Method and system for analyzing and extracting characteristics of water and soil conservation targets in remote sensing images

Abstract

The present invention provides a method and system for analyzing and extracting the characteristics of environmental protection targets in remote sensing images. The method includes: obtaining a large number of labeled remote sensing images with relevant typical ground object markers for environmental protection at different resolutions, extracting the characteristics of relevant typical ground objects from the labeled remote sensing images, comparing the image characteristics with the object characteristics in the real scene to determine the object difference characteristics at different resolutions; optimizing the target feature extraction scheme at different resolution states based on the object difference characteristics to obtain a dedicated ground object information extraction scheme for environmental protection, identifying ground objects in the target remote sensing image, and comparing the identification results with visual interpretation to generate the final interpretation of the relevant typical ground object information for environmental protection corresponding to the target remote sensing image. The present invention proposes a dedicated ground object automatic recognition scheme for environmental protection monitoring, effectively improving the accuracy and precision of environmental protection monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly relates to a method and system for analyzing and extracting characteristics of water and soil conservation targets in remote sensing images. Background Art

[0002] With the continuous increase in social power consumption, the number of transmission and transformation project construction projects has gradually increased. While promoting the rapid development of social economy, transmission and transformation projects have also brought a series of impacts on the ecological environment. Transmission and transformation projects are linear projects, which have the characteristics of involving multiple administrative regions, long line paths, and scattered land occupation in a point-like distribution. Moreover, the lack of public understanding of electromagnetic environment knowledge and the fear of ionizing radiation have increased the difficulty of environmental protection work for transmission and transformation projects. As an important means of water and soil resource protection at the present stage, water and soil conservation monitoring has realized a scientific assessment of the causes, intensity, influence range, and harm degree of soil and water loss in the region, and plays an important role in the development and protection of water and soil resources. Remote sensing technology, with its wide coverage, high resolution, high efficiency, and stability, better meets the relevant requirements of water and soil conservation monitoring and has become the mainstream means of water and soil conservation monitoring at present.

[0003] The existing remote sensing technology mainly relies on the human eye to identify and locate water and soil conservation targets on remote sensing images. However, different-resolution images have different advantages in image analysis. For example, high-resolution panchromatic images are rich in spatial structure information and can express the detailed features of ground objects in detail. Low-resolution multispectral images are rich in spectral information and are conducive to the identification and interpretation of ground objects. Moreover, the distribution rates of remote sensing images required for different actual water and soil conservation analysis requirements also vary. Due to the difference in the distribution rate of remote sensing images, the presentation effect of object images is different, which requires too much experience for the identifier and is prone to the situation of missing or misidentifying object images. When using common methods such as the maximum likelihood method, object-oriented method, and edge detection to identify the characteristics of relevant typical ground objects in water and soil conservation, due to the difference in the presentation effect of object images, when the resolution of remote sensing images changes, the situation of misidentifying object images is likely to occur when any of the above ground object feature extraction methods is used. Summary of the Invention

[0004] The present invention provides a method and system for analyzing and extracting characteristics of water and soil conservation targets in remote sensing images, generating a ground object automatic recognition scheme dedicated to water and soil conservation monitoring, which is beneficial to improving the accuracy and precision of water and soil conservation monitoring, timely discovering environmental pollution events, reducing the probability of expanding the influence range of pollution events, and at the same time realizing the combined application of ground object automatic recognition and human-computer interaction visual interpretation recognition methods, which can solve the problems of false extraction and missed extraction in the automatic extraction algorithm to a certain extent and achieve high-precision interpretation of information on relevant typical ground objects in water and soil conservation.

[0005] The present invention provides a method for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, including:

[0006] Step 1: Obtain a large number of labeled remote sensing images with relevant typical ground object markers for water and soil conservation at different resolutions, and extract the characteristics of relevant typical ground objects from the labeled remote sensing images to obtain image features;

[0007] Step 2: Compare the image features with the object image features of relevant typical ground objects for water and soil conservation in the real scene to determine the object image difference features at different resolutions;

[0008] Step 3: Optimize the target feature extraction scheme at different resolution states based on the object image difference features to obtain a dedicated ground object information extraction scheme for water and soil conservation;

[0009] Step 4: Identify the ground objects in the target remote sensing image based on the dedicated ground object information extraction scheme, and compare the recognition results with visual interpretation to generate the final interpretation of the relevant typical ground object information for water and soil conservation corresponding to the target remote sensing image.

[0010] Preferably, in a method for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, obtaining a large number of labeled remote sensing images with relevant typical ground object markers for water and soil conservation at different resolutions, and extracting the characteristics of relevant typical ground objects from the labeled remote sensing images to obtain image features, including:

[0011] Generate image labels according to the resolutions corresponding to each labeled remote sensing image and add them to each labeled remote sensing image;

[0012] Group the labeled remote sensing images based on the water and soil conservation targets captured in the images to obtain multiple groups of remote sensing images;

[0013] Extract the characteristics of relevant typical ground objects within the labeled frames on each group of remote sensing images based on a preset feature extraction model to obtain image features.

[0014] Preferably, in a method for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, comparing the image features with the object image features of relevant typical ground objects for water and soil conservation in the real scene to determine the object image difference features at different resolutions, including:

[0015] Based on the image interpretation corresponding to the labeled remote sensing image, determine the target names and shooting areas of the relevant typical ground objects marked corresponding to each labeled frame on the image;

[0016] Based on the shooting area and target name, obtain the object images and their object image features of the relevant typical ground objects marked in the real object database;

[0017] Compare the object-image features with the image features corresponding to each remote sensing image respectively to obtain the image-object differences corresponding to the remote sensing images;

[0018] Based on relevant typical land cover types, compare the image-object differences corresponding to different remote sensing images to obtain the object-image difference features of the same relevant typical land cover at different resolutions.

