Remote sensing image processing algorithm online evaluation and recommendation method

Abstract

The invention discloses a remote sensing image processing algorithm online evaluation and recommendation method, and belongs to the field of remote sensing image processing. The evaluation method comprises the following steps: constructing a standard data set required by algorithm evaluation; receiving a remote sensing image processing algorithm and a description file uploaded by a user; analyzing a description file of a remote sensing image processing algorithm; according to the algorithm type, matching a group of corresponding standard input data, taking each standard input data as an input parameter, and driving the remote sensing image processing algorithm to execute once; recording resource use information, algorithm execution time and output parameters of the algorithm, and calculating algorithm precision; and performing weighted summation on the resource use information, the algorithm execution time and the algorithm precision obtained by previous algorithm execution to obtain a comprehensive score of the algorithm. According to the recommendation method, on the basis of evaluation, an algorithm optimization deep learning model is constructed, and the most suitable algorithm is recommended, so that the image processing efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing image processing, and particularly relates to an online evaluation and recommendation method for remote sensing image processing algorithms. Background Art

[0002] With the development of fields such as commercial spaceflight, the number of on-orbit remote sensing satellites is increasing, and the quality of remote sensing data is also getting higher. All-weather, multi-angle, multi-temporal, multi-source, and multi-resolution earth observation has become a reality. Remote sensing data has been widely promoted and applied in multiple industrial fields such as military, environment, land, national defense, agriculture, forestry, water conservancy, transportation, and disaster reduction, promoting the rapid development of the national economy.

[0003] The application effectiveness of satellite remote sensing systems in various fields strongly depends on the performance of remote sensing algorithms. Due to the large amount of remote sensing data, manual interpretation alone is inefficient. The application of remote sensing data in various industries involves various complex remote sensing image processing algorithms. The effectiveness of remote sensing image processing algorithms has a direct impact on the application effect. Many current studies have shown that different remote sensing algorithms have different processing effects on different data. When processing actual data, it is necessary to optimize the algorithms according to different image processing tasks.

[0004] Currently, due to the lack of a scientific evaluation system for remote sensing image processing algorithms, the evaluation of remote sensing image processing algorithms faces problems such as insufficient objectivity and lack of unified standards, and thus there is also a lack of an algorithm optimization model, and it is impossible to optimize algorithms according to different image processing tasks. Summary of the Invention

[0005] In view of this, the present invention proposes an online evaluation and recommendation method for remote sensing image processing algorithms. The present invention classifies remote sensing image processing algorithms and constructs a standard evaluation data set, which can improve the scientificity, objectivity, and accuracy of remote sensing image processing algorithm evaluation. On this basis, an algorithm optimization deep learning model is constructed, which can recommend optimized algorithms according to different data processing tasks to improve the processing effect.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] An online evaluation method for remote sensing image processing algorithms, running on a server, includes the following steps:

[0008] Step 1, construct a standard data set required for algorithm evaluation, where the standard data set consists of standard input data and standard result data;

[0009] Step 2, receive the remote sensing image processing algorithm and description file uploaded by the user;

[0010] Step 3, perform format verification and integrity verification on the remote sensing image processing algorithm, and execute subsequent processes after passing the verification;

[0011] Step 4, parsing the description file of the remote sensing image processing algorithm, analyzing its algorithm type and required input parameters;

[0012] Step 5, according to the algorithm type, a set of corresponding standard input data is matched, each standard input data is used as an input parameter, and the remote sensing image processing algorithm is driven to execute once until all the matched standard input data are used up;

[0013] Step 6, recording resource usage information, algorithm execution time, and algorithm output parameters during each algorithm execution; wherein the resource usage information includes CPU utilization, memory utilization, hard disk utilization, and network bandwidth utilization;

[0014] Step 7, comparing the difference between the output parameters in each execution and the corresponding standard result data, and calculating the algorithm accuracy of the execution;

[0015] Step 8: Perform a weighted sum of the resource usage information, algorithm execution time, and algorithm accuracy obtained from each algorithm execution to obtain a comprehensive score for the algorithm.

