Multi-source remote sensing image change detection system, method, device, medium and product

Abstract

The invention relates to the technical field of multi-source remote sensing image change detection, in particular to a multi-source remote sensing image change detection system, method and device, a medium and a product, and the system comprises a multi-source image preprocessing module which is used for carrying out geometric registration and relative radiation correction on a multi-source image to obtain a multi-source image to be detected; the change detection module is used for carrying out change detection on the to-be-detected multi-source image; and a change detection system of a multi-source remote sensing image can be constructed.

Description

Technical Field

[0001] The present invention relates to the technical field of change detection of multi-source remote sensing images, and particularly to a change detection system, method, device, medium and product for multi-source remote sensing images. Background Art

[0002] Change detection of multi-source remote sensing images refers to the process of extracting change information using multi-temporal images from different sources. With the rapid development of aerospace technology and sensor technology, a large number of remote sensing images with different sources and types can be obtained in the field of earth observation technology, providing extremely rich data sources for change detection. In addition, the resolution of remote sensing images has also been continuously improved. Although the clarity of observation has been improved, it also poses greater demands on data processing capabilities.

[0003] There is a technical problem in this field of lacking a change detection system for multi-source remote sensing images. Summary of the Invention

[0004] The present invention provides a change detection system, method, device, medium and product for multi-source remote sensing images, and solves the technical problem of lacking a change detection system for multi-source remote sensing images.

[0005] In a first aspect, the present invention provides a change detection system for multi-source remote sensing images. The system includes: a multi-source image preprocessing module for geometric registration and relative radiometric correction of multi-source images to obtain multi-source images to be detected; a change detection module for performing change detection on the multi-source images to be detected.

[0006] In some embodiments, the change detection module includes: an unsupervised change detection module, a detection module for distinguishing change types, and / or a target change detection module.

[0007] In some embodiments, when the change detection module includes an unsupervised change detection module, performing change detection on the image to be detected includes: performing change detection on the multi-source images to be detected based on slow feature transformation function, deep feature extraction function, random multi-image function, K-means clustering function, and / or fuzzy C-means clustering function.

[0008] In some embodiments, the deep feature extraction function is written in the Python language under the theano framework; when the backend parameters of theano change, the execution of the deep feature extraction function is switched from the CPU to the GPU, or the execution of the deep feature extraction function is switched from the GPU to the CPU.

[0009] In some embodiments, when the change detection module includes a detection module for distinguishing change types, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on feature mapping transformation and / or hierarchical clustering function.

[0010] In some embodiments, when the change detection module includes a target change detection module, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on multi-feature extraction, adaptive sampling, patch post-processing, target preliminary screening, and / or target refined screening functions.

[0011] In a second aspect, the present invention provides a method for change detection of multi-source remote sensing images, the method including: performing geometric registration and relative radiometric correction on the multi-source images to obtain the multi-source image to be detected; performing change detection on the multi-source image to be detected.

[0012] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the above-mentioned method for change detection of multi-source remote sensing images.

[0013] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned method for change detection of multi-source remote sensing images are implemented.

[0014] In a fifth aspect, the present invention provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for change detection of multi-source remote sensing images are implemented.

[0015] The present invention provides a change detection system, method, device, medium, and product for multi-source remote sensing images, wherein the system includes: a multi-source image preprocessing module for performing geometric registration and relative radiometric correction on the multi-source images to obtain the multi-source image to be detected; a change detection module for performing change detection on the multi-source image to be detected; and capable of constructing a change detection system for multi-source remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Hereinafter, the present invention will be described in more detail based on embodiments and with reference to the drawings:

[0017] Figure 1 It is a schematic diagram of the architecture of a change detection system for multi-source remote sensing images provided by an embodiment of the present invention;

[0018] Figure 2 It is a schematic diagram of the interface of a change detection system for multi-source remote sensing images provided by an embodiment of the present invention;

[0019] Figure 3Schematic diagram of a multi-source image preprocessing module provided for an application example of the present invention;

[0020] Figure 4 Schematic diagram of an unsupervised change detection module provided for an application example of the present invention;

[0021] Figure 5 Schematic diagram of a detection module for distinguishing change types provided for an application example of the present invention;

[0022] Figure 6 Schematic diagram of a target change detection module provided for an application example of the present invention;

[0023] Figure 7 Schematic diagram of the flow of a change detection method for multi-source remote sensing images provided for an application example of the present invention;

[0024] Figure 8 Schematic diagram of an example of sample transfer provided for an application example of the present invention.

[0025] In the drawings, the same components are denoted by the same reference numerals, and the drawings are not drawn to actual scale. Detailed implementation manners

[0026] In order to enable those skilled in the art to better understand the solution of the present invention, and to fully understand how the present invention uses technical means to solve technical problems and the implementation process of achieving corresponding technical effects and implement accordingly, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. The embodiments of the present invention and each feature in the embodiments can be combined with each other without conflict, and the formed technical solutions are all within the protection scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0029] Change detection of multi-source remote sensing images refers to the process of extracting change information using multi-temporal images from different sources. With the rapid development of aerospace technology and sensor technology, a large number of remote sensing images with different sources and types can be obtained in the field of earth observation technology, providing extremely rich data sources for change detection. In addition, the resolution of remote sensing images has also been continuously improved. Although the clarity of observation has been improved, it has also put forward greater requirements for data processing capabilities. There is a technical problem in this field of lacking a change detection system for multi-source remote sensing images.

[0030] To solve the above technical problem of lacking a change detection system for multi-source remote sensing images, the present invention proposes a change detection system, method, device, medium and product for multi-source remote sensing images. The following details of the implementation of the present invention will be specifically described. The following content is only the implementation details provided for convenient understanding and is not necessary for implementing this solution.

[0031] Example 1

[0032] Figure 1 It is a schematic diagram of the architecture of a change detection system for multi-source remote sensing images provided by an embodiment of the present invention. Figure 2 It is a schematic diagram of the interface of a change detection system for multi-source remote sensing images provided by an embodiment of the present invention. Figure 3 It is a schematic diagram of a multi-source image preprocessing module provided by an application example of the present invention. As Figure 1 、 Figure 2 and Figure 3 shown, in the technical solution of this embodiment, a change detection system for multi-source remote sensing images is provided. The system includes: a multi-source image preprocessing module, which is used to perform geometric registration and relative radiometric correction on the multi-source images to obtain the multi-source images to be detected; a change detection module, which is used to perform change detection on the multi-source images to be detected.

