Multi-period remote sensing image automatic change detection processing method and device

Abstract

The invention relates to a multi-period remote sensing image automatic change detection processing method and device, and relates to the technical field of surveying and mapping geographic information, and the method comprises the steps: obtaining multi-period remote sensing images including a fixed base-period remote sensing image, a post-time-phase remote sensing image and a rolling base-period remote sensing image, carrying out the change extraction of the multi-period remote sensing images, and obtaining the change of the multi-period remote sensing images; the method comprises the following steps: acquiring a fixed base-period pattern spot and a rolling base-period pattern spot, and on the basis of the base-period pattern spot, performing de-duplication processing in combination with a previous pattern spot to obtain a fixed base-period pattern spot j and a target rolling base-period pattern spot, so as to perform normalized monitoring by using the extracted fixed base-period pattern spot j and the target rolling base-period pattern spot. Therefore, the suspected change pattern spots in the target range can be rapidly and accurately extracted, redundant checking work caused by repeated pattern spots extracted by multi-phase images is avoided, the requirement for rapid response of normalized monitoring is met, the change detection recall ratio is increased, and the detection efficiency is improved. And active, intelligent and comprehensive natural resource investigation and monitoring basic data is provided for natural resource investigation, monitoring and management.

Description

Technical Field

[0001] The present application relates to the field of surveying and mapping geographic information technology, and in particular to a method and device for automatic change detection and processing of multi-period remote sensing images. Background Art

[0002] In the past, surveys and monitoring of natural resources such as land mostly used remote sensing images to identify geographic elements, while the identification of changes in multi-period remote sensing images was done by "human sea tactics", and the change patches were extracted manually by going through the map one by one. There were problems such as low business collaboration, backward technical means, long monitoring cycle, and low efficiency, which could not meet the needs of natural resource business management such as farmland protection, "early detection, early warning, and early prevention" of illegal land use, and ecological protection and restoration. Under the new situation and requirements, the natural resource management business has higher requirements for survey and monitoring accuracy, more frequent requirements, and wider monitoring objects. Efficiency is the key to timely grasp the development and utilization of natural resources and support the full business and full process monitoring and supervision of natural resources.

[0003] In natural resource survey and monitoring, the existing remote sensing image change detection method mainly detects and extracts change spots by identifying and comparing fixed base period remote sensing images with the latest acquired remote sensing images. However, the base period images selected for fixed base period extraction are the change images of the previous year. Multi-source and multi-temporal change detection is prone to repeated extraction of spots, resulting in large workload and low efficiency, and cannot meet the requirements of rapid response of normalized monitoring. Summary of the invention

[0004] The present application provides a method and device for automatic change detection and processing of multi-phase remote sensing images, which utilizes multi-phase remote sensing images to quickly and accurately extract change spots within the target range, realize change detection of surface coverage, and obtain changes in important geographic elements. Routine monitoring can then be carried out to avoid redundant verification work caused by repeated spots extracted from multi-phase images, meet the requirements of rapid response in routine monitoring, and solve the problems of heavy workload and low efficiency caused by repeated extraction of spots in the prior art.

[0005] In a first aspect, the present application provides a multi-period remote sensing image automatic change detection and processing method, comprising:

[0006] Obtain fixed base period remote sensing images, rolling base period remote sensing images and the latest monitored post-phase remote sensing images of the target area;

[0007] Performing change extraction based on the fixed base period remote sensing image and the later phase remote sensing image to obtain a fixed base period image patch a, and performing change extraction based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a rolling base period image patch n;

[0008] According to the fixed base period patch a, combined with the previously verified patches, duplicate removal is performed to obtain a fixed base period patch j;

[0009] According to the rolling base period image patch n, the fixed base period image patch j and the previous online image patches are combined to perform deduplication processing to obtain the target rolling base period image patch;

[0010] Performing normalized monitoring of the target area according to the fixed base period pattern j and the target rolling base period pattern;

[0011] Among them, the fixed base period pattern j and the target rolling base period pattern are both base period change patterns.

[0012] Optionally, obtain fixed base period remote sensing images, rolling base period remote sensing images, and the latest monitored post-phase remote sensing images of the target area, including:

[0013] Extracting land change survey images of the target area from the preset land change survey database as fixed base period remote sensing images;

[0014] The initial post-phase images of the previous batch are obtained as the rolling base period remote sensing images, and the latest monitoring images are obtained as the post-phase remote sensing images.

[0015] Optionally, performing change extraction based on the rolling base period remote sensing image and the post-phase remote sensing image to obtain the rolling base period image patch n includes:

[0016] Image selection is performed based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a preferred monitoring image, wherein the preferred monitoring image is a monitoring image whose latest monitoring image phase is better than the monitoring image phase of the previous monitoring image by a first preset time;

[0017] Spot extraction is performed based on the preferred monitoring image and the rolling base period remote sensing image to obtain the rolling base period spot n.

[0018] Optionally, the fixed base period spot a is combined with the previously verified spots for deduplication to obtain the fixed base period spot j, including:

[0019] Obtain previously verified image patches and extract the change investigation scope from the preset change investigation database;

[0020] Extracting the fixed base period map b outside the scope of the change investigation from the fixed base period map a;

[0021] Extracting an image range D of the later phase remote sensing image whose phase is superior to the rolling base period remote sensing image by at least a second preset time;

[0022] According to the image range D, extracting the fixed base period spot c and the fixed base period spot d from the fixed base period spot b;

[0023] Based on the previously verified patches, the fixed base patch c and the fixed base patch d, deduplication processing is performed to obtain the fixed base patch j.

[0024] Optionally, extracting the fixed base period map spot b outside the change investigation scope from the fixed base period map spot a includes:

[0025] Through a preset semantic segmentation model, an intelligent remote sensing interpretation algorithm is used to perform full-element land class semantic segmentation on the fixed base period patch a to obtain image land class information, wherein the image land class information includes image land class and image land class range;

[0026] The image land classification information is compared with the changed survey scope, and the fixed base period patches b outside the changed survey scope are extracted from the fixed base period patches a.

[0027] Optionally, deduplication processing is performed on the previously verified patches, the fixed base patch c, and the fixed base patch d to obtain the fixed base patch j, including:

[0028] Performing summary detection based on the previously verified image spots to obtain verified reference image spots;

[0029] According to the verified reference spots, the fixed base spot c and the fixed base spot d are respectively cross-compared to obtain the fixed base spot e corresponding to the fixed base spot c and the fixed base spot f corresponding to the fixed base spot d;

[0030] Extracting an image range E whose latest monitoring image phase is different from the previous monitoring image phase by a third preset time;

[0031] According to the image range E, extracting the fixed base period spot g and the fixed base period spot h from the fixed base period spot f;

[0032] Extracting the fixed base period spots i within the cultivated land range in the image range E from the fixed base period spots g;

[0033] Merge the fixed base period spots e, h and i to obtain a fixed base period spot j after deduplication;

[0034] Among them, the fixed base period spot e is a base period spot extracted from the fixed base period spot c, the area intersection ratio of which is less than the first preset threshold, and the fixed base period spot f is a base period spot extracted from the fixed base period spot d, the area intersection ratio of which is less than the second preset threshold.