[0016] Further, in step 2, the remote sensing image processing algorithm uploaded by the user is an executable file in the format of EXE, DLL, JAR, PYTHON script file or MATLAB script file;

[0017] The types of remote sensing image processing algorithms are single value calculation, distributed intensity calculation, binary calculation or enumerable multivariate calculation; among them, single value calculation refers to an algorithm that performs single value statistical calculation on a certain type of information in a remote sensing image, distributed intensity calculation is an algorithm that performs pixel-by-pixel intensity calculation based on a certain feature in a remote sensing image, binary calculation refers to an algorithm that performs binary division calculation on a remote sensing image, and enumerable multivariate calculation refers to an algorithm that performs multivariate division calculation on a remote sensing image;

[0018] The input and output parameters are single-band images, multi-band images or numerical values. Single-band images are divided into two categories: single-band binary images and single-band intensity images; multi-band images are divided into two categories: multi-band binary images and multi-band intensity images.

[0019] The description file is an XML file, and its content includes two categories: essential elements and auxiliary elements. Among them, the essential elements include algorithm form, algorithm type, input parameters, and output parameters; the auxiliary elements include algorithm version, algorithm author, algorithm default parameters, and algorithm description.

[0020] Furthermore, the specific method of step 7 is:

[0021] For remote sensing image processing algorithms for single-value calculations: directly compare the numerical results of the algorithm to be evaluated with the standard results corresponding to the standard data , and use as a measure of the algorithm's accuracy;

[0022] For remote sensing image processing algorithms for distributed intensity calculations: the output of these algorithms is a single-band intensity image, and the algorithm accuracy calculation method is as follows:

[0023] , where , are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data respectively;

[0024] For remote sensing image processing algorithms for binary calculations: the output of these algorithms is a single-band binary image, and the algorithm accuracy calculation method is as follows:

[0025] , where , are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data respectively;

[0026] For remote sensing image processing algorithms for enumerable multi-variable calculations: the output of these algorithms is multiple binary images or an integer value image, and the algorithm accuracy calculation method is as follows:

[0027] , where , are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data respectively, and the symbol is used to determine whether two numerical values are equal, with the equal result being 0 and the unequal result being 1.

[0028] Furthermore, the specific method for step 8 is as follows:

[0029] Step 801, calculate the scores of CPU utilization, memory utilization, hard disk utilization, and network bandwidth utilization , , , , where , , , are all calculated as (1 - u) × 100, where u is the corresponding resource utilization rate;

[0030] Step 802, calculate the score of the algorithm execution time:

[0031] ,

[0032] Where t is the algorithm execution time in seconds;

[0033] Step 803: the range of the algorithm accuracy is , the score of algorithm accuracy Multiply the algorithm accuracy by 100;

[0034] Step 804, calculate the comprehensive score :

[0035] ,

[0036] in, is the weight coefficient, the sum is 1, the weight coefficient of the algorithm accuracy score Not less than 0.7.

[0037] An online recommendation method for remote sensing image processing algorithms, running in a server, comprises the following steps:

[0038] Step 1: Build an algorithm-optimized deep learning model;

[0039] Step 2: construct a training data set. The training data set is divided into multiple subsets. Each subset corresponds to a remote sensing image processing algorithm. Each training data includes an algorithm description file of the remote sensing image processing algorithm corresponding to the subset, a set of images to be processed, and a comprehensive score. The comprehensive score is obtained as follows:

[0040] Step 201, inputting an image to be processed into a remote sensing image processing algorithm, and driving the remote sensing image processing algorithm to execute once;

[0041] Step 202, repeating step 201 until all the images to be processed in the training data are used up; recording resource usage information, algorithm execution time, and output parameters of the algorithm during each algorithm execution; wherein the resource usage information includes CPU utilization, memory utilization, network bandwidth utilization, and hard disk utilization;

