A remote sensing image processing method and device supporting an expert experience fusion mechanism

Abstract

The application discloses a remote sensing image processing method and device supporting an expert experience fusion mechanism, and the method comprises the following steps: acquiring remote sensing image information to be processed; performing preprocessing on the remote sensing image information to be processed to obtain target remote sensing image information to be processed; and performing analysis and processing on the target remote sensing image information to be processed to obtain target remote sensing image analysis result information.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a remote sensing image processing method and apparatus that supports an expert experience fusion mechanism. Background Technology

[0002] Identifying ground features and facilities quickly and accurately from satellite remote sensing imagery provides crucial technical support and regulatory basis for various scenarios, including urban planning and construction, land resource planning and management, land use monitoring, and engineering surveying. However, in practical applications, available and ideal high-resolution satellite remote sensing imagery sources are often limited or even nonexistent. Under these conditions, how to efficiently and comprehensively utilize satellite optical remote sensing imagery of different resolutions to identify useful information, achieving "use what's available, effective identification, and rapid response," has become a critical issue in improving current remote sensing identification technology. This requires identification algorithms to have stronger universality across different sensors and multiple resolutions of remote sensing imagery, which is a common challenge faced by current automatic remote sensing information identification algorithms. When using multi-resolution satellite optical remote sensing imagery to identify ground features and facilities, the differences in spectral response functions of different satellite sensors lead to variations in the spectral response of different data sources for the same ground feature, increasing the difficulty of identifying information across sensors and multiple resolutions of remote sensing imagery. Through their efforts, researchers have proposed a wealth of remote sensing image classification or target extraction methods from the perspectives of spatial spectral features, features at different levels, different analysis units, and different classifiers. In the industrial application of remote sensing, software such as ENVI, ERDAS, and eCognition have been developed, providing users with supervised classification, unsupervised classification, and object-oriented classification functions, which has greatly promoted the application of medium- and low-resolution remote sensing images. However, when these methods and technical systems are applied to high-resolution remote sensing images, due to the rich details of ground features and the diversity of the images, it is difficult to obtain satisfactory extraction results. In recent years, deep learning, through its powerful ability to self-learn sample features, has overcome the limitations of traditional manual feature recognition, opening up a new avenue for intelligent extraction of remote sensing information. In the field of remote sensing interpretation research, strategies such as strengthening multi-layer sensing, spatial feature enhancement, optimizing spatial scale, and super-pixel enhancement are continuously used to improve the self-learning ability of deep learning models for spatial details and semantic target scale features of high-resolution remote sensing images. This has led to the automation of remote sensing interpretation in tasks such as water body extraction, road extraction, and land cover classification, with model accuracy in test areas approaching 90% and even exceeding 95%. However, in practical applications, remote sensing intelligent interpretation technology has not yet reached the application level of ordinary image interpretation such as face and fingerprint recognition. A significant factor contributing to this situation is that existing remote sensing deep learning generally adopts supervised deep learning or transfer learning methods, primarily training models by constructing a large number of samples. This makes the model accuracy highly dependent on the diversity and quantity of samples.When the model is applied in practice, the main difficulties are as follows: First, the number of high-resolution satellite images available for a specific area and time is often scarce, making it impossible to construct a large sample using images of the area to be surveyed. Furthermore, satellite imaging conditions are more complex in some areas, making it difficult for the sample to cover all situations. Second, compared to general remote sensing monitoring, ground feature identification applications have higher requirements in terms of accuracy and feedback timeliness. Third, due to the scarcity of images, it is necessary to make the most of multi-resolution satellite remote sensing images from different sensors. Therefore, how to conduct fast and accurate remote sensing image information identification under conditions of limited samples, cross-sensor, and multi-resolution applications has become an urgent technical need for target identification in satellite remote sensing images. This requires innovation in remote sensing deep learning models. This paper provides a remote sensing image processing method and device that supports an expert experience fusion mechanism to improve the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification of ground features from multi-resolution satellite optical remote sensing images. This addresses the challenge of achieving rapid cross-sensor, multi-resolution remote sensing identification response and application under limited data conditions. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a remote sensing image processing method and device that supports the fusion of expert experience, which is conducive to improving the accuracy of remote sensing image analysis through deep learning technology, thereby realizing the rapid and accurate identification and application of ground objects and facilities from multi-resolution satellite optical remote sensing images, so as to solve the problem of achieving rapid remote sensing identification response and application across sensors and multiple resolutions under the condition of limited data.

[0004] To address the aforementioned technical problems, a first aspect of the present invention discloses a remote sensing image processing method, the method comprising:

[0005] Acquire remote sensing image information to be processed;

[0006] The remote sensing image information to be processed is preprocessed to obtain the target processed remote sensing image information;

[0007] The remote sensing image information of the target is analyzed and processed to obtain the target remote sensing image analysis result information.

[0008] The process of analyzing and processing the remote sensing image information of the target to obtain the target remote sensing image analysis result information includes:

[0009] Based on the remote sensing image processing model and the target processed remote sensing image information, the first remote sensing image analysis result information is determined.

[0010] Based on the analysis results of the first remote sensing image, the analysis results of the target remote sensing image are determined.

[0011] The remote sensing image processing model includes a first feature extraction module, a second feature extraction module, a first feature fusion module, a second feature fusion module, a third feature fusion module, and an image segmentation module; wherein,

[0012] The input terminal of the first feature extraction module is configured to receive a first model input of the remote sensing image processing model, and the output terminal of the first feature extraction module is connected to the input terminal of the first feature fusion module; the input terminal of the first feature fusion module is also configured to receive a first model input of the remote sensing image processing model; the output terminal of the first feature fusion module is connected to the input terminal of the third feature fusion module; the input terminal of the second feature extraction module is configured to receive a second model input of the remote sensing image processing model; the output terminal of the second feature extraction module is connected to the input terminal of the second feature fusion module; the output terminal of the second feature fusion module is connected to the input terminal of the third feature fusion module; the input terminal of the third feature fusion module is also configured to receive a first model input of the remote sensing image processing model; the output terminal of the third feature fusion module is connected to the input terminal of the image segmentation module; the output terminal of the image segmentation module is configured to output the model output of the remote sensing image processing model.