[0033] The technical problem to be solved in this embodiment is how to construct a change detection system for multi-source remote sensing images. In the field of multi-source remote sensing image processing, it is necessary to perform change detection on remote sensing images with different sources, different types and continuously improving resolutions. However, there is currently a lack of a complete and efficient system for processing such image change detection. There is a technical problem in this field of lacking a unified and efficient construction scheme for multi-source remote sensing image change detection systems.

[0034] In the technical solution of this embodiment, a change detection system for multi-source remote sensing images is constructed, which includes a multi-source image preprocessing module and a change detection module. The multi-source image preprocessing module is used to perform geometric registration on the multi-source images, that is, to make images with different resolutions and different sources accurately correspond in spatial position. For example, images of the same area taken by different satellites but with different resolutions can be accurately overlapped. At the same time, relative radiometric correction is performed to ensure that the radiometric values of different images are in a comparable state, avoiding affecting the accuracy of subsequent change detection due to radiometric differences. Based on this, the change detection module further conducts change detection-related work on the multi-source images to be detected, and accurately extracts change information from the preprocessed images.

[0035] In the technical solution of this embodiment, by constructing a system including a multi-source image preprocessing module and a change detection module, the standardization and organization of the multi-source remote sensing image processing process are first realized. For multi-source images with different sources and different resolutions, the preprocessing module can effectively solve the problems of inconsistent spatial positions and radiometric values of the images, making the data basis for subsequent change detection more reliable. For example, when processing multi-resolution images of a certain urban area obtained from different periods and different sensors, after preprocessing, these originally quite different images can have a unified standard for subsequent analysis. Moreover, based on this, the change detection module can work to more accurately detect the changed areas in the images. Whether it is a change in land use type or the construction or demolition of buildings, etc., can be better detected, greatly improving the accuracy and efficiency of multi-source remote sensing image change detection, and providing a powerful tool for relevant fields to conduct change analysis using remote sensing images.

[0036] Example 2

[0037] Based on the above embodiment, the change detection module includes: an unsupervised change detection module, a detection module for distinguishing change types, and / or a target change detection module.

[0038] The technical problem to be solved in this embodiment is how to construct the change detection module of the change detection system for multi-source remote sensing images. In the multi-source remote sensing image change detection system, change detection needs to cope with various different detection requirements. A single detection method is difficult to meet the requirements of comprehensively and accurately detecting the change situations of complex multi-source images. There is a technical problem in this field that the function of the change detection module is single and cannot adapt to diverse detection requirements.

[0039] In the technical solution of this embodiment, the change detection module includes an unsupervised change detection module, a detection module for differentiating change types, and a target change detection module. The unsupervised change detection module mainly utilizes the slow feature transformation function, which can extract relatively slow but representative features from images. For example, for regions with slow changes in multi-temporal images, this function can distinguish them from regions with significant changes; the deep feature extraction function, with the help of deep learning models such as stacked denoising autoencoders, can extract deep and more discriminative image features; there are also functions such as random multi-image, which can discover image changes through different combinations and analysis methods. The detection module for differentiating change types relies on the feature mapping transformation function to establish the correlation between optical images and SAR images, and further subdivides different change types using the hierarchical clustering function, such as distinguishing whether it is a change in vegetation cover or land development. The target change detection module is based on functions such as multi-feature extraction, adaptive sampling, patch post-processing, target preliminary screening, and target refined screening, and conducts detailed change detection for specific targets, such as detecting the number and position changes of aircraft at an airport.

[0040] In the technical solution of this embodiment, by setting a change detection module containing multiple sub-modules, the functions of the entire multi-source remote sensing image change detection system are greatly enriched. When faced with a large number of unlabeled multi-source remote sensing images, the unsupervised change detection module can automatically and efficiently detect the changed areas, reducing the workload of manually labeling data and improving the processing efficiency. The detection module for differentiating change types can provide more detailed and accurate change information, no longer limited to simply knowing whether there is a change, but being able to clarify the specific type of change. For fields such as urban planning and ecological environment monitoring, corresponding strategies can be formulated more precisely based on its detection results. The target change detection module focuses on the change detection of specific targets. In scenarios such as military monitoring and air traffic management, it can timely and accurately grasp the dynamic changes of targets, comprehensively improving the ability of the multi-source remote sensing image change detection system to handle different application scenarios.

[0041] Example 3

[0042] Figure 4 It is a schematic diagram of an unsupervised change detection module provided for an application example of the present invention. As Figure 4 shown, on the basis of the above embodiment, when the change detection module includes an unsupervised change detection module, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on the slow feature transformation function, deep feature extraction function, random multi-image function, K-means clustering function, and / or fuzzy C-means clustering function.

[0043] The technical problem to be solved in this embodiment is how to construct an unsupervised change detection module for a multi-source remote sensing image change detection system. When using multi-source remote sensing images for change detection, although the unsupervised method does not require a large number of labeled samples, there are many technical problems in achieving high-precision and multi-type change detection functions. For example, how to effectively extract key features and how to perform reasonable clustering, etc. There are technical problems in this field that the function of the unsupervised change detection module is not perfect enough and it is difficult to achieve complex and accurate change detection.

[0044] In the technical solution of this embodiment, in the unsupervised change detection module, change detection is performed on the multi-source image to be detected based on the slow feature transformation function, deep feature extraction function, random multi-image function, K-means clustering function, and fuzzy C-means clustering function. The slow feature transformation function focuses on mining features in the image that change slowly over time but are important for judging changes. For example, for regions that remain unchanged in multi-temporal images for a long time, this function can extract feature information reflecting such long-term slow changes. The deep feature extraction function uses a stacked denoising autoencoder written in Python under the theano framework to extract deep features from the image and enhance the discrimination ability for different change situations. The random multi-image function constructs different representations of the image by randomly extracting features and discovers changes using the comprehensive voting method, while the K-means clustering function and the fuzzy C-means clustering function perform clustering analysis on the image based on the extracted features, thereby dividing different categories such as changed regions and unchanged regions.