[0019] Preferably, in a method for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, based on relevant typical land cover types, compare the image-object differences corresponding to different remote sensing images to obtain the object-image difference features of the same relevant typical land cover at different resolutions, including:

[0020] Based on the terrain features of the water and soil conservation targets corresponding to each remote sensing group, classify the remote sensing groups by geological types to obtain image clusters of water and soil conservation projects corresponding to different terrain environments;

[0021] Based on the relevant typical land cover markers in each remote sensing image, establish a relevant typical land cover catalog, and based on the relevant typical land cover catalog, screen the images in the image clusters of water and soil conservation projects to establish multiple same-object image sets;

[0022] Cluster the remote sensing images in the same-object image sets based on the resolution to obtain same-object same-resolution image sets, and compare the image-object differences of the same relevant typical land cover in the same-object same-resolution image sets to obtain the first resolution influence feature differences;

[0023] Based on the relevant typical land cover types, screen the same-object image sets of different image clusters of water and soil conservation projects to obtain a same-object image control group, and compare the image-object differences of the remote sensing images with the same different resolutions in the same-object image control group to obtain the terrain influence feature differences;

[0024] And compare the terrain influence feature differences corresponding to different resolutions in the same-object image control group to obtain the second resolution influence feature differences;

[0025] Based on the first resolution influence feature differences and the second resolution influence feature differences, generate the object-image difference features of relevant typical land cover.

[0026] Preferably, in a method for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, optimize the target feature extraction scheme under different resolution states based on the object-image difference features to obtain an exclusive land cover information extraction scheme for water and soil conservation, including:

[0027] Based on the object-image difference features of relevant typical land cover, combine with the relevant typical land cover catalog to generate an optimization reference table;

[0028] Based on the optimized reference table, a feature set is constructed by obtaining the image features of all relevant typical ground objects at different resolutions, and a resolution label is generated and added to the feature set.

[0029] Based on the feature set, the feature extraction process of the current target feature extraction scheme is optimized to obtain a dedicated ground object information extraction scheme for environmental protection related to water

[0030] Preferably, in a method for analyzing and extracting target features of environmental protection related to water in remote sensing images, based on the feature set, the feature extraction process of the current target feature extraction scheme is optimized to obtain a dedicated ground object information extraction scheme for environmental protection related to water, including:

[0031] Based on the feature differences between the object-image difference features corresponding to each relevant typical ground object in the feature set and the best feature extraction result identified by the current target feature extraction scheme at the resolution corresponding to the current feature set and the target object-image features corresponding to the actual ground objects, the feature extraction detail differences are obtained.

[0032] Based on the feature extraction detail differences, the reference features of the current target feature extraction scheme are corrected to obtain the best reference features, and the current target feature extraction scheme is updated to obtain a dedicated ground object information extraction sub-scheme.

[0033] The dedicated ground object information extraction sub-schemes corresponding to the resolution intervals of all feature sets are obtained, and a dedicated ground object information extraction scheme is generated. Preferably, in a method for analyzing and extracting target features of environmental protection related to water in remote sensing images, based on the dedicated ground object information extraction scheme, the ground objects in the target remote sensing image are identified, and the identification result is compared with the visual interpretation to generate the final interpretation of the relevant typical ground object information of environmental protection related to water corresponding to the target remote sensing image, including:

[0034] The resolution label of the target remote sensing image is identified, and according to the label identification result, the feature extraction result is adjusted to the best state of the dedicated ground object information extraction scheme.

[0035] Based on the dedicated ground object information extraction scheme in the best state, the relevant typical ground object names are determined based on the feature extraction result.

[0036] The relevant typical ground object names are compared with the ground object names in the visual interpretation to determine whether there are omissions or errors in the visual interpretation.

[0037] If so, the omitted names or incorrect names in the visual interpretation are modified based on the relevant typical ground object names to obtain the final interpretation of the relevant typical ground object information of environmental protection related to water corresponding to the target remote sensing image.

[0038] At the same time, a modified comparison interpretation is also generated based on the modified content of the visual interpretation and sent to the display end for comparison display.

[0039] The present invention provides a system for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, including:

[0040] An image feature extraction module, configured to obtain a large number of labeled remote sensing images with relevant typical ground object markers for water and soil conservation at different resolutions, and perform extraction of relevant typical ground object features on the labeled remote sensing images to obtain image features;

[0041] An object-image difference comparison module, configured to compare the image features with the object-image features of relevant typical ground objects for water and soil conservation in the real scene to determine the object-image difference features at different resolutions;

[0042] An exclusive solution generation module, configured to optimize the target feature extraction solution at different resolution states based on the object-image difference features to obtain an exclusive ground object information extraction solution for water and soil conservation;

[0043] A joint interpretation module, configured to perform ground object recognition on the target remote sensing image based on the exclusive ground object information extraction solution, and compare the recognition result with visual interpretation to generate a final interpretation of the relevant typical ground object information for water and soil conservation corresponding to the target remote sensing image.

[0044] Preferably, in a system for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, the image feature extraction module includes:

[0045] A label adding unit, configured to generate an image label according to the resolution corresponding to each labeled remote sensing image and add it to each labeled remote sensing image;

[0046] An image grouping unit, configured to group the labeled remote sensing images based on the water and soil conservation targets captured in the images to obtain multiple remote sensing image groups;

[0047] A feature extraction unit, configured to perform extraction of relevant typical ground object features on the relevant typical ground objects within the labeled frames of each remote sensing image group based on a preset feature extraction model to obtain image features.