[0035] Optionally, deduplication is performed on the rolling base period image patch n in combination with the fixed base period image patch j and the previous online image patches to obtain the target rolling base period image patch, including:

[0036] Perform semantic segmentation extraction based on the later-phase remote sensing image and the fixed base period patch j to obtain a fixed base period patch k and a fixed base period patch l;

[0037] Based on the summary of the previously online patches, an online reference patch is obtained, and based on the intersection comparison between the online reference patch and the fixed base patch l, a fixed base patch m is obtained;

[0038] Based on the fixed base period pattern m, the fixed base period pattern k and the rolling base period pattern n, merging and area screening are performed to obtain the target rolling base period change pattern after deduplication.

[0039] Optionally, semantic segmentation extraction is performed based on the later-phase remote sensing image and the fixed base period patch j to obtain a fixed base period patch k and a fixed base period patch l, including:

[0040] The semantic segmentation of the post-phase remote sensing image is performed by using a preset semantic segmentation model to obtain a bulldozing patch;

[0041] A fixed base period spot k intersecting with the bulldozing spot is extracted from the fixed base period spot j, and a fixed base period spot l not intersecting with the bulldozing spot is extracted from the fixed base period spot j.

[0042] Optionally, performing an intersection comparison based on the online reference pattern and the fixed base pattern l to obtain a fixed base pattern m includes:

[0043] Eliminate the easily changeable spots according to the online reference spots to obtain the target reference spots;

[0044] Calculating the area intersection ratio based on the target reference pattern and the fixed base pattern l to obtain an area intersection ratio value;

[0045] A fixed base period patch m having an area intersection ratio value less than a third preset threshold is extracted from the fixed base period patch l.

[0046] In a second aspect, the present application provides a multi-period remote sensing image automatic change detection and processing device, comprising:

[0047] Multi-period remote sensing image acquisition module, used to acquire fixed base period remote sensing images, rolling base period remote sensing images and the latest monitored post-phase remote sensing images of the target area;

[0048] A change extraction module is used to extract changes based on the fixed base period remote sensing image and the later phase remote sensing image to obtain a fixed base period image patch a, and to extract changes based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a rolling base period image patch n;

[0049] The first deduplication processing module is used to perform deduplication processing based on the fixed base period spot a and the previously verified spots to obtain the fixed base period spot j;

[0050] The second deduplication processing module is used to perform deduplication processing based on the rolling base period image patch n, in combination with the fixed base period image patch j and the previous online image patches, to obtain the target rolling base period image patch;

[0051] A normalized monitoring module, used for performing normalized monitoring on the target area according to the fixed base period pattern j and the target rolling base period pattern;

[0052] Among them, the fixed base period pattern j and the target rolling base period pattern are both base period change patterns.

[0053] In summary, the embodiment of the present application obtains multi-period remote sensing images including fixed base period remote sensing images, the latest monitored post-phase remote sensing images, and rolling base period remote sensing images, and then performs change extraction on the multi-period remote sensing images to obtain fixed base period spots and rolling base period spots. On the basis of the extracted base period spots, the spots of the previous periods are combined for deduplication processing to obtain fixed base period spots j and target rolling base period spots, so as to use the extracted fixed base period spots j and target rolling base period spots for normalized monitoring. This embodiment uses multi-phase remote sensing images to quickly and accurately extract target The suspected changed spots within the range can be detected to realize the change detection of surface cover, and the change of important geographical elements can be obtained. It can timely and accurately grasp the current status of natural resource utilization, avoid redundant verification work caused by repeated spots extracted from multiple images, meet the requirements of rapid response of routine monitoring, and improve the completeness of change detection, so as to provide active, intelligent and comprehensive natural resource survey and monitoring basic data for natural resource investigation, monitoring and management. On this basis, the national land survey results can be continuously monitored and updated to solve the problems of heavy workload and low efficiency caused by repeated extraction of spots in existing technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0056] Figure 1 A flowchart of a multi-period remote sensing image automatic change detection and processing method provided in an embodiment of the present application;

[0057] Figure 2 It is a schematic diagram of the steps of a multi-period remote sensing image automatic change detection and processing method provided by an optional embodiment of the present application;

[0058] Figure 3 It is a technical circuit diagram of a multi-period remote sensing image automatic change detection and processing method provided by an optional example of the present application;

[0059] Figure 4 It is a flow chart of spot extraction provided by an optional example of this application;

[0060] Figure 5 It is a diagram of the use of a "phase optimal selection" tool provided by an optional example of this application;

[0061] Figure 6 This is a fixed base period image spot deduplication flow chart provided by an optional example of this application;

[0062] Figure 7 It is a rolling base period image spot deduplication flow chart provided by an optional example of the present application;

[0063] Figure 8 This is a diagram of using a "deduplication model" tool provided by an optional example of this application;

[0064] Fig. 9 A structural block diagram of a multi-period remote sensing image automatic change detection and processing device provided in an embodiment of the present application;

[0065] Fig.10 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0067] To facilitate the understanding of the embodiments of the present application, further explanation will be given below in conjunction with the drawings and specific embodiments. The embodiments do not constitute a limitation on the embodiments of the present application.

[0068] Figure 1 The following is a flow chart of a multi-period remote sensing image automatic change detection and processing method provided in the embodiment of the present application. Figure 1 As shown, the multi-period remote sensing image automatic change detection processing method provided in the embodiment of the present application may specifically include the following steps:

[0069] Step 110, obtaining a fixed base period remote sensing image, a rolling base period remote sensing image, and a latest monitored later phase remote sensing image of the target area.

[0070] In this embodiment, the fixed base period remote sensing image can be understood as: the image of the land change survey in the previous year based on the current date, such as the image of the land change survey in 23 years, which is not limited in this embodiment. The rolling base period remote sensing image is also called the early monitoring image. For the post-phase remote sensing image, it usually refers to the remote sensing image acquired after the rolling base period remote sensing image. In order to improve the accuracy and efficiency of change detection, this embodiment takes the rolling base period remote sensing image as the standard, and selects the latest monitoring image that is better than the rolling base period remote sensing image from the latest monitored post-phase remote sensing image set as the post-phase remote sensing image.

[0071] It should be noted that, in the embodiments of the present application, each remote sensing image may also be referred to as an image, the front phase image may also be referred to as the front phase drop image, the back phase image may also be referred to as the back phase drop image, and the like.

[0072] Step 120, performing change extraction based on the fixed base period remote sensing image and the later phase remote sensing image to obtain a fixed base period patch a, and performing change extraction based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a rolling base period patch n.

[0073] In a specific implementation, the base period spots can also be called base period change spots, which are divided into base period spots after deduplication and base period spots without deduplication, mainly including but not limited to: fixed base period spots and rolling base period spots. This embodiment mainly uses the previous phase image and the next phase image to extract changes and obtain the corresponding base period spots.