[0045] The technical solution of this embodiment has shown significant advantages in practical applications. For multi-source remote sensing images where it is difficult to obtain a large number of labeled samples, such as when conducting long-term ecological monitoring in some remote areas, the image data volume is large but labeling is difficult. This module can automatically extract features and perform clustering analysis to accurately detect situations such as land desertification and vegetation cover changes. The slow feature transformation function ensures the effective capture of slow change situations, the deep feature extraction function improves the discriminability of features, and the clustering function reasonably classifies the image. Overall, it improves the accuracy and comprehensiveness of unsupervised change detection, enabling the efficient completion of the multi-source remote sensing image change detection task in the absence of labeled samples, providing a strong guarantee for the relevant fields to carry out large-scale and long-term remote sensing change monitoring.

[0046] Example 4

[0047] Based on the above embodiments, a deep feature extraction function is written in Python language under the Theano framework; when the backend parameters of Theano change, the execution of the deep feature extraction function is switched from the CPU to the GPU, or the execution of the deep feature extraction function is switched from the GPU to the CPU.

[0048] The technical problem to be solved in this embodiment is how to switch between the CPU and the GPU. In the process of deep feature extraction from multi-source remote sensing images, different computing devices have different performance characteristics. The CPU is suitable for processing tasks with complex logic but relatively small computational volume, while the GPU is more proficient in parallel computing and processing large-scale data operations. However, according to different working scenarios and data scales, it is necessary to flexibly switch between the two to achieve the optimal computing efficiency. There is a technical problem in this field that it is impossible to conveniently switch the deep feature extraction function between the CPU and the GPU as needed.

[0049] In the technical solution of this embodiment, the deep feature extraction function is written in Python language under the Theano framework, and by changing the backend parameters of Theano, it is possible to switch the execution of the deep feature extraction function from the CPU to the GPU, or from the GPU to the CPU. For example, when the amount of data to be processed is small and the requirement for computing speed is not extremely high, the backend parameters are set so that the deep feature extraction function is executed on the CPU, which can make full use of the advantages of the CPU to complete the task; when facing a large amount of multi-source remote sensing image data and fast extraction of deep features is required, the backend parameters are adjusted so that the function runs on the GPU. With the powerful parallel computing ability of the GPU, a large amount of data feature extraction work can be completed in a short time, greatly improving the processing efficiency.

[0050] The technical solution of this embodiment brings many benefits in the deep feature extraction of multi-source remote sensing images through a convenient CPU and GPU switching mechanism. In practical applications, different projects and different data scales can be properly handled. For example, when performing change detection on multi-source remote sensing images of a small urban area, the amount of image data is limited, and using the CPU to execute the deep feature extraction function is sufficient to meet the requirements and the resources are utilized reasonably; when analyzing multi-source remote sensing images of a large urban agglomeration or even a cross-provincial region, the amount of data involved is extremely large. By switching to the GPU to execute the deep feature extraction function, the feature extraction can be quickly completed, shortening the time of the entire change detection process, improving the overall flexibility and adaptability of the system, and ensuring that deep feature extraction work can be completed with better computing efficiency for both small-scale and large-scale multi-source remote sensing image processing tasks.

[0051] Example 5

[0052] Figure 5 Schematic diagram of a detection module for distinguishing change types provided for an application example of the present invention. As Figure 5 shown, on the basis of the above embodiment, when the change detection module includes a detection module for distinguishing change types, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on feature mapping transformation and / or hierarchical clustering function.

[0053] The technical problem to be solved in this embodiment is how to construct a detection module for distinguishing change types in a change detection system for multi-source remote sensing images. In the change detection of multi-source remote sensing images, it is not enough to only know that there are changes in the images. It is also necessary to clarify the specific change types, such as distinguishing whether the changes are caused by human construction or natural environmental evolution, etc. However, the related technologies are not perfect in this regard and it is difficult to accurately and meticulously distinguish the change types. There is a technical problem in the field that the function of the detection module for distinguishing change types is lacking and it is impossible to accurately distinguish the specific change types of multi-source remote sensing images.

[0054] In the technical solution of this embodiment, in the detection module for distinguishing change types, change detection is performed on the multi-source image to be detected based on feature mapping transformation and hierarchical clustering function. The feature mapping transformation function can establish the association between optical images and SAR images. For example, when monitoring an area with both urban buildings and natural vegetation, through this function, information such as building outlines and vegetation distributions presented in the optical image can be associated with the corresponding reflection features in the SAR image, integrating the advantageous information of different image sources. On this basis, the hierarchical clustering function further performs hierarchical analysis and clustering on the integrated features, subdividing the change areas according to different types, such as distinguishing different situations such as changes in newly built residential areas, changes in farmland converted into orchards, or natural changes such as river course changes, so as to achieve accurate distinction of change types.

[0055] The technical solution of this embodiment plays an important role in the application of change detection of multi-source remote sensing images by using the feature mapping transformation and hierarchical clustering functions to construct a detection module for distinguishing change types. In the field of urban planning, through this module, it is possible to accurately distinguish different types of changes such as which areas are newly added commercial areas and which are expanded industrial parks, providing detailed basis for the reasonable layout of urban functional areas; in the aspect of ecological environment monitoring, it is possible to clearly distinguish whether it is human-damaged changes such as deforestation and wetland reduction, or natural changes in vegetation cover caused by climate change, etc., facilitating the formulation of targeted protection and restoration measures. Accurately distinguishing change types makes the change detection results of multi-source remote sensing images more valuable, can better serve the needs of different industries for in-depth analysis of change situations, and improves the practicality and guiding significance of the entire multi-source remote sensing image change detection system in actual applications.

[0056] Example 6

[0057] Figure 6 Schematic diagram of a target change detection module provided for an application example of the present invention. As Figure 6 shown, on the basis of the above embodiment, when the change detection module includes a target change detection module, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on multi-feature extraction, adaptive sampling, patch post-processing, initial target screening, and / or fine target screening functions.