[0042] Step 203, comparing the difference between the output parameters in each execution and the corresponding standard result data, and calculating the algorithm accuracy of the execution;

[0043] Step 204, performing weighted summation of resource usage information, algorithm execution time, and algorithm accuracy obtained from previous algorithm executions to obtain a comprehensive score for the algorithm;

[0044] Step 3: Use the training data set to train the algorithm optimization deep learning model, and deploy the trained algorithm optimization deep learning model in the server;

[0045] Step 4: Receive an image to be processed uploaded by the user and algorithm description files of multiple optional remote sensing image processing algorithms, and use the algorithm optimization deep learning model for algorithm optimization recommendation.

[0046] Further, in Step 1, the algorithm optimization deep learning model is a deep learning model that uses the VGG16 model and the CBOW model. The VGG16 model is responsible for processing the input image data, and the CBOW model is responsible for processing the input algorithm description file. After the results of the VGG16 model and the CBOW model are convolved through multiple layers, they are regressed to the output layer.

[0047] Further, the input data of the algorithm optimization deep learning model is the evaluation dataset and the algorithm description file, and the output data is the algorithm evaluation result. The specific method of Step 3 is as follows:

[0048] Step 301: Set the loss function as the mean square error of the comprehensive score : , where || || represents the 2-norm, represents the evaluation result output by the algorithm optimization deep learning model;

[0049] Step 302: Each time during training, input the evaluation dataset and the algorithm description file into the model respectively, and use the evaluation result error to perform backpropagation of the gradient on the model to obtain the model parameters;

[0050] Step 303: Repeat the training process until the proportion of the loss function is less than 5%, that is , stop training.

[0051] Further, the specific method of Step 4 is as follows:

[0052] Step 401: Input the image to be processed and multiple algorithm description files into the algorithm optimization deep learning model;

[0053] Step 402: Output the comprehensive score of each algorithm under the current image to be processed through the algorithm optimization deep learning model;

[0054] Step 403: Recommend the algorithm with the highest comprehensive score to the customer.

[0055] The present invention has the following beneficial effects:

[0056] 1. The present invention constructs an algorithm classification system, and adopts different evaluation strategies according to the characteristics of different types of algorithms to achieve more scientific evaluation;

[0057] 2. The present invention constructs an algorithm optimization deep learning model, which can recommend optimized algorithms according to different image processing tasks, improving the processing efficiency.

[0058] In summary, the present invention classifies remote sensing image processing algorithms, constructs a standard evaluation data set, and constructs an optimal algorithm deep learning model on the basis of scientifically, objectively, and accurately evaluating remote sensing image processing algorithms. It can recommend optimal algorithms for different image processing tasks and improve processing efficiency. Detailed implementation manners

[0059] The following further elaborates on the present invention in conjunction with specific embodiments.

[0060] An online evaluation method for remote sensing image processing algorithms, which runs on a server and includes the following steps:

[0061] Step 1: Construct a standard data set required for algorithm evaluation. The standard data set consists of standard input data and standard result data; the standard data set adopts a classification management method, which includes standard data for different types of remote sensing image processing algorithms.

[0062] Step 2: Receive the remote sensing image processing algorithm and description file uploaded by the user; specifically, the remote sensing image processing algorithm and description file need to meet the following rules:

[0063] (1) Algorithm form: The algorithm should be an executable file, in one of the forms of EXE, DLL, JAR, PYTHON script file, and MATLAB script file;

[0064] (2) Algorithm type: Divided into four types: single-value calculation, distributed intensity calculation, binary calculation, and enumerable multi-value calculation. Among them, single-value calculation refers to an algorithm for performing single-value statistical calculation on a certain type of information in a remote sensing image, such as the area of water in an image; distributed intensity calculation is an algorithm for performing pixel-by-pixel intensity calculation based on a certain feature in a remote sensing image, such as the normalized difference vegetation index of a remote sensing image; binary calculation refers to an algorithm for performing binary division calculation on a remote sensing image, such as the extraction of roads in an image; enumerable multi-value calculation refers to an algorithm for performing multi-value division calculation on a remote sensing image, such as a land cover classification algorithm.