[0048] Preferably, in a system for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, the object-image difference comparison module includes:

[0049] An image content determination unit, configured to determine the target name and shooting area of the relevant typical ground object marked corresponding to each marked frame on the image based on the image interpretation corresponding to the labeled remote sensing image;

[0050] A data acquisition unit, configured to acquire the object-image and its object-image features of the relevant typical ground object marked in the real scene from a physical database based on the shooting area and the target name;

[0051] A feature comparison unit for comparing the object image features with the image features corresponding to each remote sensing image respectively to obtain the image-object differences corresponding to the remote sensing images;

[0052] A result output unit for comparing the image-object differences corresponding to different remote sensing images based on relevant typical land cover types to obtain the object-image difference features of the same relevant typical land cover at different resolutions.

[0053] Compared with the prior art, the present invention has at least the following beneficial effects:

[0054] The present invention first obtains a large number of labeled remote sensing images with relevant typical land cover marked with water and soil conservation at different resolutions, and extracts the relevant typical land cover features from the labeled remote sensing images to obtain image features, realizing the analysis of the image differences of relevant typical land cover at different resolutions. Then, the image features are compared with the object image features of relevant typical land cover with water and soil conservation in the real scene to determine the object-image difference features at different resolutions, realizing the difference comparison between remote sensing images and actual object images, and providing a basis for the optimization of the target feature extraction scheme for water and soil conservation. Subsequently, based on the object-image difference features, the target feature extraction scheme at different resolution states is optimized to obtain a dedicated land cover information extraction scheme for water and soil conservation, realizing the intelligent optimization of the method for identifying relevant typical land cover features in remote sensing images, obtaining a dedicated land cover information extraction scheme for water and soil conservation, which is beneficial to improving the accuracy and precision of water and soil conservation monitoring, facilitating the timely discovery of environmental pollution incidents, reducing the probability of the expansion of the impact range of pollution incidents. At the same time, the land cover in the target remote sensing image will also be identified based on the dedicated land cover information extraction scheme, and the recognition result will be compared with visual interpretation to generate the final interpretation of the relevant typical land cover information of water and soil conservation corresponding to the target remote sensing image, realizing the joint application of the automatic land cover recognition method and the human-computer interaction visual interpretation recognition method, which can solve the problems of false extraction and missed extraction in the automatic extraction algorithm to a certain extent and achieve high-precision interpretation of relevant typical land cover information of water and soil conservation.

[0055] Other features and advantages of the present invention will be described in the following specification, and some of them will become obvious from the specification or be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0056] The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0057] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, but do not constitute a limitation to the present invention. In the accompanying drawings:

[0058] Figure 1 This is a flowchart of a method for analyzing and extracting the characteristics of water and environmental protection targets in remote sensing images according to the present invention;

[0059] Figure 2 This is a flowchart of step 1 of a method for analyzing and extracting the characteristics of water and environmental protection targets in remote sensing images according to the present invention;

[0060] Figure 3 This is a flowchart of step 2 of a method for analyzing and extracting the characteristics of water and environmental protection targets in remote sensing images according to the present invention;

[0061] Figure 4 This is a flowchart of step 3 of a method for analyzing and extracting the characteristics of water and environmental protection targets in remote sensing images according to the present invention;

[0062] Figure 5 This is a flowchart of step 4 of a method for analyzing and extracting the characteristics of water and environmental protection targets in remote sensing images according to the present invention;

[0063] Figure 6 This is a structural diagram of a system for analyzing and extracting the characteristics of water and environmental protection targets in remote sensing images according to the present invention. Specific embodiments

[0064] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not intended to limit the present invention.

[0065] According to an embodiment of the present invention, a method for analyzing and extracting the characteristics of water and environmental protection targets in remote sensing images is as Figure 1 shown, and includes:

[0066] Step 1: Obtain a large number of labeled remote sensing images with relevant typical ground object markers of water and environmental protection at different resolutions, and extract the relevant typical ground object features from the labeled remote sensing images to obtain image features;

[0067] Step 2: Compare the image features with the object-image features of relevant typical ground objects of water and environmental protection in the real scene to determine the object-image difference features at different resolutions;

[0068] Step 3: Optimize the target feature extraction scheme at different resolution states based on the object-image difference features to obtain an exclusive ground object information extraction scheme for water and environmental protection;

[0069] Step 4: Identify the ground objects in the target remote sensing image based on the exclusive ground object information extraction scheme, and compare the recognition results with visual interpretation to generate the final interpretation of the relevant typical ground object information of water and environmental protection corresponding to the target remote sensing image.

