Intelligent monitoring system for urban land space change based on data processing

Abstract

The application relates to the technical field of land monitoring, in particular to an intelligent urban land space change monitoring system based on data processing, which comprises a processor and a memory, the processor executes a computer program stored in the memory to realize the following steps: acquiring a suspected land change region, screening the suspected land change region according to pixel values of pixel points in the suspected land change region and each edge section of the suspected land change region, obtaining a shadow region and a land change region on a current shadow correction image, performing shadow correction on the shadow region according to a non-shadow region adjacent to the shadow region, obtaining a current target image, and re-performing region change judgment on the current target image to obtain a land non-change region and a land change region on the current target image. The application can improve the effect of monitoring and identifying urban land space changes.

Description

Technical Field

[0001] This invention relates to the field of land monitoring technology, and in particular to an intelligent monitoring system for urban land spatial change based on data processing. Background Technology

[0002] Currently, in order to ensure the rationality of urban land planning and maintain the sustainable development of cities, it is usually necessary to continuously monitor and analyze urban land changes, or to monitor and identify areas of land that have not changed and areas of land change in cities. In other words, continuous monitoring and analysis of urban land changes, or monitoring and identifying areas of land that have not changed and areas of land change in cities, is crucial for urban planning, management and sustainable development.

[0003] Current technologies primarily rely on the color features of acquired urban remote sensing images for regional division and land type identification. For example, existing techniques such as image segmentation and classification models are used to directly segment and identify regions within the acquired urban remote sensing images, distinguishing between unchanged and changed land areas. However, the presence of tall buildings in cities creates shadow areas, which can interfere with the region identification and classification process. This can lead to incomplete or inaccurate identification of unchanged and changed land areas, such as misidentifying some unchanged areas as changed areas. Consequently, the effectiveness of urban land change monitoring is poor, impacting subsequent urban planning and management. Therefore, improving the effectiveness of urban land change monitoring is a pressing issue that needs to be addressed. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide an intelligent monitoring system for urban land spatial change based on data processing to solve the problem of poor monitoring effect of urban land change.

[0005] This invention provides an intelligent monitoring system for urban land spatial change based on data processing, including a memory, a processor, and a computer program stored in the memory and running on the processor. The system is characterized in that the processor executes the computer program to perform the following steps:

[0006] Acquire the current image of the monitored city to be shaded, the image of nearby historical targets, and the area to be analyzed on the current image to be shaded;

[0007] Based on the curvature of each edge pixel on the outer contour edge of the region to be analyzed, the outer contour edge of the region to be analyzed is segmented to obtain each edge segment of the region to be analyzed. Based on the pixel value difference between the region to be analyzed and the pixels at the same position on the neighboring historical target image, the change probability index value of the region to be analyzed is obtained. Based on the change probability index value, the region change of the current image to be shading is judged to obtain the land non-change region and the suspected land change region on the current image to be shading. Based on the pixel value of the pixel in the suspected land change region and each edge segment of the suspected land change region, the suspected land change region is filtered to obtain the shadow region and land change region on the current image to be shading.

[0008] Based on the non-shaded areas adjacent to the shaded area, shadow correction is performed on the shaded area to obtain the current target image. Then, the current target image is re-evaluated for regional changes to obtain the land-unchanged areas and land-changed areas on the current target image.

[0009] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0010] This invention first acquires the current image of the city to be shaded, the nearby historical target image, and the region to be analyzed on the current image. Then, based on the curvature of each edge pixel on the outer contour edge of the region to be analyzed, the outer contour edge of the region is segmented to obtain each edge segment of the region. Based on the pixel value difference between the region to be analyzed and the pixels at the same position in the nearby historical target image, a change probability index value for the region to be analyzed is obtained. Based on the change probability index value, the current image to be shaded is judged to determine regional changes, resulting in non-changed land areas and suspected land change areas on the current image. Based on the pixel values ​​of the pixels in the suspected land change areas and each edge segment of the suspected land change areas, the suspected land change areas are filtered to obtain the shaded areas and land change areas on the current image. Then, based on the land change areas and non-changed land areas adjacent to the shaded areas, shading correction is performed on the shaded areas to obtain the current target image. Finally, the current target image is re-judged to determine regional changes, resulting in the non-changed land areas and land change areas on the current target image. Furthermore, by accurately and reliably identifying the shadowed areas, then correcting the shadowed areas, and using the image after shadow correction to monitor changes in urban land space, this invention can improve the completeness, accuracy, and reliability of monitoring and identifying land change areas and land non-change areas in the city, or improve the effectiveness of monitoring and identifying changes in urban land space. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart of an intelligent monitoring method for urban land spatial changes based on data processing, provided in Embodiment 1 of the present invention. Detailed Implementation

[0013] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0014] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0015] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0016] The specific scenario targeted by this invention is cities that need to monitor land spatial changes. Therefore, this invention corrects the shadow areas in the collected remote sensing images of cities, thereby improving the effectiveness of urban land spatial change monitoring.

[0017] This invention provides an intelligent monitoring system for urban land spatial change based on data processing, including a processor and a memory. The processor executes a computer program stored in the memory to implement an intelligent monitoring method for urban land spatial change based on data processing. Figure 1 As shown, the method includes the following steps:

[0018] Step S001: Obtain the current image to be shading corrected for the monitored city, the nearby historical target image, and the area to be analyzed on the current image to be shading corrected.

[0019] The main purpose of this embodiment is to improve the monitoring effect of urban land change by correcting the shadow areas in the acquired remote sensing images of the city. In addition, for the convenience of subsequent description and analysis, this embodiment will take the monitoring process of land space change of any city that needs to be monitored as an example, and refer to it as the monitored city.

[0020] After identifying the cities requiring land space change monitoring, remote sensing images of these cities are acquired using optical satellite sensors. These sensors can acquire remote sensing images of the cities at any given monitoring time. This embodiment primarily monitors land space changes in the cities at the current time. Therefore, this embodiment requires acquiring remote sensing images of the cities at the current monitoring time. The remote sensing images of the cities acquired by the optical satellite sensors at the current monitoring time are recorded as the current initial remote sensing images of the cities at the current monitoring time, while the remote sensing images of the cities acquired at historical monitoring times are recorded as the historical initial remote sensing images of the cities at historical monitoring times.

[0021] In this embodiment, the monitoring frequency can be adjusted according to the actual situation such as the rate of change of the monitored city and the required monitoring accuracy. For example, since urban land changes usually require a certain construction time, meaning there will not be significant changes in a short period of time, a longer interval can be set for monitoring, such as monitoring every 10 days. Furthermore, because monitoring land changes requires comparing pixels in historical images, it is necessary to ensure that the shooting location and shooting parameters remain consistent when collecting remote sensing images of the same city at different times. Alternatively, geometric and orthorectification can be performed using existing systems. In this embodiment, the initial remote sensing image of the monitored city at any given time is the image after geometric and orthorectification. The coordinates of the same location in the monitored city are the same in remote sensing images at different times in this embodiment, as geometric and orthorectification is a known technique.

