Remote sensing intelligent application model updating method and device

Abstract

The invention discloses a remote sensing intelligent application model updating method and device. The method comprises the following steps: acquiring a remote sensing intelligent application basic model and a training data set; processing the remote sensing intelligent application basic model by utilizing the training data set to obtain a first application optimization model and a training data label set; and processing the first application optimization model by using a measured data set and the training data label set to obtain a remote sensing intelligent application optimization model. According to the method, the increment of the training sample data is controlled on the basis of the user demand and the sample positioning and classification accuracy, the intelligent model is updated and iterated by the continuously updated data, and the continuous updating and learning capabilities of the remote sensing data intelligent model are improved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and remote sensing data processing, and particularly to a method and device for updating a remote sensing intelligent application model. Background Art

[0002] The combination of artificial intelligence and remote sensing data has become the mainstream mode of current remote sensing technology applications. In practical applications, after an offline-trained remote sensing intelligent model with certain general performance is put into operation, facing different user groups and different application scenarios, the key points that the intelligent model needs to "focus on" subsequently are different. In addition, over time, the surface features are constantly changing, the remote sensing observation data is continuously updated, and the distribution of interesting objects and data features in the data will also change. This requires the remote sensing data intelligent model to have the ability of continuous update and learning based on new data. However, directly using new data to train the model will cause it to forget the knowledge of old data, and using new data and historical data to update and train the model at the same time can take into account both new and old knowledge, but as new data instances continue to accumulate, the computational cost of training continues to increase. How to implement customized update and iteration of the intelligent model based on user needs and continuously updated data has become the key to making the remote sensing intelligent model easy to use and durable. The present invention controls the increment of training sample data based on the sample location classification accuracy, proposes a set of online update methods for training data of remote sensing intelligent application models based on user needs, and uses continuously updated data to update and iterate the intelligent model, improving the continuous update and learning ability of the remote sensing data intelligent model. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method and device for updating a remote sensing intelligent application model, which improves the iterative optimization efficiency of the intelligent positioning model of remote sensing images, thereby enhancing the continuous update and learning ability of the remote sensing data intelligent model.

[0004] To solve the above technical problem, in the first aspect of the embodiments of the present invention, a method for updating a remote sensing intelligent application model is disclosed, and the method includes:

[0005] S1, obtaining a remote sensing intelligent application basic model and a training data set;

[0006] S2, processing the remote sensing intelligent application basic model by using the training data set to obtain a first application optimization model and a training data label set;

[0007] S3, processing the first application optimization model by using a measured data set and the training data label set to obtain a remote sensing intelligent application optimization model.

[0008] As an alternative implementation, in the first aspect of the embodiments of the present invention, processing the remote sensing intelligent application basic model using the training data set to obtain a first application optimization model and a training data label set includes:

[0009] S21, obtaining the training data set;

[0010] S22, dividing the training data set according to a preset ratio to obtain a data training set and a data test set;

[0011] Set the training times value to 1;

[0012] S23, using the remote sensing intelligent application basic model to process the data training set to obtain first image object information;

[0013] S24, based on the model loss function, updating the parameters of the remote sensing intelligent application basic model using the first image object information and the training data label set;

[0014] S25, determining whether the training times value is greater than a preset training times threshold to obtain a training times judgment result;

[0015] When the training times judgment result is yes, execute S23;

[0016] When the training times judgment result is no, increase the training times value by 1 and execute S26;

[0017] S26, using the remote sensing intelligent application basic model to process the data test set to obtain second image object information;

[0018] S27, comparing the second image object information with the data test label set to obtain prediction accuracy information;

[0019] S28, determining whether the prediction accuracy information is greater than a preset accuracy threshold to obtain an accuracy discrimination result;

[0020] When the accuracy discrimination result is no, execute S22;

[0021] If the accuracy discrimination result is yes, obtain a first application optimization model and execute S29;

[0022] S29, performing a fusion process on the data training label set and the data test label set to obtain a training data label set.

[0023] As an alternative implementation, in the first aspect of the embodiments of the present invention, the process of using the measured data set and the training data label set to process the first application optimization model to obtain a remote sensing intelligent application optimization model includes:

[0024] S31. Obtain a target area vector mask set;

[0025] S32. Based on the target area vector mask set, process the measured data set to obtain a first positioning and classification data set and a second positioning and classification data set;

[0026] The first positioning and classification data set includes a number of first positioning and classification data;

[0027] The second positioning and classification data set includes a number of second positioning and classification data;

[0028] S33. Use the target uncertainty acquisition model to process the training data set, the first positioning and classification data set, and the second positioning and classification data set to obtain a training data increment set;

[0029] S34. Based on the training data label set, label the training data increment set to obtain an increment data label set;

[0030] S35. Based on the increment data label set, use the training data increment set to process the first application optimization model to obtain a second application optimization model;

[0031] S36. Perform a fusion process on the increment training label set and the training data label set to obtain a historical label set;

[0032] Increase the increment processing times by 1;

[0033] S37. Determine whether the increment processing times is greater than the iteration times to obtain an iteration judgment result;

[0034] When the iteration judgment result is negative, update the training data increment set to the training data set, update the historical label set to the training data label set, and execute S32;

[0035] When the iteration judgment result is positive, update the second remote sensing image object classification optimization model to the remote sensing intelligent application optimization model.