[0070] Beneficial effects of the above technical solution: First, the present invention obtains a large number of marked remote sensing images with relevant typical ground objects of water and soil conservation at different resolutions, and extracts relevant typical ground object features from the marked remote sensing images to obtain image features, realizing the analysis of the image differences of relevant typical ground objects at different resolutions. Then, the image features are compared with the object-image features of relevant typical ground objects of water and soil conservation in the real scene to determine the object-image difference features at different resolutions, realizing the difference comparison between remote sensing images and actual object-images, providing a basis for optimizing the target feature extraction scheme of water and soil conservation. Subsequently, based on the object-image difference features, the target feature extraction scheme at different resolution states is optimized to obtain a dedicated ground object information extraction scheme for water and soil conservation, realizing the intelligent optimization of the method for identifying relevant typical ground object features in remote sensing images, obtaining a dedicated ground object information extraction scheme for water and soil conservation, which is beneficial to improving the accuracy and precision of water and soil conservation monitoring, facilitating the timely discovery of environmental pollution incidents, reducing the probability of the spread of the impact range of pollution incidents. At the same time, the ground objects of the target remote sensing image will also be identified based on the dedicated ground object information extraction scheme, and the recognition results will be compared with visual interpretation to generate the final interpretation of the relevant typical ground object information of water and soil conservation corresponding to the target remote sensing image, realizing the combined application of the ground object automatic recognition method and the human-computer interaction visual interpretation recognition method, which can solve the problems of false extraction and missed extraction in the automatic extraction algorithm to a certain extent and achieve high-precision interpretation of relevant typical ground object information of water and soil conservation.

[0071] According to an embodiment of the present invention, a method for analyzing and extracting the water and soil conservation target features of remote sensing images, step 1: Obtain a large number of marked remote sensing images with relevant typical ground objects of water and soil conservation at different resolutions, and extract relevant typical ground object features from the marked remote sensing images to obtain image features, as Figure 2 shown, including:

[0072] Step 101: Generate image labels according to the resolutions corresponding to the respective marked remote sensing images and add them to the respective marked remote sensing images;

[0073] Step 102: Group the marked remote sensing images based on the water and soil conservation targets photographed in the images to obtain a plurality of remote sensing image groups;

[0074] Step 103: Based on a preset feature extraction model, extract relevant typical ground object features from the relevant typical ground objects within the marked frames of each remote sensing image group to obtain image features.

[0075] In this embodiment, the water and soil conservation target refers to the main water and soil conservation area photographed by the marked remote sensing image, such as a certain area of a certain river, a certain forest or a certain area of the forest, etc.

[0076] Beneficial effects of the above technical solution: First, the present invention generates image tags according to the resolutions corresponding to each marked remote sensing image and adds them to each marked remote sensing image. Then, based on the water and soil conservation targets captured in the images, the marked remote sensing images are grouped to obtain multiple remote sensing image groups, realizing the automatic classification of remote sensing images and providing a basis for subsequent data comparison. Finally, based on a preset feature extraction model, relevant typical ground objects within the marked frames on each remote sensing image group are subjected to feature extraction of relevant typical ground objects to obtain image features, completing the extraction of ground object features on the remote sensing images.

[0077] According to an embodiment of the present invention, a method for analyzing and extracting water and soil conservation target features of remote sensing images, step 2: Compare the image features with the object image features of relevant typical ground objects of water and soil conservation in the real scene to determine the object image difference features at different resolutions, such as Figure 3 shown, including:

[0078] Step 201: Based on the image interpretation corresponding to the marked remote sensing image, determine the target name and shooting area of the relevant typical ground object marked corresponding to each marked frame on the image;

[0079] Step 202: Based on the shooting area and target name, obtain the object images and their object image features of the relevant typical ground object marked in the real scene from the physical database;

[0080] Step 203: Compare the object image features with the image features corresponding to each remote sensing image respectively to obtain the image object image difference corresponding to the remote sensing image;

[0081] Step 204: Based on the relevant typical ground object types, compare the image object image differences corresponding to different remote sensing images to obtain the object image difference features of the same relevant typical ground object at different resolutions.

[0082] Beneficial effects of the above technical solution: First, based on the image interpretation corresponding to the marked remote sensing image, the present invention determines the target name and shooting area of the relevant typical ground object marked corresponding to each marked frame on the remote sensing image, performs a one-to-one correspondence processing of the target name and shooting area for the relevant typical ground objects on the remote sensing image, provides a basis for the subsequent comparison of the image features and object image features of the relevant typical ground objects, and ensures an accurate one-to-one correspondence between the two. Then, the object image features are compared with the image features corresponding to each remote sensing image respectively to obtain the image object image difference corresponding to the remote sensing image, obtaining the difference between the remote sensing image and the ground object in kind. Finally, based on the relevant typical ground object types, the image object image differences corresponding to different remote sensing images are compared to obtain the object image difference features of the same relevant typical ground object at different resolutions, obtaining the image performance differences of different relevant typical ground objects at different resolutions, providing a basis for the optimization of the water and soil conservation target feature extraction scheme.

[0083] According to an embodiment of the present invention, a method for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images is provided. Based on relevant typical land cover types, the image-object differences corresponding to different remote sensing images are compared to obtain the object-image difference characteristics of the same relevant typical land cover at different resolutions, including:

[0084] Based on the terrain characteristics of the water and soil conservation targets corresponding to each remote sensing group, the remote sensing groups are classified by geological type to obtain image clusters of water and soil conservation projects corresponding to different terrain environments;

[0085] Based on the relevant typical land cover markers in each remote sensing image, a relevant typical land cover catalog is established. Based on the relevant typical land cover catalog, the images in the image clusters of water and soil conservation projects are screened to establish multiple same-object image sets;

[0086] The remote sensing images in the same-object image sets are clustered based on the resolution to obtain the same-object same-resolution image sets. The image-object differences of the same relevant typical land cover in the same-object same-resolution image sets are compared to obtain the first resolution influence characteristic differences;

[0087] Based on the types of relevant typical land cover, the same-object image sets of different image clusters of water and soil conservation projects are screened to obtain the same-object image control groups. The image-object differences of the remote sensing images with the same different resolutions in the same-object image control groups are compared to obtain the terrain influence characteristic differences;

[0088] And the terrain influence characteristic differences corresponding to different resolutions in the same-object image control groups are compared to obtain the second resolution influence characteristic differences;

[0089] Based on the first resolution influence characteristic differences and the second resolution influence characteristic differences, the object-image difference characteristics of the relevant typical land cover are generated.