[0065] (3) Algorithm input and output parameters: The input parameters and output parameters are divided in the same way, into three major types: single-band image, multi-band image, and numerical value. Among them, single-band images are divided into two categories: single-band binary images and single-band intensity images, and multi-band images are divided into two categories: multi-band binary images and multi-band intensity images.

[0066] (4) Algorithm description file: In the form of an xml file, the content includes two major categories: essential elements and auxiliary elements. Among them, the essential elements include algorithm form, algorithm type, algorithm input parameters, and algorithm output parameters; the auxiliary elements include algorithm version, algorithm author, algorithm default parameters, algorithm description, etc.

[0067] In specific implementation, based on the BS architecture, a user Web interaction interface can be provided for the client. The user uploads a remote sensing image processing algorithm through human-computer interaction, and the system transfers the remote sensing image algorithm from the client to the server.

[0068] The standard data set consists of two groups of data, namely standard input data and standard result data, and they are in one-to-one correspondence. The standard input data is mainly single-band and multi-band images, which are classified according to the classification method of input parameters. The standard result data is classified according to the classification method of output parameters. The specific quantity of the standard data set can be constructed according to the application scenario, such as standard data of vegetation index, standard data of water body index, etc.

[0069] Step 3: Perform format verification and integrity verification on the remote sensing image processing algorithm. After passing the verification, execute the subsequent process;

[0070] Step 4: Parse the description file of the remote sensing image processing algorithm, and analyze its algorithm type and the required input parameters;

[0071] Step 5: According to the algorithm type, match a group of corresponding standard input data, and use each standard input data as an input parameter to drive the remote sensing image processing algorithm to execute once until all the matched standard input data is used up;

[0072] Step 6: Record the resource usage information, algorithm execution time, and output parameters of the algorithm during each algorithm execution; among them, the resource usage information includes CPU utilization rate, memory utilization rate, hard disk utilization rate, and network bandwidth utilization rate;

[0073] Step 7: Compare the difference between the output parameters and the corresponding standard result data during each execution, and calculate the algorithm accuracy of this execution; specifically:

[0074] (1) Single-value calculation type: Directly compare the numerical result of the algorithm to be evaluated with the standard result corresponding to the standard data , and use as the measure of algorithm accuracy;

[0075] (2) Distributed intensity calculation type: The output result of this type of algorithm is a single-band intensity image, and the algorithm accuracy calculation method is as follows:

[0076] , where , are respectively the result generated by the algorithm to be evaluated and the standard result corresponding to the standard data;

[0077] (3) Binary calculation type: The output result of this type of algorithm is a single-band binary image, and the algorithm accuracy calculation method is as follows:

[0078] , where and are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data, respectively;

[0079] (4) Enumerable multi - variable calculation type: The results produced by this type of algorithm are multiple binary images or an integer - valued image. These two results are essentially equivalent. This method takes an integer - valued image as an example to illustrate the method for evaluating the algorithm accuracy, and its calculation method is as follows:

[0080] , where and are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data, respectively. The symbol is used to judge whether two numerical values are equal. If they are equal, the result is 0; if not, the result is 1.

[0081] Step 8: Perform a weighted sum of the resource usage information, algorithm execution time, and algorithm accuracy obtained from each execution of the algorithm to obtain the comprehensive score of the algorithm. Specifically, the resource occupancy situation is directly collected from the system information, and the collection parameters include CPU utilization rate, memory utilization rate, hard disk utilization rate, and network bandwidth utilization rate, all of which are percentages;

[0082] The scoring method for the resource occupancy situation is (1 - u)×100, where u is the corresponding resource utilization rate; the scores of CPU utilization rate, memory utilization rate, hard disk utilization rate, and network bandwidth utilization rate are and and and ;

[0083] The unit of the algorithm execution time t is seconds, and its score is related to the data size (unit: GB). The specific formula is ;

[0084] The algorithm accuracy value is designed to be a number in the interval, and its score is directly the algorithm accuracy multiplied by 100;

[0085] Finally, calculate the comprehensive score: , where, is the weight coefficient, and the sum is 1, which can be determined according to the actual situation. In particular, the algorithm accuracy coefficient should not be less than 0.7.