[0074] Specifically, this embodiment can pre-train the change detection model and preset the corresponding intelligent remote sensing interpretation algorithm. For the extraction of the fixed base period image a, this embodiment uses the previous year's change survey image (i.e., the fixed base period remote sensing image) as the previous phase image, combined with the latest monitored post-phase image, and uses the intelligent remote sensing interpretation algorithm and the change detection model to extract the fixed base period image.

[0075] For the extraction of rolling base period spots n, this embodiment uses the previous period of results image as the front phase image (i.e., rolling base period remote sensing image), extracts the rear phase image that is one day better than the front phase image from the latest monitored image, and uses it as the rear phase image for rolling base period spot extraction. Then, the front and rear phase images are combined, and the intelligent remote sensing interpretation algorithm is used in combination with the "change detection model" to extract the rolling base period spots n.

[0076] Step 130, performing deduplication processing based on the fixed base period patch a and the previously verified patches to obtain a fixed base period patch j.

[0077] Step 140, de-duplication processing is performed based on the rolling base period pattern n, combined with the fixed base period pattern j and the previous online patterns to obtain the target rolling base period pattern.

[0078] Among them, the fixed base period pattern j and the target rolling base period pattern are both base period change patterns.

[0079] A unified description of steps 130 to 140 is given as follows:

[0080] In this embodiment, the previously verified image spots can be obtained by merging the previously verified image spot database into one layer, wherein the previously verified image spot database mainly stores the image spots that have been verified before; the previously online image spots can be obtained by merging the previously online image spot database into one layer, wherein the previously online image spot database mainly stores the image spots that have been online before.

[0081] In a specific implementation, after the base period spots are obtained, the base period spots can be deduplicated, and the deduplication process can at least take into account the past period spots, including the past online spots and the past verified spots. This embodiment compares the base period spots and the past period spots and other spots to achieve change detection, fully considering factors such as the overlap rate between spots, and obtains the target base period spots.

[0082] In actual implementation, the deduplication processing of the base period spots in this embodiment can not only take into account the spots of the previous period, but also take into account various deduplication factors. Specifically, for the deduplication processing of the fixed base period spots a, factors such as the verified spots of the previous period, the change survey DLTB database, the acquisition time of the monitoring image, and the spot overlap rate can be taken into account. The software performs one-click deduplication processing to obtain the fixed base period spots j after deduplication. This embodiment will not be described in detail.

[0083] For the deduplication processing of the rolling base period patch n, factors such as the fixed base period patch j after deduplication, the semantic segmentation bulldozer patch of the later phase image, the patch already online in the past, the patch overlap rate, the minimum map area, etc. can be taken into consideration. The software can perform one-click deduplication processing to obtain the target rolling base period patch after deduplication. This embodiment will not describe this in detail.

[0084] Therefore, this embodiment realizes a method for rapid and precise extraction and deduplication of routine monitoring and verification spots, and adopts information technology to quickly and accurately extract change spots within the target range, avoiding redundant verification work caused by extracting duplicate spots from multiple images, and improving the completeness of change detection. It provides active, intelligent and comprehensive natural resource survey and monitoring basic data for natural resource survey, monitoring and management, and jointly forms a normalized monitoring work system serving natural resource management.

[0085] Step 150, performing normalized monitoring on the target area according to the fixed base period pattern j and the target rolling base period pattern.

[0086] In this embodiment, the change spot extraction method such as complete detection of post-phase images combining a fixed base period with a rolling base period can quickly support the routine monitoring of natural resources and achieve multiple monitoring objectives of routine monitoring, including but not limited to: real-time monitoring, monthly clearing and quarterly verification, annual summary, etc., thereby enhancing the support capacity of survey and monitoring for natural resource business management, and providing technical support for comprehensive support for precise management of natural resources and scientific decision-making.

[0087] It can be seen that the embodiment of the present application obtains multi-period remote sensing images including fixed base remote sensing images, the latest monitored post-phase remote sensing images and rolling base remote sensing images, and then performs change extraction on the multi-period remote sensing images to obtain fixed base patches and rolling base patches. On the basis of the extracted base patches, deduplication processing is performed in combination with the previous patches to obtain fixed base patches j and target rolling base patches, so as to use the extracted fixed base patches j and target rolling base patches for normalized monitoring. The present embodiment uses multi-phase remote sensing images to quickly and accurately extract suspected change patches within the target range, realize change detection of surface cover, obtain changes in important geographical elements, and can timely and accurately grasp the current status of natural resource utilization, avoid redundant verification work caused by extracting repeated patches from multi-period images, meet the requirements of rapid response of normalized monitoring, improve the completeness of change detection, and solve the problems of heavy workload and low efficiency caused by repeated extraction of patches in the prior art.

[0088] The multi-period remote sensing image automatic change detection and processing method proposed in this embodiment can continuously monitor and update the land survey results, maintain the accuracy and timeliness of basic geographic information, and realize real-time changes in land surveys by developing intelligent methods. By monitoring and analyzing the changed plots of land with multi-period remote sensing images, land violations can be quickly discovered, alleviating the difficulty of investigation; the problem of repeated extraction of change spots in multi-source and multi-phase remote sensing images and low completeness of online spots can be solved by removing overlapping processing methods.

[0089] Reference Figure 2 , shows a schematic flow chart of the steps of a multi-period remote sensing image automatic change detection processing method provided by an optional embodiment of the present application. The method may specifically include the following steps:

[0090] Step 210, obtaining a fixed base period remote sensing image, a rolling base period remote sensing image, and a latest monitored later phase remote sensing image of the target area.

[0091] Optionally, this embodiment obtains a fixed base remote sensing image, a rolling base remote sensing image, and the latest monitored post-phase remote sensing image of the target area, which may specifically include: extracting a land change survey image of the target area from a preset land change survey database as a fixed base remote sensing image; obtaining the previous batch of initial post-phase images as rolling base remote sensing images, and obtaining the latest monitoring images as post-phase remote sensing images.

[0092] In the related technology, in addition to using a fixed base period change monitoring mode to repeatedly extract spots, some existing solutions also use change monitoring based on a rolling base period. However, the base period image selected for rolling base period spot extraction is the later phase image of the previous period of normalized monitoring, which will result in a short interval between two periods of images, and it is easy to miss slowly changing spots, which will lead to a low recall rate.

[0093] In view of the problems existing in the change spot extraction scheme used in the existing routine monitoring, such as high repeated detection rate of fixed base period extraction and low recall rate of rolling base period extraction, this embodiment proposes a change spot extraction method that combines fixed base period with rolling base period to fully detect the post-phase image, so as to reduce the occurrence of spot omissions.

[0094] For example, refer to Figure 3 As shown in the technical circuit flow chart, in the data preparation stage of this embodiment, the data that needs to be obtained may include but are not limited to: the phase map before the rolling base period (i.e. the combined map of the previous period's images), the image data of the previous period's results, the phase map after the rolling base period (i.e. the combined map of the current period's images), the current period's image results data, the database of previously verified map spots, the database of previously online map spots, the DLTB layer of the previous year's land change survey data, and the previous year's land change survey images.

