Spatial spectrum fusion high-standard farmland utilization mode remote sensing monitoring method and system

Abstract

The invention provides a space-spectrum fusion high-standard farmland utilization mode remote sensing monitoring method and system, and relates to the field of remote sensing monitoring, and the method comprises the steps: carrying out the multi-temporal image synthesis and preprocessing of a remote sensing image set, and obtaining a key phenological period synthesis image set, the NDVI value of the research area in each key phenological period is calculated through the key phenological period synthetic image set; phenological characteristic analysis is performed on different farmland utilization modes based on each NDVI value, farmland utilization priori knowledge is combined to construct a cultivated land utilization map, and spectrum information of the cultivated land utilization map is acquired; performing multi-scale segmentation on the remote sensing image set to obtain field images with spatial information; and assigning spectrum information of the farmland utilization map to the field images with spatial information, and judging the farmland utilization mode of the research area. According to the invention, the spectrum information of the cultivated land utilization map and the field images with spatial information are fused, the farmland utilization mode of each field is visually displayed, and remote sensing monitoring of the farmland utilization mode at the land parcel level is realized.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing monitoring, and in particular to a space-spectrum fusion remote sensing monitoring method and system for high-standard farmland utilization. Background Art

[0002] Currently, the primary method used is to map crop distribution and monitor cropping patterns using space-air-ground remote sensing technology. Satellite remote sensing technology, with its multi-temporal and wide coverage, offers the potential for monitoring cropping patterns at regional and global scales. However, due to factors such as insufficient continuous observation data, using satellite observations for cropping pattern monitoring also faces challenges with data gaps and loss of multi-temporal information. Specifically, if the composite time period is short, data gaps are significant; if the composite time period is long, multi-temporal information is lost. This dilemma of both abundant and scarce remote sensing data is particularly prominent in monitoring cultivated land use patterns.

[0003] When cropping frequency is used to characterize multiple cropping patterns on cultivated land, the problem of mixed pixels is inevitable due to limitations on plot size, spatial and temporal image resolution, and image availability, and the diversity of cropping patterns within fields is overlooked. In southern China, double-season rice is prevalent, as are double-season cropping patterns with different phenological periods. Simply categorizing cultivated areas into single-season, double-season, and triple-season types fails to accurately represent cropping patterns. For example, due to differences in phenological periods, some double-season crops are identified as single-season crops. Existing detection methods are unable to accurately distinguish between various cropping patterns, resulting in the inability to correctly identify farmland utilization patterns. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a high-standard farmland utilization remote sensing monitoring method and system with spatial-spectral fusion, which is used to solve the problem that the existing detection methods cannot accurately distinguish various planting patterns, resulting in the inability to allocate reasonable farmland utilization methods.

[0005] The present invention provides a remote sensing monitoring method for high-standard farmland utilization patterns using space-spectrum fusion, comprising the following steps: S1: Obtain a remote sensing image collection of the study area in each key phenological period, perform multi-temporal image synthesis and preprocessing on the remote sensing image collection to obtain a composite image collection of key phenological periods, and calculate the NDVI value of the study area in each key phenological period through the composite image collection of key phenological periods; S2: Analyze the phenological characteristics of different farmland utilization patterns based on the NDVI values, construct a farmland utilization map based on prior knowledge of farmland utilization, and obtain the spectral information of the farmland utilization map; S3: Perform multi-scale segmentation on the remote sensing image collection to obtain field images with spatial information; S4: Assign the spectral information of the cultivated land use map to the field image with spatial information to determine the farmland use pattern in the study area.

[0006] Preferably, step S1 is specifically as follows: S11: The key phenological periods include: the first growing season GS1, the transition period TGS and the second growing season GS2; S12: Based on the Google Earth Engine platform, obtain a collection of all Landsat series and Sentinel-2 remote sensing images covering the study area, synthesize, remove clouds, and perform masking on each remote sensing image to obtain synthetic images of each key phenological period. S13: Use the median synthesis method to perform multi-temporal image synthesis on the synthetic images of each key phenological period, and calculate the NDVI value of each key phenological period.

