Remote sensing image change detection method and device based on split grid, and terminal

Abstract

The embodiment of the application discloses a kind of based on dissected grid remote sensing image change detection method, device, computer terminal, can let change detection precision no longer be limited to image classification precision, and can effectively avoid the influence of noise.Based on dissected grid remote sensing image change detection method, comprising: obtaining multiple remote sensing images of the pre-detection area under different time phases;According to the same dissected coding mode, multiple remote sensing images are dissected and coded to form multiple grid units, wherein each grid unit is configured with a unique code;At least one grid unit is obtained under different time phases Change situation is used to judge the image change situation in the pre-detection area.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of remote sensing change detection. More particularly, the present application relates to a remote sensing image change detection method and device based on a dissected grid, and a computer terminal. BACKGROUND

[0002] Remote sensing change detection technology is a technology for determining whether a ground object in a same region on multi-temporal remote sensing images has changed, and even detecting how the ground object has changed. Through manual and computer-assisted processing of multi-temporal remote sensing images covering the same region, change information in the images is accurately and quickly extracted, dynamic monitoring of changed ground objects and analysis of ground change trends and evolution rules are realized, and the technology plays an extremely important role in many fields such as urban expansion, land use change, forest vegetation change, ecological environment monitoring, and disaster monitoring.

[0003] Currently, the detection units in remote sensing change detection methods mainly include two types of pixels and objects: (1) based on pixels: a minimum change unit is established based on pixels of remote sensing images, a pixel direct comparison method such as a difference method, a ratio method, a regression analysis method, etc. is adopted, and a change feature map is obtained through algebraic operation on pixels at the same position on multi-temporal images, and then the change feature map is cut according to a threshold value to find a change position of the remote sensing images. The remote sensing change detection method based on pixels is mainly applied to remote sensing images with medium and low resolution, extremely rich spectral information, and obvious spectral differences between ground objects. (2) based on objects: a minimum change unit is established based on ground objects of remote sensing images, and an object is composed of mutually associated pixels. The object not only uses spectral information of the images, but also uses spatial information, and is mainly applied to a remote sensing image classification scene. Based on objects, a series of factors such as spectral statistical characteristics, shape, size, texture, and adjacent relationship are comprehensively considered to obtain a high-precision information classification result. Based on objects, the object generation method can be divided into an "object generation method" and a "graph patch generation method", the former is based on ground object classification to perform ground object segmentation on remote sensing images, and the latter is to first automatically aggregate into graph patches without classification through an algorithm, and then select a classification method or a characteristic value to classify the existing graph patches.

[0004] The above remote sensing image change detection solution has the following defects: (1) using a pixel as a remote sensing change detection unit is susceptible to image noise and has poor robustness. For comparison of high spatial resolution remote sensing images at different times, the registration accuracy of the two images is required to be higher, and the pixel is more sensitive to its true geographical position. In this case, the pixel-level comparison method is still susceptible to noise.(2) Using a feature object as a remote sensing image change detection unit must go through a remote sensing image classification process, and the image change detection accuracy is limited by the image feature classification accuracy. Image classification needs to establish a set of feature classification standards according to business needs, and the image needs to be classified according to the feature classification standard. This process consumes a lot of manpower and lacks timeliness and universality. SUMMARY

[0005] An object of the present application is to solve the above problems and provide the advantages described later.

[0006] To this end, the embodiments of the present application provide a remote sensing image change detection method and device based on a split grid, a computer terminal and a storage medium, which can make the change detection accuracy of remote sensing images no longer limited by the image classification accuracy, and can effectively avoid the influence of noise.

[0007] Specifically, the embodiments of the present application provide the following technical solutions:

[0008] In one aspect, the embodiments of the present application provide a remote sensing image change detection method based on a split grid, comprising: acquiring a plurality of remote sensing images of a pre-detection area at different times; performing split coding processing on the plurality of remote sensing images according to the same split coding mode to form a plurality of grid cells, wherein each grid cell is configured with a unique code; acquiring the change of at least one grid cell at different times, for judging the image change in the pre-detection area.

[0009] In another aspect, the embodiments of the present application provide a remote sensing image change detection method based on a split grid, comprising: acquiring a plurality of remote sensing images of a pre-detection area at different times; performing split coding processing on the plurality of remote sensing images according to the same split coding mode to form a plurality of grid cells, wherein each grid cell has a corresponding unique code and a plurality of grid images; acquiring a plurality of target grid images corresponding to the target coding corresponding grid cell at different times; judging the similarity between the plurality of target grid images; judging whether the image in the pre-detection area has changed according to the similarity.

[0010] In still another aspect, the embodiment of the present application provides a remote sensing image change detection method based on a dissected grid, comprising: selecting a pre-detection area in dynamic remote sensing images of multiple time phases; performing dissecting processing on the pre-detection area according to geographic coordinates to generate a plurality of grid cells, and performing spatial coding on each grid cell, wherein each grid cell has a corresponding unique code and a plurality of grid images; and obtaining change conditions of at least one grid cell in different time phases to determine image change conditions in the pre-detection area.