[0036] As an alternative implementation, in the first aspect of the embodiments of the present invention, the process of obtaining the target area vector mask set includes:

[0037] S311. Obtain a target area remote sensing image set;

[0038] S312. Traverse the remote sensing image set of the target area to obtain all remote sensing images of the target area;

[0039] S313. Use the first application optimization model to process any one of the remote sensing images of the target area to obtain target positioning and classification result information;

[0040] S314. Examine the target positioning and classification result information to obtain a target area vector mask;

[0041] S315. Use the area vector mask acquisition model to process the target area vector mask to obtain a target area vector mask set;

[0042] The expression of the area vector mask acquisition model is:

[0043]

[0044] where A represents the target area vector mask set; A i ; represents the i-th target area vector mask; i represents the index of the target area; U represents the union calculation.

[0045] As an optional implementation manner, in the first aspect of the embodiments of the present invention, the use of the target uncertainty acquisition model to process the training data set, the first positioning and classification data set, and the second positioning and classification data set to obtain a training data increment set includes:

[0046] S331. Process the training data set and the first positioning and classification data set to obtain a first incremental training sample subset;

[0047] The first incremental training sample subset includes a number of first incremental training samples;

[0048] S332. Based on the uncertainty calculation model, process the second positioning and classification data set to obtain a second incremental training sample subset;

[0049] The second incremental training sample subset includes a number of second incremental training samples;

[0050] S333. Arrange and combine all the first incremental training samples and the second incremental training samples in sequence to obtain a training data increment set.

[0051] As an optional implementation manner, in the first aspect of the embodiments of the present invention, the expression of the target uncertainty acquisition model is:

[0052] IA{img i} = 1 - AP50(TL(img i), PR(img i ));

[0053] Among them, i represents the index of the remote sensing image; img i is the i-th remote sensing image; IA{imgg i} represents the target uncertainty of the i-th remote sensing image; TL(img i ) is the target object label in the i-th remote sensing image; PR(img i ) is the prediction result of the i-th remote sensing image; AP50() represents calculating the first threshold accuracy of the i-th remote sensing image.

[0054] As an alternative implementation, in the first aspect of the embodiments of the present invention, the processing of the training data set and the first positioning and classification data set to obtain the first incremental training sample subset includes:

[0055] S3311, performing a union calculation on the training data set and the first positioning and classification data set to obtain a third positioning and classification data set;

[0056] S3312, traversing the third positioning and classification data set to obtain all the third positioning and classification data;

[0057] S3313, using the first application optimization model to process any one of the third positioning and classification data to obtain positioning target type information;

[0058] The positioning target type information includes target quantity information and target type information;

[0059] S3314, arranging and combining all the third positioning and classification data in order according to the amount of the target quantity information to obtain a fourth positioning and classification data set;

[0060] S3315, performing uniform sampling processing on the fourth positioning and classification data set to obtain the first incremental training sample subset.

[0061] As an alternative implementation, in the first aspect of the embodiments of the present invention, using the training data increment set to process the first application optimization model to obtain a second application optimization model includes:

[0062] S351, obtaining the training data increment set;

[0063] The training data increment set includes several incremental training data; the incremental training data is training data or first classification and positioning data or second classification and positioning data;

[0064] S352, traversing the training data increment set to obtain all the incremental training data;

[0065] S353. Parse any of the incremental training data to obtain data type information;

[0066] S354. Based on the data type information, perform a matching process on the training data label set to obtain incremental training label information;

[0067] S355. Combine all the incremental training label information in sequence to obtain an incremental training label set

[0068] S356. Based on the incremental training label set, use the training data increment set to process the first application optimization model to obtain a second application optimization model.

[0069] The second aspect of the embodiments of the present invention discloses a remote sensing intelligent application model update, and the device includes:

[0070] An acquisition module, configured to acquire a remote sensing intelligent application basic model and a training data set;

[0071] A first processing module, configured to process the remote sensing intelligent application basic model by using the training data set to obtain a first application optimization model and a training data label set;

[0072] A second processing module, configured to process the first application optimization model by using an actual measurement data set and the training data label set to obtain a remote sensing intelligent application optimization model.

[0073] The third aspect of the present invention discloses another remote sensing intelligent application model update, and the device includes:

[0074] A memory storing executable program code;

[0075] A processor coupled to the memory;

[0076] The processor calls the executable program code stored in the memory and executes some or all of the steps in the remote sensing intelligent application model update method disclosed in the first aspect of the embodiments of the present invention.