[0090] Beneficial effects of the above technical solution: First, based on the topographic features of the water and soil conservation targets corresponding to each remote sensing group, the invention classifies the remote sensing groups by geological type to obtain image clusters of water and soil conservation projects corresponding to different topographic environments. Then, based on the relevant typical ground object markers in each remote sensing image, a relevant typical ground object catalog is established. Based on the relevant typical ground object catalog, the images in the image clusters of water and soil conservation projects are screened to establish multiple same-ground-object image sets. The remote sensing images in the same-ground-object image sets are clustered by resolution to obtain the same-object-and-same-resolution image sets. The image-object differences of the images of the same relevant typical ground object in the same-object-and-same-resolution image sets are compared to obtain the first resolution influence feature differences. Then, based on the types of relevant typical ground objects, the same-ground-object image sets of different water and soil conservation project image clusters are screened to obtain the same-object image control group. The image-object differences of the remote sensing images with different resolutions in the same-object image control group are compared to obtain the topographic influence feature differences. The topographic influence feature differences corresponding to different resolutions in the same-object image control group are compared to obtain the second resolution influence feature differences. Finally, based on the first resolution influence feature differences and the second resolution influence feature differences, the image-object difference features of relevant typical ground objects are generated. While analyzing the image differences of the same ground object at different resolutions, the image differences of the same ground object under different topographies are also analyzed, and the image differences affected by topography are combined with the resolution influence. Then, according to the combined influence of topography and resolution and the individual influence results of resolution, the image-object difference features of a certain relevant typical ground object are generated, providing a basis for optimizing the water and soil conservation target feature extraction scheme and facilitating the accuracy of feature extraction of relevant typical ground objects at different resolutions and in different environments.

[0091] According to an embodiment of the present invention, a method for analyzing and extracting water and soil conservation target features of remote sensing images, step 3: Optimize the target feature extraction scheme in different resolution states based on the image-object difference features to obtain a dedicated ground object information extraction scheme for water and soil conservation, as Figure 4 shown, including:

[0092] Step 301: Generate an optimization reference table based on the image-object difference features of relevant typical ground objects in combination with the relevant typical ground object catalog;

[0093] Step 302: Based on the optimization reference table, obtain a feature set constructed from the image features of all relevant typical ground objects at different resolutions, and add a resolution label to the feature set;

[0094] Step 303: Optimize the feature extraction process of the current target feature extraction scheme based on the feature set to obtain a dedicated ground object information extraction scheme for water and soil conservation.

[0095] Advantages of the above technical solution: By using the image difference features of relevant typical ground objects and combining with the relevant typical ground object catalog, the present invention generates an optimized reference table. Then, based on the image resolution and the optimized reference table, it obtains the image difference features of different relevant typical ground objects in the same image, establishes an image object difference set, combines it with the on-site physical object difference feature set, determines the image features of all relevant typical ground objects on remote sensing images with different resolutions. Subsequently, based on the feature set constructed from the image features of all relevant typical ground objects with different resolutions, a resolution label is generated and added to the feature set. Finally, based on the feature set, the feature extraction process of the current target feature extraction scheme is optimized to obtain a dedicated ground object information extraction scheme for environmental protection related to water, realizing the intelligent optimization of the method for identifying the features of relevant typical ground objects in remote sensing images, obtaining a dedicated ground object information extraction scheme for environmental protection related to water, which is beneficial to improving the accuracy and precision of environmental protection monitoring related to water, facilitating the timely discovery of environmental pollution incidents, and reducing the probability of the expansion of the impact range of pollution incidents.

[0096] According to an embodiment of the present invention, a method for analyzing and extracting environmental protection related to water target features from remote sensing images, which optimizes the feature extraction process of the current target feature extraction scheme based on a feature set to obtain a dedicated ground object information extraction scheme for environmental protection related to water, includes:

[0097] Based on the feature difference between the image difference feature corresponding to each relevant typical ground object in the feature set and the best feature extraction result identified by the current target feature extraction scheme at the resolution corresponding to the current feature set and the target image feature corresponding to the actual ground object, the feature extraction detail difference is obtained.

[0098] Based on the feature extraction detail difference, the reference feature of the current target feature extraction scheme is corrected to obtain the best reference feature, and the current target feature extraction scheme is updated to obtain a dedicated ground object information extraction sub-scheme.

[0099] Obtain the dedicated ground object information extraction sub-schemes corresponding to the resolution intervals of all feature sets, and generate a dedicated ground object information extraction scheme.

[0100] Beneficial effects of the above technical solution: Based on the feature differences between the object image difference features corresponding to each relevant typical ground object in the feature set and the best feature extraction results identified by the current target feature extraction scheme at the resolution corresponding to the current feature set and the target object image features corresponding to the real scene ground objects, the feature extraction detail differences are obtained; based on the feature extraction detail differences, the reference features of the current target feature extraction scheme are corrected to obtain the best reference features, and the current target feature extraction scheme is updated to obtain a dedicated ground object information extraction sub-scheme; the dedicated ground object information extraction sub-schemes corresponding to the resolution intervals of all feature sets are obtained, and a dedicated ground object information extraction scheme is generated, realizing the dedicated customization of the ground object recognition feature extraction of remote sensing images with different resolutions. Each resolution of remote sensing image has its corresponding target feature scheme, which can more precisely identify the ground objects in the image, effectively improving the ground object recognition ability of low-resolution remote sensing images while enhancing the ground object recognition ability of high-resolution remote sensing images, and effectively improving the accuracy of ground object recognition in remote sensing images.