[0095] It should be noted that Figure 3 And in the subsequent attached figures: "23-year change survey image" is the fixed base period remote sensing image, "latest monitoring image" is the later phase remote sensing image, and "early monitoring image" is the rolling base period remote sensing image.

[0096] In actual implementation, the acquired image data, including the image combination file of the pre-phase drop image and the post-phase drop image, can adopt a fixed and consistent format. Preferably, the shapefile format can be adopted to ensure that the mathematical basis is consistent with the image. The data structure can be shown in Table 1 below:

[0097]

[0098] Table 1 Data structure of the combined image of the anterior / posterior phase

[0099] Step 220, performing change extraction based on the fixed base period remote sensing image and the later phase remote sensing image to obtain a fixed base period patch a, and performing change extraction based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a rolling base period patch n.

[0100] Exemplary, combined Figure 3 See the technical circuit diagram shown in Figure 4As shown in the flowchart of image patch extraction, this embodiment uses the "change detection model" of the intelligent remote sensing interpretation algorithm production and application platform to extract fixed base period image patches and rolling base period image patches.

[0101] Optionally, the above-mentioned change extraction based on the rolling base period remote sensing image and the post-phase remote sensing image to obtain the rolling base period spot n may specifically include: performing image selection based on the rolling base period remote sensing image and the post-phase remote sensing image to obtain a preferred monitoring image, wherein the preferred monitoring image is a monitoring image whose latest monitoring image phase is better than the first preset time of the previous monitoring image phase; performing spot extraction based on the preferred monitoring image and the rolling base period remote sensing image to obtain the rolling base period spot n.

[0102] In this embodiment, preferably, the first preset time is 1 day, that is, in the process of extracting the rolling base period spot n in this embodiment, the previous monitoring image can be used as the base period image, and the latest monitoring image that is 1 day or more better than the previous monitoring image can be obtained as the post-phase image to extract the rolling base period change spot.

[0103] In actual implementation, this embodiment can realize the extraction of spots according to actual spot extraction needs in combination with the "change detection model" of the intelligent remote sensing interpretation algorithm production and application platform. Specifically, the spot extraction includes: ① taking the previous year's change survey image as the base image, and the latest monitoring image as the post-phase image, and using the "change detection model" of the intelligent remote sensing interpretation algorithm production and application platform to extract the "fixed base spot"; ② taking the previous monitoring image as the base image, and obtaining the latest monitoring image that is 1 day or more better than the previous monitoring image as the post-phase image, and using the "change detection model" of the intelligent remote sensing interpretation algorithm production and application platform to extract the "rolling base spot", and extract the rolling base change spots.

[0104] In actual implementation, this embodiment can construct a "phase optimal selection" tool. Figure 5 As shown in the figure, by using the "Optimal Phase Selection" tool to select the corresponding related images and input parameters (including but not limited to the number of days for image drop), the optimal image can be extracted with one click. For the post-phase image drop of the rolling base period image extraction: use the "Optimal Phase Selection" tool. The method of using this tool is as follows Figure 5 The usage steps are as follows: ① Select "Previous image combination map" in "Phase before input"; ② Select "Current image combination map" in "Phase after input"; ③ Fill in "1" for the number of days for the image drop, which means that the current image combination map is better than the previous image by 1 day or more; ④ Output the latest drop map, and it is "a later phase drop map that is better than the previous image by 1 day", and the result image corresponding to this drop map is "a later phase image that is better than the previous image by 1 day".

[0105] Furthermore, for one-click extraction of optimal images, this embodiment can use the model builder in ArcGIS (a geographic data processing software) to write an image selection system as a "phase optimal selection" tool, and embed it into the ArcGIS toolbox for execution. In this way, in normalized monitoring work, the image data required for extracting the change map spots can be selected with one click.

[0106] Step 230, obtaining previously verified image patches and extracting the change investigation scope from a preset change investigation database.

[0107] Step 240, extracting the fixed base period image spots b outside the changed survey range from the fixed base period image spots a, and extracting the image range D of the later phase remote sensing image whose phase is superior to the rolling base period remote sensing image by at least a second preset time.

[0108] A unified description of steps 230 to 240 is given as follows:

[0109] In this embodiment, in addition to preparing the before / after phase images, the deduplication processing of the base period spots can also prepare separate layer data, where the extraction of separate layer data includes but is not limited to: merging the database of previously verified spots into one layer, such as "previous verified spots"; merging the database of previously online spots into one layer, such as "previous online verified spots"; extracting the cultivated land layer of the change survey data DLTB, such as "DLTB_cultivated land"; extracting the garden and forest land layers of the change survey data DLTB, such as "DLTB_garden land". This embodiment determines the change survey scope by extracting the corresponding layers from the change survey database, and the layer categories include but are not limited to: cultivated land layer, forest land layer, and garden land layer.

[0110] In a specific implementation, for a fixed base period spot a, this embodiment can identify the land type (such as cultivated land, forest land, and garden land, etc.) contained in the base period spot a, and then extract the fixed base period spot b outside the changed survey scope (such as outside the garden land and forest land) in combination with the changed survey scope, such as Figure 3 and Figure 6 shown.

[0111] In an optional embodiment, the embodiment of the present application extracts the fixed base period patch b outside the change survey range from the fixed base period patch a, which may specifically include: performing full-element land class semantic segmentation on the fixed base period patch a using an intelligent remote sensing interpretation algorithm through a preset semantic segmentation model to obtain image land class information, wherein the image land class information includes image land class and image land class range; comparing the image land class information with the change survey range, and extracting the fixed base period patch b outside the change survey range from the fixed base period patch a.

[0112] In this embodiment, the land category range can be understood as the corresponding range of the image land category in the fixed base period patch a. For example, when the fixed base period patch a includes a garden, the garden is in the specific range of the fixed base period patch a.

[0113] In a specific implementation, this embodiment can input the fixed base period patch a into the semantic segmentation model, and combine the intelligent remote sensing interpretation algorithm to perform full-element land semantic segmentation on the fixed base period patch a, so as to extract the fixed base period patch b outside the change survey range from the fixed base period patch a, such as Figure 3 and Figure 6 shown.

[0114] Among them, semantic segmentation extraction mainly uses the "semantic segmentation model" of the intelligent remote sensing interpretation algorithm production and application platform to extract the semantics of all elements of land categories from the current image. The semantic segmentation code table involved in the image deduplication processing in this application is shown in Table 2 below:

[0115] Serial number Semantic Code Semantic Class 1 1 arable land 2 2 Garden, young forest 3 3 Water 4 4 grassland 5 5 building 6 6 the way 7 7 Structures 8 9 Pushing fill soil, bare soil

[0116] Table 2 Semantic segmentation code table The semantic segmentation results are in the "GDB" database format, and the data structure is shown in Table 3:

[0117] Serial number Field Name Field Code Field Type Field length Decimal Places Remark 1 post post Char 10

[0118] Table 3. Data structure of semantic segmentation results

[0119] Step 250: extracting fixed base period spot c and fixed base period spot d from the fixed base period spot b according to the image range D.