[0013] The first feature fusion module includes a first convolutional unit, a second convolutional unit, a first normalization unit, a second normalization unit, a first activation unit, a second activation unit, and a first fusion unit; wherein,

[0014] The input of the first convolutional unit is connected to the output of the first feature extraction module; the first convolutional unit, the first normalization unit, the first activation unit, and the first fusion unit are connected in sequence; the input of the first fusion unit is also configured to receive the first model input of the remote sensing image processing model; the first fusion unit, the second convolutional unit, the second normalization unit, and the second activation unit are connected in sequence; the output of the second activation unit is connected to the input of the third feature fusion module.

[0015] The third feature fusion module includes a second fusion unit, a third fusion unit, a fourth fusion unit, a first pooling unit, a third convolution unit, a third normalization unit, and a third activation unit; wherein,

[0016] The input of the third fusion unit is connected to the output of the first feature fusion module; the input of the third fusion unit is also configured to receive the first model input of the remote sensing image processing model; the output of the third fusion unit is connected to the input of the first pooling unit; the output of the first pooling unit is connected to the input of the second fusion unit; the input of the second fusion unit is also connected to the output of the second feature fusion module; the output of the second fusion unit is connected to the input of the fourth fusion unit; the input of the fourth fusion unit is also connected to the output of the second feature fusion module; the fourth fusion unit, the third convolution unit, the third normalization unit, and the third activation unit are connected sequentially; the output of the third activation unit is connected to the input of the image segmentation module.

[0017] The first feature extraction module includes a fourth convolutional unit, a fifth convolutional unit, a sixth convolutional unit, a seventh convolutional unit, a fourth normalization unit, a fifth normalization unit, a sixth normalization unit, a seventh normalization unit, an eighth normalization unit, a ninth normalization unit, a tenth normalization unit, an eleventh normalization unit, a twelfth normalization unit, a thirteenth normalization unit, a fourth activation unit, a fifth activation unit, a sixth activation unit, a seventh activation unit, an eighth activation unit, a ninth activation unit, a tenth activation unit, an eleventh activation unit, a first connection unit, a second pooling unit, a third pooling unit, and a fourth pooling unit; wherein,

[0018] The input of the fourth convolutional unit is configured as the input of the first feature extraction module; the fourth convolutional unit, the fourth normalization unit, the fourth activation unit, the fifth convolutional unit, the fifth normalization unit, the fifth activation unit, the sixth normalization unit, the sixth activation unit, the seventh normalization unit, the second pooling unit, the eighth normalization unit, the seventh activation unit, the sixth convolutional unit, the ninth normalization unit, the eighth activation unit, the tenth normalization unit, the ninth activation unit, the eleventh normalization unit, the third pooling unit, the twelfth normalization unit, the tenth activation unit, the seventh convolutional unit, the thirteenth normalization unit, the eleventh activation unit, the fourth pooling unit, and the first connection unit are connected sequentially; the output of the first connection unit is configured as the output of the first feature extraction module;

[0019] The image segmentation module includes a fourth fusion unit, an eighth convolution unit, a fourteenth normalization unit, a twelfth activation unit, a thirteenth activation unit, and a second connection unit; wherein,

[0020] The input of the fourth fusion unit is connected to the output of the third feature fusion module and the output of the thirteenth activation unit, respectively; the fourth fusion unit, the eighth convolution unit, the fourteenth normalization unit, the twelfth activation unit, the second connection unit, and the fourteenth activation unit are connected in sequence; the output of the fourteenth activation unit is configured to output the model output of the remote sensing image processing model.

[0021] The step of determining the target remote sensing image analysis result information based on the first remote sensing image analysis result information includes:

[0022] Based on the remote sensing image processing model, the target processing remote sensing image information collected at each time moment is processed to obtain the corresponding first remote sensing image analysis results.

[0023] The analysis results of the first remote sensing images at all times are fused and calculated to obtain the target remote sensing image analysis result information;

[0024] The expression for the fusion calculation is:

[0025]

[0026] Among them, y i The result of the analysis of the first remote sensing image at time i is shown. y0 and y0 represent the maximum, minimum, and average values ​​of the first remote sensing image analysis results at all times, respectively, while y represents the target remote sensing image analysis result information.

[0027] The fusion calculation expression utilizes the maximum, minimum, and average values ​​of the first remote sensing image analysis results at all times to construct weighting coefficients. This allows for adaptive adjustment of the weights of results at each time point based on the data distribution. For values ​​close to the average and located in the middle of the data distribution, the weights are relatively large, highlighting stable and reliable analysis results. Conversely, outliers deviating significantly from the average have their interference with the final results reduced due to lower weights, effectively filtering noise and abnormal fluctuations, thus improving the accuracy and stability of the analysis results. When processing remote sensing images acquired at different times, this formula fully considers the characteristics of time-series data. By fusing results from multiple times, it avoids analytical biases caused by weather changes, sensor errors, and other factors in single-time images. Integrating information from multiple times provides a more comprehensive and objective reflection of the true state and changing trends of ground features, making it particularly suitable for dynamic monitoring scenarios such as vegetation growth changes and urban expansion monitoring. This makes the target remote sensing image analysis results more closely aligned with reality, enhancing the reliability of remote sensing image time-series analysis. Compared to traditional simple averaging or weighted averaging methods, this fusion calculation expression constructs nonlinear weights based on the statistical characteristics of the data, enabling the discovery of potential correlations and patterns between data points and mathematically optimizing the output of the remote sensing image processing model. By fusing the analysis results from different time points in a more scientific manner, it provides more accurate data support for subsequent decision-making based on target remote sensing image analysis results (such as resource assessment and disaster early warning), enhancing the practicality and application value of the entire remote sensing image processing method.

[0028] A second aspect of this invention discloses a remote sensing image processing apparatus, the apparatus comprising:

[0029] The acquisition module is used to acquire remote sensing image information to be processed;

[0030] The first processing module is used to preprocess the remote sensing image information to be processed to obtain the target processed remote sensing image information.

[0031] The second processing module is used to analyze and process the target remote sensing image information to obtain target remote sensing image analysis result information.

[0032] A third aspect of the present invention discloses another remote sensing image processing apparatus, the apparatus comprising:

[0033] Memory containing executable program code;

[0034] A processor coupled to the memory;

[0035] The processor calls the executable program code stored in the memory to execute some or all of the steps in the remote sensing image processing method disclosed in the first aspect of the present invention.