[0022] Furthermore, due to the influence of sunlight conditions, the brightness and color temperature of remote sensing images acquired at different times may vary. To facilitate comparison between historical and current images, or to ensure the reliability and accuracy of subsequent analysis of current urban land changes based on historical land conditions, this embodiment requires color correction of the acquired initial remote sensing images. Since this embodiment monitors urban land spatial changes in real time, color and shadow corrections should also be performed in real time. Therefore, before correcting the current initial remote sensing image of the monitored city at the current monitoring time, or before acquiring the current initial remote sensing image, color and shadow corrections will be performed on the historical initial remote sensing images. This embodiment records the image after color and shadow correction as the target image. Therefore, before performing color correction on the current initial remote sensing image of the monitored city at the current monitoring time, this embodiment can obtain the image from before the current monitoring time. Historical target images of the monitored city at all historical monitoring times; the methods for color correction and shadow correction of any initial remote sensing image in this embodiment are the same as the methods for color correction and shadow correction of the current initial remote sensing image of the monitored city at the current monitoring time described later in this embodiment. That is, this embodiment will describe the process of color correction and shadow correction of the current initial remote sensing image of the monitored city at the current monitoring time as an example; in addition, it should be noted that the images in this embodiment are all RGB images. The pixel value of the pixel point on the RGB image is the RGB pixel value. The RGB image has three channels, which are called the red channel, green channel and blue channel respectively. In this embodiment, the red channel, green channel and blue channel are denoted as R channel, G channel and B channel respectively. That is to say, each pixel point on the image in this embodiment corresponds to the pixel value of the three channels, namely the R channel pixel value, G channel pixel value and B channel pixel value.

[0023] Since most of the land in two consecutively acquired images should not change significantly under normal circumstances, this embodiment, when performing color correction on the initial remote sensing image of the monitored city at a certain monitoring time, should refer to the image after color correction and shadow correction at the previous monitoring time. That is, when performing color correction on the current initial remote sensing image of the monitored city at the current monitoring time, this embodiment needs to refer to the historical target image of the monitored city at the previous adjacent historical monitoring time, and record the historical target image of the monitored city at the previous adjacent historical monitoring time as the neighboring historical target image at the current monitoring time. In other words, this embodiment will subsequently perform color correction on the current initial remote sensing image based on the difference in the same channel pixel values ​​between pixels at the same location on the current initial remote sensing image and the neighboring historical target image, obtaining the current shadow-corrected image of the monitored city at the current monitoring time. The specific process of this embodiment, based on the difference in the same channel pixel values ​​between pixels at the same location on the current initial remote sensing image and the neighboring historical target image, to obtain the current shadow-corrected image of the monitored city at the current monitoring time, is as follows:

[0024] First, calculate the in-channel pixel value difference between pixels at the same location in the current initial remote sensing image and the neighboring historical target image. The vector formed by these in-channel pixel value differences is denoted as the relative difference vector for each pixel in the current initial remote sensing image. The relative difference vector for the *a*-th pixel in the current initial remote sensing image consists of the R-channel relative difference, G-channel relative difference, and B-channel relative difference of the *a*-th pixel. The R-channel relative difference of the *a*-th pixel is the result of subtracting the R-channel target pixel value of the historical pixel at the same location from the initial R-channel pixel value of the *a*-th pixel. The G-channel relative difference of the *a*-th pixel is the result of subtracting the G-channel target pixel value of the historical pixel at the same location from the initial G-channel pixel value of the *a*-th pixel. The B-channel relative difference of the *a*-th pixel is the result of subtracting the R-channel target pixel value of the historical pixel at the same location from the initial B-channel pixel value of the *a*-th pixel. The result of the B-channel target pixel value of the historical co-position pixels of the pixel is obtained. The position coordinates of any pixel in the current initial remote sensing image are the same as the position coordinates of the historical co-position pixels of that pixel. The historical co-position pixels of each pixel in the current initial remote sensing image are all located on the nearest historical target images. The first relative difference in any relative difference vector is the R-channel relative difference, the second relative difference is the G-channel relative difference, and the third relative difference is the B-channel relative difference. Then, based on the relative difference vectors of each pixel in the current initial remote sensing image, DBSCAN clustering is performed on the pixels in the current initial remote sensing image, and the clusters obtained by clustering are all recorded as pixel clusters. Here, the distance between two pixels during clustering is the Euclidean distance between the relative difference vectors of the two pixels. The process of clustering the pixels in the current initial remote sensing image using the DBSCAN clustering algorithm based on the relative difference vectors of the pixels is a well-known technique.Since the relative difference vector corresponding to the cluster center of the largest pixel cluster among all obtained pixel clusters can represent the color difference between the current initial remote sensing image and the nearby historical target image, this embodiment selects the largest pixel cluster from all pixel clusters after obtaining the pixel clusters. The largest pixel cluster is the cluster with the most pixels. The relative difference vector corresponding to the cluster center of the selected largest pixel cluster is recorded as the target relative difference vector. Then, based on the difference between the initial pixel value of the pixel in the current initial remote sensing image and the relative difference in the same channel in the target relative difference vector, the color correction value of the pixel in the current initial remote sensing image is obtained, and the R of the a-th pixel in the current initial remote sensing image is... The channel color correction value is the result of subtracting the relative difference of the R channel in the target relative difference vector from the initial pixel value of the R channel of the a-th pixel. The color correction value of the G channel of the a-th pixel in the current initial remote sensing image is the result of subtracting the relative difference of the G channel in the target relative difference vector from the initial pixel value of the G channel of the a-th pixel. The color correction value of the B channel of the a-th pixel in the current initial remote sensing image is the result of subtracting the relative difference of the B channel in the target relative difference vector from the initial pixel value of the B channel of the a-th pixel. The R-channel color correction value, G-channel color correction value, and B-channel color correction value are all color correction values. Finally, the initial pixel value of each pixel in the current initial remote sensing image is replaced with the color correction value of the corresponding pixel, and the replaced image is recorded as the current shadow correction image.

[0025] In this embodiment, the pixel value of a pixel in the target image is called the target pixel value. The pixel value of any pixel in the target image includes the R-channel target pixel value, the G-channel target pixel value, and the B-channel target pixel value. The pixel value of a pixel in the initial remote sensing image is called the initial pixel value. The pixel value of any pixel in the initial remote sensing image includes the R-channel initial pixel value, the G-channel initial pixel value, and the B-channel initial pixel value. The pixel value of the image to be shading corrected is called the pixel value to be corrected. The pixel value of any pixel in the image to be shading corrected includes the R-channel pixel value to be corrected, the G-channel pixel value to be corrected, and the B-channel pixel value to be corrected.