[0058] The technical problem to be solved in this embodiment is how to construct a target change detection module for a change detection system of multi-source remote sensing images. In the application of multi-source remote sensing images, sometimes it is necessary to perform detailed change detection on specific targets, such as monitoring the change in the number and position of airplanes at an airport, the dynamic change of ships in a port, etc. However, there are many challenges in constructing a module that can effectively achieve such target change detection. For example, how to accurately extract target-related features and how to reasonably screen targets. There are technical problems in the art that the function of the target change detection module is imperfect and it is difficult to accurately complete the specific target change detection task.

[0059] In the technical solution of this embodiment, in the target change detection module, change detection is performed on the multi-source image to be detected based on multi-feature extraction, adaptive sampling, patch post-processing, initial target screening, and fine target screening functions. The multi-feature extraction function extracts features in multiple aspects such as spatial features and texture features. For example, for an airplane target, the spatial features corresponding to its outer contour and the texture features on the fuselage surface are extracted to enhance the recognition of the target. The adaptive sampling function is completed by combining and compiling the IR-MAD transform and the adaptive sampling selection algorithm, and can select appropriate samples according to the actual situation of the image to improve the accuracy of detection. Patch post-processing includes operations such as dilation and erosion, which can optimize the initially detected target patches to make them more conform to the actual target shape. The initial target screening and fine target screening functions, on the basis of introducing a deep learning-based target recognition method, first initially screen out possible targets, and then further accurately determine the targets, such as accurately finding the position and status changes of each airplane in the airport remote sensing image.

[0060] The technical solution of this embodiment constructs a target change detection module by applying these functions, which has outstanding effects in actual application scenarios. In the aspect of air traffic management, through the analysis of multi-source remote sensing images of the airport area, it can accurately detect the increase or decrease of aircraft, changes in parking positions, etc., providing real-time and accurate data support for the dispatching command and safety monitoring of the airport; in the field of military monitoring, for the change detection of specific military targets such as ships and war vehicles, this module can accurately grasp their dynamic changes such as quantity and distribution, which helps to make strategic decisions in a timely manner. Moreover, the entire module integrates deep learning algorithms, further improving the accuracy of target change detection, so that in the face of complex backgrounds and multi-source images, it can still efficiently and accurately complete the change detection task of specific targets, greatly expanding the application ability of the multi-source remote sensing image change detection system in the monitoring of specific targets.

[0061] Example 7

[0062] Figure 7 is a schematic flow chart of a method for change detection of multi-source remote sensing images provided by an embodiment of the present application. As Figure 7 shown, in the technical solution of this embodiment, a method for change detection of multi-source remote sensing images is provided. The method includes: performing geometric registration and relative radiometric correction on the multi-source images to obtain the multi-source images to be detected; performing change detection on the multi-source images to be detected.

[0063] The technical problem to be solved in this embodiment is how to construct a method for change detection of multi-source remote sensing images. In the field of multi-source remote sensing image processing, it is necessary to perform change detection on remote sensing images from different sources, of different types, and with continuously increasing resolutions. However, there is currently a lack of a complete and efficient method for processing such image change detection, and there is a technical problem in this field of lacking a unified and efficient construction scheme for multi-source remote sensing image change detection methods.

[0064] In the technical solution of this embodiment, a method for change detection of multi-source remote sensing images is constructed, which includes a multi-source image preprocessing method and a change detection method. The multi-source image preprocessing method is used to perform geometric registration on the multi-source images, that is, to accurately correspond the images with different resolutions and different sources in terms of spatial position. For example, it can make the images of the same area taken by different satellites but with different resolutions accurately coincide; at the same time, perform relative radiometric correction to ensure that the radiometric values of different images are in a comparable state, avoiding affecting the accuracy of subsequent change detection due to radiometric differences. The change detection method is to further carry out relevant work on change detection of the multi-source images to be detected on this basis, and accurately extract change information from the preprocessed images.

[0065] The technical solution of this embodiment, by constructing a method including a multi-source image preprocessing method and a change detection method, first realizes the standardization and organization of the multi-source remote sensing image processing process. For multi-source images with different sources and different resolutions, the preprocessing method can effectively solve the problem of inconsistent image spatial position and radiation value, making the data basis for subsequent change detection more reliable. For example, when processing multi-resolution images of a certain urban area obtained from different periods and different sensors, after preprocessing, these originally large-scale images can have a unified standard for subsequent analysis. Moreover, the change detection method works based on this, and can more accurately find the changed area in the image, whether it is a change in land use type or a change in the construction and demolition of a building, etc., which can be better detected, greatly improving the accuracy and efficiency of multi-source remote sensing image change detection, and providing a powerful tool for related fields to use remote sensing images for change analysis.

[0066] Based on the above embodiments, the change detection method includes: an unsupervised change detection method, a detection method for distinguishing change types and / or a target change detection method.

[0067] The technical problem to be solved in this embodiment is how to construct a change detection method for multi-source remote sensing images. In the change detection method for multi-source remote sensing images, change detection needs to cope with a variety of different detection requirements. A single detection method is difficult to meet the requirements of comprehensive and accurate detection of changes in complex multi-source images. In this field, there are technical problems that the change detection method has a single function and cannot adapt to diversified detection requirements.

[0068] In the technical solution of this embodiment, the change detection method includes an unsupervised change detection method, a detection method for distinguishing change types, and a target change detection method. The unsupervised change detection method mainly uses the slow feature transformation function to dig out relatively slow but representative features from the image. For example, for the slowly changing areas in the multi-phase image, this function can distinguish them from the areas with significant changes; the deep feature extraction function uses the deep learning model, such as the stacked denoising autoencoder, to extract deep and more discriminative image features; there are also random multi-image functions, etc., which can discover image changes through different combinations and analysis methods. The detection method for distinguishing the type of change relies on the feature mapping transformation function to establish the association between optical images and SAR images, and uses the hierarchical clustering function to further subdivide different types of changes, such as distinguishing whether it is a vegetation cover change or a land development change. The target change detection method is based on multi-feature extraction, adaptive samples, spot post-processing, target initial screening and target fine screening functions, and performs detailed change detection for specific targets, such as the detection of changes in the number and position of aircraft at the airport.