[0011] In still another aspect, the embodiment of the present application provides a remote sensing image change detection device based on a dissected grid, comprising: an image acquisition module configured to acquire a plurality of remote sensing images of a pre-detection area in different time phases; a dissecting and coding processing module configured to perform dissecting and coding processing on the plurality of remote sensing images according to the same dissecting and coding mode to form a plurality of grid cells, wherein each grid cell has a corresponding unique code and a plurality of grid images; and a change detection module configured to obtain change conditions of at least one grid cell in different time phases to determine image change conditions in the pre-detection area.

[0012] In still another aspect, the embodiment of the present application provides a remote sensing image change detection device based on a dissected grid, comprising: a pre-detection area rotating module configured to select a pre-detection area in dynamic remote sensing images of multiple time phases; a dissecting and coding processing module configured to perform dissecting processing on the pre-detection area according to geographic coordinates to generate a plurality of grid cells, and perform spatial coding on each grid cell; wherein each grid cell has a corresponding unique code and a plurality of grid images; and a change detection module configured to obtain change conditions of at least one grid cell in different time phases to determine image change conditions in the pre-detection area.

[0013] In still another aspect, the embodiment of the present application provides a computer terminal, comprising: a memory, a processor, and a computer instruction program stored in the memory and executable on the processor, wherein the computer instruction program is executed by the processor to implement the steps of the remote sensing image change detection method based on a dissected grid as described in the foregoing aspects.

[0014] The technical effects of the embodiment of the present application include:

[0015] (1) The remote sensing monitoring method, device, terminal and medium of the embodiment of the present application can make up for the defects existing in the two detection modes of pixel and object. Specifically, the remote sensing change detection unit formed based on the split grid is an abstraction of a region and can have multiple characteristic values at the same time; therefore, unlike the object detection unit, it does not need to consider the ground object properties carried by the object, so it does not need to perform remote sensing image change detection comparison after classification; therefore, the remote sensing image change detection method of the present application can skip the image classification step and directly perform comparison detection, so that the change detection accuracy is no longer limited by the image classification accuracy. In addition, for remote sensing image, especially high spatial resolution remote sensing image change detection, the remote sensing monitoring method provided by the embodiment of the present application replaces the pixel with the grid unit for the remote sensing change detection unit, overcomes the defects brought by the pixel, so as to effectively avoid the influence of noise and improve the accuracy and usability of the detection result.

[0016] (2) Since the commonly used remote sensing image is L2 level data, which has been geometrically coarsely corrected. Therefore, based on this basis, the embodiment of the present application proposes a "grid automatic micro-correction" method, which moves another grid according to a certain range by fixing a grid, compares the Euclidean distance value of each movement, completes grid correction when the Euclidean distance is the smallest, that is, completes position registration, which can solve the problem of difficulty in realizing geometric fine correction of high spatial resolution remote sensing image, and effectively reduce the workload of manual registration of high spatial resolution image.

[0017] (3) The embodiment of the present application also enables the change result to have the ability to interact with external data through the split grid coding mode, and gives new vitality to the change detection process.

[0018] Other advantages, objects, and features of the present application will be apparent from the following description, and will be understood by persons skilled in the art. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 FIG. 1 is a flowchart of a remote sensing image change detection method based on split grid according to an embodiment of the present application;

[0020] Figure 2 FIG. 2 is a remote sensing image split coding processing schematic diagram in the remote sensing image change detection method based on split grid according to an embodiment of the present application;

[0021] Figure 3 FIG. 3 is another remote sensing image split coding processing schematic diagram in the remote sensing image change detection method based on split grid according to an embodiment of the present application;

[0022] Figure 4Figure 1 is a schematic diagram of a grid cell in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0023] Figure 5(a) is a schematic diagram of a remote sensing image in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0024] Figure 5(b) is a schematic diagram of another remote sensing image in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0025] Figure 6 Figure 6 is a schematic diagram of another remote sensing image in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0026] Figure 7 Figure 7 is a schematic diagram of another remote sensing image in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0027] Figure 8 Figure 8 is a schematic diagram of another remote sensing image in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0028] Figure 9 Figure 9 is a schematic diagram of another remote sensing image in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0029] Figure 10 Figure 10 is a schematic diagram of a flow in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0030] Figure 11 Figure 11 is a schematic diagram of a flow in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0031] Figure 12 Figure 12 is a schematic diagram of a flow in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0032] Figure 13 Figure 13 is a schematic diagram of a flow in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0033] Figure 14 Figure 14 is a schematic diagram of a flow in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0034] Figure 15 Figure 15 is a schematic diagram of a remote sensing image in a split-grid based remote sensing image change detection method according to an embodiment of the present application;

[0035] Figure 16Another flowchart of the method for detecting changes in remote sensing images based on a dissected grid according to an embodiment of the present application is shown in FIG. 6.

[0036] Figure 17 Another flowchart of the method for detecting changes in remote sensing images based on a dissected grid according to an embodiment of the present application is shown in FIG. 6.

[0037] Figure 18 Another flowchart of the method for detecting changes in remote sensing images based on a dissected grid according to an embodiment of the present application is shown in FIG. 6.

[0038] Figure 19 Another flowchart of the method for detecting changes in remote sensing images based on a dissected grid according to an embodiment of the present application is shown in FIG. 6.

[0039] Figure 20 Another flowchart of the method for detecting changes in remote sensing images based on a dissected grid according to an embodiment of the present application is shown in FIG. 6. DETAILED DESCRIPTION

[0040] The present application will be further described in details with reference to the accompanying drawings.