[0077] The fourth aspect of the present invention discloses a computer-readable storage medium, and the computer-readable storage medium stores computer instructions, which are used to execute some or all of the steps in the remote sensing intelligent application model update method disclosed in the first aspect of the embodiments of the present invention when called.

[0078] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0079] In the embodiments of the present invention, the training data of the remote sensing intelligent application model is updated online based on user requirements and usage feedback, and the intelligent model is updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0081] Figure 1 It is a schematic diagram of the scenario of a remote sensing intelligent application model update system disclosed in the embodiments of the present invention;

[0082] Figure 2 It is a schematic flowchart of a method for updating a remote sensing intelligent application model disclosed in the embodiments of the present invention;

[0083] Figure 3 It is a schematic structural diagram of a device for updating a remote sensing intelligent application model disclosed in the embodiments of the present invention;

[0084] Figure 4 It is a schematic structural diagram of another device for updating a remote sensing intelligent application model disclosed in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0085] In order to enable those skilled in the art to better understand the solutions of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0086] The terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product, or equipment that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or equipment.

[0087] References to "embodiments" in this specification mean that a particular feature, structure, or characteristic described in connection with the embodiments can be included in at least one embodiment of the invention. The phrase appears in various places in the specification and is not necessarily referring to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive of other embodiments. Those skilled in the art will understand explicitly and implicitly that the embodiments described herein can be combined with other embodiments.

[0088] In this application, the term "exemplary" is used to mean "serving as an example, instance, or illustration". Any embodiment described as "exemplary" in this application is not necessarily to be construed as more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. In the following description, details are set forth for the purpose of explanation. It should be understood that those of ordinary skill in the art can recognize that this application can be implemented without these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of this application with unnecessary details. Therefore, this application is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0089] It should be noted that since the method of the embodiments of this application is executed in a computer device, the processing objects of each computer device exist in the form of data or information. For example, time, which is actually time information. It can be understood that in subsequent embodiments, if dimensions, quantities, positions, etc. are mentioned, they are all corresponding data existences for the computer device to process, and specific details are not elaborated here.

[0090] It should be noted that a brief description of the artificial intelligence-related technologies that may be involved in this application is provided. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines to enable the machines to have the functions of perception, reasoning, and decision-making.

[0091] Artificial intelligence technology is an interdisciplinary subject that involves a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0092] Computer Vision Technology (CV) Computer vision is a science that studies how to enable machines to "see". More specifically, it refers to using cameras and computers to replace human eyes to perform machine vision such as target recognition and measurement, and further perform graphic processing to make the computer process images that are more suitable for human eyes to observe or be transmitted to instruments for detection. As a scientific discipline, computer vision studies related theories and technologies and attempts to build artificial intelligence systems that can obtain information from images or multi-dimensional data. Computer vision technology usually includes technologies such as image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping, and also includes common biometric recognition technologies such as face recognition and fingerprint recognition.

[0093] Single-modal information is data of only one type, such as one of the data information of text, image, audio, video, electromagnetic signal, etc. Multi-modal information is data information that includes at least 2 types of single-modal information. Further, multi-modal information is applicable to complex tasks that require integrating multiple information sources, such as sentiment analysis, robot interaction, and autonomous driving. By integrating information of multiple modalities, higher performance and accuracy can usually be achieved in tasks.

[0094] A large model refers to an artificial neural network model with a very large number of parameters. In the field of artificial intelligence, a large model usually refers to a model with hundreds of millions to trillions of parameters. The model usually needs to be trained on a large-scale dataset and requires a large amount of computing resources for optimization and adjustment. Large models are usually used to solve complex natural language processing, computer vision, and speech recognition tasks. Generative AI is a type of AI that can create new content and ideas, including conversations, stories, images, videos, and music. In the embodiments of this application, the large model can be large language models such as ChatGPT, BERT, XLNet, Zhipu Model, Claude, Moonshot AI Model, ChatGLM Model, Qianyitongwen Model, MiniMax Model, Spark Model, Llama Model, 360GPT Model, Qwen Model, Baichuan Model, Lark Model, vivoLM Model, and Wenxin Yiyan, and the embodiments of this application do not make any limitations.

[0095] The embodiments of the present application provide a method, device, computer device and computer-readable storage medium for updating a remote sensing intelligent application model, which will be described in detail below.

[0096] Please refer to Figure 1 , Figure 1 , which is a schematic diagram of the application scenario of the remote sensing intelligent application model updating system provided by the embodiments of the present application in a remote sensing positioning system. The remote sensing positioning system may include a computer device 100, and the remote sensing intelligent application model updating system is integrated in the computer device 100, such as Figure 1 the computer device in

[0097] In the embodiments of the present application, the computer device 100 may be an independent server, or a server network or server cluster composed of servers. For example, the computer device 100 described in the embodiments of the present application includes, but is not limited to, a computer, a network host, a single network server, a set of multiple network servers, or a cloud server composed of multiple servers. Among them, the cloud server is composed of a large number of computers or network servers based on cloud computing.