[0026] Therefore, this embodiment completes the color correction of the current initial remote sensing image, obtaining the current shadow image to be corrected after color correction and the neighboring historical target image at the current monitoring time. The neighboring historical target image is also needed for subsequent shadow area identification. Furthermore, because tall buildings exist in cities, they easily create building shadow areas. The existence of shadow areas can interfere with the area identification and classification process, leading to incomplete or inaccurate identification of land-unchanged and land-changed areas. For example, some land-unchanged areas may be identified as land-changed areas, resulting in poor urban land change monitoring and affecting subsequent urban planning and management. Therefore, after obtaining the current shadow image to be corrected, this embodiment will perform shadow correction on it. That is, the monitoring of urban land spatial changes at the current monitoring time will be based on the shadow-corrected image. However, before performing shadow correction, the shadow area needs to be obtained first. Therefore, this embodiment needs to divide the current shadow image to be corrected into regions to obtain the region to be analyzed on the current shadow image. The region to be analyzed is the basis for subsequent shadow area identification. The specific process for obtaining the region to be analyzed on the current shadow image is as follows:

[0027] First, based on the pixel values ​​to be corrected in the current image to be shading corrected, k-means clustering is performed on the pixels in the current image to be shading corrected. All the resulting clusters are denoted as feature clusters. Pixels in the same cluster are pixels with similar colors. Here, the distance between two pixels is measured by the Euclidean distance between the pixel values ​​to be corrected. The process of obtaining clustering results using the k-means clustering algorithm under the premise of knowing the clustering objects is a well-known technique. Next, region growing is performed on the current image to be shading corrected. Each grown region is denoted as the region to be analyzed in the current image to be shading corrected. The growth criteria for region growing include spatial adjacency and feature similarity. Spatial adjacency refers to belonging to the eight-neighborhood of a pixel, and feature similarity refers to belonging to the same feature cluster. That is, for pixel 1 and pixel 2, only when the pixel values ​​are similar in color can the pixel values ​​be considered similar. Pixels 1 and 2 are classified into the same region only when they belong to the same feature cluster and Pixel 2 is also located within the eight-neighborhood of Pixel 1. In other words, this embodiment merges and classifies spatially adjacent (within the eight-neighborhood) pixels belonging to the same feature cluster in the current image to be shaded into the same region. Each region obtained by merging and dividing is the region to be analyzed in the current image to be shaded. The process of growing regions on the image to obtain the corresponding growing regions on the image under the premise of known growth criteria is a well-known technique. The reason for determining the region to be analyzed based on the clustering results is that land areas of the same type have high color similarity in remote sensing images, while land areas of different types have obvious color differences. For example, green areas, buildings, and highways have obvious color differences in remote sensing images. Therefore, belonging to the same cluster is used as the growth criterion for region growth. Furthermore, if a pixel is not adjacent to any pixel in its feature cluster, it may be caused by interference factors such as local reflection. To avoid such pixels interfering with region division, median filtering is used to filter such pixels before they are assigned to the corresponding region. In other words, in this embodiment, the initial pixel value of the pixel in the current initial remote sensing image can be replaced with the color correction value of the corresponding pixel before filtering the replaced image. The filtered image is then used as the current shadow correction image. The above can be used to divide the current shadow correction image obtained after replacing the color correction value and filtering. That is, if a pixel is not adjacent to any pixel in its feature cluster, this embodiment can replace the current shadow correction image with the filtered current shadow correction image.

[0028] Therefore, this embodiment can obtain the region to be analyzed on the current image to be shaded through the above process, and then further analyze the region to be analyzed to identify the shadow region.

[0029] Step S002: Based on the curvature of each edge pixel on the outer contour edge of the region to be analyzed, the outer contour edge of the region to be analyzed is segmented to obtain each edge segment of the region to be analyzed. Based on the pixel value difference between the region to be analyzed and the pixels at the same position on the nearby historical target image, the change probability index value of the region to be analyzed is obtained. Based on the change probability index value, the region change of the current image to be shading is judged to obtain the land non-change region and the suspected land change region on the current image to be shading. Based on the pixel value of the pixel in the suspected land change region and each edge segment of the suspected land change region, the suspected land change region is filtered to obtain the shadow region and land change region on the current image to be shading.

[0030] After obtaining the region to be analyzed, this embodiment needs to identify and analyze the region to obtain the shadow region within it, and then perform shadow correction on the shadow region. Before identifying the shadow region, this embodiment needs to obtain the edge segments of each region to be analyzed. The edge segments of the region to be analyzed are mainly used for shadow region identification and correction. The specific process of obtaining the edge segments of each region to be analyzed on the current image to be shadow corrected is as follows:

[0031] First, all outer contour edge pixels in each region to be analyzed are obtained. For any pixel in a region to be analyzed, if not all pixels belonging to its eight-neighborhood are located within that region, then that pixel is considered a contour edge pixel of that region. Then, all outer contour edge pixels in each region to be analyzed are connected to obtain the outer contour edges of each region. Points on the outer contour edges are recorded as edge pixels. Here, a known connection strategy is used to connect the outer contour edge pixels in the region to be analyzed. Next, the curvature values ​​of each edge pixel on the outer contour edge of the region to be analyzed are calculated. The calculation process for the curvature of edge pixels on the edge line is well-known. Each edge pixel on the outer contour edge of the region to be analyzed is then marked, obtaining the marked values. For any region to be analyzed, the specific process for obtaining the marked values ​​of each edge pixel on the outer contour edge of the region to be analyzed is as follows: Any edge pixel on the outer contour edge of the region to be analyzed is selected as the starting point, and from... Starting from the starting point, the edge pixels on the outer contour edge of the region to be analyzed are traversed and marked sequentially in a clockwise direction until all edge pixels on the outer contour edge of the region to be analyzed are marked. The edge pixel reached on the g-th traversal when traversing clockwise from the starting point is marked as g. In other words, the edge pixel reached on the g-th traversal when traversing clockwise from the starting point is marked as g. The edge pixels on the outer contour edge of the region to be analyzed that belong to the starting point... The edge pixel is the edge pixel encountered in the first traversal. For example, if edge pixel g1 on the outer contour edge of the region to be analyzed is the starting point, then the mark value of edge pixel g1 is 1. Starting from the starting point, the traversal proceeds in a clockwise direction. If the first edge pixel encountered is edge pixel g2 on the outer contour edge of the region to be analyzed, then the mark value of edge pixel g2 is 2. The traversal continues. If edge pixel g3 on the outer contour edge of the region to be analyzed is encountered after edge pixel g2 and before it is encountered, then the mark value of edge pixel g3 is 3.Then, based on the curvature and marker values ​​of the edge pixels on the outer contour edges of each region to be analyzed, segmentation points on the outer contour edges of each region to be analyzed are obtained. These segmentation points are then used to segment the outer contour edges of the corresponding regions to be analyzed, and each segmented edge is recorded as an edge segment of the corresponding region to be analyzed. The specific process for obtaining the segmentation points on the outer contour edges of any region to be analyzed, F, and its edge segments is as follows: A mapping space for region to be analyzed, F, is constructed. The horizontal axis of the mapping space represents the marker value, and the vertical axis represents the curvature value. The curvature and marker values ​​of each edge pixel on the outer contour edges of region to be analyzed are mapped to the mapping space of region to be analyzed, resulting in a mapping point corresponding to each edge pixel on the outer contour edges of region to be analyzed. That is, the points in the mapping space are mapping points. The horizontal coordinate of the mapping point corresponding to an edge pixel is the marker value of the corresponding edge pixel, and the vertical coordinate is the curvature value of the corresponding edge pixel. The mapping points in the mapping space are connected sequentially in ascending order of the marker values, and the connection result is recorded as the region to be analyzed, F. The feature curve is obtained by acquiring the inflection points on the feature curve. The acquisition of inflection points is a known technique. The edge pixels corresponding to the inflection points on the feature curve are all recorded as segmentation points on the outer contour edge of the region to be analyzed, F. If the mapping point corresponding to the edge pixel f0 on the outer contour edge of the region to be analyzed is an inflection point on the feature curve, then the edge pixel f0 is a segmentation point on the outer contour edge of the region to be analyzed, F. After obtaining all the segmentation points on the outer contour edge of the region to be analyzed, the outer contour edge of the region to be analyzed is segmented using the segmentation points on the outer contour edge of the region to be analyzed. The segmented edge segments are all recorded as edge segments of the region to be analyzed, and the edge line segments between adjacent segmentation points on the outer contour edge of the region to be analyzed, or the line segments formed by the edge pixels between adjacent segmentation points on the outer contour edge of the region to be analyzed, are edge segments of the region to be analyzed. All edge segments of the region to be analyzed belong to the outer contour edge of the region to be analyzed, and the edge segments formed by the edge pixels between adjacent segmentation points on the outer contour edge of the region to be analyzed include the corresponding segmentation points.