[0036] The fourth aspect of the present invention discloses a computer-readable storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the remote sensing image processing method disclosed in the first aspect of the present invention. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a scene diagram of the remote sensing image processing system provided in an embodiment of the present invention;

[0039] Figure 2 This is a flowchart illustrating a remote sensing image processing method disclosed in an embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of the structure of a remote sensing image processing device disclosed in an embodiment of the present invention;

[0041] Figure 4 This is a schematic diagram of another remote sensing image processing device disclosed in an embodiment of the present invention;

[0042] Figure 5 This is a schematic diagram of the structure of a remote sensing image processing model disclosed in an embodiment of the present invention;

[0043] Figure 6 This is a schematic diagram of the structure of a first feature extraction module disclosed in an embodiment of the present invention;

[0044] Figure 7 This is a schematic diagram of the structure of an image segmentation module disclosed in an embodiment of the present invention. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0047] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0048] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being 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. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0049] It should be noted that since the method in this application embodiment is executed in a computer device, the processing objects of each computer device exist in the form of data or information, such as time, which is essentially time information. It is understood that if size, quantity, position, etc. are mentioned in subsequent embodiments, they are all corresponding data that exist so that the computer device can process them. Specific details will not be elaborated here.

[0050] It should be noted that the artificial intelligence-related technologies that may be involved in this application will be briefly described. Artificial intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal 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 kind of intelligent machine that can react in a way similar to human intelligence. Artificial intelligence is the study of the design principles and implementation methods of various intelligent machines, enabling machines to have the functions of perception, reasoning, and decision-making.

[0051] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0052] Computer vision (CV) is the science that studies how to enable machines to "see." More specifically, it refers to machine vision, which uses cameras and computers to replace human eyes in recognizing and measuring targets, and then performs image processing to create images more suitable for human observation or transmission to instruments. As a scientific discipline, computer vision studies related theories and technologies, attempting to build artificial intelligence systems capable of extracting information from images or multidimensional data. Computer vision technologies typically include image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, 3D object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping (SLAM), and common biometric recognition technologies such as facial recognition and fingerprint recognition.

[0053] Monomodal information refers to data of only one type, such as text, images, audio, video, or electromagnetic signals. Multimodal information refers to data that includes at least two types of monomodal information. Furthermore, multimodal information is suitable for complex tasks that require the integration of multiple information sources, such as sentiment analysis, robot interaction, and autonomous driving. By integrating information from multiple modalities, higher performance and accuracy can usually be achieved in these tasks.

[0054] Large models refer to artificial neural network models with a very large number of parameters. In the field of artificial intelligence, large models typically refer to models with hundreds of millions to trillions of parameters. These models usually need to be trained on large-scale datasets and require a significant amount of computing resources for optimization and tuning. Large models are commonly used to solve complex tasks such as natural language processing, computer vision, and speech recognition. Generative AI is a type of AI that can create new content and ideas, including dialogues, stories, images, videos, and music. In this embodiment, the large model can be a language model of the scale of ChatGPT, BERT, XLNet, Zhipu model, Claude, Moonshot AI model, ChatGLM model, Tongwen Qianyi model, MiniMax model, Xinghuo model, Llama model, 360GPT model, Qwen model, Baichuan model, Yunque model, vivoLM model, and Wenxin Yiyan, etc., and this embodiment does not limit the scope of the large model.

[0055] This application provides a remote sensing image processing method, apparatus, computer equipment, and computer-readable storage medium, which will be described in detail below.

[0056] Please see Figure 1 , Figure 1 This is a schematic diagram of a remote sensing image processing system provided in an embodiment of this application. The remote sensing image processing system may include a computer device 100, which integrates a remote sensing image processing unit, such as... Figure 1 Computer equipment in the country.

[0057] In this embodiment of the application, the computer device 100 is mainly used to acquire remote sensing image information to be processed;

[0058] The remote sensing image information to be processed is preprocessed to obtain the target processed remote sensing image information;

[0059] The remote sensing image information of the target is analyzed and processed to obtain the target remote sensing image analysis result information.

[0060] It can improve the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0061] In this embodiment, the computer device 100 can be a standalone server, a server network, or a server cluster. For example, the computer device 100 described in this embodiment 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. The cloud server is composed of a large number of computers or network servers based on cloud computing.

[0062] It is understood that the computer device 100 used in the embodiments of this application can be a device that includes both receiving and transmitting hardware, that is, a device having receiving and transmitting hardware capable of performing bidirectional communication on a bidirectional communication link. Such a device may include: cellular or other communication devices having a single-line display, 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 may also be one of a mobile phone, tablet computer, laptop computer, etc.

[0063] Those skilled in the art will understand that Figure 1 The application environment shown is merely one application scenario of the solution in this application and does not constitute a limitation on the application scenario of the solution in this application. Other application environments may include those that are more specific to this application. Figure 1 The number of computer devices shown is more or less, for example Figure 1 Only one computer device is shown in the diagram. It is understood that the remote sensing image processing system may also include one or more other services, which are not limited here.

[0064] In addition, such as Figure 1 As shown, the remote sensing image processing system may also include a memory 200 for storing data, such as image data and location information.

[0065] It should be noted that, Figure 1 The schematic diagram of the remote sensing image processing system shown is merely an example. The remote sensing image processing system and scenario described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of remote sensing image processing systems and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0066] This invention discloses a remote sensing image processing method and apparatus that supports an expert experience fusion mechanism. This facilitates the improvement of the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This addresses the challenge of achieving rapid, multi-resolution remote sensing identification and application across sensors under limited data conditions. Detailed descriptions follow.

[0067] Example 1

[0068] Please see Figure 2 , Figure 2 This is a schematic flowchart of a remote sensing image processing method disclosed in an embodiment of the present invention. Wherein, Figure 2 The described remote sensing image processing method is applied in a management system, such as a local server or cloud server for management, and the embodiments of the present invention are not limited thereto. Figure 2 As shown, the remote sensing image processing method may include the following operations:

[0069] 101. Obtain remote sensing image information to be processed.

[0070] 102. Preprocess the remote sensing image information to be processed to obtain the target processed remote sensing image information.

[0071] 103. Analyze and process the remote sensing image information of the target to obtain the target remote sensing image analysis results.