[0120] In a specific implementation, after extracting the image range D whose latest monitoring image phase is superior to the previous monitoring image phase by a second preset time (preferably, the second preset time is 10 days), the fixed base period spot c falling within the image range D can be extracted from the fixed base period spot b, and the fixed base period spot d not falling within the range D can be extracted from the fixed base period spot b, such as Figure 3 and Figure 6 shown.

[0121] Step 260, performing deduplication processing based on the previously verified patches, the fixed base patch c and the fixed base patch d to obtain the fixed base patch j.

[0122] In the specific implementation, for the deduplication of fixed base period spots, this embodiment mainly uses the previously verified spots, fixed base period spots c and fixed base period spots d for deduplication. For the deduplication of fixed base period spots, this embodiment fully considers factors such as the change survey DLTB database, the acquisition time of monitoring images, and the spot overlap rate to solve the problems of the prior art caused by repeated extraction of spots, the increase in verification workload, the large number of pseudo-changes in spots, and the redundancy of verification work.

[0123] Optionally, the above-mentioned deduplication processing is performed based on the previously verified spots, the fixed base spot c and the fixed base spot d to obtain the fixed base spot j, which may specifically include the following sub-steps:

[0124] Sub-step 2601, performing summary detection based on the previously verified image spots to obtain verified reference image spots.

[0125] Sub-step 2602, based on the verified reference pattern, respectively perform intersection comparison on the fixed base pattern c and the fixed base pattern d to obtain the fixed base pattern e corresponding to the fixed base pattern c and the fixed base pattern f corresponding to the fixed base pattern d.

[0126] Among them, the fixed base period spot e is a base period spot extracted from the fixed base period spot c, the area intersection ratio of which is less than the first preset threshold, and the fixed base period spot f is a base period spot extracted from the fixed base period spot d, the area intersection ratio of which is less than the second preset threshold.

[0127] Sub-steps 2601 and 2602 are described uniformly:

[0128] Preferably, the first preset threshold is 50%, and the second preset threshold is 20%.

[0129] In the specific implementation, the previously verified patches refer to the previously verified change detection reference patches. By summarizing all previously verified change detection reference patches, the verified reference patches are obtained, and then the change detection reference patches are used to perform intersection comparison on the fixed base patches. The intersection comparison process mainly includes calculating the area intersection ratio of the two patches. Specifically, refer to Figure 3 and Figure 6 As shown, this embodiment calculates the intersection ratio of the reference spot and the fixed base spot c, extracts the fixed base spot e whose area intersection ratio in the fixed base spot c is less than 50%, and calculates the intersection ratio of the reference spot and the fixed base spot d, extracts the fixed base spot f whose area intersection ratio in the fixed base spot d is less than 20%.

[0130] Sub-step 2603, extracting the image range E whose time phase of the latest monitoring image differs from the time phase of the previous monitoring image by a third preset time.

[0131] Sub-step 2604, extracting fixed base period spot g and fixed base period spot h from the fixed base period spot f according to the image range E.

[0132] Sub-step 2605, extracting the fixed base period spots i within the cultivated land range in the image range E from the fixed base period spots g.

[0133] Sub-step 2606, merging the fixed base period spots e, the fixed base period spots h and the fixed base period spots i to obtain the fixed base period spots j after deduplication.

[0134] Sub-steps 2603 to 2606 are described uniformly:

[0135] Preferably, the third preset time may be 30 days.

[0136] Reference Figure 3 and Figure 6 As shown, this embodiment extracts the image range E whose latest monitoring image phase is 30 days different from the previous monitoring image phase, so as to compare the fixed base period spot f with the image range E, and extracts the fixed base period spot g that falls within the image range E and the fixed base period spot h that does not fall within the image range E from the fixed base period spot f. Then, the fixed base period spot i that falls within the cultivated land range of the land change in the previous year is extracted from the fixed base period spot g, so that the present embodiment obtains the fixed base period spot e, the fixed base period spot h, and the fixed base period spot i after deduplication using the monitoring image acquisition time, the spot overlap rate and other factors. Finally, the fixed base period spot e, the fixed base period spot h, and the fixed base period spot i are merged to obtain the fixed base period change spot j after deduplication.

[0137] In the related technologies, the existing technologies still have the problem of backward deduplication mode of change detection spots. Specifically, the previous deduplication mode is to directly erase the newly extracted spots and the previously online spots. This mode is prone to generate a large number of fragmented spots, and is prone to missing continuously changing spots and key monitoring spots.

[0138] In order to solve the problems caused by the backward deduplication mode of the existing technology, this embodiment proposes a deduplication method that takes into account factors such as the online spots, image resolution, image acquisition time, and spot overlap rate, which greatly reduces the amount of spot overlap and solves the problem of repeated extraction of change spots in multi-source and multi-temporal remote sensing images and low recall rate of online spots. Compared with the method without deduplication and fixed base period, this embodiment is 2 times more efficient, and the subsequent rolling base period spot recall rate is increased to 85%.

[0139] Step 270, performing semantic segmentation extraction based on the later phase remote sensing image and the fixed base period patch j to obtain a fixed base period patch k and a fixed base period patch l.

[0140] In the specific implementation, this embodiment takes into account the fixed base period change pattern j during the deduplication process of the rolling base period pattern. Figure 3 and Figure 7 As shown, this embodiment performs semantic segmentation on the latest monitoring image B to extract semantically segmented patches, such as semantically segmented bulldozing patches, and then combines the semantically segmented bulldozing patches and fixed base period patches j to perform patch extraction to obtain fixed base period patches k and fixed base period patches l.

[0141] In an optional embodiment, this embodiment performs semantic segmentation extraction based on the post-phase remote sensing image and the fixed base period patch j to obtain a fixed base period patch k and a fixed base period patch l, which may specifically include: performing semantic segmentation on the post-phase remote sensing image through a preset semantic segmentation model to obtain a bulldozing patch; extracting a fixed base period patch k that intersects with the bulldozing patch from the fixed base period patch j, and extracting a fixed base period patch l that does not intersect with the bulldozing patch from the fixed base period patch j.

[0142] Among them, the bulldozing patch is also called semantic segmentation bulldozing patch.

[0143] Step 280, obtaining online reference spots based on the aggregation of the previously online spots, and obtaining fixed base spot m based on the intersection comparison of the online reference spots and the fixed base spot l.

[0144] In a specific implementation, this embodiment can summarize all the transformation detection reference spots that have been online in the past to obtain the online reference spots, and then use the online spots and the fixed base period spots l to perform intersection comparison to obtain the fixed base period spots m.

[0145] Optionally, the above-mentioned intersecting comparison based on the online reference pattern and the fixed base pattern l to obtain the fixed base pattern m may specifically include the following sub-steps:

[0146] Sub-step 2801, removing easily changeable spots based on the online reference spots to obtain target reference spots.