[0086] This method constructs an algorithm classification system. According to the characteristics of different types of algorithms, different evaluation strategies are adopted, realizing a more scientific evaluation of remote - sensing image - processing algorithms.

[0087] By using the scoring method in the above evaluation method, an online recommendation method for remote sensing image processing algorithm can also be provided. The method runs in a server and includes the following steps:

[0088] Step 1, construct an algorithm-optimized deep learning model; the algorithm-optimized deep learning model is a deep learning model that uses the VGG16 model and the CBOW model, where the VGG16 model is responsible for processing the input image data, and the CBOW model is responsible for processing the input algorithm description file; the results of the VGG16 model and the CBOW model are regressed to the output layer after multiple layers of convolution.

[0089] Step 2: construct a training data set. The training data set is divided into multiple subsets. Each subset corresponds to a remote sensing image processing algorithm. Each training data includes an algorithm description file of the remote sensing image processing algorithm corresponding to the subset, a set of images to be processed, and a comprehensive score. The comprehensive score is obtained as follows:

[0090] Step 201, inputting an image to be processed into a remote sensing image processing algorithm, and driving the remote sensing image processing algorithm to execute once;

[0091] Step 202, repeating step 201 until all the images to be processed in the training data are used up; recording resource usage information, algorithm execution time, and output parameters of the algorithm during each algorithm execution; wherein the resource usage information includes CPU utilization, memory utilization, network bandwidth utilization, and hard disk utilization;

[0092] Step 203, comparing the difference between the output parameters in each execution and the corresponding standard result data, and calculating the algorithm accuracy of the execution;

[0093] Step 204, performing weighted summation of resource usage information, algorithm execution time, and algorithm accuracy obtained from previous algorithm executions to obtain a comprehensive score for the algorithm;

[0094] Step 3: Use the training data set to train the algorithm optimization deep learning model, and deploy the trained algorithm optimization deep learning model on the server; the specific method is:

[0095] (1) Definition of model input and output data: In the training phase, the model input data is the evaluation data set and algorithm description document, and the output data is the algorithm evaluation result. In the inference phase, the model input data is the data set to be processed and the algorithm, and the output is the algorithm optimization result;

[0096] (2) Model composition: A deep learning model that combines the VGG16 model and CBOW is adopted. The VGG16 model is responsible for processing the input image data, and the CBOW model is responsible for processing the input algorithm description document. After the results of VGG16 and CBOW are convolved through multiple layers, they are regressed to the output layer, where the loss function is the mean square error of the algorithm score: , where || || represents the 2-norm, represents the evaluation result output by the algorithm-optimized deep learning model;

[0097] (3) Model training: Different algorithms are used to process the same image data to obtain training data consisting of images, algorithms, and evaluation results. The algorithm-optimized deep learning model is used to regress the evaluation results. Each time during training, the evaluation dataset and the algorithm description file are input into the model respectively, and the model is backward gradient propagated using the evaluation result error to obtain the model parameters. The training process is repeated until the proportion of the loss function is less than 5%, that is when the training is stopped to obtain the final model parameters.

[0098] Step 4: Receive an image to be processed uploaded by the user and the algorithm description files of multiple optional remote sensing image processing algorithms, and use the algorithm-optimized deep learning model for algorithm-optimized recommendation; the specific method is as follows:

[0099] Step 401: Input the image to be processed and multiple algorithm description files into the algorithm-optimized deep learning model;

[0100] Step 402: Output the comprehensive score of each algorithm under the current image to be processed through the algorithm-optimized deep learning model;

[0101] Step 403: Recommend the algorithm with the highest comprehensive score to the customer.