[0072] It should be noted that the remote sensing image information to be processed mentioned above includes remote sensing image information that integrates expert experience and remote sensing image information taken by satellites, and the embodiments of the present invention do not limit it.

[0073] Furthermore, the target processing of remote sensing image information includes first remote sensing image information obtained by processing remote sensing image information that integrates expert experience and second remote sensing image information obtained by processing remote sensing image information captured by satellite. This embodiment of the invention does not limit the scope of the information.

[0074] Furthermore, the first model input is the first remote sensing image information, and the second model input represents the second remote sensing image information.

[0075] Furthermore, the above-mentioned preprocessing of the remote sensing image information to be processed includes first performing mean filtering on the image, then smoothing the image by using Gaussian filtering to remove noise in the image, thereby improving image quality, and finally adjusting the image size. This embodiment of the invention does not limit the scope of the invention.

[0076] It should be noted that the method in this application, based on obtaining a basic feature segmentation model and a key facility and target detection model that meet the longitude requirements, solidifies the model algorithm in a Python + TensorFlow environment. In the VS2015 integrated development environment, using C++, the QGIS secondary development package, TensorFlow library, GDAL library, and OpenCV library are imported, and the trained remote sensing image processing model is called to achieve the identification of key facilities based on multi-resolution satellite optical remote sensing imagery. This invention does not limit the scope of the embodiments described.

[0077] It is evident that implementing the remote sensing image processing method described in the embodiments of the present invention is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0078] In an optional embodiment, the above-described analysis and processing of the target remote sensing image information to obtain target remote sensing image analysis result information includes:

[0079] Based on the remote sensing image processing model and the target-processed remote sensing image information, the analysis results of the first remote sensing image are determined.

[0080] Based on the analysis results of the first remote sensing image, the analysis results of the target remote sensing image were determined.

[0081] It should be noted that the above-mentioned determination of the target remote sensing image analysis result information based on the first remote sensing image analysis result information is achieved by using a target detection model to identify the target object from the first remote sensing image analysis result information, thereby obtaining the target remote sensing image analysis result information with the identified target object. This embodiment of the invention does not limit this. Furthermore, the above-mentioned target detection model can be built based on the YOLO series models or based on a large model, and this embodiment of the invention does not limit this.

[0082] It is evident that implementing the remote sensing image processing method described in the embodiments of the present invention is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0083] In another alternative embodiment, such as Figure 5 As shown, the remote sensing image processing model includes a first feature extraction module, a second feature extraction module, a first feature fusion module, a second feature fusion module, a third feature fusion module, and an image segmentation module; wherein,

[0084] The input of the first feature extraction module is configured to receive the first model input of the remote sensing image processing model, and the output of the first feature extraction module is connected to the input of the first feature fusion module; the input of the first feature fusion module is also configured to receive the first model input of the remote sensing image processing model; the output of the first feature fusion module is connected to the input of the third feature fusion module; the input of the second feature extraction module is configured to receive the second model input of the remote sensing image processing model; the output of the second feature extraction module is connected to the input of the second feature fusion module; the output of the second feature fusion module is connected to the input of the third feature fusion module; the input of the third feature fusion module is also configured to receive the first model input of the remote sensing image processing model; the output of the third feature fusion module is connected to the input of the image segmentation module; the output of the image segmentation module is configured to output the model output of the remote sensing image processing model.

[0085] It should be noted that the aforementioned remote sensing image processing model supports the fusion of prior knowledge from external experts. This is primarily used for the fusion of external prior knowledge, performing feature-intensive comparisons between query images and supporting images, and learning and simulating the ability of humans to analyze similar and dissimilar features when recognizing objects. This embodiment of the invention does not limit this. Furthermore, the remote sensing image processing model uses a cross-validation training method, where the training and validation sets are input into the model simultaneously. Each training iteration randomly selects a batch size from the validation set to calculate the loss and accuracy, optimizing the model's training. A binary classification cross-entropy loss function is used, and the stochastic gradient descent (SGD) optimization algorithm is employed to optimize the objective function. Regarding image input size and batch size, since model training places high demands on the computer's GPU memory, the method of this invention uses images of size 512*512 as network input, sets the batch size to 4, sets the number of rounds to 50, the number of iterations per round to 4000, and sets the initial learning rate to 1e-4. In addition, in order to better train the model, the learning rate will be automatically adjusted as the number of training rounds increases, that is, the learning rate will decrease by a factor of 10 every 20 rounds. Combined with the optimizer, the convergence of the network can be accelerated. This embodiment of the invention does not limit the scope of the invention.

[0086] Furthermore, the training samples used to train the remote sensing image processing model can be obtained in the following way:

[0087] Sample library collection principles:

[0088] Due to the influence of various factors such as imaging conditions, sensor characteristics, weather conditions, climate conditions, solar altitude, background environment, and high light reflectivity, the target representation of remote sensing images changes. Furthermore, the different satellite sensors involved in multi-resolution imagery significantly increase the difficulty of identifying important facilities and targets within multi-resolution satellite remote sensing images. Therefore, to improve model accuracy, when utilizing existing sample libraries and constructing a target identification sample library using self-collected data, it is essential to comprehensively cover all types of elements and complex conditions.

[0089] Training library construction principles: Sample selection should be representative, and the samples for each facility type and facility element should be diverse, covering different situations such as different resolutions, different sensors, different imaging conditions, and different backgrounds; the number of samples for each target type and covering various typical situations should reach a certain number; the number of samples covering various typical situations should be distributed as evenly as possible.

[0090] Sample library acquisition methods:

[0091] This invention produces conventional samples according to the aforementioned sample collection methods and principles. To further transform this conventional sample dataset into a sample set that meets the requirements for training small-sample models, this invention proposes a remote sensing sample construction method that incorporates expert interpretation experience expression and transmission.

[0092] Sample preparation process:

[0093] Unlike the creation of conventional samples, the creation of multi-resolution samples that express and transmit interpretation experience involves first analyzing expert interpretation experience and patterns based on a conventional sample database. Then, it selects methods such as saliency feature maps and exponential feature maps to represent these patterns, forming a multi-channel expert interpretation experience map. Finally, this map, along with the original remote sensing imagery and the labels from the remote sensing human interpretation, serves as training samples to provide prior knowledge for the deep learning model.