[0098] It can be understood that the computer device 100 used in the embodiments of the present application may be a device that includes both receiving and transmitting hardware, that is, a device that has receiving and transmitting hardware capable of performing two-way communication on a two-way communication link. Such a device may include: a cellular or other communication device, which has a single-line display or a multi-line display or a cellular or other communication device without a multi-line display. Specifically, the computer device 100 may be a desktop terminal or a mobile terminal, and the computer device 100 may specifically be one of a mobile phone, a tablet computer, a laptop computer, etc.

[0099] Those skilled in the art can understand that Figure 1 the application environment shown in Figure 1 is only an application scenario of the solution of the present application, and does not constitute a limitation on the application scenario of the solution of the present application. Other application environments may also include more or fewer computer devices than those shown in Figure 1 . For example, only 1 computer device is shown in

[0100] . It can be understood that the system may further include one or more other services, which are not specifically limited here. Figure 1 In addition, as shown in

[0101] It should be noted that Figure 1The scene schematic diagram of the remote sensing positioning system shown is only an example. The multi-sample simulation control system and the scene for air-ground game described in the embodiments of this application are for more clearly explaining the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As can be known to those of ordinary skill in the art, with the evolution of the simulation control management system and the emergence of new business scenarios, the technical solutions provided by the embodiments of this application are equally applicable to similar technical problems.

[0102] The present invention discloses a method and device for updating a remote sensing intelligent application model. The present invention performs online update on the training data of the remote sensing intelligent application model based on user requirements and usage feedback, and uses the continuously updated data to update and iterate the intelligent model, improving the continuous update and learning ability of the remote sensing data intelligent model. The following will be described in detail respectively.

[0103] Embodiment 1

[0104] Please refer to Figure 2 , Figure 2 which is a schematic flowchart of a method for updating a remote sensing intelligent application model disclosed in an embodiment of the present invention. Among them, Figure 2 The described method for updating the remote sensing intelligent application model is applied to a remote sensing object positioning application system, such as a local server or a cloud server for managing the remote sensing object positioning system, etc., and the embodiments of the present invention do not make limitations. As Figure 2 shown, the update of the remote sensing intelligent application model may include the following operations:

[0105] To solve the above technical problems, a first aspect of an embodiment of the present invention discloses a method for updating a remote sensing intelligent application model, the method including:

[0106] S1, obtaining a remote sensing intelligent application basic model and a training data set;

[0107] S2, processing the remote sensing intelligent application basic model using the training data set to obtain a first application optimization model and a training data label set;

[0108] S3, processing the first application optimization model using the measured data set and the training data label set to obtain a remote sensing intelligent application optimization model.

[0109] It can be seen that by implementing the method for updating the remote sensing intelligent application model described in the embodiments of the present invention, the training data of the remote sensing intelligent application model can be updated online according to user requirements, and the intelligent model can be updated and iterated using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0110] In an optional embodiment, in the above step S2, the process of using the training data set to process the remote sensing intelligent application basic model to obtain a first application optimization model and a training data label set includes:

[0111] S21, obtain the training data set;

[0112] S22, divide the training data set according to a preset ratio to obtain a data training set and a data test set;

[0113] It should be noted that in this embodiment, the preset ratio is 4:1;

[0114] Set the training times value to 1;

[0115] S23, use the remote sensing intelligent application basic model to process the data training set to obtain first image object information;

[0116] S24, based on the model loss function, use the first image object information and the training data label set to update the parameters of the remote sensing intelligent application basic model;

[0117] S25, determine whether the training times value is greater than a preset training times threshold to obtain a training times judgment result;

[0118] When the training times judgment result is yes, execute S23;

[0119] When the training times judgment result is no, increase the training times value by 1 and execute S26;

[0120] S26, use the remote sensing intelligent application basic model to process the data test set to obtain second image object information;

[0121] S27, compare the second image object information with the data test label set to obtain prediction accuracy information;

[0122] S28, determine whether the prediction accuracy information is greater than a preset accuracy threshold to obtain an accuracy discrimination result;

[0123] When the accuracy discrimination result is no, execute S22;

[0124] If the accuracy discrimination result is yes, obtain a first application optimization model and execute S29;

[0125] S29, perform a fusion process on the data training label set and the data test label set to obtain a training data label set;

[0126] It should be noted that the first application optimization model refers to the first remote sensing intelligent application optimization model obtained after training the remote sensing intelligent application basic model.

[0127] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the training dataset is used to process the remote sensing intelligent application basic model to obtain the first application optimization model. On this basis, the training data of the remote sensing intelligent application model can be updated online according to user needs, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0128] In an optional embodiment, in the above step S24, the remote sensing intelligent application basic model includes a model input end, a feature extraction module, a feature fusion module, and a prediction output end;

[0129] The model input end, the feature extraction module, the feature fusion module, and the prediction output end are sequentially connected by data;

[0130] The expression of the model loss function is:

[0131] L = A + αB + βC;

[0132] Where A represents the classification loss; B represents the position loss; C represents the confidence loss; α represents the first weight coefficient; β represents the second weight coefficient;

[0133] It should be noted that in this embodiment, the first weight coefficient is set to 1;

[0134] It should be noted that in this embodiment, the second weight coefficient is set to 1.