[0102] This method uses a deep learning model that combines the VGG16 model and CBOW to perform regression calculations on the algorithm evaluation results, and can automatically recommend optimized algorithms for different data processing tasks to improve processing efficiency.

[0103] In summary, the present invention selects corresponding standard data for evaluation according to the type of algorithm, adopts different evaluation methods for different algorithm types, and realizes scientific, objective, and accurate evaluation of remote sensing image processing algorithms. On the basis of algorithm evaluation, an algorithm-optimized deep learning model is constructed to regress the algorithm evaluation results. This model preferably recommends the most suitable algorithm for a specified image processing task to improve image processing efficiency.

Claims

1. An online evaluation method for remote sensing image processing algorithms, characterized in that, Running on the server includes the following steps: Step 1: Construct a standard data set required for algorithm evaluation. The standard data set consists of standard input data and standard result data. Step 2, receiving the remote sensing image processing algorithm and description file uploaded by the user; Step 3, perform format verification and integrity verification on the remote sensing image processing algorithm, and execute subsequent processes after passing the verification; Step 4, parsing the description file of the remote sensing image processing algorithm, analyzing its algorithm type and required input parameters; Step 5, according to the algorithm type, a set of corresponding standard input data is matched, each standard input data is used as an input parameter, and the remote sensing image processing algorithm is driven to execute once until all the matched standard input data are used up; Step 6, recording resource usage information, algorithm execution time, and algorithm output parameters during each algorithm execution; wherein the resource usage information includes CPU utilization, memory utilization, hard disk utilization, and network bandwidth utilization; Step 7, comparing the difference between the output parameters in each execution and the corresponding standard result data, and calculating the algorithm accuracy of the execution; Step 8: Perform a weighted sum of the resource usage information, algorithm execution time, and algorithm accuracy obtained from each algorithm execution to obtain a comprehensive score for the algorithm.

2. The online evaluation method for a remote sensing image processing algorithm according to claim 1, characterized in that In step 2, the remote sensing image processing algorithm uploaded by the user is an executable file in the format of EXE, DLL, JAR, PYTHON script file or MATLAB script file; The types of remote sensing image processing algorithms are single value calculation, distributed intensity calculation, binary calculation or enumerable multivariate calculation; among them, single value calculation refers to an algorithm that performs single value statistical calculation on a certain type of information in a remote sensing image, distributed intensity calculation is an algorithm that performs pixel-by-pixel intensity calculation based on a certain feature in a remote sensing image, binary calculation refers to an algorithm that performs binary division calculation on a remote sensing image, and enumerable multivariate calculation refers to an algorithm that performs multivariate division calculation on a remote sensing image; The input and output parameters are single-band images, multi-band images or numerical values. Single-band images are divided into two categories: single-band binary images and single-band intensity images; multi-band images are divided into two categories: multi-band binary images and multi-band intensity images. The description file is an XML file, and its content includes two categories: essential elements and auxiliary elements. Among them, the essential elements include algorithm form, algorithm type, input parameters, and output parameters; the auxiliary elements include algorithm version, algorithm author, algorithm default parameters, and algorithm description.

3. An online evaluation method for a remote sensing image processing algorithm according to claim 1, characterized in that, The specific method of step 7 is: For single-value calculation remote sensing image processing algorithms: directly compare the numerical results of the algorithm to be evaluated with the standard results corresponding to the standard data , and use as a measure of the algorithm's accuracy; For distributed intensity calculation remote sensing image processing algorithms: the output of this type of algorithm is a single-band intensity image, and the algorithm accuracy is calculated as follows: , where and are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data, respectively; For binary computing remote sensing image processing algorithms: the result of this type of algorithm is a single-band binary image, and the algorithm accuracy is calculated as follows: , where and are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data, respectively; For enumerable multivariate computing remote sensing image processing algorithms: the results of this type of algorithm are multiple binary images or an integer value image. The algorithm accuracy is calculated as follows: , where and are the results generated by the algorithm to be evaluated and the standard results corresponding to the standard data respectively, The symbol is used to determine whether two numerical values are equal. If they are equal, the result is 0; if not, the result is 1.