[0147] Sub-step 2802, calculating the area intersection ratio based on the target reference pattern and the fixed base pattern l to obtain an area intersection ratio value.

[0148] Sub-step 2803: extracting, from the fixed base period patch l, the fixed base period patch m whose area intersection ratio is less than a third preset threshold.

[0149] Sub-steps 2801 to 2803 are described uniformly:

[0150] Preferably, the third preset threshold is 50%.

[0151] In the specific implementation, refer to Figure 3 and Figure 7 As shown, this embodiment performs easy-to-change elimination based on the online reference spots, and the easy-to-change elimination process includes eliminating easy-to-change spots such as bulldozing in the online reference spots to obtain the target reference spots. Then, the intersection ratio of the target reference spots and the fixed base spot l is calculated, and the fixed base spot m whose area intersection ratio in the spot l is less than 50% is extracted.

[0152] Step 290, based on the fixed base period pattern m, the fixed base period pattern k and the rolling base period pattern n, merge and screen the area to obtain the target rolling base period change pattern after deduplication.

[0153] In the specific implementation, this embodiment merges the fixed base period spot m, the fixed base period spot k, and the rolling base period change spot n to obtain the change spot, and then performs area screening on the merged change spot. The area screening mainly eliminates the spots with an area less than 200 square meters, and finally obtains the final deduplicated spots, that is, the deduplicated rolling base period change spots.

[0154] In actual implementation, this embodiment can use the packaged "de-duplication model" to perform one-click de-duplication processing, thereby achieving de-duplication of fixed base period spots and rolling base period spots. Exemplarily, using the packaged "de-duplication model", input relevant vector data in each input condition. The tool is used as follows: Figure 8 As shown, the usage steps are as follows: ① Select "Last period image combination map" in "Enter the previous phase map_rolling base period"; ② Select "Current period image combination map" in "Enter the phase map after input"; ③ Select "Fixed base period map" in "Enter the fixed base period map"; ④ Select "Rolling base period map" in "Enter the rolling base period map"; ⑤ Select "Previous period verification map" in "Enter the previous period BHJC map"; ⑥ Select "Previous period online map" in "Enter the previous period online map"; ⑦ Select "YYFG map" in "Enter YYFG results"; ⑧ Select "DLTB_cultivated land" in "Enter DLTB_NMK_01"; ⑨ Select "DLTB_garden land" in "Enter DLTB_NMK_garden"; ⑩ Enter "BHJC map after deduplication" and the storage location in "Enter the map after deduplication".

[0155] Furthermore, for one-click deduplication, this embodiment can use the model builder in ArcGIS to write a deduplication model system and embed it into the ArcGIS toolbox for execution. In the routine monitoring work, one-click operation can be used to remove the overlap of the extracted spots with the previous spots to obtain the suspected change spots that need to be manually checked, thus achieving one-click deduplication.

[0156] Therefore, this embodiment uses information technology to quickly and accurately extract the change spots within the target range, avoid redundant verification work caused by extracting repeated spots from multiple images, and improve the completeness of change detection.

[0157] Step 300, performing normalized monitoring on the target area according to the fixed base period pattern j and the target rolling base period pattern.

[0158] This embodiment uses the deduplicated fixed base period map j and the target rolling base period map to carry out routine monitoring of the target area, which has a strong supporting role in routine monitoring, and achieves the monitoring objectives of routine monitoring "real-time monitoring, monthly clearing, seasonal verification, and annual summary", enhances the support capacity of survey and monitoring for natural resource business management, and provides technical support for comprehensive support for precise management of natural resources and scientific decision-making.

[0159] In summary, the embodiment of the present application obtains multiple remote sensing images such as fixed base period remote sensing images, rolling base period remote sensing images and the latest monitored post-phase remote sensing images of the target area, so as to combine the multiple remote sensing images and extract changes respectively, thereby obtaining fixed base period spots a and rolling base period spots n. Then, the fixed base period spots a are deduplicated, by extracting fixed base period spots b outside the change survey range from the fixed base period spots a, and extracting the image range D of the post-phase remote sensing image whose phase is better than the rolling base period remote sensing image by at least a second preset time, and according to the image range D, extracting fixed base period spots c falling into the image range D and fixed base period spots d not falling into the image range D from the fixed base period spots b, and deduplicating according to the previously verified spots, fixed base period spots c and fixed base period spots d, to obtain the deduplicated fixed base period spots j. For deduplication of rolling base patches, this embodiment performs semantic segmentation extraction based on the later-phase remote sensing images and the fixed base patches j to obtain fixed base patches k and fixed base patches l, then summarizes the online patches in the previous periods to obtain the online reference patches, and performs intersection comparison based on the online reference patches and the fixed base patches l to obtain the fixed base patches m, merges and screens the areas of the fixed base patches m, the fixed base patches k and the rolling base patches n to obtain the deduplicated target rolling base change patches, thereby achieving deduplication of the fixed base change patches and the rolling base change patches, and further realizing a method and system for rapid and precise extraction and deduplication of normalized monitoring and verification patches, avoiding redundant verification work caused by extracting duplicate patches from multiple periods of images, improving the completeness of change detection, and solving various problems existing in the prior art in the change detection of base patches.

[0160] In addition, the embodiment of the present application conducts regular monitoring of the target area based on the fixed base period map j and the target rolling base period map extracted by deduplication, so as to provide strong support for regular monitoring, and provide active, intelligent and comprehensive natural resource survey and monitoring basic data for natural resource survey, monitoring and management, and jointly form a regular monitoring work system serving natural resource management.

[0161] It should be noted that, for the purpose of simple description, the method embodiments are expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited to the described order of actions, because according to the embodiments of the present application, certain steps may be performed in other orders or simultaneously.

[0162] like Fig. 9 As shown, the embodiment of the present application also provides a multi-period remote sensing image automatic change detection and processing device 900, comprising:

[0163] The multi-period remote sensing image acquisition module 910 is used to acquire the fixed base period remote sensing image, the rolling base period remote sensing image and the latest monitored post-phase remote sensing image of the target area;

[0164] The change extraction module 920 is used to extract changes based on the fixed base period remote sensing image and the later phase remote sensing image to obtain a fixed base period image patch a, and to extract changes based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a rolling base period image patch n;

[0165] The first deduplication processing module 930 is used to perform deduplication processing based on the fixed base period image patch a and the previously verified image patches to obtain a fixed base period image patch j;

[0166] The second deduplication processing module 940 is used to perform deduplication processing based on the rolling base period image patch n, in combination with the fixed base period image patch j and the previous online image patches, to obtain a target rolling base period image patch;

[0167] The normalized monitoring module 950 is used to perform normalized monitoring on the target area according to the fixed base period pattern j and the target rolling base period pattern; wherein the fixed base period pattern j and the target rolling base period pattern are both base period change patterns.