[0135] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the training data of the remote sensing intelligent application model can be updated online according to user needs, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0136] In an optional embodiment, in the above step S28, the fusion processing of the training data training set label information and the training data test set label information to obtain the training data label set includes:

[0137] S281, traverse the training data training set to obtain all historical training data;

[0138] S282, perform parsing processing on any one of the historical training data to obtain the training data type information and the training data label information;

[0139] S283. Traverse the training data test set to obtain all historical test data;

[0140] S284. Parse and process any of the historical test data to obtain test data type information and test data label information;

[0141] S285. Remove duplicates from the training data type information, the training data label information, the test data type information, and the test data label information to obtain a training data label set;

[0142] It should be noted that the duplicate removal process means only retaining one for multiple data of the same type and label.

[0143] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the training data of the remote sensing intelligent application model can be updated online according to user needs, and the intelligent model can be updated and iterated using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0144] In an optional embodiment, in the above step S3, the process of using the measured data set and the training data label set to process the first application optimization model to obtain a remote sensing intelligent application optimization model includes:

[0145] S31. Obtain a target area vector mask set;

[0146] S32. Based on the target area vector mask set, process the measured data set to obtain a first positioning and classification data set and a second positioning and classification data set;

[0147] It should be noted that the first positioning and classification data set includes several first positioning and classification data;

[0148] It should be noted that the first classification and positioning data refers to the measured remote sensing image data with accurate positioning and classification;

[0149] It should be noted that the second positioning and classification data set includes several second positioning and classification data;

[0150] It should be noted that the second classification and positioning data refers to the measured remote sensing image data with positioning and classification errors, missed detections, and false alarms;

[0151] It should be noted that the measured data set is the combination of the first classification and positioning data set and the second classification and positioning data set;

[0152] S33. Use the target uncertainty acquisition model to process the training data set, the first positioning and classification data set, and the second positioning and classification data set to obtain a training data increment set;

[0153] S34. Based on the training data label set, annotate the training data increment set to obtain an increment data label set;

[0154] S35. Based on the increment data label set, use the training data increment set to process the first application optimization model to obtain a second application optimization model;

[0155] It should be noted that the second application optimization model refers to a second remote sensing intelligent application optimization model obtained after iterative processing of the first application optimization model;

[0156] S36. Perform a fusion process on the increment training label set and the training data label set to obtain a historical label set;

[0157] Increase the increment processing count by 1;

[0158] S37. Determine whether the increment processing count is greater than the iteration count to obtain an iteration determination result;

[0159] When the iteration determination result is no, update the training data increment set to the training data set, update the historical label set to the training data label set, and execute S32;

[0160] When the iteration determination result is yes, update the second remote sensing image object classification optimization model to the remote sensing intelligent application optimization model.

[0161] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, using the measured data set and the training data label set to update and iterate the first application optimization model, the continuous update and learning ability of the remote sensing data intelligent model is improved.

[0162] In an alternative embodiment, in the above step 31, the obtaining of the target area vector mask set includes:

[0163] S311. Obtain a target area remote sensing image set;

[0164] S312. Traverse the target area remote sensing image set to obtain all target area remote sensing images;

[0165] S313. Use the first application optimization model to process any one of the target area remote sensing images to obtain target positioning classification result information;

[0166] S314. Examine the target positioning classification result information to obtain a target area vector mask;

[0167] It should be noted that the target location classification result information represents the location classification result of the model object within the region of interest.

[0168] It should be noted that the inspection means using the known target information database to perform matching and verification processing on the target location classification result information, eliminating the false alarm target area data, and obtaining the target area vector mask.

[0169] S315. Use the region vector mask acquisition model to process the target area vector mask to obtain a set of target area vector masks.

[0170] The expression of the region vector mask acquisition model is:

[0171]

[0172] where A represents the set of target area vector masks; A i ; represents the i-th target area vector mask; i represents the index of the target area; U represents the union calculation.

[0173] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the training data of the remote sensing intelligent application model can be updated online according to user needs, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0174] In an alternative embodiment, in step 33 above, using the target uncertainty acquisition model to process the training data set, the first location classification data set, and the second location classification data set to obtain a training data increment set includes:

[0175] S331. Process the training data set and the first location classification data set to obtain a first incremental training sample subset.

[0176] The first incremental training sample subset includes a number of first incremental training samples.

[0177] S332. Based on the uncertainty calculation model, process the second location classification data set to obtain a second incremental training sample subset.

[0178] The second incremental training sample subset includes a number of second incremental training samples.