[0101] According to an embodiment of the present invention, a method for analyzing and extracting the target features of environmental protection related to water in remote sensing images, step 4: Based on the dedicated ground object information extraction scheme, perform ground object recognition on the target remote sensing image, and compare the recognition result with the visual interpretation to generate the final interpretation of the relevant typical ground object information of the environmental protection related to water corresponding to the target remote sensing image, as Figure 5 shown, including:

[0102] Step 401: Identify the resolution label of the reference remote sensing image, and adjust the dedicated ground object information extraction scheme to the optimal state according to the label recognition result and the feature extraction result;

[0103] Step 402: Based on the dedicated ground object information extraction scheme in the optimal state and the feature extraction result, determine the names of relevant typical ground objects;

[0104] Step 403: Compare the names of relevant typical ground objects with the names of ground objects in the visual interpretation to determine whether there are omissions or errors in the visual interpretation;

[0105] If so, modify the omitted names or incorrect names in the visual interpretation based on the names of relevant typical ground objects to obtain the final interpretation of the relevant typical ground object information of the environmental protection related to water corresponding to the target remote sensing image;

[0106] At the same time, a modified comparison interpretation is also generated based on the modified content of the visual interpretation and sent to the display end for comparison display.

[0107] In this embodiment, adjusting the dedicated ground object information extraction scheme to the optimal state means automatically switching the target feature extraction scheme to the dedicated ground object information extraction sub-scheme applicable to the target remote sensing image.

[0108] Beneficial effects of the above technical solution: First, identify the resolution labels of the remote sensing images, and adjust the exclusive feature extraction scheme of the ground object information to the optimal state according to the label recognition results, effectively improving the accuracy of ground object recognition in the target remote sensing images. Then, perform ground object recognition on the target remote sensing images based on the exclusive ground object information extraction scheme, and compare the recognition results with visual interpretation to generate the final interpretation of the relevant typical ground object information of the water and soil conservation corresponding to the target remote sensing images. The combined application of the automatic ground object recognition and the human-computer interaction visual interpretation recognition method can solve the problems of false extraction and missed extraction in the automatic extraction algorithm to a certain extent, and achieve high-precision interpretation of the relevant typical ground object information of the water and soil conservation.

[0109] The present invention provides a system for analyzing and extracting the characteristics of water and soil conservation targets in remote sensing images, as Figure 6 shown, including:

[0110] An image feature extraction module, configured to obtain a large number of labeled remote sensing images with different resolutions and having relevant typical ground object markings of water and soil conservation, and perform relevant typical ground object feature extraction on the labeled remote sensing images to obtain image features;

[0111] An object-image difference comparison module, configured to compare the image features with the object-image features of the relevant typical ground objects of water and soil conservation in the real scene to determine the object-image difference features at different resolutions;

[0112] An exclusive scheme generation module, configured to optimize the target feature extraction scheme in different resolution states based on the object-image difference features to obtain an exclusive ground object information extraction scheme for water and soil conservation;

[0113] A combined interpretation module, configured to perform ground object recognition on the target remote sensing images based on the exclusive ground object information extraction scheme, and compare the recognition results with visual interpretation to generate the final interpretation of the relevant typical ground object information of the water and soil conservation corresponding to the target remote sensing images.

[0114] Beneficial effects of the above technical solution: The present invention obtains a large number of labeled remote sensing images with relevant typical ground objects of water and soil conservation with different resolutions through the image feature extraction module, and extracts relevant typical ground object features from the labeled remote sensing images to obtain image features, realizing the analysis of the image differences of relevant typical ground objects at different resolutions. Then, through the object-image difference comparison module, the image features are compared with the object-image features of relevant typical ground objects of water and soil conservation in the real scene to determine the object-image difference features at different resolutions, realizing the difference comparison between remote sensing images and actual object-images, providing a basis for the optimization of the target feature extraction scheme for water and soil conservation. Subsequently, through the exclusive scheme generation module, the target feature extraction scheme under different resolution states is optimized based on the object-image difference features to obtain an exclusive ground object information extraction scheme for water and soil conservation, realizing the intelligent optimization of the method for identifying relevant typical ground object features in remote sensing images, obtaining an exclusive ground object information extraction scheme for water and soil conservation, which is beneficial to improving the accuracy and precision of water and soil conservation monitoring, facilitating the timely discovery of environmental pollution incidents, reducing the probability of the expansion of the impact range of pollution incidents. At the same time, through the joint interpretation module, the ground objects of the target remote sensing image are identified based on the exclusive ground object information extraction scheme, and the recognition results are compared with visual interpretation to generate the final interpretation of the relevant typical ground object information of water and soil conservation corresponding to the target remote sensing image, realizing the joint application of the ground object automatic recognition method and the human-computer interaction visual interpretation recognition method, which can solve the problems of false extraction and missed extraction in the automatic extraction algorithm to a certain extent and achieve high-precision interpretation of relevant typical ground object information of water and soil conservation.

[0115] According to an embodiment of the present invention, a system for analyzing and extracting target features of water and soil conservation in remote sensing images, the image feature extraction module includes:

[0116] A label adding unit for generating image labels according to the resolutions corresponding to each labeled remote sensing image and adding them to each labeled remote sensing image;

[0117] An image grouping unit for grouping the labeled remote sensing images based on the water and soil conservation targets captured in the images to obtain multiple remote sensing image groups;

[0118] A feature extraction unit for extracting relevant typical ground object features from the relevant typical ground objects within the labeled frames of each remote sensing image group based on a preset feature extraction model to obtain image features.