[0007] Preferred: The formula of the median composite method is as follows:

[0008] in, Indicates the median composite image at pixel position The pixel value of median represents the median operation, that is, all remote sensing images in the key phenological period are selected at the pixel position The middle value of the pixel values. If the number of remote sensing images is even, the average of the two middle pixel values is selected. Indicates the pixel position of the composite image of the Nth key phenological period The pixel value of .

[0009] Preferably, step S2 is specifically as follows: S21: Quantify the NDVI values of each key phenological period, and the NDVI values conform to the normal distribution N (0.2, 0.05) is defined as a low value, and the NDVI value conforms to the normal distribution. N (0.4, 0.05) is defined as the median value, and the NDVI value conforms to the normal distribution. N (0.6, 0.05) was defined as a high value, and each quantitative NDVI value was obtained; S22: Fill each quantitative NDVI value into the probability condition table, arrange and combine the quantitative NDVI values of each key phenological period, remove invalid combinations, further arrange and combine the existing combinations and establish a mapping relationship with the farmland utilization mode to obtain the cultivated land utilization map.

[0010] Preferably, step S3 is specifically as follows: S31: Obtain multispectral Sentinel-2 images and Gaofen-6 images from the remote sensing image collection, fuse the panchromatic bands of the multispectral Sentinel-2 images and Gaofen-6 images using the Gram-Schmidt pan-sharpening algorithm to obtain a fused image. S32: performing multi-scale segmentation on the fused image to obtain an initial segmented image, where the initial segmented image includes multiple field areas; S33: Calculating the similarity of indicators of adjacent field areas, and merging two adjacent field areas whose indicator similarity is greater than a preset value into one field area; S34: Repeat step S33 until the similarity of the indicators of each adjacent field area is no greater than a preset value, thereby obtaining a field image with spatial information, where the spatial information includes the position of each field area in the field image.

[0011] Preferred: The calculation formula of indicator similarity is:

[0012] in, Indicates field area With field area The similarity of indicators between Indicates field area With field area The spatial similarity between Indicates field area With field area The spectral similarity between and is the weight parameter.

[0013] Preferably, step S4 is specifically as follows: S41: Mark each farmland utilization type in the spectrum information as a different color, and fill each pixel position in the study area with the color corresponding to the farmland utilization type to obtain a filled image; S42: Overlaying the field image with spatial information and the filling image to obtain the color of each field area; S43: Visualize the farmland utilization pattern of each field area through the color of each field area.

[0014] A spatial-spectral fusion high-standard farmland utilization mode remote sensing monitoring system includes: a processor and a storage medium; the processor loads and executes instructions and data in the storage medium to implement the spatial-spectral fusion high-standard farmland utilization mode remote sensing monitoring method.

[0015] The present invention has the following beneficial effects: 1. Based on dense time-series remote sensing imagery and prior knowledge of farmland use patterns (farmland planting information collected through field surveys), key phenological characteristics are synthesized. Phenological characteristics of different farmland use patterns (double-season cropping, single-season cropping, etc.) are analyzed to identify differences in phenological characteristics between different cropping patterns. The process of vegetation index state changes and spatial differences in farmland use patterns are then integrated into a single map to construct a farmland use map encompassing all farmland use patterns. 2. Utilizing the fusion of multispectral Sentinel-2 and GF-6 images, we perform object-oriented multiresolution image segmentation to extract field images. Field images contain not only rich spatial information but also spectral information (with four bands: blue, green, red, and near-infrared). Spectral difference segmentation is performed on the initial segmentation results to obtain more accurate field boundary information.