[0041] It should be noted that, in this document, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of steps or units does not necessarily comprise only those steps or units but can include other steps or units not expressly listed or inherent to such process, method, article, or apparatus.

[0042] In the present application, the word "exemplary" or "for example" is used to mean serving as an example, instance, or illustration. Any implementation described as "exemplary" or "for example" is not necessarily to be construed as preferred or advantageous over other implementations. The word "exemplary" or "for example" is intended to present concepts in a concrete manner. In addition, it is also necessary to emphasize that the term "embodiment" mentioned in this document means that the specific features, structures, or characteristics described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase at various places in the specification does not necessarily mean that the same embodiment is referred to at all places, nor does it mean that other embodiments are mutually exclusive or alternative to the embodiment. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0043] <First aspect>

[0044] As Figure 1As shown in the first aspect, the embodiments of the present application provide a remote sensing image change detection method based on a dissected grid, comprising the following steps:

[0045] S101, acquiring multiple remote sensing images of a pre-detection area at different time phases.

[0046] Specifically, the remote sensing photographing device can be used to capture images in the pre-detection area at different times, so as to acquire multiple remote sensing images. For example, the remote sensing photographing device can be used to capture images in the pre-detection area at T1, T2, T3, …, and Tn, so as to acquire remote sensing images corresponding to different time phases. It should be noted that the remote sensing photographing device can be a known device, such as a camera, a satellite, etc. After acquiring multiple remote sensing images, the remote sensing images can be stored in a pre-set image database for standby use, and the remote sensing images can be acquired from the image database when needed.

[0047] S102, performing dissected coding processing on the multiple remote sensing images according to the same dissected coding mode, to form multiple grid cells, wherein each grid cell has a corresponding unique code and multiple grid images.

[0048] Here, the multiple remote sensing images can be simultaneously subjected to the dissected coding processing. Alternatively, a remote sensing image at one time can be selected as a reference image, the reference image can be subjected to the dissected coding processing, and then the remote sensing images at other times can be subjected to the dissected coding processing.

[0049] As shown in Figure 2 and 3 , in some embodiments, the dissected coding processing on the remote sensing images at T1 and T2 is respectively shown. In Figure 2 , each grid image corresponds to a code, i.e., “1-16” shown in the figure. Based on the foregoing, it can be easily understood that the multiple remote sensing images can be subjected to the dissected coding processing, to form multiple grid cells, each of which has a corresponding unique code and multiple grid images. Figure 4 As shown in Figure 2 and 3 , a schematic diagram of the relationship between the grid cell with the code 1 and the multiple grid images at T1-Tn is shown.

[0050] Similarly, for the dissected coding physical schematic diagram described above, reference can be made to FIG. 5. FIG. 5(a) and (b) respectively show remote sensing images of the same pre-detection area at different time phases. According to FIG. 5(a) and (b), it can be seen that the multiple remote sensing images have been dissected and processed to form multiple grid cells, each of which has a corresponding grid image and code.

[0051] S103, acquire the change condition of at least one grid unit in different time phases, for judging the image change condition in the pre-detection area.

[0052] It should be noted that the image can be the image of the ground object in the pre-detection area. For example, when the pre-detection area is a wasteland area, the ground object can be land; when the pre-detection area is a water flow area, the ground object can be a river channel, water flow, etc. According to the above method, for the same grid unit, as long as it is judged whether the corresponding grid image of the grid in different time phases changes, for example, the similarity between the multiple grid images of the same grid unit is judged, it can be judged whether the image in the pre-detection area changes. More specifically, as shown in FIGS. 5(a) and (b), for the first grid unit (i.e., the first row and first column grid unit shown in the figure) therein, the similarity between the grid images in FIGS. 5(a) and (b) is not high, and there is a relatively obvious change, which indicates that the part corresponding to the first grid unit in the pre-detection area may have an image change.

[0053] The embodiments of the present application have the following beneficial effects relative to the prior art:

[0054] The remote sensing monitoring method of the embodiments of the present application can make up for the defects existing in the two detection methods based on pixels and objects. Specifically, the remote sensing change detection unit formed based on the split grid is an abstraction of a region and can have multiple characteristic values at the same time; therefore, unlike the object detection unit, it does not need to consider the ground object properties carried by the object, so it does not need to perform remote sensing image change detection comparison after classification; therefore, the remote sensing image change detection method of the present application can skip the image classification step and directly perform comparison detection, so that the change detection precision is no longer limited by the image classification precision. In addition, for remote sensing image, especially high spatial resolution remote sensing image change detection, the remote sensing monitoring method provided by the embodiments of the present application replaces the pixel with the grid unit for the remote sensing change detection unit, which overcomes the defects brought by the pixel, so it can effectively avoid the influence of noise and improve the accuracy and usability of the detection result.

[0055] It should be noted that there are multiple ways to perform split grid and encoding processing on remote sensing images, for example: Figure 6 The "standard map split encoding" based on different scale standard maps, in which "A1, A2, B1, B2" in the figure is the encoding of the grid; Figure 7 The "quadtree split encoding" based on projection; Figure 8 The "GeoSOT encoding"; Figure 9The diagram shows the "latitude and longitude kilometer grid" and the "BeiDou grid code," etc. Among them, the "standard map sheet subdivision code" and "quadritree subdivision code" are custom subdivision grids and codes; the "latitude and longitude kilometer grid" and "BeiDou grid code" are standard subdivision grids and codes.