[0069] The technical solution of this embodiment greatly enriches the functions of the entire multi-source remote sensing image change detection method by setting a change detection method including multiple sub-methods. When faced with a large number of unlabeled multi-source remote sensing images, the unsupervised change detection method can automatically and efficiently detect the changed areas, reducing the workload of manually labeling data and improving the processing efficiency. The detection method that distinguishes change types can give more detailed and accurate change information. It is no longer limited to simply knowing whether there is a change, but can clearly identify the specific type of change. For fields such as urban planning and ecological environment monitoring, more precise countermeasures can be formulated based on its detection results. The target change detection method focuses on the change detection of specific targets. In scenarios such as military monitoring and air traffic management, it can timely and accurately grasp the dynamic changes of targets, comprehensively improving the ability of the multi-source remote sensing image change detection method to handle different application scenarios.

[0070] Based on the above embodiment, when the change detection method includes an unsupervised change detection method, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on the slow feature transformation function, deep feature extraction function, random multi-image function, K-means clustering function, and / or fuzzy C-means clustering function.

[0071] The technical problem to be solved in this embodiment is how to construct an unsupervised change detection method for the multi-source remote sensing image change detection method. When using multi-source remote sensing images for change detection, although the unsupervised method does not require a large number of labeled samples, there are many technical problems in achieving high-precision and multi-type change detection functions, such as how to effectively extract key features and how to perform reasonable clustering. There are technical problems in this field that the function of the unsupervised change detection method is not perfect enough and it is difficult to achieve complex and accurate change detection.

[0072] In the technical solution of this embodiment, in the unsupervised change detection method, change detection is performed on the multi-source image to be detected based on the slow feature transformation function, deep feature extraction function, random multi-image function, K-means clustering function, and fuzzy C-means clustering function. The slow feature transformation function focuses on mining features in the image that change relatively slowly over time but are important for judging changes. For example, for areas that remain unchanged in multi-temporal images for a long time, this function can extract feature information reflecting such long-term slow changes. The deep feature extraction function uses a stacked denoising autoencoder written in Python under the theano framework to extract deep features from the image, enhancing the ability to distinguish different change situations. The random multi-image function constructs different representations of the image by randomly extracting features and uses the comprehensive voting method to discover changes, while the K-means clustering function and the fuzzy C-means clustering function perform clustering analysis on the image based on the extracted features, thereby dividing different categories such as changed areas and unchanged areas.

[0073] The technical solution of this embodiment has shown significant advantages in practical applications by constructing an unsupervised change detection method. For multi-source remote sensing images where it is difficult to obtain a large number of labeled samples, such as when conducting long-term ecological monitoring in some remote areas, although the image data volume is large but labeling is difficult, this method can automatically extract features and perform clustering analysis to accurately detect situations such as land desertification and vegetation cover changes. The slow feature transformation function ensures the effective capture of slow-changing situations, the deep feature extraction function enhances the discriminability of features, and the clustering function reasonably classifies the images, overall improving the accuracy and comprehensiveness of unsupervised change detection, enabling the efficient completion of the change detection task for multi-source remote sensing images even in the absence of labeled samples, providing a strong guarantee for carrying out large-scale and long-term remote sensing change monitoring in related fields.

[0074] Based on the above embodiment, the deep feature extraction function is written in Python under the Theano framework; when the backend parameters of Theano change, the execution of the deep feature extraction function can be switched from the CPU to the GPU, or the execution of the deep feature extraction function can be switched from the GPU to the CPU.

[0075] The technical problem to be solved in this embodiment is how to switch between the CPU and the GPU. During the deep feature extraction process of multi-source remote sensing images, different computing devices have different performance characteristics. The CPU is suitable for processing tasks with complex logic but relatively small computational volume, while the GPU is better at parallel computing and processing large-scale data operations. However, according to different working scenarios and data scales, it is necessary to flexibly switch between the two to achieve the optimal computing efficiency. There is a technical problem in this field that it is not convenient to realize the on-demand switching of the deep feature extraction function between the CPU and the GPU.

[0076] In the technical solution of this embodiment, the deep feature extraction function is written in Python under the Theano framework, and by changing the backend parameters of Theano, the execution of the deep feature extraction function can be switched from the CPU to the GPU, or from the GPU to the CPU. For example, when the amount of data to be processed is small and the requirement for computing speed is not extremely high, the backend parameters are set so that the deep feature extraction function is executed on the CPU, making full use of the advantages of the CPU to complete the task; while when facing a large amount of multi-source remote sensing image data and fast extraction of deep features is required, the backend parameters are adjusted to make this function run on the GPU. With the powerful parallel computing ability of the GPU, a large amount of data feature extraction work can be completed in a short time, greatly improving the processing efficiency.

[0077] The technical solution of this embodiment brings many benefits in the deep feature extraction of multi-source remote sensing images through a convenient CPU and GPU switching mechanism. In practical applications, different projects and different data scales can be properly addressed. For example, when performing change detection on multi-source remote sensing images of a small urban area, the amount of image data is limited, and using the CPU to execute the deep feature extraction function is sufficient to meet the requirements, and the resource utilization is reasonable; while when analyzing multi-source remote sensing images of a large urban agglomeration or even a cross-provincial area, the amount of data involved is extremely large. By switching to the GPU to execute the deep feature extraction function, the feature extraction can be quickly completed, shortening the time of the entire change detection process, improving the overall flexibility and adaptability of the method, and ensuring that the deep feature extraction work can be completed with better computational efficiency for both small-scale and large-scale multi-source remote sensing image processing tasks.

[0078] Based on the above embodiment, when the change detection method includes a detection method for distinguishing change types, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on the feature mapping transformation and / or hierarchical clustering function.

[0079] The technical problem to be solved in this embodiment is how to construct a detection method for distinguishing change types in the change detection method of multi-source remote sensing images. In the change detection of multi-source remote sensing images, it is not enough to only know that there are changes in the images. It is also necessary to clarify the specific change types, such as distinguishing whether the changes are caused by human construction or natural environmental evolution, etc. However, the related technologies are not perfect in this regard and it is difficult to accurately and carefully distinguish the change types. There is a technical problem in this field that the function of the detection method for distinguishing change types is lacking and it is impossible to accurately distinguish the specific change types of multi-source remote sensing images.