[0016] 3. Integrate the spectral information of the cultivated land use map and the field images with spatial information to visualize the farmland use patterns of each field and realize remote sensing monitoring of farmland use patterns at the field level. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a method according to an embodiment of the present invention; Figure 2 for the arable land use atlas; Figure 3 The flowchart of obtaining field images, where (a) is the Sentinel-2 image, (b) is the fused image, (c) is the initial segmentation result, and (d) is the spectral difference segmentation result; Figure 4 Flowchart for assigning spectral information to field images; Figure 5 Schematic diagram of the remote sensing model of spatial-spectral fusion, where (a) is the NDVI of GS1, (b) is the NDVI of TG1, (c) is the NDVI of GS2, and (d) is the spatial-spectral fusion effect diagram; Figure 6 The diagram shows the farmland utilization pattern, where (a) is the spring and summer double season, and (b) is the summer single season; The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] Reference Figure 1 The present invention provides a remote sensing monitoring method for high-standard farmland utilization patterns using spatial-spectral fusion, comprising the following steps: S1: Obtain a remote sensing image collection of the study area in each key phenological period, perform multi-temporal image synthesis and preprocessing on the remote sensing image collection to obtain a composite image collection of key phenological periods, and calculate the NDVI value of the study area in each key phenological period through the composite image collection of key phenological periods; As an example: Step S1 is specifically as follows: S11: The key phenological periods include: the first growing season GS1, the transition period TGS and the second growing season GS2; Specifically, assume that the study area has two growing seasons in a natural year. Crop growth has three main phenological stages: the first growing season (GS1), the second growing season (GS2), and the transition between the two growing seasons (TGS). In the Jianghan Plain, the first growing season is primarily winter crops (including winter rapeseed and winter wheat). These crops are typically planted in late autumn or early winter of the previous year, with the vegetation index peaking from mid-March to early April. Harvested in late May to early June, the second growing season begins around the same time, when the vegetation index is at its lowest. The second growing season is typically rice or soybeans, with a typical growing period of approximately three months. During this period, light and heat conditions are excellent, marking the peak growth season for vegetation, with the vegetation index peaking from mid-July to early August.

[0020] S12: Based on the Google Earth Engine platform, obtain a collection of all Landsat series and Sentinel-2 remote sensing images covering the study area, synthesize, remove clouds, and perform masking on each remote sensing image to obtain synthetic images of each key phenological period. S13: Use the median synthesis method to perform multi-temporal image synthesis on the synthetic images of each key phenological period, and calculate the NDVI value of each key phenological period.

[0021] Specifically, based on the phenological characteristics of crops in the region, the median composite method was used to perform multi-temporal image synthesis to obtain the time series Normalized Difference Vegetation Index (NDVI) for three key phenological periods (GS1, TG1, and GS2). The formula of the median composite method is as follows:

[0022] in, Indicates the median composite image at pixel position The pixel value of median represents the median operation, that is, all remote sensing images in the key phenological period are selected at the pixel position The middle value of the pixel values. If the number of remote sensing images is even, the average of the two middle pixel values is selected. Indicates the pixel position of the composite image of the Nth key phenological period The pixel value of .

[0023] S2: Analyze the phenological characteristics of different farmland utilization patterns based on the NDVI values, construct a farmland utilization map based on prior knowledge of farmland utilization, and obtain the spectral information of the farmland utilization map; As an example: Step S2 is specifically as follows: S21: Quantify the NDVI values of each key phenological period, and the NDVI values conform to the normal distribution N (0.2, 0.05) is defined as a low value, and the NDVI value conforms to the normal distribution. N (0.4, 0.05) is defined as the median value, and the NDVI value conforms to the normal distribution. N (0.6, 0.05) was defined as a high value, and each quantitative NDVI value was obtained; Specifically, these three phenological periods are all in the peak growth season of vegetation. Due to the phenological differences in different planting patterns, the vegetation index of cultivated land is either very high (at the peak) or very low (at the valley, the rotation period between two crops or the preparation period for the second season of farming). For the time being, "high value" or "low value" is used to qualitatively describe the status of the vegetation index.

[0024] When crop NDVI is at its valley value, meaning that the land is transitioning between two crops or preparing for the second cropping season, the soil background is dominant, and the NDVI value conforms to a normal distribution (N(0.2, 0.05)). When the crops are in their peak growing season, the NDVI is at its peak, and the NDVI value conforms to a normal distribution (N(0.6, 0.05)). This pattern is used to quantify the status of the vegetation index ("high," "medium," or "low").