[0032] Therefore, this embodiment can obtain all edge segments of the region to be analyzed. After obtaining the edge segments of the region to be analyzed, this embodiment identifies the shadow region based on the pixel value difference between the region to be analyzed and the pixels at the same position in the nearby historical target image, as well as the edge segments of the region to be analyzed, to obtain the shadow region on the current image to be shadowed. The specific process of obtaining the shadow region is as follows:

[0033] First, based on the pixel value differences between the corresponding pixels in the current image to be shaded and the historical target images, the change probability index value of each region to be analyzed is obtained. The change probability index value reflects the probability that the region to be analyzed is a non-changing region, a changing region, or a shaded region. The specific process for obtaining the change probability index value of any region F to be analyzed is as follows: In the historical target images, obtain the corresponding pixels of each pixel in region F to be analyzed. The corresponding pixel of the b-th pixel in region F belongs to the historical target image and has the same coordinate position as the b-th pixel. That is, the coordinates of the b-th pixel in the current image to be shaded and the coordinates of the corresponding pixel of the b-th pixel in the historical target image are consistent. Calculate the mean of the absolute values ​​of the same-channel pixel value differences between each pixel in region F and its corresponding corresponding pixel, and record this as the corresponding pixel difference, i.e., the change probability index value of the region to be analyzed. The pixel difference at the same position of the b-th pixel in the analysis area is the average of the R-channel difference, G-channel difference, and B-channel difference of the b-th pixel in the analysis area. The R-channel difference of the b-th pixel in the analysis area is the absolute value of the difference between the R-channel pixel value to be corrected and the R-channel target pixel value of the b-th pixel at the same position. The G-channel difference of the b-th pixel is the absolute value of the difference between the G-channel pixel value to be corrected and the G-channel target pixel value of the b-th pixel at the same position. The B-channel difference of the b-th pixel is the absolute value of the difference between the B-channel pixel value to be corrected and the B-channel target pixel value of the b-th pixel at the same position. The average of the pixel differences at the same position of all pixels in the analysis area F is calculated and denoted as the probability index value of change of the analysis area F. The specific calculation expression for the probability index value of change of the analysis area F is as follows:

[0034]

[0035] in, Let F be the probability index value of change in the region to be analyzed, and B be the total number of pixels in the region to be analyzed, where B is the number of pixels in F. Let be the pixel difference at the same position of the b-th pixel in the area to be analyzed. The color difference between areas where the land has not changed in the current shadow correction image and the corresponding areas in the historical target image should be small, while areas where the land has changed will have a larger color difference compared to the corresponding pixels in the historical target image. Therefore, in this embodiment, based on the color difference between each area to be analyzed in the current shadow correction image and the corresponding pixels in the historical target image, the probability that each area to be analyzed is a land-unchanged area or a suspected land-changed area is calculated. It can reflect the probability that the area F to be analyzed is a land-unchanged area or a suspected land-changed area. Subsequently, based on the probability that each area to be analyzed is a land-unchanged area or a suspected land-changed area, land-unchanged areas and suspected land-changed areas can be identified. After identifying land-unchanged areas and suspected land-changed areas, the suspected land-changed areas will be analyzed again to identify the shaded areas.

[0036] After obtaining the change probability index values ​​for each region to be analyzed, the current shading correction image is assessed for regional changes based on these values. This process identifies the land-unchanged areas and suspected land-change areas on the current shading correction image. The specific process involves: Based on the change probability index values ​​for each region to be analyzed on the current shading correction image, performing a D... BSCAN clustering categorizes the clustering results as region clusters. Here, the absolute value of the difference in the probability of change index between the regions under analysis is used to measure the distance between them when clustering the region to be analyzed using the DBSCAN clustering algorithm. The clustering process of the DBSCAN algorithm is well-known. Since the colors of non-changeable regions remain largely unchanged, while the color change of changeable regions may vary due to different directions of change, the probability of change index values ​​of changeable regions are significantly different from those of non-changeable regions. Furthermore, in continuous images, due to the short intervals between images, most areas... All regions should be in a non-changing state. Therefore, among the clusters obtained above, the region to be analyzed within the cluster containing the most pixels or the largest area is the land non-changing region. The regions to be analyzed within other region clusters besides the one containing the most pixels or the largest area are suspected land change regions. That is, in this embodiment, the sum of the areas of all regions to be analyzed in each region cluster is recorded as the area of ​​the corresponding region cluster. The area of ​​the region to be analyzed is measured by the number of pixels in the corresponding region. The region cluster with the largest area is selected and recorded as the non-changing region cluster. All regions to be analyzed in the clusters are denoted as land-unchanged regions on the current image to be shaded. All regions to be analyzed in the clusters other than the land-unchanged regions are denoted as suspected land change regions on the current image to be shaded. In other words, all regions to be analyzed on the current image to be shaded except the land-unchanged regions are denoted as suspected land change regions on the current image to be shaded. Land-unchanged regions refer to regions that have changed compared to history, while suspected land change regions refer to regions that may change compared to history. That is, suspected land change regions may be shaded regions or land change regions, and further analysis is required.

[0037] As mentioned above, the shadowed areas in the image to be shaded will also be identified as suspected land change areas during the analysis process. Therefore, this embodiment needs to distinguish shadowed areas from the suspected land change areas, perform correction on the shadowed areas, and then analyze the land change based on the corrected image. In other words, this embodiment will next filter the suspected land change areas based on the pixel values ​​of the suspected land change areas in the image to be shaded and the edge segments of the suspected land change areas, thus obtaining the shadowed areas and land change areas in the image to be shaded. The specific process is as follows:

[0038] First, based on the pixel values ​​of suspected land change areas in the current image to be shaded and the edge segments of these areas, the shadow probability index value for each suspected land change area in the current image to be shaded is obtained. The shadow probability index value is crucial for filtering and identifying shadowed areas from suspected land change areas. Therefore, the specific process for obtaining the shadow probability index value for each suspected land change area in the current image to be shaded is as follows: For any suspected land change area Q:

[0039] First, the sum of the average values ​​of all channels of the suspected land change area Q is added to a preset constant, and then the normalized result of the reciprocal is taken as the first probability index value of the suspected land change area Q. The specific calculation expression for the first probability index value of the suspected land change area Q is as follows:

[0040]

[0041] in, The first probability index value for the suspected land change area Q is given by Norm(), which is the normalization function. This represents the mean value of the R channel pixel values ​​to be corrected for all pixels in the suspected land change area Q. The mean value of the pixel values ​​to be corrected in the G channel of all pixels in the suspected land change area Q. This represents the mean value of the B channel pixel values ​​to be corrected for all pixels in the suspected land change area Q. As a preset constant, This is to prevent the denominator from being zero, which would make the calculation meaningful. In specific applications, the implementer can set it according to the actual situation. For example, in this embodiment, it can be... Set to 0.01. And the shadow area is usually close to black, meaning the values ​​of all three channels are relatively small, so when The smaller, that is The larger the value, the closer the color of the suspected land change area Q is to black, and the greater the probability that the suspected land change area Q is a shaded area. Conversely, when the value is smaller... When it is larger, that is The smaller the value, the less likely the color of the suspected land change area Q is to be black, and the less likely the suspected land change area Q is to be a shaded area.