[0119] Beneficial effects of the above technical solution: First, the present invention generates image tags according to the resolution corresponding to each marked remote sensing image through the tag adding unit and adds them to each marked remote sensing image. Then, the image grouping unit groups the marked remote sensing images based on the water and soil conservation targets captured in the images to obtain multiple remote sensing image groups, realizing the automatic classification of remote sensing images and providing a basis for subsequent data comparison. Finally, the feature extraction unit extracts the features of relevant typical ground objects within the marked frames on each remote sensing image group based on a preset feature extraction model to obtain image features, completing the extraction of ground object features on the remote sensing images.

[0120] According to an embodiment of the present invention, a system for analyzing and extracting water and soil conservation target features of remote sensing images, the object-image difference comparison module includes:

[0121] An image content determination unit for determining, based on the image interpretation corresponding to the marked remote sensing image, the target name and the shooting area of the relevant typical ground object marked corresponding to each marked frame on the image;

[0122] A data acquisition unit for acquiring the object image and its object image features of the relevant typical ground object marked in the real scene in the physical database based on the shooting area and the target name;

[0123] A feature comparison unit for comparing the object image features with the image features corresponding to each remote sensing image respectively to obtain the image-object difference corresponding to the remote sensing image;

[0124] A result output unit for comparing the image-object differences corresponding to different remote sensing images based on the relevant typical ground object type to obtain the object-image difference features of the same relevant typical ground object at different resolutions.

[0125] Beneficial effects of the above technical solution: First, the image content determination unit of the present invention determines the target name and shooting area of the relevant typical ground objects corresponding to each marking box on the image based on the image interpretation of the marked remote sensing image, and performs a one-to-one correspondence processing of the target name and shooting area for the relevant typical ground objects on the remote sensing image, providing a basis for the subsequent comparison of the image features and object-image features of the relevant typical ground objects to ensure their accurate one-to-one correspondence. Then, through the data acquisition unit and the feature comparison unit, the object-image features are acquired and compared with the image features corresponding to each remote sensing image respectively to obtain the image-object differences corresponding to the remote sensing images, and the differences between the remote sensing images and the ground objects are obtained. Finally, the result output unit compares the image-object differences corresponding to different remote sensing images based on the relevant typical ground object types to obtain the object-image difference features of the same relevant typical ground object at different resolutions and the image performance differences of different relevant typical ground objects at different resolutions, providing a basis for the optimization of the target feature extraction scheme for environmental protection around water bodies.

[0126] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A method for analyzing and extracting characteristics of environmental and water conservation targets from remote sensing images, characterized in that: include: Obtain a large number of marked remote sensing images with different resolutions and related typical land object marks for water and environmental protection, and extract relevant typical land object features from the marked remote sensing images to obtain image features; Compare the image features with the image features of typical objects related to water and soil conservation in real scenes to determine the difference features of the images at different resolutions; Based on the difference characteristics of objects and images, the target feature extraction schemes under different resolution states are optimized to obtain a unique ground feature information extraction scheme for environmental and water conservation; Based on the exclusive ground object information extraction scheme, the target remote sensing image is identified, and the identification results are compared with the visual interpretation to generate the final interpretation of the typical ground object information related to environmental and water conservation corresponding to the target remote sensing image; Among them, based on the difference characteristics of objects and images, the target feature extraction schemes under different resolution states are optimized to obtain the exclusive ground feature information extraction scheme for environmental and water conservation, including: Based on the difference characteristics of the images of relevant typical objects and in combination with the catalog of relevant typical objects, an optimized reference table is generated; Based on the optimized reference table, a feature set constructed by image features of all relevant typical objects at different resolutions is obtained, and a resolution label is generated and added to the feature set; Based on the feature set, the feature extraction process of the current target feature extraction scheme is optimized to obtain a unique ground feature information extraction scheme for environmental and water conservation; Among them, the feature extraction process of the current target feature extraction scheme is optimized based on the feature set to obtain an exclusive ground feature information extraction scheme for environmental and water conservation, including: Based on the object image difference feature corresponding to each relevant typical object in the feature set, the feature difference between the best feature extraction result identified by the current target feature extraction scheme at the resolution corresponding to the current feature set and the target object image feature corresponding to the real scene object is compared to obtain the feature extraction detail difference; Based on the differences in feature extraction details, the reference features of the current target feature extraction scheme are corrected to obtain the best reference features, and the current target feature extraction scheme is updated to obtain a dedicated ground feature information extraction sub-scheme; Obtain the exclusive ground feature information extraction sub-scheme corresponding to the resolution interval corresponding to all feature sets, and generate an exclusive ground feature information extraction scheme.

2. The method for analyzing and extracting environmental and water conservation target features of remote sensing images according to claim 1 is characterized in that: Obtain a large number of marked remote sensing images with different resolutions and related typical land features of environmental and water conservation, and extract relevant typical land features from the marked remote sensing images to obtain image features, including: Generate image labels according to the resolutions corresponding to each labeled remote sensing image, and add them to each labeled remote sensing image; Based on the environmental and water conservation targets captured in the images, the marked remote sensing images are grouped to obtain a plurality of remote sensing image groups; Based on the preset feature extraction model, the features of relevant typical objects in the marked boxes on each remote sensing image group are extracted to obtain image features.