[0179] S333. Arrange and combine all the first incremental training samples and the second incremental training samples in sequence to obtain a training data increment set.

[0180] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the target uncertainty acquisition model is used to process the training data set, the first positioning and classification data set, and the second positioning and classification data set to obtain a training data increment set. On this basis, the training data of the remote sensing intelligent application model can be updated online according to user needs, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0181] In an optional embodiment, in the above step S331, the expression of the target uncertainty acquisition model is:

[0182] IA{img i} = 1 - AP50(TL(img i ),PR(img i ));。

[0183] Where, i represents the index of the remote sensing image; img i is the i-th remote sensing image; IA{img i} represents the target uncertainty of the i-th remote sensing image; TL(img i ) is the target object label in the i-th remote sensing image; PR(img i ) is the prediction result of the i-th remote sensing image; AP60() represents calculating the first threshold accuracy of the i-th remote sensing image.

[0184] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the target uncertainty can be obtained according to user needs, the training data of the remote sensing intelligent application model can be updated online based on the target uncertainty, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0185] In an optional embodiment, in the above step S331, the processing of the training data set and the first positioning and classification data set to obtain the first incremental training sample subset includes:

[0186] S3311, perform a union calculation on the training data set and the first positioning and classification data set to obtain a third positioning and classification data set;

[0187] S3312, traverse the third positioning and classification data set to obtain all the third positioning and classification data;

[0188] S3313, use the first application optimization model to process any of the third positioning and classification data to obtain the positioning target type information;

[0189] The positioning target type information includes target quantity information and target type information;

[0190] S3314. For all the third positioning classification data, perform sequential permutation and combination according to the amount of the target quantity information to obtain a fourth positioning classification data set;

[0191] S3315. Perform uniform sampling processing on the fourth positioning classification data set to obtain a first incremental training sample subset.

[0192] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the training data of the remote sensing intelligent application model can be updated online according to user requirements, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0193] In an optional embodiment, in the above step S35, using the training data increment set to process the first application optimization model to obtain a second application optimization model includes:

[0194] S351. Obtain the training data increment set;

[0195] The training data increment set includes several pieces of incremental training data; the incremental training data is training data or first classification positioning data or second classification positioning data;

[0196] S352. Traverse the training data increment set to obtain all the incremental training data;

[0197] S353. Perform parsing processing on any one of the incremental training data to obtain data type information;

[0198] It should be noted that the parsing processing means obtaining the data type information according to the value of the data type identification bit;

[0199] S354. Based on the data type information, perform matching processing on the training data label set to obtain incremental training label information;

[0200] It should be noted that the matching processing means obtaining the training data label in the training data label set that is consistent with the data type;

[0201] S355. Combine all the incremental training label information in sequence to obtain an incremental training label set

[0202] S356. Based on the incremental training label set, use the training data increment set to process the first application optimization model to obtain a second application optimization model.

[0203] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the model for obtaining the target uncertainty is used to process the training data set, the first positioning and classification data set, and the second positioning and classification data set to obtain a training data increment set. On this basis, the training data of the remote sensing intelligent application model can be updated online according to user requirements, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0204] In an optional embodiment, in the above step S353, the process of using the training data increment set to process the first application optimization model based on the incremental training label set to obtain a second application optimization model includes:

[0205] S3561, traverse the training data increment set to obtain all incremental training data;

[0206] S3562, perform parsing processing on any one of the incremental training data to obtain an incremental training sample type and incremental training sample data;

[0207] S3563, determine whether the incremental training sample type is equal to a specified type value to obtain an incremental type determination result;

[0208] It should be noted that in this embodiment, the specified type value is set to 1;

[0209] S3564, when the incremental type determination result is yes, based on the first loss function, input the incremental training sample data into the first application optimization model to obtain third image object classification information;

[0210] The expression of the first loss function is:

[0211] L ROI = LClass + αL locate + βL confi

[0212] where L ROI represents the model loss value of the first application optimization model; LClass represents the classification loss value of the first application optimization model; L locate represents the position loss value of the first application optimization model; L confi represents the confidence loss value of the first application optimization model; α represents the third weight coefficient; β represents the fourth weight coefficient;

[0213] It should be noted that in this embodiment, the third weight coefficient is set to 1;

[0214] It should be noted that in this embodiment, the fourth weight coefficient is set to 1.