[0042] Then, the orientation angles of suspected land change areas on the current image to be shading corrected are obtained. The orientation angle of any edge segment is the angle between the fitted line obtained by fitting the edge segment using the least squares method and the horizontal axis of the current image to be shading corrected. The horizontal axis of the current image to be shading corrected is horizontal. That is, the orientation angle of any edge segment is the angle of the fitted line obtained by rotating counterclockwise from the positive direction of the horizontal axis of the current image to be shading corrected to the edge segment. Then, a set consisting of other suspected land change areas besides suspected land change area Q is obtained on the current image to be shading corrected, and is denoted as the reference set of suspected land change area Q. Then, the difference between the orientation angles of each edge segment of suspected land change area Q and the orientation angles of each suspected land change area in the reference set of suspected land change area Q is used to determine the orientation angle of suspected land change area Q. The absolute value and the first probability index value of each suspected land change area in the reference set of suspected land change area Q are used to obtain the weighted directional consistency index value between suspected land change area Q and each suspected land change area in the reference set of suspected land change area Q. The method for obtaining the first probability index value of any suspected land change area is the same as the method for obtaining the first probability index value of suspected land change area Q. The weighted directional consistency index value is the key to subsequently obtaining the second probability index value of suspected land change area Q. Therefore, the specific process for obtaining the weighted directional consistency index value between suspected land change area Q and each suspected land change area in the reference set of suspected land change area Q is as follows: For suspected land change area Q and the c-th suspected land change area in the reference set of suspected land change area Q:

[0043] Based on the absolute values ​​of the differences between the directional angles of each edge segment of the suspected land change area Q and the directional angles of each edge segment of the c-th suspected land change area, a set of directional angle differences corresponding to each edge segment of the suspected land change area Q is obtained. The directional angle differences in the set corresponding to any edge segment of the suspected land change area Q are composed of the absolute values ​​of the directional angle differences between that edge segment and each edge segment of the c-th suspected land change area. That is, the h-th directional angle difference in the set corresponding to any edge segment k0 of the suspected land change area Q is the absolute value of the difference between the directional angle of edge segment k0 and the directional angle of the h-th edge segment of the c-th suspected land change area. The number of directional angle differences in the set corresponding to edge segment k0 is equal to the number of edge segments in the c-th suspected land change area. Since the number of edge segments in the region is consistent, the smallest directional angle difference in the set of directional angle differences corresponding to each edge segment of the suspected land change region Q is taken as the representative directional angle difference of the corresponding edge segment. The representative directional angle difference of the edge segment represents the degree of proximity of the corresponding edge segment to the c-th suspected land change region in terms of direction. The mean of the representative directional angle differences of all edge segments of the suspected land change region Q is calculated and denoted as the mean of characteristic differences. A negative correlation mapping is performed on the mean of characteristic differences, and the mapping result is denoted as the initial direction consistency index value between the suspected land change region Q and the c-th suspected land change region. Here, the negative correlation mapping means taking the reciprocal of the mean of characteristic differences, and the calculation expression of the initial direction consistency index value between the suspected land change region Q and the c-th suspected land change region is:

[0044]

[0045] in, Let be the initial directional consistency index value between suspected land change area Q and the c-th suspected land change area in the reference set of suspected land change area Q. The min() function takes the minimum value of the total number of edge segments of the suspected land change area Q. This is the set of orientation angle differences corresponding to the nth edge segment of the suspected land change area Q. To obtain the minimum value of the direction angle difference set corresponding to the nth edge segment, which also represents the edge segment whose direction is closest to the nth edge segment of the suspected land change area Q among all edge segments of the cth suspected land change area. Since the shape of the shadow is often related to the angle of sunlight, or in other words, the tilt angles of all building shadows in the same image usually have a high degree of similarity, the larger the initial direction consistency index value between the suspected land change area Q and the cth suspected land change area, or in other words... The smaller the value, the more consistent the slope direction of the edge segment between the suspected land change area Q and the edge segment of the c-th suspected land change area. This also indicates that the suspected land change area Q is more likely to be a shaded area.

[0046] Since a higher first probability index value for suspected land change areas in the reference set of suspected land change areas Q indicates a more reliable and valuable initial direction consistency index value between suspected land change area Q and the suspected land change area Q, this embodiment uses the first probability index value for suspected land change areas in the reference set of suspected land change areas Q to weight the initial direction consistency index values ​​between suspected land change areas Q and each suspected land change area in the reference set of suspected land change areas Q, resulting in a weighted direction consistency index value between suspected land change areas Q and each suspected land change area in the reference set of suspected land change areas Q. The weighted direction consistency index value between suspected land change areas Q and the c-th suspected land change area in the reference set of suspected land change areas Q is the result of multiplying the initial direction consistency index value between suspected land change areas Q and the c-th suspected land change area by the first probability index value of the c-th suspected land change area. In other words, the weighted direction consistency index value between suspected land change areas Q and the c-th suspected land change area in the reference set of suspected land change areas Q is... , The first probability index value is for the c-th suspected land change area.

[0047] Next, the normalized mean of the weighted direction consistency index values ​​among the suspected land change areas Q and the reference set of suspected land change areas Q is calculated and denoted as the second probability index value of suspected land change area Q. The specific calculation expression for the second probability index value of suspected land change area Q is as follows:

[0048]

[0049] in, The second probability indicator value for Q, which represents a suspected land change area. This represents the total number of suspected land change areas in the reference set of suspected land change areas Q. Let be the initial directional consistency index value between suspected land change area Q and the c-th suspected land change area in the reference set of suspected land change area Q. The first probability index value is the c-th suspected land change area in the reference set of suspected land change areas Q; and When it is larger, The larger, and The larger the value of Q, the greater the likelihood that the suspected land change area Q is a shaded area, and vice versa. The smaller the value, the less likely the suspected land change area Q is to be a shaded area.

[0050] After obtaining the first and second probability index values ​​of the suspected land change area Q, the shadow probability index value of the suspected land change area Q is obtained by combining the first and second probability index values ​​of the suspected land change area Q. In this embodiment, the average of the first and second probability index values ​​of the suspected land change area Q is used as the shadow probability index value of the suspected land change area Q. The larger the shadow probability index value of the suspected land change area Q, the greater the probability that the suspected land change area Q is a shadow area.