[0056] In some embodiments, such as Figure 10 As shown, step S103, which involves obtaining the variation of at least one grid cell under different time phases, includes:

[0057] S1031. Determine at least one grid cell to be detected based on multiple grid cells;

[0058] Those skilled in the art can select at least one grid cell with detection from multiple grid cells as needed, for example, by screening grid cells according to the criteria of image noise level.

[0059] S1032. Obtain the similarity between multiple grid images corresponding to each grid cell to be detected at different time phases.

[0060] Similarity can be divided into two categories: similar and dissimilar. It can also be classified according to probability, such as similarity of 50%, 60%, 100%, etc.

[0061] The similarity here can be achieved using existing technology. There are already many methods for determining whether images are similar, such as image-aware hashing algorithms and face recognition algorithms.

[0062] S1033. Determine the change status of the corresponding grid cell to be detected based on the similarity.

[0063] It is easy to understand that after obtaining the similarity of the grid cells to be detected, it is possible to determine the changes of the corresponding grid image at different times.

[0064] For example, if multiple denoised grid images corresponding to the same grid at different time phases have a high similarity, it indicates that the images recorded in these multiple grid images are roughly the same and have not changed. In other words, the image corresponding to the grid cell in the pre-detection area has not changed. Therefore, based on the similarity, the change of the corresponding grid cell can be determined, and thus the change of the image in the pre-detection area can be known.

[0065] In some implementations, such as Figure 11 As shown, the step S103 of obtaining the change of at least one grid cell under different time phases further includes: S1031a, performing position registration on multiple grid images to be detected corresponding to the grid cell to be detected under different time phases.

[0066] In some embodiments, step S1031a comprises: taking one of the grid images to be detected as a reference image; moving the other grid image to be detected within a predetermined range for a predetermined number of times, and forming an Euclidean distance value each time the grid image is moved, thereby obtaining a plurality of Euclidean distance values; comparing the sizes of the plurality of Euclidean distance values; and if the Euclidean distance value is the smallest, the positions of the two grid images corresponding to the movement are registered. It is easy to understand that the position registration of the plurality of grid images can be sequentially implemented by analogy with the above method.

[0067] Commonly used remote sensing images are mostly L2-level data, which have been geometrically coarsely corrected. Based on this, the above-mentioned "automatic grid micro-correction" method proposed in the present application fixes one grid, moves another grid within a certain range, compares the Euclidean distance values of each movement, and completes grid correction when the Euclidean distance is the smallest, i.e. the position registration is completed. The above-mentioned method can solve the problem of difficulty in geometrically fine correction of high spatial resolution remote sensing images, and can effectively reduce the workload of manual registration of high spatial resolution images.

[0068] In some embodiments, step S1031 comprises determining at least one grid unit to be detected according to the plurality of grid units, comprising: removing interference grid units in all grid units to obtain the at least one grid unit to be detected. Because some grid units contain grid images that are noise interference images, such as cloud images and the like, and cannot reflect the changes of the ground object images in the pre-detection area, they need to be removed.

[0069] In some embodiments, as shown in FIG. 10B, step S1032 comprises: Figure 12

[0070] S10321, obtaining the attribute features of each grid image to be detected.

[0071] The attribute features are used to reflect the properties or characteristics of the grid images, so as to facilitate the similarity comparison of two or more images. Exemplarily, the attribute features can be shape, color and the like, so that the similarity comparison can be performed through the shape, color and the like of the images. Further, the shape, color and other features of the images can be presented through different feature expression methods, such as HOG texture features, image perceptual hash value features, spectral curve features, image structure similarity features and the like.

[0072] S10322, judging the similarity according to the attribute features.

[0073] ​There are various methods for judging the similarity according to the attribute features. For example, the Euclidean distance method, and the specific method described below.

[0074] In some embodiments, the attribute features include at least one of a HOG texture feature, an image perceptual hash value feature, a spectral curve feature, and an image structural similarity feature. It should be noted that the similarity judgment by the HOG texture, hash value, spectral curve, and image structural similarity features is a method for denoising and purifying the original image. If the original image (i.e., the image without the HOG texture feature, image perceptual hash value feature, spectral curve feature, and image structural similarity feature) is directly compared, there will be a problem of too much noise, and a good result is often not obtained. Only by denoising through the above-mentioned related features and comparing through the features can a more accurate similarity be obtained.

[0075] It should be noted that the HOG texture feature can only focus on the corner information of the image, and is used to outline the contour information of the image. The hash value is equivalent to the fingerprint information of the image. The spectral curve can effectively express the spectral information of the image. The SSIM image structural similarity feature mainly represents the spatial structure information of the image. More specifically, the HOG texture feature, image perceptual hash value feature, spectral curve feature, and image structural similarity feature of the aforementioned grid image can be extracted through an image algorithm module. For example, the image perceptual hash value feature can be extracted through a perceptual hash algorithm. The spectral curve feature can be extracted through a principal curve algorithm.