3. The method for analyzing and extracting environmental and water conservation target features of remote sensing images according to claim 1 is characterized in that: Compare the image features with the image features of typical objects related to water and environmental protection in real scenes to determine the difference features of the images at different resolutions, including: Based on the image interpretation corresponding to the marked remote sensing image, the mark is to be determined according to the target name and shooting area of ​​the marked relevant typical landform corresponding to each marked frame on the image; Based on the shooting area and the target name, the image and image features of the marked relevant typical objects in the real scene are obtained in the physical database; Comparing the object image features with the image features corresponding to each remote sensing image, respectively, to obtain the image object image difference corresponding to the remote sensing image; Based on the types of relevant typical objects, the image differences corresponding to different remote sensing images are compared to obtain the image difference characteristics of the same relevant typical object at different resolutions.

4. The method for analyzing and extracting environmental and water conservation target features of remote sensing images according to claim 3 is characterized in that: Based on the relevant typical land object types, the image differences corresponding to different remote sensing images are compared to obtain the image difference characteristics of the same relevant typical land object at different resolutions, including: Based on the terrain characteristics of the environmental and water conservation targets corresponding to each remote sensing group, the remote sensing groups are classified into geological types to obtain image clusters of environmental and water conservation projects corresponding to different terrain environments; Based on the relevant typical ground object tags in each remote sensing image, a relevant typical ground object catalog is established. Based on the relevant typical ground object catalog, images in the environmental and water conservation project image cluster are screened to establish multiple sets of images with the same ground object. Based on the resolution, the remote sensing images in the image set of the same ground object are clustered to obtain the image set of the same object and the same resolution, and the image differences of the same related typical ground objects in the image set of the same object and the same resolution are compared to obtain the difference of the first resolution-affecting feature; Based on the relevant typical landform types, the same landform image sets of different environmental and water conservation project image clusters were screened to obtain the same-object image control group. The image differences of remote sensing images with the same and different resolutions in the same-object image control group were compared to obtain the differences in terrain impact characteristics. And the differences of terrain influence features corresponding to different resolutions in the control group of the same object image are compared to obtain the differences of the second resolution influence features; Based on the first resolution-affecting feature difference and the second resolution-affecting feature difference, an object-image difference feature of the relevant typical ground object is generated.

5. The method for analyzing and extracting environmental and water conservation target features of remote sensing images according to claim 1 is characterized in that: Based on the exclusive ground object information extraction scheme, the target remote sensing image is identified, and the identification results are compared with the visual interpretation to generate the final interpretation of the typical ground object information related to environmental and water conservation corresponding to the target remote sensing image, including: Identify the resolution labels of the remote sensing images, and adjust the exclusive ground feature information extraction scheme to the best state based on the feature extraction results of the label identification results; Based on the best state of exclusive feature information extraction scheme, determine the names of relevant typical features based on the feature extraction results; Compare the names of relevant typical features with those in the visual interpretation to determine whether there are omissions or errors in the visual interpretation; If it exists, the omitted or incorrect names in the visual interpretation are modified based on the names of the relevant typical features to obtain the final interpretation of the relevant typical features information of environmental and water conservation corresponding to the target remote sensing image; At the same time, a modified comparison interpretation is generated based on the visual interpretation of the modified content and sent to the display end for comparison display.

6. A system for analyzing and extracting environmental and water conservation target features from remote sensing images, used to implement a method for analyzing and extracting environmental and water conservation target features from remote sensing images as described in any one of claims 1 to 5, characterized in that: include: An image feature extraction module is used to obtain a large number of marked remote sensing images with different resolutions and related typical land object marks for water and environmental protection, and extract relevant typical land object features from the marked remote sensing images to obtain image features; The object-image difference comparison module is used to compare the image features with the image features of typical objects related to environmental and water conservation in real scenes, and determine the image difference features at different resolutions; The exclusive solution generation module is used to optimize the target feature extraction schemes under different resolutions based on the difference characteristics of the object and image, and obtain the exclusive ground object information extraction scheme for environmental and water conservation; The joint interpretation module is used to identify the objects in the target remote sensing image based on a dedicated object information extraction scheme, and compare the identification results with the visual interpretation to generate the final interpretation of the typical object information related to water and environmental protection corresponding to the target remote sensing image.

7. The system for analyzing and extracting environmental and water conservation target features from remote sensing images according to claim 6 is characterized in that: Image feature extraction module, including: A label adding unit, used for generating image labels according to the resolution corresponding to each labeled remote sensing image, and adding the labels to each labeled remote sensing image; An image grouping unit, used to group the marked remote sensing images based on the environmental and water conservation targets captured in the images to obtain a plurality of remote sensing image groups; The feature extraction unit is used to extract the features of the relevant typical objects in the marking frame on each remote sensing image group based on a preset feature extraction model to obtain image features.

8. The system for analyzing and extracting environmental and water conservation target features from remote sensing images according to claim 6 is characterized in that: Object-image difference comparison module, including: An image content determination unit, for determining, based on the image interpretation corresponding to the marked remote sensing image, the target name and the shooting area of ​​the marked relevant typical landform corresponding to each marked frame on the image; A data acquisition unit, used to acquire the image and image features of the marked relevant typical ground objects in the real scene from the physical object database based on the shooting area and the target name; A feature comparison unit, used to compare the object image feature with the image features corresponding to each remote sensing image, to obtain the image object image difference corresponding to the remote sensing image; The result output unit is used to compare the image differences corresponding to different remote sensing images based on the relevant typical land object types, and obtain the image difference characteristics of the same relevant typical land object at different resolutions.

Images