[0051] Therefore, this embodiment can obtain the shadow probability index value of each suspected land change area through the above process. After obtaining the shadow probability index value of the suspected land change area, the shadow area is identified based on the shadow probability index value. For example, for a suspected land change area Q, it is determined whether the shadow probability index value of the suspected land change area Q is greater than a preset shadow threshold. If so, it indicates that the suspected land change area Q is a shadow area, and then the suspected land change area Q is recorded as a shadow area. Otherwise, it indicates that the suspected land change area Q is not a shadow area, but a land change area, and then the suspected land change area Q is recorded as a land change area. In this embodiment... In this embodiment, the implementer can set a preset shadow threshold based on the actual situation, such as the range of shadow probability index values. For example, a shadow area has obvious shadow characteristics in both color and edge direction, while other areas where land changes are difficult to satisfy the shadow area characteristics in both color and edge direction. Therefore, the shadow probability index value of the shadow area is significantly different from that of the actual area where land changes occur. That is, the shadow probability index value of the shadow area and the shadow probability index value of the area where land changes occur are usually distributed at both ends of the range of shadow probability index values. Therefore, in this embodiment, the preset shadow threshold is set to the median value of the range of shadow probability index values, that is, the preset shadow threshold is set to 0.5.

[0052] Therefore, this embodiment can obtain the shadow area, land change area and land non-change area on the current image to be shadowed through the above process.

[0053] Step S003: Based on the non-shaded areas adjacent to the shaded area, perform shadow correction on the shaded area to obtain the current target image, and re-determine the regional changes of the current target image to obtain the land non-change areas and land change areas on the current target image.

[0054] In this embodiment, after obtaining the shadow region on the current image to be shadow-corrected, the next step is to perform shadow correction on the shadow region on the current image to be shadow-corrected based on the non-shadow region adjacent to the shadow region, to obtain the current target image. The current target image is the image after shadow correction of the current image to be shadow-corrected. The specific process of performing shadow correction on the shadow region of the current image to be shadow-corrected to obtain the current target image is as follows:

[0055] For any shadow region T in the current image to be shaded: First, obtain the non-shaded regions adjacent to the shadow region T in the current image to be shaded. The non-shaded regions include land change areas and land unchanged areas. The set of all non-shaded regions adjacent to the shadow region T in the current image to be shaded is denoted as the set of adjacent regions of the shadow region T. All regions in the set of adjacent regions are denoted as adjacent regions. Since the real pixels of the shadow region T are usually adjacent to the shadow region T and are generally not the building areas that form the shadow region T, this embodiment needs to exclude the building areas that form the shadow region T when correcting the shadow region T. That is, in the set of adjacent regions of the shadow region T, obtain the adjacent region corresponding to the maximum value of the initial direction consistency index value between the shadow region T and the shadow region T, and denot it as the adjacent region. The building area of ​​the shadow region T is defined as follows: If the initial direction alignment index value between adjacent areas t and the shadow region T is the largest, then adjacent area t is the building area of ​​the shadow region T, and thus constitutes the building area that forms the shadow region T. Then, other non-shadow areas adjacent to the shadow region T, excluding the building areas of the shadow region T, are obtained and recorded as correction reference areas for the shadow region T. Finally, based on the pixel values ​​to be corrected in all correction reference areas of the shadow region T, the replacement pixel values ​​for each pixel in the shadow region T are obtained. The specific process for obtaining the replacement pixel values ​​for each pixel in the shadow region T based on the pixel values ​​to be corrected in all correction reference areas of the shadow region T is as follows:

[0056] Since urban land is typically divided in a relatively regular manner, the pixels in each edge segment of the correction reference region of the shaded region T are fitted using the least squares method based on their coordinates. The fitted line of each edge segment in the shaded region is thus the edge extension line. In other words, this embodiment uses the least squares method to fit each edge segment of the correction reference region, and the fitted line obtained is recorded as the edge line corresponding to the edge segment. That is, the fitted line obtained by fitting any edge segment is the edge line corresponding to that edge segment. To prevent the influence of the edge shape of some regions on subsequent segmentation, the extension lines that cross their own regions need to be filtered out to obtain all the dividing edge lines corresponding to the shaded region T. That is, to determine the edge line of any edge segment of any correction reference region. The system checks whether the extension of the corresponding edge line crosses the correction reference area. If it does not, the edge line corresponding to the edge segment is recorded as a dividing edge line corresponding to the shadow area T. If it does cross, the edge line corresponding to the edge segment is determined not to be a dividing edge line corresponding to the shadow area T. Then, all the dividing edge lines corresponding to the obtained shadow area T are extended towards the shadow area T until they cover the entire range of the shadow area T. The extended dividing edge lines will divide the shadow area T. In this embodiment, the sub-regions obtained by dividing the shadow area T by the extended dividing edge lines are all recorded as the divided sub-regions in the shadow area T. During the division, the edge extension line can only divide the shadow part adjacent to its own region and cannot cross other non-shadow areas. Next, the shaded region is segmented, and the correction reference region to which the edge line containing the boundary line of the segmented sub-region belongs is obtained. Within the correction reference region to which the edge line containing the boundary line of the segmented sub-region belongs, the correction reference region adjacent to the corresponding sub-region is obtained and recorded as the replacement region corresponding to the corresponding sub-region. That is, if the dividing line or boundary line of a certain sub-region is the extension of the edge line corresponding to the edge segment of a correction reference region belonging to the shaded region T, and the sub-region is also adjacent to that correction reference region, then the sub-region and the correction reference region are classified as the same region. In other words, the correction reference region is recorded as the replacement region corresponding to that sub-region. Then, each sub-region in the shaded region T is obtained. The average value of the pixel to be corrected in the replacement region corresponding to each sub-region is used as the average value of the pixel to be corrected in the replacement region corresponding to each sub-region. That is, for any sub-region of the shadow region T, the average value of the pixel to be corrected in the R channel, the average value of the pixel to be corrected in the G channel, and the average value of the pixel to be corrected in the B channel of each pixel in the replacement region corresponding to the sub-region are respectively the R channel replacement pixel value, G channel replacement pixel value, and B channel replacement pixel value of each pixel in the sub-region.This yields the replacement pixel values ​​for each pixel in the shaded region T.

[0057] After obtaining the replacement pixel values ​​of each pixel in each shadow region of the current image to be shaded, the pixel values ​​to be corrected in each shadow region of the current image to be shaded are replaced with the replacement pixel values ​​of the corresponding pixels, and the image after the replacement is recorded as the current target image, thus completing the shadow correction of the current image to be shaded. That is, for any pixel in the shadow region T, the pixel value to be corrected in the R channel of the pixel is replaced with the replacement pixel value in the R channel of the pixel, the pixel value to be corrected in the G channel of the pixel is replaced with the replacement pixel value in the G channel of the pixel, and the pixel value to be corrected in the B channel of the pixel is replaced with the replacement pixel value in the B channel of the pixel.