[0080] In the technical solution of this embodiment, in the detection method for distinguishing change types, change detection is performed on the multi-source image to be detected based on the feature mapping transformation and hierarchical clustering function. The feature mapping transformation function can establish the association between optical images and SAR images. For example, when monitoring an area with both urban buildings and natural vegetation, through this function, the information such as building outlines and vegetation distributions presented in the optical image can be associated with the corresponding reflection features in the SAR image, integrating the advantageous information of different image sources. On this basis, the hierarchical clustering function further performs hierarchical analysis and clustering on the integrated features, subdividing the change areas according to different types, such as distinguishing the changes in newly built residential areas, the conversion of farmland into orchards, or natural changes such as river course changes, etc., so as to achieve accurate distinction of change types.

[0081] The technical solution of this embodiment plays an important role in the change detection application of multi-source remote sensing images by constructing a detection method for distinguishing change types using feature mapping transformation and hierarchical clustering functions. In the field of urban planning, this method can accurately distinguish different types of changes, such as newly added commercial areas and expanded industrial parks, providing detailed basis for the rational layout of urban functional areas; in the aspect of ecological environment monitoring, it can clearly distinguish whether it is human-damaged changes such as deforestation and wetland reduction, or natural changes in vegetation cover caused by climate change, facilitating the formulation of targeted protection and restoration measures. Precise distinction of change types makes the change detection results of multi-source remote sensing images more valuable, better serving the needs of different industries for in-depth analysis of change situations, and enhancing the practicality and guiding significance of the entire multi-source remote sensing image change detection method in practical applications.

[0082] Based on the above embodiment, when the change detection method includes a target change detection method, performing change detection on the image to be detected includes: performing change detection on the multi-source image to be detected based on multi-feature extraction, adaptive sampling, patch post-processing, target preliminary screening, and / or target fine screening functions.

[0083] The technical problem to be solved in this embodiment is how to construct a target change detection method for the change detection method of multi-source remote sensing images. In the application of multi-source remote sensing images, sometimes it is necessary to conduct detailed change detection for specific targets, such as monitoring the changes in the number and position of aircraft at airports and the dynamic changes of ships in ports. However, constructing a method that can effectively achieve such target change detection faces many challenges, such as how to accurately extract target-related features and how to reasonably screen targets. There are technical problems in this field, such as the imperfect function of the target change detection method and the difficulty in accurately completing the specific target change detection task.

[0084] In the technical solution of this embodiment, in the target change detection method, multi-source images to be detected are subjected to change detection based on functions such as multi-feature extraction, adaptive sampling, patch post-processing, initial target screening, and precise target screening. The multi-feature extraction function extracts features in multiple aspects such as spatial features and texture features. For example, for an aircraft target, spatial features corresponding to its shape outline and texture features on the fuselage surface are extracted to enhance the recognition of the target. The adaptive sampling function is completed by combining and compiling the IR-MAD transform and the adaptive sampling selection algorithm, which can select appropriate samples according to the actual situation of the image to improve the accuracy of detection. Patch post-processing includes operations such as dilation and erosion, which can optimize the initially detected target patches to make them more conform to the actual target shape. The initial target screening and precise target screening functions, on the basis of introducing a deep learning-based target recognition method, first initially screen out possible targets and then further accurately determine the targets. For example, accurately find the position and status changes of each aircraft in airport remote sensing images.

[0085] In the technical solution of this embodiment, a target change detection method is constructed by using these functions, which has outstanding effects in actual application scenarios. In terms of air traffic management, through the analysis of multi-source remote sensing images of the airport area, it is possible to accurately detect the increase and decrease of aircraft, changes in parking positions, etc., providing real-time and accurate data support for airport dispatching, safety monitoring, etc.; in the field of military monitoring, for the change detection of specific military targets such as ships and war vehicles, this method can accurately grasp their dynamic changes such as quantity and distribution, helping to make strategic decisions in a timely manner. Moreover, the entire method integrates deep learning algorithms, further improving the accuracy of target change detection, enabling efficient and accurate completion of the change detection task for specific targets even in the face of complex backgrounds and multi-source images, greatly expanding the application ability of the multi-source remote sensing image change detection method in the monitoring of specific targets.

[0086] Example 8

[0087] In the technical solution of this embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps of the change detection method for multi-source remote sensing images in the above embodiment.

[0088] In the technical solution of this embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps of the change detection method for multi-source remote sensing images in the above embodiment are implemented.

[0089] In the technical solution of this embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps of the change detection method for multi-source remote sensing images in the above embodiment are implemented.

[0090] The processor may include, but is not limited to, for example, one or more processors or microprocessors, etc. Each processor may be implemented by an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic components, and is used to execute the methods in the above embodiments. The computer-readable storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof. The computer-readable storage medium may include, but is not limited to, for example, Random Access Memory (RAM), Read Only Memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (such as hard disks, floppy disks, solid state drives, removable disks, CD-ROMs, DVD-ROMs, Blu-ray discs, etc.).

[0091] The computer-readable storage medium may also store at least one computer-executable program / instructions, and the computer-executable program / instructions are, for example, computer-readable instructions. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, Random Access Memory (RAM) and / or cache, etc. The computer-readable storage medium may include, for example, Read Only Memory (ROM), hard disks, flash memory, etc. For example, the non-transitory computer-readable storage medium may be connected to a computing device such as a computer. Then, when the computing device runs the computer-readable instructions stored on the computer-readable storage medium, the various methods described above may be performed.

[0092] In addition, the computer device may also include (but is not limited to) a data bus, an Input / Output (I / O) bus, a display, and input / output devices (such as a keyboard, a mouse, a speaker, etc.). The processor may communicate with external devices via the I / O bus through a wired or wireless network. In one embodiment, the at least one computer-executable instruction may also be compiled into or form a software product / computer program product, and when one or more computer-executable instructions are run by the processor, the steps of the various functions and / or methods described in the embodiments of the present technology are executed.

[0093] Example 9

[0094] Based on the above embodiments, this embodiment provides an application example.

[0095] This application example provides a change detection system for high-resolution multi-source remote sensing images.

[0096] Change detection of multi-source remote sensing images refers to the process of extracting change information using multi-temporal images from different sources. With the rapid development of aerospace technology and sensor technology, a large number of remote sensing images with different sources and types can be obtained in the field of earth observation technology, providing extremely rich data sources for change detection. In addition, the resolution of remote sensing images is also constantly improving. Although it improves the clarity of observation, it also poses greater demands on data processing capabilities.