[0168] Optionally, the multi-period remote sensing image acquisition module 910 includes:

[0169] The fixed base period remote sensing image extraction submodule is used to extract the land change survey image of the target area from the preset land change survey database as the fixed base period remote sensing image;

[0170] The rolling base period remote sensing image acquisition submodule is used to obtain the initial post-phase images of the previous batch as rolling base period remote sensing images;

[0171] The post-phase remote sensing image acquisition submodule is used to obtain the latest monitoring image as the post-phase remote sensing image.

[0172] Optionally, the change extraction module 920 includes:

[0173] An image selection submodule is used to select images based on the rolling base period remote sensing images and the later phase remote sensing images to obtain a preferred monitoring image, wherein the preferred monitoring image is a monitoring image whose latest monitoring image phase is better than the monitoring image phase of the previous monitoring image by a first preset time;

[0174] The spot extraction submodule is used to extract spots based on the preferred monitoring image and the rolling base period remote sensing image to obtain the rolling base period spot n.

[0175] Optionally, the first deduplication processing module 930 includes:

[0176] The verified image spot acquisition submodule is used to obtain the verified image spots in the past;

[0177] A change investigation scope extraction submodule is used to extract the change investigation scope from a preset change investigation database;

[0178] A fixed base period spot b extraction submodule is used to extract the fixed base period spot b outside the change investigation range from the fixed base period spot a;

[0179] An image range D extraction submodule, used for extracting an image range D of the later-phase remote sensing image whose phase is superior to the rolling base period remote sensing image by at least a second preset time;

[0180] A patch c and patch d extraction submodule extracts a fixed base patch c and a fixed base patch d from the fixed base patch b according to the image range D;

[0181] The first deduplication processing submodule is used to perform deduplication processing based on the previously verified image spots, the fixed base period image spots c and the fixed base period image spots d to obtain the fixed base period image spots j.

[0182] Optionally, the fixed base period spot b extraction submodule includes:

[0183] A first semantic segmentation unit is used to perform full-element land class semantic segmentation on the fixed base period patch a by using a preset semantic segmentation model and an intelligent remote sensing interpretation algorithm to obtain image land class information, wherein the image land class information includes image land class and image land class range;

[0184] The first comparison unit is used to compare the image land type information with the changed survey range, and extract the fixed base period patch b outside the changed survey range from the fixed base period patch a.

[0185] Optionally, the first deduplication processing submodule includes:

[0186] A summary detection unit, used for performing summary detection based on the previously verified image spots to obtain verified reference image spots;

[0187] An intersection comparison unit is used to perform intersection comparison on the fixed base period spot c and the fixed base period spot d according to the verified reference spot, to obtain a fixed base period spot e corresponding to the fixed base period spot c and a fixed base period spot f corresponding to the fixed base period spot d;

[0188] An image range E extraction unit, used for extracting an image range E whose time phase of the latest monitoring image differs from that of the previous monitoring image by a third preset time;

[0189] A patch g and patch h extraction unit, configured to extract a fixed base patch g and a fixed base patch h from the fixed base patch f according to the image range E;

[0190] A patch i extraction unit is used to extract the fixed base patch i within the cultivated land range in the image range E from the fixed base patch g;

[0191] The first deduplication unit is used to merge the fixed base spot e, the fixed base spot h and the fixed base spot i to obtain the fixed base spot j after deduplication; wherein the fixed base spot e is a base spot extracted from the fixed base spot c, the area intersection ratio of which is less than a first preset threshold, and the fixed base spot f is a base spot extracted from the fixed base spot d, the area intersection ratio of which is less than a second preset threshold.

[0192] Optionally, the second deduplication processing module 940 includes:

[0193] A semantic segmentation submodule, used for performing semantic segmentation extraction based on the later phase remote sensing image and the fixed base period patch j to obtain a fixed base period patch k and a fixed base period patch l;

[0194] An online reference image patch summary submodule is used to obtain online reference image patches by summarizing the previously online image patches;

[0195] An intersection comparison submodule is used to perform an intersection comparison based on the online reference pattern and the fixed base pattern l to obtain a fixed base pattern m;

[0196] The merging and area screening submodule is used to perform merging and area screening based on the fixed base period pattern m, the fixed base period pattern k and the rolling base period pattern n to obtain the target rolling base period change pattern after deduplication.

[0197] Optionally, the semantic segmentation submodule includes:

[0198] A second semantic segmentation unit is used to perform semantic segmentation on the later-phase remote sensing image by using a preset semantic segmentation model to obtain a bulldozing patch;

[0199] The patch k and patch l extraction unit is used to extract the fixed base patch k intersecting with the bulldozing patch from the fixed base patch j, and to extract the fixed base patch l not intersecting with the bulldozing patch from the fixed base patch j.

[0200] Optionally, the intersection comparison submodule includes:

[0201] A volatile elimination unit, used for eliminating volatile spots according to the online reference spots to obtain target reference spots;

[0202] An intersection ratio calculation unit, used to calculate the area intersection ratio based on the target reference pattern and the fixed base pattern l to obtain an area intersection ratio value;

[0203] The patch m extraction unit is used to extract the fixed base patch m whose area intersection ratio is less than a third preset threshold from the fixed base patch l.

[0204] It should be noted that the multi-period remote sensing image automatic change detection and processing device provided in the embodiment of the present application can execute the multi-period remote sensing image automatic change detection and processing method provided in any embodiment of the present application, and has the corresponding functions and beneficial effects of executing the multi-period remote sensing image automatic change detection and processing method.

[0205] In a specific implementation, the above-mentioned multi-period remote sensing image automatic change detection processing device can be integrated into a device, so that the device can use information technology to quickly and accurately extract the change spots within the target range from the multi-period remote sensing images, and remove duplicates, so as to avoid redundant verification work caused by extracting duplicate spots from multi-period images. As an electronic device, it can reduce the occurrence of missed spots and improve the completeness of change detection. The electronic device can be composed of two or more physical entities, or it can be composed of one physical entity. For example, the electronic device can be a personal computer (PC), a computer, a server, etc., and the embodiments of the present application do not impose specific restrictions on this.

[0206] like Fig.10As shown, an embodiment of the present application provides an electronic device, including a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114; the memory 113 is used to store computer programs; the processor 111 is used to implement the steps of the multi-period remote sensing image automatic change detection processing method provided by any of the aforementioned method embodiments when executing the program stored in the memory 113. Exemplarily, the steps of the multi-period remote sensing image automatic change detection and processing method may include the following steps: acquiring a fixed base period remote sensing image, a rolling base period remote sensing image and a newly monitored post-phase remote sensing image of the target area; performing change extraction based on the fixed base period remote sensing image and the post-phase remote sensing image to obtain a fixed base period spot a, and performing change extraction based on the rolling base period remote sensing image and the post-phase remote sensing image to obtain a rolling base period spot n; performing deduplication processing based on the fixed base period spot a in combination with previously verified spots to obtain a fixed base period spot j; performing deduplication processing based on the rolling base period spot n in combination with the fixed base period spot j and previously online spots to obtain a target rolling base period spot; performing regular monitoring of the target area based on the fixed base period spot j and the target rolling base period spot; wherein, the fixed base period spot j and the target rolling base period spot are both base period change spots.

