Crop missing strip satellite remote sensing monitoring method, system, device and medium

Abstract

The application discloses a kind of crop missing strip satellite remote sensing monitoring method, system, equipment and medium. Obtain multispectral satellite remote sensing image and plot vector, determine effective monitoring range;According to seedling strip texture determines ridge direction and ridge direction center line phase, obtains ridge distance, and generates ridge direction center line set accordingly;Extract the candidate object that vegetation characteristic value is lower than surrounding local area, extends along ridge direction and crosses multiple center lines, constructs the reference band that transverse projection range is same before and after it in its along ridge direction;Screen reference band effective pixel and determine reference band effectiveness, subtract the vegetation characteristic value of candidate object from the lower vegetation characteristic value in two effective reference bands, obtain vegetation deficiency amount;When candidate object is located between two reference bands and vegetation deficiency amount is greater than 0, output the planar vector information of suspected missing strip object.The application can improve the spatial comparability of reference information, reduce the misjudgment caused by local interference.

Description

Technical Field

[0001] This invention relates to the field of agricultural remote sensing monitoring technology, and in particular to a satellite remote sensing monitoring method, system, equipment and medium for crop seedling loss and crop strip breakage. Background Technology

[0002] After corn seedlings emerge, gaps or breaks in seedling growth may occur in some areas due to factors such as sowing quality, soil moisture, pests and diseases, and field management. Timely identification of the location and extent of these gaps is crucial for replanting, reseeding, and subsequent field management. Traditional manual inspections require checking each section along the planting ridges, which is labor-intensive and insufficient for monitoring large areas of cornfields. With the improved spatial resolution of satellite remote sensing imagery, using multispectral satellite remote sensing images to identify corn seedling strips and abnormal areas has become a key technology for monitoring crop growth in farmland.

[0003] Existing remote sensing monitoring methods typically extract areas with low vegetation characterization values ​​within a plot based on vegetation indices, spectral reflectance characteristics, or changes between images from different periods, and identify these areas as suspected seedling-deficient regions. However, maize seedling strips exhibit a continuous spatial distribution along planting ridges, and the spectral characteristics of seedling-deficient areas are easily affected by factors such as soil background, field moisture, shadows, bright spots, and plot edges. When using a uniform threshold for the entire plot or a single local reference for judgment, it is difficult to distinguish between genuine seedling-deficient strips and differences in normal growth or local environmental disturbances.

[0004] Furthermore, within the same plot, there may be variations in ridge direction, misalignment of seedling strips, or inconsistencies in local planting directions. If the candidate and reference areas are not determined by combining the seedling strip direction and the spatial relationship between the ridge direction and the seedling strip direction, the reference area may correspond to different seedling strips or planting positions than the area to be identified, resulting in a lack of comparability in their vegetation characterization values. Therefore, existing methods still suffer from misjudgment and omission when monitoring areas with missing or broken maize seedling strips in complex plots. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that existing satellite remote sensing monitoring methods are unable to obtain vegetation reference information comparable to the area to be identified under conditions of ridge direction changes, seedling strip misalignment, and local environmental interference, which leads to the easy misjudgment or omission of crop seedling gaps and strip breaks. The invention provides a satellite remote sensing monitoring method, system, equipment, and medium for crop seedling gaps and strip breaks.

[0006] The first aspect of this invention provides a satellite remote sensing monitoring method for crop seedling gaps and breaks, comprising: acquiring multispectral satellite remote sensing images and plot vectors of a target crop plot, and determining an effective monitoring range based on the plot vectors; determining the ridge direction and ridge centerline phase based on the seedling strip texture in the multispectral satellite remote sensing images, obtaining the ridge spacing, generating a set of ridge centerlines based on the ridge direction, the ridge centerline phase, and the ridge spacing, and establishing local ridge coordinates including those along the ridge direction and laterally, wherein the lateral coordinates are perpendicular to the ridge direction; and extracting candidate objects from the effective monitoring range, wherein the candidate objects are those with low vegetation characterization values. The vegetation characterization values ​​of the local area surrounding the candidate object, the planar objects extending along the ridge direction and crossing multiple ridge centerlines, and the ridge length and lateral coverage of each candidate object are determined; for each candidate object, based on the ridge length and lateral coverage, a forward reference zone and a backward reference zone are constructed in front of and behind the candidate object along the ridge direction, respectively, so that the forward reference zone and the backward reference zone have the same lateral projection range as the candidate object and cover the ridge centerline covered by the candidate object in the lateral direction; respectively from the forward reference... Pixels located outside the effective monitoring range, within other candidate objects, with invalid image data, not in the geometric seedling strip position corresponding to any of the ridge-direction center lines, or inconsistent in phase with the ridge-direction center line covered by the current candidate object, are removed from the image strip and the backward reference strip. The remaining pixels are determined as valid pixels. When both the number and percentage of valid pixels in the corresponding reference strip reach a preset threshold, the corresponding reference strip is determined to be valid. When both the forward and backward reference strips are valid, the same vegetation characterization value calculation method is used, based on the candidate objects... The pixels determine the vegetation characterization value of the candidate object, and the vegetation characterization values ​​of the forward reference zone and the backward reference zone are determined according to the effective pixels in each reference zone. The lower of the vegetation characterization values ​​of the two reference zones is determined as the weaker reference value, and the difference between the weaker reference value and the vegetation characterization value of the candidate object is determined as the vegetation deficit. When the candidate object is located between the forward reference zone and the backward reference zone and the vegetation deficit is greater than 0, the candidate object is determined as a suspected seedling gap object, and the areal vector information representing the boundary and position of the suspected seedling gap object is output.