[0097] Although many studies have proposed change detection methods for high-resolution multi-source remote sensing images, they often only focus on the algorithm level and lack the support of a mature software platform. The problems solved by different methods also have different focuses. Some use unsupervised methods to achieve fully automatic change detection, while others achieve supervised change detection with certain sample support. Some commercial remote sensing software also provides a change detection module, but the methods used are often relatively simple, usually mainly based on common difference methods and ratio methods, and do not integrate more complex and advanced change detection algorithms, which limits the processing ability for high-resolution multi-source remote sensing images.

[0098] Change detection for high-resolution multi-source remote sensing images involves multiple steps such as multi-source image preprocessing, feature extraction, multi-feature fusion, feature transformation, unsupervised clustering, supervised classification, and post-processing, and has certain requirements for the functions and performance of the system.

[0099] Currently, the change detection functions of remote sensing software on the market are often relatively simple and direct, lacking the support of advanced algorithms, and have limited processing ability for high-resolution multi-source remote sensing images. For developers of change detection algorithms, they often need to start from basic algorithm research, which greatly increases the entry cost of research. For practitioners of change detection, they usually need to switch between multiple software platforms when processing high-resolution multi-source remote sensing images, making it significantly more difficult to get started than other types of remote sensing image processing tasks.

[0100] In general remote sensing image processing software, ENVI is a unified platform developed using the Interactive Data Language (IDL) with many basic remote sensing image processing functions. Although it does not integrate relevant algorithms and functions for high-resolution multi-source remote sensing image change detection, it is easy to perform secondary development to meet user requirements. Therefore, the present invention adopts the method of IDL secondary development to build a change detection system for high-resolution multi-source remote sensing images, which not only has mature and efficient traditional remote sensing image processing functions but also can specifically process high-resolution multi-source remote sensing images.

[0101] ENVI (The Environment for Visualizing Images) is a powerful set of remote sensing image processing software developed by scientists in the remote sensing field using the Interactive Data Language (IDL).

[0102] In terms of functions, in the preprocessing of multi-source images, it is necessary to implement geometric registration between images of different resolutions and relative radiometric correction between images of different sources; feature extraction includes both conventional image feature extraction, such as spatial features and texture features, and feature extraction based on deep learning, such as stacked denoising autoencoders; unsupervised clustering mainly includes K-means clustering and fuzzy C-clustering, and selection is made according to different application requirements; supervised classification includes both traditional classification algorithms such as BP neural networks and support vector machines, and should also have classification algorithms based on deep learning, such as convolutional neural networks; in post-processing, the system should mainly have functions such as convolutional filtering, clustering processing, filtering processing, and vectorization of changed patches.

[0103] In terms of performance, since the system not only involves traditional remote sensing image processing but also needs to integrate certain deep learning algorithms, it is necessary to ensure that traditional image processing algorithms can be executed maturely and efficiently and also have hardware acceleration capabilities for deep learning.

[0104] According to the above functional requirements and performance requirements, the requirements for each step of the high-resolution multi-source remote sensing image change detection system are uniformly planned and arranged, and the overall system architecture is designed, as Figure 1 shown. The main interface of the system is as Figure 2 shown.

[0105] According to the system framework designed by the present invention, the system mainly includes four modules: multi-source image preprocessing, unsupervised change detection, detection for differentiating change types, and target change detection. The specific implementation process and functions are as follows:

[0106] ① Multi-source image preprocessing module

[0107] The two functions of image relative registration and relative radiometric correction are the most basic preprocessing functions for change detection. Therefore, these two functions are first integrated into the multi-source image preprocessing module. In addition, to facilitate the operation of subsequent modules, the present invention also integrates functions such as image cropping, band stacking, band operation, spectral operation, and image segmentation into the preprocessing module. Among them, the registration method can either select the manual point selection method or the automatic registration method based on SIFT feature points; the relative radiometric correction method is the histogram method.

[0108] This module mainly integrates relevant functional functions in ENVI. Each function is mature, efficient, and stable in operation. The developed multi-source image preprocessing module is as shown.

[0109] ② Unsupervised Change Detection Module

[0110] This module mainly uses IDL language to write functional functions such as slow feature transformation, deep feature extraction, random multi-image, K-means clustering, and fuzzy C-means clustering, and can implement a variety of advanced unsupervised change detection algorithms, such as multi-scale fusion change detection method based on slow feature analysis, change detection method combining slow feature analysis and deep feature learning, and change detection method based on random multi-image, etc., as Figure 4 shown. Among them, the deep feature extraction function is written in Python language under the theano framework, and the model used is a stacked denoising autoencoder. By modifying the backend parameters of theano, the mode switch of the deep feature extraction function from CPU or GPU execution can be realized.

[0111] ③ Detection Module for Distinguishing Change Types

[0112] This module mainly includes functions such as feature mapping transformation and hierarchical clustering, and can implement a change detection method for distinguishing change types, that is, establish the correlation between optical images and SAR images through feature mapping transformation, and distinguish different change types through hierarchical clustering, as Figure 5 shown. The deep feature extraction step in this method also uses a stacked denoising autoencoder, so this step can be implemented using the deep feature extraction function in the unsupervised change detection module.

[0113] ④ Target Change Detection Module

[0114] This module mainly integrates or writes functional functions such as multi-feature extraction, adaptive sampling, patch post-processing, target initial screening, and target refined screening, and can implement a target change detection method combining multi-feature fusion and deep learning, such as Figure 6As shown in the figure. Among them, multi-feature extraction includes spatial feature extraction, texture feature extraction, etc., and post-processing of image patches includes dilation, erosion, etc. These functions are all integrated and implemented by relevant functions in ENVI. After multi-feature extraction, feature combination and calculation of feature difference vectors are required, and the two can be realized through band overlay and spectral operation functions in the multi-source image preprocessing module respectively. The adaptive sample function is completed by combining and compiling the IR-MAD transform and the adaptive sample selection algorithm. In the compilation process of the target initial screening function and the target refined screening function, a target recognition method based on deep learning needs to be introduced. Therefore, the corresponding model is compiled and trained in the caffe framework using the python language in this system, and this function is realized through the hybrid programming of IDL+python.