[0094] Experts interpret the expression of empirical rules:

[0095] When performing visual interpretation of remote sensing images, interpretation professionals can accurately identify ground features and facilities by using shape, texture, color characteristics, and different spectral combinations. To transform this expert visual interpretation experience into prior knowledge, express it, and pass it to a deep learning model to guide its learning and convergence, this invention primarily utilizes existing texture feature extraction methods, geometric morphology feature extraction methods, color saliency-based extraction methods, and exponential factor extraction methods to extract the perceptual experience of texture, geometry, color, and spectrum features from expert interpretation. These features are then normalized and synthesized into a multi-channel expert interpretation experience map.

[0096] Furthermore, the aforementioned remote sensing image processing model can effectively integrate expert knowledge into the model training and inference process to improve the model's ability to understand and analyze remote sensing images. In this way, the model can better capture subtle changes in remote sensing images, improve the accuracy of change detection, and adapt to different application scenarios, such as urban planning and environmental monitoring. This embodiment of the invention does not limit these applications.

[0097] It should be noted that the model architecture of the first feature extraction module and the second feature extraction module described above is consistent, and this embodiment of the invention does not impose any limitations on them. Furthermore, the first feature extraction module and the second feature extraction module described above are twin structures. Further, the twin structure feature extraction module is mainly used to extract common features of intermediate-level features that do not have category discrimination function, and for learning and comparing objects. This embodiment of the invention does not impose any limitations on it. Furthermore, the first feature extraction module and the second feature extraction module of the two twin structures share the exact same system architecture and weights. The first feature extraction module and the second feature extraction module in the twin network have the same parameters and weights. During network training, the parameters are updated jointly on both the first feature extraction module and the second feature extraction module. Twin neural networks help discover similarities and correlations between different structures. The shared weights of the first feature extraction module and the second feature extraction module in the twin network mean that training requires fewer parameters, that is, less data and is less prone to overfitting. This embodiment of the invention does not impose any limitations on it.

[0098] It should be noted that the model architecture of the first feature fusion module and the second feature fusion module described above is consistent, and this embodiment of the invention does not limit the scope of the module. Furthermore, the first feature fusion module and the second feature fusion module are mainly used to fuse the prior knowledge extracted by experts with the intermediate-level features extracted by the Siamese network, providing rich shared knowledge and patterns for feature comparison; this embodiment of the invention does not limit the scope of the module.

[0099] It should be noted that the aforementioned third feature fusion module is mainly used to learn and compare similar features of the target category in the supporting image and the query image, and this embodiment of the invention does not limit it.

[0100] It should be noted that, due to the influence of external factors such as imaging conditions, there are certain differences in the spectrum and appearance of the same category in remote sensing images. Dense comparison can only match a part of the object, which is insufficient to accurately segment or identify the object in the image. However, the initial segmentation or identification result can provide important coarse spatial location information of the target. Therefore, the image segmentation module will continuously use iterative optimization to achieve segmentation. Since the residual structure in the residual network has strong feature reuse and information transmission capabilities, the recursive iterative prediction segmentation submodule mainly uses the residual structure for implementation. This embodiment of the invention does not limit this.

[0101] It is evident that implementing the remote sensing image processing method described in the embodiments of the present invention is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0102] In yet another alternative embodiment, such as Figure 5 As shown, the first feature fusion module includes a first convolutional unit, a second convolutional unit, a first normalization unit, a second normalization unit, a first activation unit, a second activation unit, and a first fusion unit; wherein,

[0103] The input of the first convolutional unit is connected to the output of the first feature extraction module; the first convolutional unit, the first normalization unit, the first activation unit, and the first fusion unit are connected in sequence; the input of the first fusion unit is also configured to receive the first model input of the remote sensing image processing model; the first fusion unit, the second convolutional unit, the second normalization unit, and the second activation unit are connected in sequence; the output of the second activation unit is connected to the input of the third feature fusion module.

[0104] It should be noted that the convolution kernels of the first and second convolution units can be one of 1×1, 3×1, 3×3, 5×1, 5×3, 5×5, 7×5 and 7×7, with a stride of 1. This embodiment of the invention does not limit the specific kernels.

[0105] It should be noted that the first normalization unit and the second normalization unit mentioned above are constructed based on the batch normalization layer, and this embodiment of the present invention does not limit them.

[0106] It should be noted that the first activation unit and the second activation unit mentioned above are constructed based on the ReLU activation function, and this embodiment of the invention does not limit them.

[0107] It should be noted that the first fusion unit described above is constructed based on splicing operations, and this embodiment of the present invention does not limit it.

[0108] It is evident that implementing the remote sensing image processing method described in the embodiments of the present invention is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0109] In yet another alternative embodiment, such as Figure 5As shown, the third feature fusion module includes a second fusion unit, a third fusion unit, a fourth fusion unit, a first pooling unit, a third convolution unit, a third normalization unit, and a third activation unit; wherein,

[0110] The input of the third fusion unit is connected to the output of the first feature fusion module; the input of the third fusion unit is also configured to receive the first model input of the remote sensing image processing model; the output of the third fusion unit is connected to the input of the first pooling unit; the output of the first pooling unit is connected to the input of the second fusion unit; the input of the second fusion unit is also connected to the output of the second feature fusion module; the output of the second fusion unit is connected to the input of the fourth fusion unit; the input of the fourth fusion unit is also connected to the output of the second feature fusion module; the fourth fusion unit, the third convolution unit, the third normalization unit, and the third activation unit are connected sequentially; the output of the third activation unit is connected to the input of the image segmentation module.

[0111] It should be noted that the second, third, and fourth fusion units mentioned above are constructed based on splicing operations, and this embodiment of the invention does not impose any limitations on them.

[0112] It should be noted that the first pooling unit mentioned above is constructed based on the max pooling layer, and this embodiment of the invention does not limit it.

[0113] It should be noted that the convolution kernel of the third convolution unit mentioned above can be one of 1×1, 3×1, 3×3, 5×1, 5×3, 5×5, 7×5 and 7×7, with a stride of 1. This embodiment of the invention does not limit the type of kernel.

[0114] It should be noted that the third normalization unit mentioned above is constructed based on the batch normalization layer, and this embodiment of the present invention does not limit it.