[0076] In addition, the embodiments of the present application can also combine the feature extraction function with the data reading mode to visualize, and visualize the two-dimensional data into a one-dimensional curve. The image information of the grid unit is read in the array thinning, sequential reading, and Z sequence reading modes, and is curve-shaped, so that the image features of the grid unit can be better reflected. Specifically, an image is composed of pixels in two directions of length and width, and different effects can be obtained by different reading modes of image pixels. The following examples are given: (1) Sequential reading: reading the pixels of the image row by row or column by column; (2) Z sequence reading: dividing the image into a plurality of small cells, and dividing each cell into two cells until the cell cannot be divided, and reading the image pixels by the cell, so that the spatial distribution information of the image can be ensured; (3) Array thinning: reading the image pixels by sampling, for example, reading the image data by sampling one point every 10 pixels.

[0077] Further, in some embodiments, the number of the plurality of grid images to be detected is 2, the number of the attribute features included is at least 1, and step S10322 of judging the similarity between the plurality of grid images to be detected according to the attribute features includes the following steps, such asFigure 13 As shown:

[0078] S103221. Obtain the difference value of each attribute feature of two grid images to be detected at different time phases, and generate at least one difference value.

[0079] Specifically, taking the attribute features as the perceptual hash value features as an example, we can first obtain the perceptual hash value feature values ​​of the grid image to be detected at times T1 and T2 respectively, and then subtract the two perceptual hash value feature values ​​to obtain the difference value of the corresponding perceptual hash value features.

[0080] In some implementations, the attribute features include image perceptual hash value features, spectral curve features, and structural similarity features, with a grid cell count of 21 and corresponding codes ranging from 1 to 21. At least one difference value derived therefrom can be found in Table 1.

[0081] Table 1

[0082]

[0083]

[0084] S103222. Generate a new feature value for the corresponding grid cell to be detected based on the at least one difference value.

[0085] In some implementations, the new feature value can be specifically calculated according to the following formula (1):

[0086]

[0087] Where n is the number of attribute features, x i ω is the difference value of the i-th attribute feature. i F(x) represents the weight of the difference value of the i-th attribute feature. F(x) represents the new feature value formed by combining all the attribute features of the grid cell. For example, for grid 1 in Table 1, x1, as the first feature value, can be set as the difference value of the perceptual hash value feature, and W1 is the weight of the difference value of the perceptual hash value feature among all feature values; similarly, the difference value of the spectral curve feature can be set as the second feature value.

[0088] S103223. Determine the similarity based on the new feature value.

[0089] Specifically, a threshold can be set. When the aforementioned F(x) is less than the threshold, it indicates that the similarity is not high, which means that the grid cell to be detected has changed.

[0090] In other embodiments, S10322, judging the similarity between the plurality of grid images to be detected according to the attribute characteristics can also be achieved by using the "Euclidean distance method". Unlike the traditional pixel algebraic operation, the "Euclidean distance method" needs to read the spectral characteristics in Z sequence and generate a feature curve by thinning and denoising, and then compare the feature curves by using the "grid automatic micro-correction" and "brightness matching" algorithms. The smaller the Euclidean distance, the more similar the images. Further, in some embodiments, as shown in Figure 14 The split-grid-based remote sensing image change detection method further includes the following steps:

[0091] S104, judging whether there is a changed grid cell in the at least one grid cell, and if there is, obtaining the change type of the changed grid cell.

[0092] According to the change detection result generated by the changed grid cell, the user can preliminarily judge the change type of the changed grid cell in combination with external data. If the change type can be determined, the attribute of the changed grid cell can be directly modified. For the grid cell whose change type is not clear, the "on-site visit" method can be used to determine the change type. According to the detection results in different periods, longitudinal tracking can also be performed to realize historical monitoring of the changed plot and obtain the change rule of various ground objects in the region, thereby playing a data supporting role for the management, governance and decision-making of the region.

[0093] Each grid cell on the earth represents a unique geographical space. Each geographical space can be spatially encoded as an index to record the information that has occurred in the space, which is the index function of split coding. For example, by using the Beidou grid code to split and code the remote sensing image map, the obtained remote sensing image becomes one of the numerous information of the space. After grid change detection, it can be known that which grid has changed. By using the coding of the changed grid cell, it can be known whether the space contains information of spatial change. For example, the space has carried out an old factory area reconstruction project in this time period, so it can be known that the change type of the grid cell is "building reconstruction". In this way, the existing prior knowledge can be combined to assist in judging the change type of the grid cell.

[0094] There can be many grid change types, which are illustrated by examples as follows: taking land as an example, there can be the following changes: (1) bare land -> forest land; (2) water area -> building group; (3) wasteland -> grassland. Taking the degree of air pollution as an example, there can be the following changes: CO concentration exceeds the standard -> SO concentration exceeds the standard.

[0095] In general, the change type of the grid cell is various, depending on the business requirement. The change of the grid can be determined by the attribute of the grid cell, and there are two types of change and non-change. More specific change type needs to be determined by combining multi-source data or other machine models.

[0096] In the embodiment of the present application, the grid and external multi-source data can be fused by using the encoding of the grid cell as an index, and the type of the grid cell can be determined with the aid of prior knowledge.

[0097] For example Figure 15 As shown in one application example, the external information of the changed grid cell can be viewed through the Beidou grid code, and the earthmoving situation of the grid cell can be known through the external information, so that the change type of the grid cell can be determined.

