be evaluated according to the evaluation plan, and calculate the evaluation index; Compare evaluation indicators to optimize remote sensing algorithms.

2. The remote sensing algorithm evaluation method with rapid deployment and automatic optimization functions according to claim 1 is characterized in that: The method for constructing the virtual mirror environment specifically includes: Install Docker Engine; Add the current user to the Docker group; Create a Dockerfile, configure the operating environment of the Docker container, use an orchestration tool to combine multiple Docker containers according to the operating environments and dependencies of multiple devices on which the remote sensing algorithms to be evaluated depend, and package them to form the virtual image environment; The virtual mirror environment is stored.

3. The remote sensing algorithm evaluation method with rapid deployment and automatic optimization functions according to claim 1 is characterized in that: The standardized interface is used to unify the specifications: input data and path, output data and path, data set naming, and remote sensing algorithm scheduling parameters.

4. The remote sensing algorithm evaluation method with rapid deployment and automatic optimization functions according to claim 1 is characterized in that: The evaluation plan includes evaluation reference data, time and space scope, time and space matching plan and the evaluation indicators.

5. The remote sensing algorithm evaluation method with rapid deployment and automatic optimization functions according to claim 4 is characterized in that: The batch scheduling task is specifically used for: Acquisition of remote sensing data; Standardizing the remote sensing data to obtain standard data; Using the time-space matching scheme to screen the standard data within the time-space range to obtain remote sensing data that meets the remote sensing algorithm to be evaluated; The evaluation index is calculated using the remote sensing data that conforms to the remote sensing algorithm to be evaluated.

6. A remote sensing algorithm evaluation device with rapid deployment and automatic optimization functions, characterized in that: include: An image building unit: used for building multiple virtual image environments according to the device operating environments and dependencies on which multiple remote sensing algorithms to be evaluated depend; Interface construction unit: used to build standardized interfaces; Deployment unit: used for deploying the corresponding remote sensing algorithm to be evaluated in each of the virtual mirror environments; Evaluation unit: used to obtain the evaluation plan; create a batch scheduling task according to the evaluation plan; the batch scheduling task is used to obtain remote sensing data that meets the remote sensing algorithm to be evaluated according to the evaluation plan, and calculate the evaluation index; compare the evaluation index to achieve the optimization of the remote sensing algorithm.

7. The remote sensing algorithm evaluation device with rapid deployment and automatic optimization functions according to claim 6, characterized in that: The mirror image building unit is specifically used for: Install Docker Engine; Add the current user to the Docker group; Create a Dockerfile, configure the operating environment of the Docker container, use an orchestration tool to combine multiple Docker containers according to the operating environments and dependencies of multiple devices on which the remote sensing algorithms to be evaluated depend, and package them to form the virtual image environment; The virtual mirror environment is stored.

8. The remote sensing algorithm evaluation device with rapid deployment and automatic optimization functions according to claim 6, characterized in that: The standardized interface is used to unify the specifications: input data and path, output data and path, data set naming, and remote sensing algorithm scheduling parameters.

9. The remote sensing algorithm evaluation device with rapid deployment and automatic optimization functions according to claim 6, characterized in that: The evaluation plan includes evaluation reference data, time and space scope, time and space matching plan and the evaluation indicators.

10. The remote sensing algorithm evaluation device with rapid deployment and automatic optimization functions according to claim 9, characterized in that: The batch scheduling task is specifically used for: Acquisition of remote sensing data; Standardizing the remote sensing data to obtain standard data; Using the time-space matching scheme to screen the standard data within the time-space range to obtain remote sensing data that meets the remote sensing algorithm to be evaluated; The evaluation index is calculated using the remote sensing data that conforms to the remote sensing algorithm to be evaluated.

Images