[0115] It should be noted that the third activation unit mentioned above is constructed based on the ReLU activation function, and this embodiment of the invention does not limit it.

[0116] It is evident that implementing the remote sensing image processing method described in the embodiments of the present invention is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0117] In an optional embodiment, such as Figure 6As shown, the aforementioned first feature extraction module includes a fourth convolutional unit, a fifth convolutional unit, a sixth convolutional unit, a seventh convolutional unit, a fourth normalization unit, a fifth normalization unit, a sixth normalization unit, a seventh normalization unit, an eighth normalization unit, a ninth normalization unit, a tenth normalization unit, an eleventh normalization unit, a twelfth normalization unit, a thirteenth normalization unit, a fourth activation unit, a fifth activation unit, a sixth activation unit, a seventh activation unit, an eighth activation unit, a ninth activation unit, a tenth activation unit, an eleventh activation unit, a first connection unit, a second pooling unit, a third pooling unit, and a fourth pooling unit; wherein,

[0118] The input of the fourth convolutional unit is configured as the input of the first feature extraction module; the fourth convolutional unit, the fourth normalization unit, the fourth activation unit, the fifth convolutional unit, the fifth normalization unit, the fifth activation unit, the sixth normalization unit, the sixth activation unit, the seventh normalization unit, the second pooling unit, the eighth normalization unit, the seventh activation unit, the sixth convolutional unit, the ninth normalization unit, the eighth activation unit, the tenth normalization unit, the ninth activation unit, the eleventh normalization unit, the third pooling unit, the twelfth normalization unit, the tenth activation unit, the seventh convolutional unit, the thirteenth normalization unit, the eleventh activation unit, the fourth pooling unit, and the first connection unit are connected sequentially; the output of the first connection unit is configured as the output of the first feature extraction module.

[0119] It should be noted that the convolution kernels of the fourth, fifth, sixth, and seventh convolution units mentioned above can be one of 1×1, 3×1, 3×3, 5×1, 5×3, 5×5, 7×5, and 7×7, with a stride of 1. This embodiment of the invention does not limit the specific kernels.

[0120] It should be noted that the above-mentioned fourth, fifth, sixth, seventh, eighth, ninth, tenth, eleventh, twelfth, and thirteenth normalization units are constructed based on the batch normalization layer, and this embodiment of the invention does not limit them.

[0121] It should be noted that the above-mentioned fourth activation unit, fifth activation unit, sixth activation unit, seventh activation unit, eighth activation unit, ninth activation unit, tenth activation unit, and eleventh activation unit are constructed based on the ReLU activation function, and the embodiments of the present invention are not limited thereto.

[0122] It should be noted that the first connection unit mentioned above is constructed based on a fully connected layer, and this embodiment of the present invention does not limit it.

[0123] It should be noted that the second, third, and fourth pooling units described above are constructed based on the max pooling layer, and this embodiment of the invention does not impose any limitations on them.

[0124] It should be noted that the module consisting of (normalization unit-activation unit-convolution unit-normalization unit-activation unit) is a basic feature extraction module, while the unit module consisting of (normalization unit-activation unit-normalization unit-pooling unit) is a connecting unit connecting the basic feature extraction module to reduce the size and number of feature maps. This embodiment of the invention does not limit this approach. Furthermore, this multi-level feature extraction method gives the model strong feature accumulation and propagation capabilities and multi-level feature integration capabilities. This embodiment of the invention does not limit this approach. Furthermore, to maximize the transmission of feature information between two neural layers, the first feature extraction module uses a deep convolutional neural network connection method that connects all layers together. This allows each layer during training to access the loss gradients formed at the end of the model and the beginning of the structure. Utilizing a dense connection pattern effectively improves the information flow between layers, making the model easier to train. This embodiment of the invention does not limit this approach.

[0125] It is evident that implementing the remote sensing image processing method described in the embodiments of the present invention is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0126] In another alternative embodiment, such as Figure 7 As shown, the image segmentation module includes a fourth fusion unit, an eighth convolution unit, a fourteenth normalization unit, a twelfth activation unit, a thirteenth activation unit, and a second connection unit; wherein,

[0127] The input of the fourth fusion unit is connected to the output of the third feature fusion module and the output of the thirteenth activation unit, respectively; the fourth fusion unit, the eighth convolution unit, the fourteenth normalization unit, the twelfth activation unit, the second connection unit and the fourteenth activation unit are connected in sequence; the output of the fourteenth activation unit is configured to output the model output of the remote sensing image processing model.

[0128] It should be noted that the fourth fusion unit described above is constructed based on splicing operations, and this embodiment of the invention does not limit it.

[0129] It should be noted that the convolution kernel of the eighth convolution unit can be one of 1×1, 3×1, 3×3, 5×1, 5×3, 5×5, 7×5 and 7×7, with a stride of 1. This embodiment of the invention does not limit the type of kernel.

[0130] It should be noted that the above-mentioned fourteenth normalization unit is constructed based on the batch normalization layer, and this embodiment of the invention does not limit it.

[0131] It should be noted that the twelfth and thirteenth activation units mentioned above are constructed based on the ReLU activation function, and this embodiment of the invention does not limit them.

[0132] It should be noted that the second connection unit mentioned above is constructed based on a fully connected layer, and this embodiment of the present invention does not limit it.

[0133] It is evident that implementing the remote sensing image processing method described in the embodiments of the present invention is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0134] Example 2

[0135] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a remote sensing image processing device disclosed in an embodiment of the present invention. Figure 3 The described apparatus can be applied in management systems, such as local servers or cloud servers for management, and the embodiments of the present invention are not limited thereto. Figure 3 As shown, the device may include:

[0136] The acquisition module 201 is used to acquire remote sensing image information to be processed;

[0137] The first processing module 202 is used to preprocess the remote sensing image information to be processed to obtain the target remote sensing image information;

[0138] The second processing module 203 is used to analyze and process the remote sensing image information of the target to obtain the target remote sensing image analysis result information.

[0139] It is evident that implementation Figure 3 The described remote sensing image processing device is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0140] In another alternative embodiment, such as Figure 3 As shown, the remote sensing image information of the target is analyzed and processed to obtain the target remote sensing image analysis results, including:

[0141] Based on the remote sensing image processing model and the target-processed remote sensing image information, the analysis results of the first remote sensing image are determined.