[0025] S22: Fill each quantitative NDVI value into the probability condition table, arrange and combine the quantitative NDVI values of each key phenological period, remove invalid combinations, further arrange and combine the existing combinations and establish a mapping relationship with the farmland utilization mode to obtain the cultivated land utilization map.

[0026] Specifically, knowledge-based probabilistic coding is used to fill in the Conditional Probability Table (CPT). Precise eigenvalues are not required; as long as the state of the planting pattern at the corresponding eigenvalue (high, medium, or low) is correct, the CPT can be manually filled in based on prior knowledge. Only the planting patterns relevant to the node are filled in, making the CPT a sparse table (nodes not involved in the decision remain blank).

[0027] Different arable land use patterns have their own unique phenological rhythms, which are reflected in the key phenological period images as their own unique curves. For example, the spring and summer seasons show a "high-low-high" characteristic in the key phenological period, while the summer season shows a "low-low-high" characteristic. The states of the three phenological periods are arranged and combined, invalid combinations are removed, and the existing combinations are further arranged and mapped to the arable land use patterns to obtain the arable land use map, such as Figure 2 shown.

[0028] S3: Perform multi-scale segmentation on the remote sensing image collection to obtain field images with spatial information; As an example: Step S3 is specifically as follows: S31: Obtain multispectral Sentinel-2 images and Gaofen-6 images from the remote sensing image collection, fuse the panchromatic bands of the multispectral Sentinel-2 images and Gaofen-6 images using the Gram-Schmidt pan-sharpening algorithm to obtain a fused image. Specifically, multi-resolution segmentation is performed based on the fusion of Sentinel-2 images and GF-6 images to construct homogeneous areas as the basic units of cultivated land use patterns.

[0029] The GF-6 panchromatic band has a spatial resolution of 2 meters, which can reveal spatial details of ground features, but it lacks spectral information. Sentinel-2 images have rich spectral information, but cannot accurately identify field boundaries. The Gram-Schmidt pan-sharpening algorithm is used to fuse the multispectral Sentinel-2 imagery with the GF-6 panchromatic band.

[0030] S32: performing multi-scale segmentation on the fused image to obtain an initial segmented image, where the initial segmented image includes multiple field areas; Specifically, the fusion result has a spatial resolution of 2m and spectral information of eight bands. The boundaries of the plots are clearer than those of the Sentinel-2 image, and the feature information of the land objects is more obvious than that of the Sentinel-6 image, which is more conducive to the segmentation and extraction of the plots.

[0031] S33: Calculating the similarity of indicators of adjacent field areas, and merging two adjacent field areas whose indicator similarity is greater than a preset value into one field area; Specifically, the fused image is segmented at multiple scales to generate homogeneous regions (patches) with low heterogeneity and high homogeneity. This method first calculates the homogeneity criterion for adjacent pixels based on the multi-scale segmentation. If the calculated metric is highly similar, the pixels are merged; otherwise, they are not merged.

[0032] The calculation formula of indicator similarity is:

[0033] in, Indicates field area With field area The similarity of indicators between Indicates field area With field area The spatial similarity between Indicates field area With field area The spectral similarity between and is the weight parameter.

[0034] S34: Repeat step S33 until the similarity of the indicators of each adjacent field area is no greater than a preset value, thereby obtaining a field image with spatial information, where the spatial information includes the position of each field area in the field image.

[0035] Specifically, based on the fused image, eCognition's multi-resolution segmentation algorithm was used for initial segmentation. The image domain was set to pixel level, and the scale parameter, shape parameter, and compactness parameters were optimized to obtain the initial patch. On top of the initial patch, land cover and cultivated land mask data were superimposed, and spectral difference segmentation was performed again (maximum spectral difference was set to 10) to obtain field information, such as Figure 3 shown.