4. A method for online evaluation of a remote sensing image processing algorithm according to claim 1, characterized in that, The specific method of step 8 is: Step 801, calculate the scores of CPU utilization rate, memory utilization rate, hard disk utilization rate, and network bandwidth utilization rate , , , , where , , , are all calculated as (1 - u) × 100, where u is the corresponding resource utilization rate; Step 802, calculate the score of the algorithm execution time: ， Where t is the algorithm execution time in seconds; Step 803, the value range of the algorithm accuracy is , and the score of the algorithm accuracy is the algorithm accuracy multiplied by 100; Step 804, calculate the comprehensive score : ， Among them, is the weight coefficient, and the sum is 1. The weight coefficient of the algorithm accuracy score is not less than 0.

7.

5. An online recommendation method for remote sensing image processing algorithms, characterized in that, Running on the server includes the following steps: Step 1, construct an algorithm-optimized deep learning model; Step 2, construct a training data set. The training data set is divided into multiple subsets, each subset corresponding to a remote sensing image processing algorithm. Each training data includes the algorithm description file of the remote sensing image processing algorithm corresponding to the subset it belongs to, a set of images to be processed, and a comprehensive score. The way to obtain the comprehensive score is as follows: Step 201, input an image to be processed into the remote sensing image processing algorithm to drive the remote sensing image processing algorithm to execute once; Step 202, repeat Step 201 until all the images to be processed in the training data are used up; record the resource usage information, algorithm execution time, and output parameters of the algorithm during each algorithm execution. The resource usage information includes CPU utilization, memory utilization, network bandwidth utilization, and hard disk utilization; Step 203, compare the difference between the output parameters and the corresponding standard result data during each execution, and calculate the algorithm accuracy of this execution; Step 204, perform weighted summation on the resource usage information, algorithm execution time, and algorithm accuracy obtained from each algorithm execution to obtain the comprehensive score of the algorithm; Step 3, use the training data set to train the algorithm-optimized deep learning model, and deploy the trained algorithm-optimized deep learning model on the server; Step 4, receive an image to be processed uploaded by the user and the algorithm description files of multiple optional remote sensing image processing algorithms, and use the algorithm-optimized deep learning model to perform algorithm-optimized recommendation.

6. The online recommendation method for a remote sensing image processing algorithm according to claim 5, characterized in that, In Step 1, the algorithm-optimized deep learning model is a deep learning model using the VGG16 model and the CBOW model. The VGG16 model is responsible for processing the input image data, and the CBOW model is responsible for processing the input algorithm description file. The results of the VGG16 model and the CBOW model are convolved through multiple layers and then regressed to the output layer.

7. An online recommendation method for a remote sensing image processing algorithm according to claim 5, characterized in that The input data of the algorithm-optimized deep learning model is the evaluation data set and the algorithm description file, and the output data is the algorithm evaluation result. The specific method of Step 3 is as follows: Step 301, set the loss function as the mean squared error of the comprehensive score : , where || || represents the 2-norm, represents the evaluation result output by the algorithm-optimized deep learning model; Step 302, during each training, input the evaluation data set and the algorithm description file into the model respectively, and use the evaluation result error to perform backpropagation of the gradient on the model to obtain the model parameters; Step 303. Repeat the training process until the proportion of the loss function is less than 5%, that is when the training is stopped.

8. The online recommendation method for a remote sensing image processing algorithm according to claim 5, characterized in that The specific method of Step 4 is as follows: Step 401, input the image to be processed and multiple algorithm description files into the algorithm-optimized deep learning model; Step 402, output the comprehensive score of each algorithm under the current image to be processed through the algorithm-optimized deep learning model; Step 403, recommend the algorithm with the highest comprehensive score to the customer.

Images