[0098] In some embodiments, the change detection method based on the split grid remote sensing image further includes: merging the two grid cells that have changed, then labeling the change category and making machine learning training data, and establishing a machine learning model for determining whether the two different grid cells have changed.

[0099] <Second aspect>

[0100] Based on the same idea as the first aspect, as Figure 16 shown, the second aspect provides a change detection method based on a split grid remote sensing image, including the following steps:

[0101] S201, acquiring a plurality of remote sensing images of a pre-detection area at different time phases.

[0102] Specifically, the remote sensing image of the pre-detection area can be captured by a remote sensing photographing device at different times, so as to acquire a plurality of remote sensing images. For example, the remote sensing image of the pre-detection area can be captured at T1, T2 and T3, so as to obtain the remote sensing image corresponding to different time phases. It should be noted that the remote sensing photographing device can be a known device, such as a camera. After acquiring a plurality of remote sensing images, they can be stored in a pre-set image database for standby, and when needed, the remote sensing image can be acquired from the image database.

[0103] S202, performing a split coding processing on the plurality of remote sensing images according to the same split coding mode, to form a plurality of grid cells, wherein each grid cell has a corresponding unique code and a plurality of grid images. In some embodiments, a correspondence between each remote sensing image and the plurality of grid cells, and a correspondence between the grid cells and the code and the plurality of grid images can also be established, and the various correspondences and the grid cells, the code and the grid images are stored in an image database. In addition, an index can also be established according to the correspondence, and is also stored in the image database.

[0104] S203, obtaining a plurality of target grid images corresponding to the grid cell with the target code at different time phases. It is easy to understand that how to obtain the corresponding grid image according to the code can be realized by the prior art, so it will not be repeated here.

[0105] S204, judging the similarity between the plurality of target grid images. The similarity between the plurality of target grid images can be calculated according to the prior art, for example, a known image similarity algorithm. In addition, the similarity between the plurality of target grid images can also be understood or implemented according to the part of judging the similarity in the embodiment of the first aspect.

[0106] S205, judging whether the image in the pre-detection area changes according to the similarity. Since the plurality of target grid images are obtained from the plurality of remote sensing images at different time phases, the plurality of target grid images can reflect the image in the pre-detection area corresponding to the target grid at different time phases. Specifically, the degree of change of the image is inversely proportional to the similarity, that is, if the similarity between the plurality of target grid images is higher, the degree of change of the image is lower, and vice versa. In this way, the similarity can be used to judge whether the image changes.

[0107] According to the method of the embodiment of the present application, the change of the image in the part of the pre-detection area corresponding to each grid cell is judged one by one, and the overall change of the image in the pre-detection area can be known. Thus, it can be judged whether the ground object in the pre-detection area changes in a period of time.

[0108] It should be noted that the image can be the image of the ground object in the pre-detection area. For example, when the pre-detection area is a wasteland area, the ground object can be land; when the pre-detection area is a water flow area, the ground object can be a river channel, water flow, etc.

[0109] In some embodiments, the split grid-based remote sensing image change detection method further comprises: S203a, performing position registration on the plurality of target grid images. This step is performed before step S204.

[0110] In some embodiments, step S203a comprises: taking one of the target grid images as a reference image; moving the other target grid image to be registered within a predetermined range for a predetermined number of times, each movement forming an Euclidean distance value, thereby obtaining a plurality of Euclidean distance values; comparing the sizes of the plurality of Euclidean distance values; and if the Euclidean distance value is the smallest, the positions of the two grid images corresponding to the movement are registered. It is easy to understand that the position registration of the plurality of target grid images can be sequentially implemented by analogy with the above method.

[0111] In some embodiments, the method for detecting changes in remote sensing images based on a dissected grid further comprises the following step: S206, judging whether the target grid unit is a changed grid unit, and if so, obtaining the change type of the target grid unit.

[0112] According to the change detection result of the target grid unit and in combination with external data, the user can preliminarily judge the change type of the target grid unit indoors, and if the type is clear, the attribute of the grid unit can be directly modified, and for the grid unit whose change type is unclear, the change type can be determined by means of outdoor "field visit". According to the detection results in different periods, longitudinal tracking can be performed to realize historical monitoring of the change patches and obtain the change rules of various ground objects in the region, thereby playing a data supporting role for the management, governance and decision-making of the region.

[0113] <Third aspect>

[0114] Based on the same idea as the first aspect, the third aspect provides a method for detecting changes in remote sensing images based on a dissected grid, comprising the following steps:

[0115] S301, selecting a pre-detection area in dynamic remote sensing images of multiple time phases; specifically, the dynamic remote sensing images comprise a plurality of remote sensing images, and the time phases of the plurality of remote sensing images are different. For example, the dynamic remote sensing images of multiple time phases comprise a plurality of remote sensing images from time TI to TN.

[0116] S302, dissecting the pre-detection area according to geographic coordinates to generate a plurality of grid units, and spatially encoding each grid unit; wherein each grid unit has a corresponding unique code and a plurality of grid images. The code can be allocated to each grid unit of the earth space grid system by using the structured indexing technology of geographic space position, thereby realizing the unified identification of the position of the grid unit. More specifically, the dissecting and encoding method herein can be understood or implemented by referring to the description of the dissecting and encoding part in the first aspect or the second aspect.