[0007] A second aspect of the present invention provides a satellite remote sensing monitoring system for crop seedling gaps and missing strips, comprising: a monitoring range determination module, used to acquire multispectral satellite remote sensing images and plot vectors of a target crop plot, and determine an effective monitoring range based on the plot vectors; a ridge direction space construction module, used to determine the ridge direction and ridge direction centerline phase based on the seedling strip texture in the multispectral satellite remote sensing images, obtain the ridge spacing, generate a set of ridge direction centerlines based on the ridge direction, the ridge direction centerline phase, and the ridge spacing, and establish ridge direction local coordinates including those along the ridge direction and laterally, wherein the lateral direction is perpendicular to the ridge direction; and an object processing module, used to extract candidate objects from the effective monitoring range. The candidate objects are planar objects whose vegetation characterization values ​​are lower than those of the surrounding local areas, extending along the ridge direction and spanning multiple ridge centerlines. The ridge length and lateral coverage of each candidate object are determined. A reference strip processing module is used to construct a forward reference strip and a backward reference strip in front of and behind each candidate object along the ridge direction, based on the candidate object's ridge length and lateral coverage. The forward and backward reference strips have the same lateral projection range as the candidate object and laterally cover the ridge centerline covered by the candidate object. The vegetation deficiency determination module is used to remove pixels from the forward reference zone and the backward reference zone that are located outside the effective monitoring range, within other candidate objects, have invalid image data, are not located at the geometric seedling strip position corresponding to any of the ridge-direction center lines, or are inconsistent in phase with the ridge-direction center line covered by the current candidate object. The retained pixels are determined as valid pixels. When the number of valid pixels and the proportion of valid pixels in the corresponding reference zone both reach the corresponding preset thresholds, the corresponding reference zone is determined to be valid. The vegetation deficiency determination module is used to calculate vegetation characteristic values ​​using the same method when both the forward and backward reference zones are valid. The vegetation characterization value of the candidate object is determined based on the pixels in the candidate object, and the vegetation characterization values ​​of the forward reference zone and the backward reference zone are determined based on the effective pixels in each reference zone. The lower of the vegetation characterization values ​​of the two reference zones is determined as the weaker reference value, and the difference between the weaker reference value and the vegetation characterization value of the candidate object is determined as the vegetation deficit. The output module is used to determine the candidate object as a suspected seedling gap object when the candidate object is located between the forward reference zone and the backward reference zone and the vegetation deficit is greater than 0, and output the areal vector information characterizing the boundary and position of the suspected seedling gap object.

[0008] A third aspect of the present invention provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to implement the method described in the first aspect.

[0009] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the method described in the first aspect.

[0010] This invention determines the ridge direction and ridge centerline phase based on the seedling strip texture in multispectral satellite remote sensing imagery. It then extracts candidate objects extending along the ridge direction and spanning multiple ridge centerlines in local ridge coordinates, ensuring these candidate objects correspond to the spatial distribution characteristics of crop seedling strips. By constructing reference zones with the same lateral projection range and covering the same ridge centerline in front of and behind the candidate objects along the ridge direction, the planting positions of the candidate objects and reference zones are aligned. Pixels outside the effective monitoring range, within other candidate objects, with invalid image data, not located at any geometric seedling strip position corresponding to any ridge centerline, or whose phase is inconsistent with the ridge centerline covered by the current candidate object are eliminated. The validity of the reference zones is then assessed, reducing interference from plot edges, other abnormal areas, and seedling strip misalignment. Furthermore, using the lower vegetation characterization value from the two effective reference zones to determine vegetation deficit reduces the impact of unilateral reference zone vegetation anomalies on the identification results, thereby improving the reliability of crop seedling gap identification and obtaining the boundary and location information of suspected seedling gap objects. Attached Figure Description

[0011] Figure 1 This is a flowchart of a satellite remote sensing monitoring method for crop seedling loss and crop belt gaps according to an embodiment of the present invention; Figure 2 It is a comparison image of a drone image with obvious missing seedlings, weak seedlings or local low growth and normal seedlings according to an embodiment of the present invention. Figure 3 This is a comparison diagram of a suspected object in satellite imagery and UAV imagery at the same location according to an embodiment of the present invention; Figure 4 This is a framework diagram of a satellite remote sensing monitoring system for crop seedling loss and crop belt gaps according to an embodiment of the present invention; Figure 5 This is a framework diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0012] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0013] The technical solution of the present invention is illustrated below through specific embodiments. It should be understood that the one or more steps mentioned in the present invention do not preclude the existence of other methods and steps before or after the combined steps, or that other methods and steps may be inserted between these explicitly mentioned steps. It should also be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Unless otherwise stated, the numbering of each method step is only for the purpose of identifying each method step, and not for limiting the order of each method or limiting the scope of the present invention. Changes or adjustments to their relative relationships, without substantial changes to the technical content, can also be considered as within the scope of the present invention.