[0207] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the multi-period remote sensing image automatic change detection processing method provided in any of the aforementioned method embodiments are implemented.

[0208] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0209] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest range consistent with the principles and novel features applied for herein.

Claims

1. A multi-period remote sensing image automatic change detection and processing method, characterized in that: include: Obtain fixed base period remote sensing images, rolling base period remote sensing images and the latest monitored post-phase remote sensing images of the target area; Performing change extraction based on the fixed base period remote sensing image and the later phase remote sensing image to obtain a fixed base period image patch a, and performing change extraction based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a rolling base period image patch n; According to the fixed base period patch a, combined with the previously verified patches, duplicate removal is performed to obtain a fixed base period patch j; According to the rolling base period image patch n, the fixed base period image patch j and the previous online image patches are combined to perform deduplication processing to obtain the target rolling base period image patch; Performing normalized monitoring of the target area according to the fixed base period pattern j and the target rolling base period pattern; Among them, the fixed base period pattern j and the target rolling base period pattern are both base period change patterns.

2. The method according to claim 1, characterized in that Obtain fixed base period remote sensing images, rolling base period remote sensing images and the latest monitored post-phase remote sensing images of the target area, including: Extracting land change survey images of the target area from the preset land change survey database as fixed base period remote sensing images; The initial post-phase images of the previous batch are obtained as the rolling base period remote sensing images, and the latest monitoring images are obtained as the post-phase remote sensing images.

3. The method according to claim 1, characterized in that The rolling base period patch n is obtained by performing change extraction based on the rolling base period remote sensing image and the post-phase remote sensing image, including: Image selection is performed based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a preferred monitoring image, wherein the preferred monitoring image is a monitoring image whose latest monitoring image phase is better than the monitoring image phase of the previous monitoring image by a first preset time; Spot extraction is performed based on the preferred monitoring image and the rolling base period remote sensing image to obtain the rolling base period spot n.

4. The method according to claim 1, characterized in that: According to the fixed base period patch a, combined with the verified patches in the past, duplicate removal is performed to obtain the fixed base period patch j, including: Obtain previously verified image patches and extract the change investigation scope from the preset change investigation database; Extracting the fixed base period spots b outside the change survey range from the fixed base period spots a, and extracting the image range D of the later phase remote sensing image whose phase is superior to the rolling base period remote sensing image by at least a second preset time; According to the image range D, extracting the fixed base period spot c and the fixed base period spot d from the fixed base period spot b; Based on the previously verified patches, the fixed base patch c and the fixed base patch d, deduplication processing is performed to obtain the fixed base patch j.

5. The method according to claim 4, characterized in that Extracting the fixed base period map b outside the scope of the change investigation from the fixed base period map a includes: Through a preset semantic segmentation model, an intelligent remote sensing interpretation algorithm is used to perform full-element land class semantic segmentation on the fixed base period patch a to obtain image land class information, wherein the image land class information includes image land class and image land class range; The image land classification information is compared with the changed survey scope, and the fixed base period patches b outside the changed survey scope are extracted from the fixed base period patches a.

6. The method according to claim 4, characterized in that Based on the previously verified spots, the fixed base spot c and the fixed base spot d, duplicate removal is performed to obtain the fixed base spot j, including: Performing summary detection based on the previously verified image spots to obtain verified reference image spots; According to the verified reference spots, the fixed base spot c and the fixed base spot d are respectively cross-compared to obtain the fixed base spot e corresponding to the fixed base spot c and the fixed base spot f corresponding to the fixed base spot d; Extracting an image range E whose latest monitoring image phase is different from the previous monitoring image phase by a third preset time; According to the image range E, extracting the fixed base period spot g and the fixed base period spot h from the fixed base period spot f; Extracting the fixed base period spots i within the cultivated land range in the image range E from the fixed base period spots g; Merge the fixed base period spots e, h and i to obtain a fixed base period spot j after deduplication; Among them, the fixed base period spot e is a base period spot extracted from the fixed base period spot c, the area intersection ratio of which is less than the first preset threshold, and the fixed base period spot f is a base period spot extracted from the fixed base period spot d, the area intersection ratio of which is less than the second preset threshold.

7. The method according to claim 1, characterized in that According to the rolling base period image n, the fixed base period image j and the previous online image are combined to perform deduplication processing to obtain the target rolling base period image, including: Perform semantic segmentation extraction based on the later-phase remote sensing image and the fixed base period patch j to obtain a fixed base period patch k and a fixed base period patch l; Based on the summary of the previously online patches, an online reference patch is obtained, and based on the intersection comparison between the online reference patch and the fixed base patch l, a fixed base patch m is obtained; Based on the fixed base period pattern m, the fixed base period pattern k and the rolling base period pattern n, merging and area screening are performed to obtain the target rolling base period change pattern after deduplication.

8. The method according to claim 7, characterized in that Semantic segmentation and extraction are performed according to the later phase remote sensing image and the fixed base period patch j to obtain the fixed base period patch k and the fixed base period patch l, including: The semantic segmentation of the post-phase remote sensing image is performed by using a preset semantic segmentation model to obtain a bulldozing patch; A fixed base period spot k intersecting with the bulldozing spot is extracted from the fixed base period spot j, and a fixed base period spot l not intersecting with the bulldozing spot is extracted from the fixed base period spot j.

9. The method according to claim 7, characterized in that: Based on the intersection comparison between the online reference patch and the fixed base patch l, a fixed base patch m is obtained, including: Eliminate the easily changeable spots according to the online reference spots to obtain the target reference spots; Calculating the area intersection ratio based on the target reference pattern and the fixed base pattern l to obtain an area intersection ratio value; A fixed base period patch m having an area intersection ratio value less than a third preset threshold is extracted from the fixed base period patch l.

10. A multi-period remote sensing image automatic change detection and processing device, characterized in that: include: The multi-period remote sensing image acquisition module is used to acquire fixed base period remote sensing images, rolling base period remote sensing images and the latest monitored post-phase remote sensing images of the target area; A change extraction module is used to extract changes based on the fixed base period remote sensing image and the later phase remote sensing image to obtain a fixed base period image patch a, and to extract changes based on the rolling base period remote sensing image and the later phase remote sensing image to obtain a rolling base period image patch n; The first deduplication processing module is used to perform deduplication processing based on the fixed base period spot a and the previously verified spots to obtain the fixed base period spot j; The second deduplication processing module is used to perform deduplication processing based on the rolling base period image patch n, in combination with the fixed base period image patch j and the previous online image patches, to obtain the target rolling base period image patch; A normalized monitoring module, used for performing normalized monitoring on the target area according to the fixed base period pattern j and the target rolling base period pattern; Among them, the fixed base period pattern j and the target rolling base period pattern are both base period change patterns.

Images