[0117] S303, obtain the change situation of at least one grid unit in different time phases, for judging the image change situation in the pre-detection area. Specifically, it can be understood or implemented by referring to the relevant records in the first aspect or the second aspect described above. In some embodiments, the remote sensing image change detection method based on the split grid further comprises the following steps:

[0118] S304, judge whether there is a change grid unit in at least one grid unit, if there is, obtain the change type of the change grid unit. Specifically, it can be understood or implemented by referring to the relevant records in the first aspect or the second aspect described above.

[0119] <Fourth aspect>

[0120] Based on the same idea as the first aspect, the fourth aspect also provides a remote sensing image change detection device based on split grid, for executing the remote sensing image change detection method based on split grid of the first aspect, the remote sensing image change detection device based on split grid comprises:

[0121] An image acquisition module is configured to acquire a plurality of remote sensing images of a pre-detection area in different time phases.

[0122] A split coding processing module is configured to perform split coding processing on the plurality of remote sensing images according to the same split coding manner, to form a plurality of grid units, wherein each grid unit has a corresponding unique code and a plurality of grid images.

[0123] A change detection module is configured to obtain the change situation of at least one grid unit in different time phases, for judging the image change situation in the pre-detection area.

[0124] <Fifth aspect>

[0125] Based on the same idea as the second aspect, the fifth aspect also provides a remote sensing image change detection device based on split grid, for executing the remote sensing image change detection method based on split grid of the second aspect, the remote sensing image change detection device based on split grid comprises:

[0126] An image acquisition module is configured to acquire a plurality of remote sensing images of a pre-detection area in different time phases.

[0127] A split coding processing module is configured to perform split coding processing on the plurality of remote sensing images according to the same split coding manner, to form a plurality of grid units, wherein each grid unit has a corresponding unique code and a plurality of grid images.

[0128] A target grid image acquisition module is configured to obtain a plurality of target grid images corresponding to the grid unit corresponding to the target code in different time directions.

[0129] The similarity judgment module is configured to judge the similarity between the plurality of target grid images.

[0130] The change detection module is configured to judge whether the image in the pre-detection area has changed according to the similarity.

[0131] In some embodiments, the split-grid-based remote sensing image change detection device further comprises:

[0132] The image registration module is configured to perform position registration on the plurality of target grid images.

[0133] In some embodiments, the change detection module is further configured to judge whether the target grid unit is a changed grid unit, and if so, to obtain the change type of the target grid unit.

[0134] <Sixth aspect>

[0135] Based on the same concept as the third aspect, the sixth aspect further provides a split-grid-based remote sensing image change detection device for executing the split-grid-based remote sensing image change detection method of the third aspect, the split-grid-based remote sensing image change detection device comprising:

[0136] The pre-detection area selection module is configured to select a pre-detection area in the multi-temporal dynamic remote sensing image.

[0137] The split coding processing module is configured to perform split processing on the pre-detection area according to geographic coordinates, to generate a plurality of grid units, and to perform spatial coding on each grid unit; wherein each grid unit has a corresponding unique code and a plurality of grid images.

[0138] The change detection module is configured to obtain the change of at least one grid unit at different time phases, and to judge the image change in the pre-detection area.

[0139] <Seventh aspect>

[0140] On the basis of the foregoing, the seventh aspect further provides a computer terminal comprising a memory, a processor, and a computer instruction program stored on the memory and executable on the processor, wherein the computer instruction program, when executed by the processor, implements the steps of the split-grid-based remote sensing image change detection method as described in the foregoing embodiments.

[0141] <Eighth aspect>

[0142] Based on the same idea, the eighth aspect also provides a computer readable storage medium, wherein a computer instruction program is stored on the computer readable storage medium, and the computer instruction program is executed by a processor to implement the steps of the method for detecting changes in remote sensing images based on a split grid as described in the foregoing aspects and embodiments.

[0143] While the embodiments of the application have been disclosed as above, they are not limited to the use listed in the specification and embodiments, and can be fully applied to various fields suitable for the application, and additional modifications can be easily made by those skilled in the art, and therefore the application is not limited to specific details and the figures shown and described herein, without departing from the general concept defined by the claims and the equivalent scope.

Claims

1. A method for detecting changes in remote sensing images based on a grid, characterized in that, include: Acquire multiple remote sensing images of the pre-detection area at different time phases; Multiple remote sensing images are segmented and coded using the same segmentation and coding method to form multiple grid cells, where each grid cell is configured with a unique code; The changes of at least one grid cell in different time phases are obtained to determine the image changes in the pre-detection area; The step of obtaining the variation of at least one grid cell under different time phases includes: Remove interfering grid cells from multiple grid cells to identify at least one grid cell to be detected; The method for obtaining the similarity between multiple grid images corresponding to each grid cell under different time phases includes: obtaining multiple attribute features of each grid image, and judging the similarity based on the multiple attribute features. Each attribute feature is different and is selected from any one of HOG texture features, image perceptual hash value features, spectral curve features, and image structure similarity features. The changes in the corresponding grid cells to be detected are determined based on the similarity. Similarity is determined based on multiple attribute features, including: Obtain the difference value of each attribute feature of the grid image to be detected, and obtain multiple difference values; Generate new feature values ​​for the corresponding grid cells to be detected based on all the difference values; The similarity is determined based on the new feature value; The new eigenvalue is calculated according to the following formula (1): Where n is the number of attribute features, x i ω is the difference value of the i-th attribute feature. i F(x) represents the difference weight of the i-th attribute feature, and F(x) represents the new feature value formed by combining all the attribute features of the grid cell to be detected.