[0215] It should be noted that when the incremental type determination result is yes, it indicates that the incremental training sample data is training data or first classification and positioning data; the incremental training sample data is a remote sensing image of the target area;

[0216] When the incremental type determination result is no, based on the second loss function, the incremental training sample data is input into the first application optimization model to obtain third image object classification information;

[0217] The expression of the second loss function is:

[0218] L non-ROI = L locate + yL confi

[0219] where L non-ROI represents the model loss value of the second application optimization model; L locate represents the position loss value of the first application optimization model; L confi represents the confidence loss value of the first application optimization model; Y represents the fifth weight coefficient;

[0220] It should be noted that in this embodiment, the fifth weight coefficient is set to 0.5;

[0221] It should be noted that when the incremental type determination result is no, it indicates that the incremental training sample data is second classification and positioning data; the incremental training sample data is a remote sensing image of a non-target area;

[0222] S3565. Calculate and process the third image object classification information and the incremental training label set to obtain classification accuracy information;

[0223] It should be noted that the calculation and processing means calculating using the average precision calculation method of the COCO dataset;

[0224] S3566. Determine whether the classification accuracy information is greater than a preset accuracy threshold to obtain a classification accuracy discrimination result;

[0225] When the classification accuracy discrimination result is no, execute S3562;

[0226] If the classification accuracy discrimination result is yes, obtain the second application optimization model;

[0227] It should be noted that in this embodiment, the accuracy threshold is set to 0.5;

[0228] It can be seen that by implementing the remote sensing intelligent application model update method described in the embodiments of the present invention, the training data of the remote sensing intelligent application model can be updated online according to user requirements, and the intelligent model can be updated iteratively using the continuously updated data, improving the continuous update and learning ability of the remote sensing data intelligent model.

[0229] Embodiment 2

[0230] Please refer to Figure 3 , Figure 3 which is a schematic structural diagram of a remote sensing intelligent application model update device disclosed in the embodiments of the present invention. Among them, Figure 3 the described device can be applied to a remote sensing object positioning system, such as a local server or a cloud server for remote sensing object positioning, etc., which is not limited in the embodiments of the present invention. As Figure 3 shown, the device may include:

[0231] An acquisition module 101, configured to acquire a remote sensing intelligent application basic model and a training data set;

[0232] A first processing module 102, configured to process the remote sensing intelligent application basic model using the training data set to obtain a first application optimization model and a training data label set;

[0233] A second processing module 103, configured to process the first application optimization model using a measured data set and the training data label set to obtain a remote sensing intelligent application optimization model.

[0234] Embodiment 3

[0235] Please refer to Figure 4 , Figure 4 which is a schematic structural diagram of another remote sensing intelligent application model update device disclosed in the embodiments of the present invention. Among them, Figure 4 the described device can be applied to a satellite observation system, such as a local server or a cloud server for satellite observation, etc., which is not limited in the embodiments of the present invention. As Figure 4 shown, the device may include:

[0236] A memory 201 storing executable program code;

[0237] A processor 202 coupled to the memory;

[0238] The processor 202 calls the executable program code stored in the memory 201 to execute the steps in the remote sensing intelligent application model update method described in Embodiment 1.

[0239] Embodiment 4

[0240] An embodiment of the present invention discloses a computer-readable storage medium that stores a computer program for electronic data exchange. The computer program causes a computer to execute the steps in the remote sensing intelligent application model updating method described in Embodiment 1.

[0241] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0242] Through the specific descriptions of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, and the storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disk memories, tape memories, or any other computer-readable medium capable of carrying or storing data.

[0243] Finally, it should be noted that what is disclosed by a remote sensing intelligent application model updating method and device disclosed in the embodiments of the present invention is only the preferred embodiments of the present invention, and is only used to illustrate the technical solutions of the present invention, rather than limiting it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for updating a remote sensing intelligent application model, characterized in that, The method includes: S1. Obtain a remote sensing intelligent application basic model and a training data set; S2. Process the remote sensing intelligent application basic model by using the training data set to obtain a first application optimization model and a training data label set; S3. Process the first application optimization model by using an actual measurement data set and the training data label set to obtain a remote sensing intelligent application optimization model.

2. The method for updating a remote sensing intelligent application model according to claim 1, wherein The process of using the training data set to process the remote sensing intelligent application basic model to obtain a first application optimization model and a training data label set includes: S21. Obtain the training data set; S22. Divide the training data set according to a preset ratio to obtain a data training set and a data test set; Set the number of training times value to 1; S23. Process the data training set by using the remote sensing intelligent application basic model to obtain first image object information; S24. Update the parameters of the remote sensing intelligent application basic model based on a model loss function by using the first image object information and the training data label set; S25. Judge whether the number of training times value is greater than a preset training times threshold to obtain a training times judgment result; When the training times judgment result is yes, execute S23; When the training times judgment result is no, increase the number of training times value by 1 and execute S26; S26. Process the data test set by using the remote sensing intelligent application basic model to obtain second image object information; S27. Compare the second image object information with the data test label set to obtain prediction accuracy information; S28. Judge whether the prediction accuracy information is greater than a preset accuracy threshold to obtain an accuracy discrimination result; When the accuracy discrimination result is no, execute S22; If the accuracy discrimination result is yes, obtain a first application optimization model and execute S29; S29. Perform a fusion process on the data training label set and the data test label set to obtain a training data label set.