[0142] Based on the analysis results of the first remote sensing image, the analysis results of the target remote sensing image were determined.

[0143] It is evident that implementation Figure 3 The described remote sensing image processing device is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0144] In yet another alternative embodiment, such as Figure 3 As shown, the remote sensing image processing model includes a first feature extraction module, a second feature extraction module, a first feature fusion module, a second feature fusion module, a third feature fusion module, and an image segmentation module; wherein,

[0145] The input of the first feature extraction module is configured to receive the first model input of the remote sensing image processing model, and the output of the first feature extraction module is connected to the input of the first feature fusion module; the input of the first feature fusion module is also configured to receive the first model input of the remote sensing image processing model; the output of the first feature fusion module is connected to the input of the third feature fusion module; the input of the second feature extraction module is configured to receive the second model input of the remote sensing image processing model; the output of the second feature extraction module is connected to the input of the second feature fusion module; the output of the second feature fusion module is connected to the input of the third feature fusion module; the input of the third feature fusion module is also configured to receive the first model input of the remote sensing image processing model; the output of the third feature fusion module is connected to the input of the image segmentation module; the output of the image segmentation module is configured to output the model output of the remote sensing image processing model.

[0146] It is evident that implementation Figure 3 The described remote sensing image processing device is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0147] In yet another alternative embodiment, such as Figure 3 As shown, the first feature fusion module includes a first convolutional unit, a second convolutional unit, a first normalization unit, a second normalization unit, a first activation unit, a second activation unit, and a first fusion unit; wherein,

[0148] The input of the first convolutional unit is connected to the output of the first feature extraction module; the first convolutional unit, the first normalization unit, the first activation unit, and the first fusion unit are connected in sequence; the input of the first fusion unit is also configured to receive the first model input of the remote sensing image processing model; the first fusion unit, the second convolutional unit, the second normalization unit, and the second activation unit are connected in sequence; the output of the second activation unit is connected to the input of the third feature fusion module.

[0149] It is evident that implementation Figure 3 The described remote sensing image processing device is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0150] In yet another alternative embodiment, such as Figure 3 As shown, the third feature fusion module includes a second fusion unit, a third fusion unit, a fourth fusion unit, a first pooling unit, a third convolution unit, a third normalization unit, and a third activation unit; wherein,

[0151] The input of the third fusion unit is connected to the output of the first feature fusion module; the input of the third fusion unit is also configured to receive the first model input of the remote sensing image processing model; the output of the third fusion unit is connected to the input of the first pooling unit; the output of the first pooling unit is connected to the input of the second fusion unit; the input of the second fusion unit is also connected to the output of the second feature fusion module; the output of the second fusion unit is connected to the input of the fourth fusion unit; the input of the fourth fusion unit is also connected to the output of the second feature fusion module; the fourth fusion unit, the third convolution unit, the third normalization unit, and the third activation unit are connected sequentially; the output of the third activation unit is connected to the input of the image segmentation module.

[0152] It is evident that implementation Figure 3 The described remote sensing image processing device is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0153] In yet another alternative embodiment, such as Figure 3As shown, the first feature extraction module includes a fourth convolutional unit, a fifth convolutional unit, a sixth convolutional unit, a seventh convolutional unit, a fourth normalization unit, a fifth normalization unit, a sixth normalization unit, a seventh normalization unit, an eighth normalization unit, a ninth normalization unit, a tenth normalization unit, an eleventh normalization unit, a twelfth normalization unit, a thirteenth normalization unit, a fourth activation unit, a fifth activation unit, a sixth activation unit, a seventh activation unit, an eighth activation unit, a ninth activation unit, a tenth activation unit, an eleventh activation unit, a first connection unit, a second pooling unit, a third pooling unit, and a fourth pooling unit; wherein,

[0154] The input of the fourth convolutional unit is configured as the input of the first feature extraction module; the fourth convolutional unit, the fourth normalization unit, the fourth activation unit, the fifth convolutional unit, the fifth normalization unit, the fifth activation unit, the sixth normalization unit, the sixth activation unit, the seventh normalization unit, the second pooling unit, the eighth normalization unit, the seventh activation unit, the sixth convolutional unit, the ninth normalization unit, the eighth activation unit, the tenth normalization unit, the ninth activation unit, the eleventh normalization unit, the third pooling unit, the twelfth normalization unit, the tenth activation unit, the seventh convolutional unit, the thirteenth normalization unit, the eleventh activation unit, the fourth pooling unit, and the first connection unit are connected sequentially; the output of the first connection unit is configured as the output of the first feature extraction module.

[0155] It is evident that implementation Figure 3 The described remote sensing image processing device is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0156] In yet another alternative embodiment, such as Figure 3 As shown, the image segmentation module includes a fourth fusion unit, an eighth convolution unit, a fourteenth normalization unit, a twelfth activation unit, a thirteenth activation unit, and a second connection unit; wherein,

[0157] The input of the fourth fusion unit is connected to the output of the third feature fusion module and the output of the thirteenth activation unit, respectively; the fourth fusion unit, the eighth convolution unit, the fourteenth normalization unit, the twelfth activation unit, the second connection unit and the fourteenth activation unit are connected in sequence; the output of the fourteenth activation unit is configured to output the model output of the remote sensing image processing model.

[0158] It is evident that implementation Figure 3The described remote sensing image processing device is beneficial for improving the accuracy of remote sensing image analysis through deep learning technology, thereby enabling rapid and accurate identification and application of ground features and facilities from multi-resolution satellite optical remote sensing images. This solves the problem of achieving rapid remote sensing identification response and application across sensors and at multiple resolutions when data conditions are limited.

[0159] Example 3

[0160] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of another remote sensing image processing device disclosed in an embodiment of the present invention. Figure 4 The described apparatus can be applied in management systems, such as local servers or cloud servers for management, and the embodiments of the present invention are not limited thereto. Figure 4 As shown, the device may include:

[0161] Memory 301 storing executable program code;

[0162] Processor 302 coupled to memory 301;

[0163] The processor 302 calls the executable program code stored in the memory 301 to execute the steps in the remote sensing image processing method described in Embodiment 1.

[0164] Example 4

[0165] This invention discloses a computer-readable storage medium that stores a computer program for electronic data interchange, wherein the computer program causes a computer to perform the steps in the remote sensing image processing method described in Embodiment 1.