[0036] S4: Assign the spectral information of the cultivated land use map to the field image with spatial information to determine the farmland use pattern in the study area.

[0037] As an example: Step S4 is specifically as follows: S41: Mark each farmland utilization type in the spectrum information as a different color, and fill each pixel position in the study area with the color corresponding to the farmland utilization type to obtain a filled image; Specifically, the spatial information of high-resolution images (fields or planting structure units) and the spectral information of time series remote sensing images (crop phenology information and different planting patterns) are integrated to build a remote sensing model of spatial-spectral fusion to achieve remote sensing monitoring of cultivated land use patterns at the plot level. Figure 5 As shown in the figure, unlike the traditional fusion method that derives a new raster layer, this model assigns the spectral information of the time series remote sensing image to the vector layer with spatial information, indirectly realizing the spatial-spectral fusion of geological information and remote sensing data. Specific technologies include: (1) Overlay of raster layer (time series remote sensing data) and vector layer (field data): unifying the projection and coordinate base of data from different sources to achieve accurate overlay of field scale information. (2) Synchronization of "spatial"-"spectral" information of different granularities. Based on the spectral information of time series remote sensing data and the field information of high-resolution images, the spectral information (based on pixels) and field information (based on patches) are synchronized, as shown in the figure. Figure 4 (3) Using statistical methods to assign spectral information to patches: Based on the spectral information of cultivated land use patterns at the pixel scale on time series images, statistical methods are used to assign “spectral” information to patches, thereby restoring the cultivated land planting conditions and idle types at the field scale.

[0038] The spatial distribution information (patches) of the fields was overlaid with the multi-temporal Sentinel-2 imagery (three key phenological periods, GS1, TG1, and GS2). Zonal statistics were used to summarize the NDVI values for the key phenological periods within the patch. The spectral information of Sentinel-2 was then assigned to the patches to obtain the cultivated land use patterns at the field scale. For zonal statistics, the statistical type was selected as "Majority," assigning the most frequently occurring grayscale value in the corresponding time series image to each patch. The vector layers of the three key phenological periods were converted to raster data and then false-color composited. The resulting filled image was a false-color composite image that mimicked the three-temporal Sentinel-2 images but with different granularity. For some fields that were too small to be correctly assigned, the value of the nearest patch was used based on the principle of spectral similarity.

[0039] S42: Overlaying the field image with spatial information and the filling image to obtain the color of each field area; S43: Visualize the farmland utilization pattern of each field area through the color of each field area.

[0040] Specifically, based on the developed visualization platform, a visual query of farmland plot utilization can be realized. Click any plot on the map, and the spectrum information of the plot (displayed in different colors) will be highlighted, such as Figure 6 shown.

[0041] A spatial-spectral fusion high-standard farmland utilization mode remote sensing monitoring system includes: a processor and a storage medium; the processor loads and executes instructions and data in the storage medium to implement the spatial-spectral fusion high-standard farmland utilization mode remote sensing monitoring method.

[0042] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0043] The serial numbers of the embodiments of the present invention are for descriptive purposes only and do not represent superiority or inferiority of the embodiments. In a unit claim that lists several means, several of these means may be embodied by the same item of hardware. The use of the terms first, second, and third, etc., does not denote any order and should be construed as identifiers.

[0044] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A remote sensing monitoring method for high-standard farmland utilization patterns based on spatial-spectral fusion, characterized in that: Including steps: S1: Obtain a remote sensing image collection of the study area in each key phenological period, perform multi-temporal image synthesis and preprocessing on the remote sensing image collection to obtain a composite image collection of key phenological periods, and calculate the NDVI value of the study area in each key phenological period through the composite image collection of key phenological periods; S2: Analyze the phenological characteristics of different farmland utilization patterns based on the NDVI values, construct a farmland utilization map based on prior knowledge of farmland utilization, and obtain the spectral information of the farmland utilization map; S3: Perform multi-scale segmentation on the remote sensing image collection to obtain field images with spatial information; S4: Assign the spectral information of the cultivated land use map to the field image with spatial information to determine the farmland use pattern in the study area.