[0115] After each module is compiled separately, the main menu of the system is designed, and each module is embedded separately to achieve the integration of each module of the system.

[0116] Figure 8 This is a schematic diagram of a sample transfer example provided for the application example of the present invention. Considering that there is still a certain amount of data interaction between each module, the present invention designs the transfer rules for corresponding files, parameters, etc. Among them, data such as images and feature sets mainly adopt the file format of ENVI itself to achieve the unity of file formats between each module. In addition, the realization of some functions also requires the use of sample files generated by other functions, so the sample transfer format between each function is also designed. Considering from aspects such as implementation efficiency and versatility, this system uses the.txt file format encoded in ASCII to achieve sample transfer between each function, and its schematic diagram is as Figure 8 shown.

[0117] The present invention realizes a change detection system for high-resolution multi-source remote sensing images, which is characterized in that based on the general remote sensing image processing software architecture, a multi-source image preprocessing module, an unsupervised change detection module, a detection module for distinguishing change types, and a target change detection module are designed, and different types of multi-source remote sensing image processing algorithms can be realized.

[0118] In practical applications, it helps operators to conveniently realize the change detection tasks of different types of data, and also supports developers to perform secondary development and function realization of various change detection algorithms based on this system.

[0119] The multi-source image preprocessing module is an essential processing step for all change detection tasks. It mainly has functions such as image registration, image cropping, and radiometric correction to ensure the comparability of multi-source images in terms of space and radiometric values. In addition, the multi-source image preprocessing module also integrates basic functions such as certain band stacking and band operations, and can set formulas according to user needs to perform operations on multi-band images or data. Since these functions already exist in general remote sensing image processing software, only relevant function calls are required, and the implementation process is simple and straightforward, facilitating operation.

[0120] The unsupervised change detection module mainly has functions such as slow feature transformation, deep feature extraction, and random multi-image, so as to realize relatively advanced multi-scale fusion change detection methods based on slow feature analysis, change detection methods combining slow feature analysis and deep feature learning, and change detection methods based on random multi-image. Since these methods can be automatically implemented without manual intervention, it greatly improves the automatic operation ability of this system.

[0121] The detection module for distinguishing change types obtains more detailed change detection results. It can not only distinguish between changed areas and unchanged areas, but also further subdivide different change types. At the same time, it has the ability to detect changes in optical images and SAR images, and is mainly composed of functions such as feature mapping transformation and hierarchical clustering. Therefore, this module is more suitable for processing different types of multi-source images.

[0122] The target change detection module includes functions such as multi-feature extraction, multi-feature fusion change detection, post-processing of image patches, initial screening of aircraft targets, and fine screening of aircraft targets, and can realize change detection tasks at the target level. A systematic change detection function implementation process is designed mainly for aircraft targets. This module especially integrates deep learning algorithms, which helps to improve the accuracy of target change detection.

[0123] By integrating each independent module, the present invention successfully constructs a change detection system for high-resolution multi-source remote sensing images, and endows the system with the following characteristics:

[0124] 1. The system is developed based on general remote sensing image processing software, has strong applicability, can process various types of remote sensing data, including optical images, SAR images, etc., and can efficiently complete the preprocessing of these data.

[0125] 2. The system has a high degree of automation, flexible processing methods, simple system operation, modularly integrates each processing step, and requires fewer parameters to be set.

[0126] 3. It has a user-friendly design and strong scalability, and new functions can be continuously developed based on application needs. In particular, the application of deep learning algorithms is well implemented in the general remote sensing image processing platform, and other remote sensing applications based on deep learning can be further integrated into this system.

[0127] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0128] It should be noted that in the present invention, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, the element limited by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0129] Although the disclosed embodiments of the present invention are as above, the above content is only an embodiment adopted for the convenience of understanding the present invention and is not intended to limit the present invention. Any person skilled in the art within the technical field to which the present invention pertains can make any modifications and changes in the form of implementation and details without departing from the spirit and scope disclosed by the present invention. However, the scope of patent protection of the present invention shall still be subject to the scope defined by the appended claims.

Claims

1. A change detection system for multi-source remote sensing images, characterized in that: The system comprises: The multi-source image preprocessing module is used to perform geometric registration and relative radiation correction on the multi-source images to obtain the multi-source images to be detected; The change detection module is used to perform change detection on the multi-source images to be detected.

2. The change detection system of multi-source remote sensing images according to claim 1, characterized in that: The change detection module comprises: An unsupervised change detection module, a detection module that distinguishes change types, and / or a target change detection module.

3. The change detection system of multi-source remote sensing images according to claim 2, characterized in that: In the case where the change detection module includes an unsupervised change detection module, performing change detection on the image to be detected includes: Based on the slow feature transformation function, deep feature extraction function, random multi-image function, K-means clustering function and / or fuzzy C-means clustering function, change detection is performed on the multi-source images to be detected.

4. The change detection system of multi-source remote sensing images according to claim 3 is characterized in that: Write the deep feature extraction function in Python under the Theano framework; When the backend parameters of theano are changed, the execution of the deep feature extraction function is switched from the CPU to the GPU, or the execution of the deep feature extraction function is switched from the GPU to the CPU.

5. The change detection system of multi-source remote sensing images according to claim 2, characterized in that: In the case where the change detection module includes a detection module for distinguishing change types, performing change detection on the image to be detected includes: Based on feature map transformation and / or hierarchical clustering functions, change detection is performed on the multi-source images to be detected.

6. The change detection system of multi-source remote sensing images according to claim 2, characterized in that: In the case where the change detection module includes a target change detection module, performing change detection on the image to be detected includes: Based on multi-feature extraction, adaptive sampling, spot post-processing, target preliminary screening and / or target fine screening functions, change detection is performed on the multi-source images to be detected.

7. A method for detecting changes in multi-source remote sensing images, characterized in that: The method comprises: Perform geometric registration and relative radiation correction on multi-source images to obtain multi-source images to be detected; Perform change detection on the multi-source images to be detected.

8. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the multi-source remote sensing image change detection method according to claim 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting changes in multi-source remote sensing images as claimed in claim 7 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for detecting changes in multi-source remote sensing images as claimed in claim 7 are implemented.

Images