[0058] In this embodiment, after obtaining the current target image after shadow correction, the regional changes of the current target image are re-evaluated to obtain the land-unchanged areas and land-changed areas on the current target image. The specific process is as follows:

[0059] First, the current target image is divided into regions to obtain the target regions. The method for dividing the current target image into regions to obtain the target regions is the same as the method described above for dividing the current image to be shaded to obtain the regions to be analyzed in the current image to be shaded to be corrected, so it will not be described in detail. Then, based on the pixel value difference between the target regions in the current target image and the pixels at the same position in the nearby historical target images, the change probability index value of the target regions is obtained. The method for obtaining the change probability index value of the target regions based on the pixel value difference between the target regions in the current target image and the pixels at the same position in the nearby historical target images is the same as the method described above for obtaining the change probability index value of each region to be analyzed based on the pixel value difference between each region to be analyzed in the current image to be shaded to be corrected and the pixels at the same position in the nearby historical target images, so it will not be described in detail. Then, based on the change probability index value of the target regions, the current target image is judged for regional changes to obtain the land-unchanged regions and land-changed regions in the current target image. The process of judging regional changes in the current target image based on the change probability index value of the target regions to obtain the land-unchanged regions and land-changed regions in the current target image is as follows:

[0060] Based on the change probability index values ​​of each target region in the current target image, DBSCAN clustering is performed on all target regions in the current target image. The clustering results are all recorded as target clusters. Here, the absolute value of the difference in the change probability index values ​​between target regions is used to measure the distance between target regions when clustering them using the DBSCAN clustering algorithm. The clustering process of the DBSCAN algorithm is well-known. The sum of the areas of all target regions in each target cluster is recorded as the area of ​​the corresponding target cluster. The area of ​​a target region is measured by the number of pixels in the corresponding target region. The target cluster with the largest area is selected and recorded as the non-change target cluster. All target regions in the non-change target cluster are recorded as the land non-change areas in the current target image, which are also the land non-change areas in the monitored city at the current monitoring time. Target regions within all target clusters other than the non-change target clusters are recorded as the land change areas in the current target image, or in other words, all target regions in the current target image except for the land non-change areas are recorded as the current land change areas. The land change areas on the target image are also the land change areas in the monitored city at the current monitoring time. That is, the process of determining the land non-change areas and land change areas on the current target image based on the change probability index value of the target area in this embodiment is the same as the process of determining the land non-change areas and suspected land change areas on the current shadow correction image based on the change probability index value of each area to be analyzed. The only difference is the determination of the area type. For the process of obtaining the land non-change areas and land change areas on the current target image, all target areas in the target clusters other than the largest target cluster are recorded as land change areas on the current target image. For the process of obtaining the land non-change areas and land change areas on the current shadow correction image, all areas to be analyzed in the clusters other than the largest area are recorded as suspected land change areas on the current shadow correction image. The other processes are similar.

[0061] Thus, this embodiment completes the intelligent monitoring of urban land spatial changes, obtaining the land change areas and non-change areas in the monitored city at the current monitoring time. Subsequently, existing urban land classification models can be used to identify the land types in the land change areas. Combining the area of ​​the land change area with the land increase of that type of land, and combining it with the corresponding land type in historical target images, the decrease of the corresponding land type can be obtained. Therefore, the changing trend of each type of land area can be obtained, providing a basis for urban planning. Furthermore, how to monitor and analyze newly added construction land and farmland loss based on the land change areas and non-change areas in the city at the current monitoring time is a well-known technology. In addition, this embodiment improves the completeness, accuracy, and reliability of monitoring and identifying land change areas and non-change areas in the city by accurately and reliably identifying the shaded areas, then performing shadow correction, and using the corrected image to monitor changes in urban land space. This can improve the effectiveness of monitoring and identifying land change areas and non-change areas in the city.

[0062] In summary, this embodiment first acquires the current image to be shading corrected for the monitored city, the nearby historical target image, and the region to be analyzed on the current image to be shading corrected. Then, based on the curvature of each edge pixel on the outer contour edge of the region to be analyzed, the outer contour edge of the region to be analyzed is segmented to obtain each edge segment of the region to be analyzed. Based on the pixel value difference between the pixel at the same position in the region to be analyzed and the nearby historical target image, the change probability index value of the region to be analyzed is obtained. Based on the change probability index value, the region change of the current image to be shading corrected is judged to obtain the land non-change region and the suspected land change region on the current image to be shading corrected. Based on the pixel value of the pixel in the suspected land change region and each edge segment of the suspected land change region, the suspected land change region is filtered to obtain the shading region and the land change region on the current image to be shading corrected. Then, based on the non-shading region adjacent to the shading region, shading correction is performed on the shading region to obtain the current target image, and the region change of the current target image is judged again to obtain the land non-change region and the land change region on the current target image. Furthermore, this embodiment, by accurately and reliably identifying the shadowed areas, then correcting the shadowed areas, and using the image after shadow correction to monitor changes in urban land space, can improve the completeness, accuracy, and reliability of monitoring and identifying land change areas and land non-change areas in the city, or improve the effectiveness of monitoring and identifying changes in urban land space.

[0063] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A smart monitoring system for urban land spatial change based on data processing, comprising a processor and a memory, characterized in that, The processor executes the computer program stored in the memory to perform the following steps: Color correction is performed on the current initial remote sensing image to obtain the current image to be shading corrected; Based on the pixel values ​​of the pixels in the current image to be shaded, the pixels in the current image to be shaded are clustered to obtain each feature cluster. The pixels in the current image to be shaded and adjacent to each other belonging to the same feature cluster are divided into the same region, and each region obtained by division is recorded as the region to be analyzed in the current image to be shaded. Based on the curvature of each edge pixel on the outer contour edge of the region to be analyzed, the outer contour edge of the region to be analyzed is segmented to obtain each edge segment of the region to be analyzed. Based on the pixel value difference between the region to be analyzed and the pixels at the same position on the nearby historical target image, the change probability index value of the region to be analyzed is obtained. Based on the change probability index value, the region change of the current image to be shadowed is judged to obtain the land non-change region and the suspected land change region on the current image to be shadowed. Based on the pixel value of the pixel in the suspected land change region and each edge segment of the suspected land change region, the suspected land change region is filtered to obtain the shadow region and land change region on the current image to be shadowed. Based on the non-shaded areas adjacent to the shaded area, shadow correction is performed on the shaded area to obtain the current target image. Then, the current target image is re-evaluated for regional changes to obtain the land-unchanged areas and land-changed areas on the current target image.

2. The intelligent monitoring system for urban land spatial change based on data processing according to claim 1, characterized in that, The method for acquiring the current image to be shading corrected includes: The remote sensing image of the monitored city acquired at the current monitoring time is recorded as the current initial remote sensing image. The pixel values ​​on the current initial remote sensing image are initial pixel values. The vector formed by the difference in pixel values ​​of the same position and the same channel between the current initial remote sensing image and the neighboring historical target image is recorded as the relative difference vector of corresponding pixels on the current initial remote sensing image. The first relative difference in the relative difference vector is the R channel relative difference, the second relative difference is the G channel relative difference, and the third relative difference is the B channel relative difference. Based on the relative difference vector... The pixels in the current initial remote sensing image are clustered to obtain pixel clusters. The relative difference vector corresponding to the cluster center point of the largest pixel cluster is selected and recorded as the target relative difference vector. Based on the difference between the initial pixel value of the pixel in the current initial remote sensing image and the relative difference of the same channel in the target relative difference vector, the color correction value of the pixel in the current initial remote sensing image is obtained. The pixel value of the pixel in the current initial remote sensing image is replaced with the color correction value of the corresponding pixel. The replaced image is recorded as the current image to be shading corrected.