2. The high-standard farmland utilization pattern remote sensing monitoring method based on spatial-spectral fusion according to claim 1 is characterized in that: Step S1 is specifically as follows: S11: The key phenological periods include: the first growing season GS1, the transition period TGS and the second growing season GS2; S12: Based on the Google Earth Engine platform, obtain a collection of all Landsat series and Sentinel-2 remote sensing images covering the study area, synthesize, remove clouds, and perform masking on each remote sensing image to obtain synthetic images of each key phenological period. S13: Use the median synthesis method to perform multi-temporal image synthesis on the synthetic images of each key phenological period, and calculate the NDVI value of each key phenological period.

3. The high-standard farmland utilization pattern remote sensing monitoring method based on spatial-spectral fusion according to claim 2 is characterized by: The formula of the median composite method is as follows: in, Indicates the median composite image at pixel position The pixel value of median represents the median operation, that is, all remote sensing images in the key phenological period are selected at the pixel position The middle value of the pixel values. If the number of remote sensing images is even, the average of the two middle pixel values is selected. Indicates the pixel position of the composite image of the Nth key phenological period The pixel value of .

4. The method for remote sensing monitoring of high-standard farmland utilization patterns based on spatial-spectral fusion according to claim 1 is characterized in that: Step S2 is specifically as follows: S21: Quantify the NDVI values of each key phenological period, and the NDVI values conform to the normal distribution N (0.2, 0.05) is defined as a low value, and the NDVI value conforms to the normal distribution. N (0.4, 0.05) is defined as the median value, and the NDVI value conforms to the normal distribution. N (0.6, 0.05) was defined as a high value, and each quantitative NDVI value was obtained; S22: Fill each quantitative NDVI value into the probability condition table, arrange and combine the quantitative NDVI values of each key phenological period, remove invalid combinations, further arrange and combine the existing combinations and establish a mapping relationship with the farmland utilization mode to obtain the cultivated land utilization map.

5. The remote sensing monitoring method for high-standard farmland utilization patterns based on spatial-spectral fusion according to claim 1 is characterized in that: Step S3 is specifically as follows: S31: Obtain multispectral Sentinel-2 images and Gaofen-6 images from the remote sensing image collection, fuse the panchromatic bands of the multispectral Sentinel-2 images and Gaofen-6 images using the Gram-Schmidt pan-sharpening algorithm to obtain a fused image. S32: performing multi-scale segmentation on the fused image to obtain an initial segmented image, where the initial segmented image includes multiple field areas; S33: Calculate and obtain the index similarity of each adjacent field area, and merge two adjacent field areas whose index similarity is greater than a preset value into one field area; S34: Repeat step S33 until the similarity of the indicators of each adjacent field area is no greater than a preset value, thereby obtaining a field image with spatial information, where the spatial information includes the position of each field area in the field image.

6. The method for remote sensing monitoring of high-standard farmland utilization patterns using spatial-spectral fusion according to claim 5, characterized in that: The calculation formula of indicator similarity is: in, Indicates field area With field area The similarity of indicators between Indicates field area With field area The spatial similarity between Indicates field area With field area The spectral similarity between and is the weight parameter.

7. The method for remote sensing monitoring of high-standard farmland utilization patterns based on spatial-spectral fusion according to claim 1, characterized in that: Step S4 is specifically as follows: S41: Mark each farmland utilization type in the spectrum information as a different color, and fill each pixel position in the study area with the color corresponding to the farmland utilization type to obtain a filled image; S42: Overlaying the field image with spatial information and the filling image to obtain the color of each field area; S43: Visualize the farmland utilization pattern of each field area through the color of each field area.

8. A high-standard farmland utilization remote sensing monitoring system using spatial-spectral fusion, characterized by: include: Processor and storage medium; the processor loads and executes instructions and data in the storage medium to implement the high-standard farmland utilization mode remote sensing monitoring method with spatial-spectral fusion as described in any one of claims 1 to 7.

Images