[0166] Example 5

[0167] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps in the remote sensing image processing method described in Embodiment 1.

[0168] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0169] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, 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. This computer software product can be stored in a computer-readable storage medium, including 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 storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0170] Finally, it should be noted that the remote sensing image processing method and apparatus supporting expert experience fusion mechanism disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A remote sensing image processing method, characterized in that, The method includes: Acquire remote sensing image information to be processed; The remote sensing image information to be processed is preprocessed to obtain the target processed remote sensing image information; The remote sensing image information of the target is analyzed and processed to obtain target remote sensing image analysis result information, including: Based on the remote sensing image processing model and the target processed remote sensing image information, the first remote sensing image analysis result information is determined. Based on the analysis results of the first remote sensing image, the analysis results of the target remote sensing image are determined. The remote sensing image processing model includes a first feature extraction module, a second feature extraction module, a first feature fusion module, a second feature fusion module, a third feature fusion module, and an image segmentation module; wherein, The input terminal of the first feature extraction module is configured to receive a first model input of the remote sensing image processing model, and the output terminal of the first feature extraction module is connected to the input terminal of the first feature fusion module; the input terminal of the first feature fusion module is also configured to receive a first model input of the remote sensing image processing model; the output terminal of the first feature fusion module is connected to the input terminal of the third feature fusion module; the input terminal of the second feature extraction module is configured to receive a second model input of the remote sensing image processing model; the output terminal of the second feature extraction module is connected to the input terminal of the second feature fusion module; the output terminal of the second feature fusion module is connected to the input terminal of the third feature fusion module; the input terminal of the third feature fusion module is also configured to receive a first model input of the remote sensing image processing model; the output terminal of the third feature fusion module is connected to the input terminal of the image segmentation module; the output terminal of the image segmentation module is configured to output the model output of the remote sensing image processing model.

2. The remote sensing image processing method according to claim 1, characterized in that, The first feature fusion module includes a first convolutional unit, a second convolutional unit, a first normalization unit, a second normalization unit, a first activation unit, a second activation unit, and a first fusion unit; wherein, The input of the first convolutional unit is connected to the output of the first feature extraction module; the first convolutional unit, the first normalization unit, the first activation unit, and the first fusion unit are connected in sequence; the input of the first fusion unit is also configured to receive the first model input of the remote sensing image processing model; the first fusion unit, the second convolutional unit, the second normalization unit, and the second activation unit are connected in sequence; the output of the second activation unit is connected to the input of the third feature fusion module.

3. The remote sensing image processing method according to claim 1, characterized in that, The third feature fusion module includes a second fusion unit, a third fusion unit, a fourth fusion unit, a first pooling unit, a third convolution unit, a third normalization unit, and a third activation unit; wherein, The input of the third fusion unit is connected to the output of the first feature fusion module; the input of the third fusion unit is also configured to receive the first model input of the remote sensing image processing model; the output of the third fusion unit is connected to the input of the first pooling unit; the output of the first pooling unit is connected to the input of the second fusion unit; the input of the second fusion unit is also connected to the output of the second feature fusion module; the output of the second fusion unit is connected to the input of the fourth fusion unit; the input of the fourth fusion unit is also connected to the output of the second feature fusion module; the fourth fusion unit, the third convolution unit, the third normalization unit, and the third activation unit are connected sequentially; the output of the third activation unit is connected to the input of the image segmentation module.

4. The remote sensing image processing method according to claim 1, characterized in that, The first feature extraction module includes a fourth convolutional unit, a fifth convolutional unit, a sixth convolutional unit, a seventh convolutional unit, a fourth normalization unit, a fifth normalization unit, a sixth normalization unit, a seventh normalization unit, an eighth normalization unit, a ninth normalization unit, a tenth normalization unit, an eleventh normalization unit, a twelfth normalization unit, a thirteenth normalization unit, a fourth activation unit, a fifth activation unit, a sixth activation unit, a seventh activation unit, an eighth activation unit, a ninth activation unit, a tenth activation unit, an eleventh activation unit, a first connection unit, a second pooling unit, a third pooling unit, and a fourth pooling unit; wherein, The input of the fourth convolutional unit is configured as the input of the first feature extraction module; the fourth convolutional unit, the fourth normalization unit, the fourth activation unit, the fifth convolutional unit, the fifth normalization unit, the fifth activation unit, the sixth normalization unit, the sixth activation unit, the seventh normalization unit, the second pooling unit, the eighth normalization unit, the seventh activation unit, the sixth convolutional unit, the ninth normalization unit, the eighth activation unit, the tenth normalization unit, the ninth activation unit, the eleventh normalization unit, the third pooling unit, the twelfth normalization unit, the tenth activation unit, the seventh convolutional unit, the thirteenth normalization unit, the eleventh activation unit, the fourth pooling unit, and the first connection unit are connected sequentially; the output of the first connection unit is configured as the output of the first feature extraction module; The image segmentation module includes a fourth fusion unit, an eighth convolution unit, a fourteenth normalization unit, a twelfth activation unit, a thirteenth activation unit, and a second connection unit; wherein, The input of the fourth fusion unit is connected to the output of the third feature fusion module and the output of the thirteenth activation unit, respectively; the fourth fusion unit, the eighth convolution unit, the fourteenth normalization unit, the twelfth activation unit, the second connection unit and the fourteenth activation unit are connected in sequence; the output of the fourteenth activation unit is configured to output the model output of the remote sensing image processing model.

5. The remote sensing image processing method according to claim 1, characterized in that, The step of determining the target remote sensing image analysis result information based on the first remote sensing image analysis result information includes: Based on the remote sensing image processing model, the target processing remote sensing image information collected at each time moment is processed to obtain the corresponding first remote sensing image analysis results. The analysis results of the first remote sensing images at all times are fused and calculated to obtain the target remote sensing image analysis result information; The expression for the fusion calculation is: ， in, The result of the analysis of the first remote sensing image at time i is shown. , and , representing the maximum, minimum, and average values ​​of the first remote sensing image analysis results at all times, respectively, and y represents the target remote sensing image analysis results information.

6. A remote sensing image processing device, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the remote sensing image processing method as described in any one of claims 1-5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when invoked, are used to execute the remote sensing image processing method as described in any one of claims 1-5.

Images