3. The intelligent monitoring system for urban land spatial change based on data processing according to claim 2, characterized in that, The method for obtaining the color correction values ​​of pixels in the current initial remote sensing image includes: For the a-th pixel in the current initial remote sensing image, the result of subtracting the relative difference of the R channel in the target relative difference vector from the initial pixel value of the R channel of the a-th pixel is recorded as the R channel color correction value of the a-th pixel. The result of subtracting the relative difference of the G channel in the target relative difference vector from the initial pixel value of the G channel of the a-th pixel is recorded as the G channel color correction value of the a-th pixel. The result of subtracting the relative difference of the B channel in the target relative difference vector from the initial pixel value of the B channel of the a-th pixel is recorded as the B channel color correction value of the a-th pixel. The R channel color correction value, G channel color correction value, and B channel color correction value are all color correction values.

4. The intelligent monitoring system for urban land spatial change based on data processing according to claim 1, characterized in that, The method for obtaining each edge segment of the region to be analyzed includes: Select any edge pixel on the outer contour edge of the region to be analyzed as the starting point. Starting from the starting point, mark the edge pixels on the outer contour edge of the region to be analyzed in a clockwise direction to obtain the marked value of each edge pixel on the outer contour edge of the region to be analyzed. Construct a mapping space, where the horizontal axis of the mapping space represents the marked value and the vertical axis represents the curvature value. Map the curvature and marked value of each edge pixel on the outer contour edge of the region to be analyzed to the mapping space. Connect the points in the mapping space in ascending order of marked value to obtain a feature curve. Obtain the inflection point on the feature curve. Record the edge pixel corresponding to the inflection point as a segmentation point. Use the segmentation points to segment the outer contour edge of the region to be analyzed to obtain each edge segment of the region to be analyzed.

5. The intelligent monitoring system for urban land spatial change based on data processing according to claim 1, characterized in that, The method for obtaining the change probability index value of the region to be analyzed includes: Obtain the pixel difference at the same position of each pixel in the region to be analyzed. The pixel difference at the same position of the b-th pixel in the region to be analyzed is the average of the absolute values ​​of the differences in pixel values ​​in the same channel between the b-th pixel and the pixel at the same position of the b-th pixel. The pixel at the same position of the b-th pixel belongs to the neighboring historical target image and has the same position coordinates as the b-th pixel. Use the average of the pixel differences at the same position of all pixels in the region to be analyzed as the change probability index value of the region to be analyzed.

6. The intelligent monitoring system for urban land spatial change based on data processing according to claim 1, characterized in that, The method for obtaining the land non-change areas and suspected land change areas on the current image to be shading corrected includes: Based on the change probability index value of the area to be analyzed, the area to be analyzed is clustered to obtain each area cluster. The areas to be analyzed in the area cluster with the largest total area are all recorded as land-unchanged areas, and the areas to be analyzed other than the land-unchanged areas are all recorded as suspected land-change areas.

7. The intelligent monitoring system for urban land spatial change based on data processing according to claim 1, characterized in that, The method for obtaining the shadowed areas and land change areas on the current image to be shaded includes: Regarding any suspected land change area Q: The sum of the average values ​​of all channels of all pixels in the suspected land change area Q, plus a preset constant, and then the normalized result of taking the reciprocal, is used as the first probability index value of the suspected land change area Q. The set of other suspected land change areas besides the suspected land change area Q is denoted as the reference set; Based on the absolute value of the difference between the direction angle of each edge segment of the suspected land change area Q and the direction angle of each suspected land change area in the reference set, and the first probability index value of each suspected land change area in the reference set, the weighted direction consistency index value between the suspected land change area Q and each suspected land change area in the reference set is obtained. The direction angle of the edge segment is the angle between the fitted line obtained by fitting the corresponding edge segment and the horizontal axis of the image. The normalized result of the mean of the weighted direction consistency index values ​​between the suspected land change area Q and the suspected land change areas in the reference set is denoted as the second probability index value of the suspected land change area Q; the mean of the first probability index value and the second probability index value of the suspected land change area Q is denoted as the shadow probability index value of the suspected land change area Q. Determine whether the shadow probability index value of the suspected land change area Q is greater than a preset shadow threshold. If so, the suspected land change area Q is recorded as a shadow area; otherwise, the suspected land change area Q is recorded as a land change area.

8. The intelligent monitoring system for urban land spatial change based on data processing according to claim 7, characterized in that, The method for obtaining the weighted direction consistency index value between the suspected land change area Q and each suspected land change area in the reference set includes: Regarding the suspected land change area Q and the c-th suspected land change area in the reference set: Obtain the set of directional angle differences corresponding to each edge segment of the suspected land change area Q. The directional angle difference in the set of directional angle difference corresponding to any edge segment of the suspected land change area Q is composed of the absolute value of the directional angle difference between the corresponding edge segment and each edge segment of the c-th suspected land change area. Use the negative correlation mapping value of the mean of the representative directional angle differences of each edge segment of the suspected land change area Q as the initial directional consistency index value between the suspected land change area Q and the c-th suspected land change area in the reference set. Multiply the initial directional consistency index value between the suspected land change area Q and the c-th suspected land change area by the first probability index value of the c-th suspected land change area as the weighted directional consistency index value between the suspected land change area Q and the c-th suspected land change area. The representative directional angle difference of the edge segment is the minimum value in the set of directional angle differences corresponding to the edge segment.

9. The intelligent monitoring system for urban land spatial change based on data processing according to claim 8, characterized in that, A method for obtaining a current target image by performing shadow correction on the shadowed region based on the non-shadowed region adjacent to the shadowed region includes: For any shaded area, the set of non-shaded areas adjacent to the shaded area is denoted as the neighboring area set. The non-shaded areas include land change areas and land non-change areas. The neighboring area corresponding to the largest initial direction consistency index value among the initial direction consistency index values ​​between each neighboring area in the neighboring area set and the shaded area is denoted as the building area of ​​the shaded area. All other non-shaded areas adjacent to the shaded area, excluding the building areas of the shaded area, are denoted as the correction reference area of ​​the shaded area. The fitted line obtained by fitting each edge segment of the correction reference area of ​​the shaded area is denoted as the edge line corresponding to the corresponding edge segment. The edge segments of the correction reference area are then judged to correspond to... If the extension of the edge line does not cross the corresponding correction reference area, then the corresponding edge line is recorded as the dividing edge line corresponding to the shadow area. All dividing edge lines corresponding to the shadow area are extended to the shadow area until they cover the entire range of the shadow area. The sub-regions obtained by dividing the shadow area with the extended dividing edge lines are all recorded as the dividing sub-regions in the shadow area. The correction reference area adjacent to the corresponding dividing sub-region in the correction reference area to which the edge line containing the boundary line of the dividing sub-region belongs is recorded as the replacement area corresponding to the dividing sub-region. The average pixel value of the replacement area corresponding to the dividing sub-region is recorded as the replacement pixel value of each pixel in the dividing sub-region. Replace the pixel values ​​of the pixels in the shadow area of ​​the current image to be shadowed with the replacement pixel values ​​of the corresponding pixels, and record the replaced image as the current target image.

Images