3. The remote sensing intelligent application model updating method according to claim 1, characterized in that, The process of using an actual measurement data set and the training data label set to process the first application optimization model to obtain a remote sensing intelligent application optimization model includes: S31. Obtain a target area vector mask set; S32. Process the actual measurement data set based on the target area vector mask set to obtain a first positioning and classification data set and a second positioning and classification data set; The first positioning and classification data set includes a number of first positioning and classification data; The second positioning and classification data set includes a number of second positioning and classification data; S33. Process the training data set, the first positioning and classification data set and the second positioning and classification data set by using a target uncertainty acquisition model to obtain a training data increment set; S34. Label the training data increment set based on the training data label set to obtain an increment data label set; S35. Process the first application optimization model by using the training data increment set based on the increment data label set to obtain a second application optimization model; S36. The incremental training label set and the training data label set are fused to obtain a historical label set; Increase the number of incremental processing times by 1; S37. Determine whether the number of incremental processing times is greater than the number of iterations to obtain an iteration judgment result; When the iteration judgment result is no, the training data increment set is updated to the training data set, and the historical label set is updated to the training data label set, and S32 is executed; When the iteration judgment result is yes, the second remote sensing image object classification optimization model is updated to the remote sensing intelligent application optimization model.

4. The remote sensing intelligent application model updating method according to claim 3, characterized in that The expression of the target uncertainty acquisition model is: IA{img i} = 1 - AP50(TL(img i ), PR(img i )); where i represents the index of the remote sensing image; img i is the i-th remote sensing image; IA{img i} represents the target uncertainty of the i-th remote sensing image; TL(img i ) is the target object label in the i-th remote sensing image; PR(img i ) is the prediction result of the i-th remote sensing image; AP50() represents calculating the first threshold accuracy of the i-th remote sensing image.

5. The method for updating a remote sensing intelligent application model according to claim 3, wherein, The obtaining of the target area vector mask set includes: S311. Obtain a set of remote sensing images of the target area; S312. Traverse the set of remote sensing images of the target area to obtain all remote sensing images of the target area; S313. Use the first application optimization model to process any one of the remote sensing images of the target area to obtain target positioning classification result information; S314. Check the target positioning classification result information to obtain a target area vector mask; S315. Use the area vector mask acquisition model to process the target area vector mask to obtain a target area vector mask set; The expression of the area vector mask acquisition model is: Among them, A represents the set of target region vector masks; A i ; represents the i-th target region vector mask; i represents the index of the target region; U represents the union calculation.

6. The method for updating the remote sensing intelligent application model according to claim 3, wherein The use of the target uncertainty acquisition model to process the training data set, the first positioning classification data set and the second positioning classification data set to obtain a training data increment set includes: S331. Process the training data set and the first positioning classification data set to obtain a first incremental training sample subset; The first incremental training sample subset includes a number of first incremental training samples; S332. Based on the uncertainty calculation model, process the second positioning classification data set to obtain a second incremental training sample subset; The second incremental training sample subset includes a number of second incremental training samples; S333. All the first incremental training samples and the second incremental training samples are arranged and combined in sequence to obtain a training data increment set.

7. The method for updating the remote sensing intelligent application model according to claim 6, wherein The processing of the training data set and the first positioning classification data set to obtain a first incremental training sample subset includes: S3311. Perform a union calculation on the training data set and the first positioning classification data set to obtain a third positioning classification data set; S3312. Traverse the third positioning classification data set to obtain all third positioning classification data; S3313. Use the first application optimization model to process any one of the third positioning classification data to obtain positioning target type information; The positioning target type information includes target quantity information and target type information; S3314. Arrange and combine all the third positioning classification data in order according to the amount of the target quantity information to obtain a fourth positioning classification data set; S3315. Perform uniform sampling processing on the fourth positioning classification data set to obtain a first incremental training sample subset.

8. The method for updating the remote sensing intelligent application model according to claim 3, wherein Processing the first application optimization model by using the training data increment set based on the increment data tag set to obtain a second application optimization model, including: S351. Obtain the training data increment set; The training data increment set includes a number of increment training data; the increment training data is training data or first classification and positioning data or second classification and positioning data; S352. Traverse the training data increment set to obtain all the increment training data; S353. Perform parsing processing on any one of the increment training data to obtain data type information; S354. Based on the data type information, perform matching processing on the training data tag set to obtain increment training tag information; S355. Combine all the increment training tag information in sequence to obtain an increment training tag set; S356. Based on the increment training tag set, process the first application optimization model by using the training data increment set to obtain a second application optimization model.

9. A remote sensing intelligent application model updating device, characterized in that The device includes: An acquisition module, configured to acquire a remote sensing intelligent application basic model and a training data set; A first processing module, configured to process the remote sensing intelligent application basic model by using the training data set to obtain a first application optimization model and a training data tag set; A second processing module, configured to process the first application optimization model by using a measured data set and the training data tag set to obtain a remote sensing intelligent application optimization model.

10. A remote sensing intelligent application model updating device, characterized in that, The device includes: A memory storing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory and executes the remote sensing intelligent application model update method according to any one of claims 1-8.

Images