2. The remote sensing image change detection method based on subdivided grids according to claim 1, characterized in that, Also includes: Determine whether there is a changing grid cell in at least one grid cell. If so, obtain the change type of the changing grid cell.

3. A method for detecting changes in remote sensing images based on a grid, characterized in that, include: Selecting pre-detection areas in multi-temporal dynamic remote sensing images; The pre-detection area is segmented based on geographic coordinates to generate multiple grid cells, and each grid cell is spatially encoded; each grid cell has a corresponding unique code and multiple grid images. The changes of at least one grid cell in different time phases are obtained to determine the image changes in the pre-detection area; The step of obtaining the variation of at least one grid cell under different time phases includes: Remove interfering grid cells from multiple grid cells to identify at least one grid cell to be detected; The method for obtaining the similarity between multiple grid images corresponding to each grid cell under different time phases includes: obtaining multiple attribute features of each grid image, and judging the similarity based on the multiple attribute features. Each attribute feature is different and is selected from any one of HOG texture features, image perceptual hash value features, spectral curve features, and image structure similarity features. The changes in the corresponding grid cells to be detected are determined based on the similarity. Similarity is determined based on multiple attribute features, including: Obtain the difference value of each attribute feature of the grid image to be detected, and obtain multiple difference values; Generate new feature values ​​for the corresponding grid cells to be detected based on all the difference values; The similarity is determined based on the new feature value; The new eigenvalue is calculated according to the following formula (1): Where n is the number of attribute features, x i ω is the difference value of the i-th attribute feature. i F(x) represents the difference weight of the i-th attribute feature, and F(x) represents the new feature value formed by combining all the attribute features of the grid cell to be detected.

4. A remote sensing image change detection device based on a grid, characterized in that, include: The image acquisition module is used to acquire multiple remote sensing images of the pre-detection area at different time phases; The segmentation and coding processing module is used to segment and code multiple remote sensing images according to the same segmentation and coding method to form multiple grid units, wherein each grid unit has a corresponding unique code and multiple grid images; The change detection module is used to acquire the change of at least one grid cell under different time phases, and to determine the image change in the pre-detection area; The step of obtaining the variation of at least one grid cell under different time phases includes: Remove interfering grid cells from multiple grid cells to identify at least one grid cell to be detected; The method for obtaining the similarity between multiple grid images corresponding to each grid cell under different time phases includes: obtaining multiple attribute features of each grid image, and judging the similarity based on the multiple attribute features. Each attribute feature is different and is selected from any one of HOG texture features, image perceptual hash value features, spectral curve features, and image structure similarity features. The changes in the corresponding grid cells to be detected are determined based on the similarity. Similarity is determined based on multiple attribute features, including: Obtain the difference value of each attribute feature of the grid image to be detected, and obtain multiple difference values; Generate new feature values ​​for the corresponding grid cells to be detected based on all the difference values; The similarity is determined based on the new feature value; The new eigenvalue is calculated according to the following formula (1): Where n is the number of attribute features, x i ω is the difference value of the i-th attribute feature. i F(x) represents the difference weight of the i-th attribute feature, and F(x) represents the new feature value formed by combining all the attribute features of the grid cell to be detected.

5. A remote sensing image change detection device based on a grid, characterized in that, include: The pre-detection region rotation module is used to select pre-detection regions in multi-temporal dynamic remote sensing images; The segmentation and encoding processing module is used to segment the pre-detection area according to geographic coordinates, generate multiple grid cells, and perform spatial encoding on each grid cell; wherein, each grid cell has a corresponding unique code and multiple grid images; The change detection module is used to acquire the change of at least one grid cell under different time phases, and to determine the image change in the pre-detection area; The step of obtaining the variation of at least one grid cell under different time phases includes: Remove interfering grid cells from multiple grid cells to identify at least one grid cell to be detected; The method for obtaining the similarity between multiple grid images corresponding to each grid cell under different time phases includes: obtaining multiple attribute features of each grid image, and judging the similarity based on the multiple attribute features. Each attribute feature is different and is selected from any one of HOG texture features, image perceptual hash value features, spectral curve features, and image structure similarity features. The changes in the corresponding grid cells to be detected are determined based on the similarity. Similarity is determined based on multiple attribute features, including: Obtain the difference value of each attribute feature of the grid image to be detected, and obtain multiple difference values; Generate new feature values ​​for the corresponding grid cells to be detected based on all the difference values; The similarity is determined based on the new feature value; The new eigenvalue is calculated according to the following formula (1): Where n is the number of attribute features, x i ω is the difference value of the i-th attribute feature. i F(x) represents the difference weight of the i-th attribute feature, and F(x) represents the new feature value formed by combining all the attribute features of the grid cell to be detected.

6. A computer terminal, characterized in that, include: The memory, the processor, and the computer instruction program stored in the memory and executable on the processor, wherein the computer instruction program, when executed by the processor, implements the steps of the remote sensing image change detection method based on any one of claims 1-3.

Images