[0014] In this embodiment of the invention, the satellite remote sensing monitoring method for crop seedling gaps and broken stripes is applied to row-planted crop fields with striped seedling strip textures. It is used to identify areas within the field suspected of having missing seedling strips or broken stripes after crop emergence using satellite remote sensing imagery. The implementing entity of this method can be a satellite remote sensing monitoring system for crop seedling gaps and broken stripes. The following description, in conjunction with optional image processing algorithms and a set of implementation parameters, illustrates the satellite remote sensing monitoring method for crop seedling gaps and broken stripes in this embodiment; the listed parameters are used to illustrate the implementation of the corresponding technical means and do not constitute a limitation on the scope of protection of this invention.

[0015] like Figure 1 As shown, the satellite remote sensing monitoring method for crop seedling gaps and crop strip breaks can specifically include the following steps: S101: Acquire multispectral satellite remote sensing images and plot vectors of the target crop plots, and determine the effective monitoring range based on the plot vectors.

[0016] The target crop plots are the crop planting areas requiring monitoring of missing seedlings and broken planting zones. Multispectral satellite remote sensing imagery contains image information of the target crop plots across multiple spectral bands, reflecting the vegetation status and spatial distribution within the plots. Plot vectors are areal vector data characterizing the boundaries and spatial locations of the target crop plots.

[0017] The plot regions in the multispectral satellite remote sensing imagery are defined by plot vectors, and the areas within these regions used for monitoring missing seedlings and broken strips are determined as the effective monitoring range. Subsequent seedling strip texture analysis, candidate object extraction, and two-sided reference comparison are all performed within this effective monitoring range.

[0018] S102, determine the ridge direction and ridge centerline phase based on the seedling strip texture in the multispectral satellite remote sensing image, obtain the ridge spacing, generate a set of ridge centerlines based on the ridge direction, ridge centerline phase and ridge spacing, and establish local ridge coordinates including along the ridge direction and laterally, with the lateral direction perpendicular to the ridge direction.

[0019] Seedling strip texture is the strip-like spatial texture formed in multispectral satellite remote sensing imagery by crops distributed along planting ridges. The ridge direction indicates the extension direction of the seedling strip within the plot, and the ridge centerline indicates the center position of each seedling strip in the image. The ridge centerline phase represents the lateral offset of the set of ridge centerlines relative to a reference position within the plot. For example, although multiple sets of parallel lines extend in the same direction, only one set of parallel lines corresponds to the center position of each seedling strip; the lateral offset of this set of parallel lines is the ridge centerline phase.

[0020] Based on the ridge direction, ridge spacing, and ridge centerline phase, a set of ridge centerlines corresponding to the spatial position of each seedling strip is generated. Using the ridge direction as the ridge-side direction and the direction perpendicular to the ridge direction as the lateral direction, local ridge-side coordinates are established. This allows for the description of the image object's position along the seedling strip's extension direction and its position perpendicular to the seedling strip's extension direction, respectively.

[0021] S103. Extract candidate objects from the effective monitoring range. Candidate objects are areal objects whose vegetation characterization values ​​are lower than the vegetation characterization values ​​of the local area surrounding the candidate object, which extend along the ridge direction and cross multiple ridge center lines. Determine the ridge length and lateral coverage of each candidate object.

[0022] Vegetation characterization values ​​are numerical values ​​determined based on multispectral satellite remote sensing imagery and used to characterize the state of crop vegetation. The local area surrounding a candidate object is a region near the candidate object used to characterize a locally normal vegetation state. By comparing the vegetation characterization values ​​of the areal object and its surrounding local area, areal objects with relatively low vegetation status can be identified from the effective monitoring range.

[0023] In the local coordinates along the ridge direction, the spatial morphology of the corresponding areal objects and their relationship with the set of ridge centerlines are determined. Areal objects that extend along the ridge direction and cross multiple ridge centerlines laterally are identified as candidate objects. For example, if a region with low vegetation status extends continuously along the direction of the seedling strip and covers multiple adjacent seedling strips laterally, then this region can be considered a candidate object.

[0024] The length along the ridge of a candidate object is its extension along the ridge direction, and the lateral coverage is its projection in the lateral direction. Based on the lateral coverage, the centerline of the ridge covered by the candidate object can be determined.

[0025] S104. For each candidate object, based on the length of the candidate object along the ridge and the lateral coverage range, construct a forward reference zone and a backward reference zone in front of and behind the candidate object along the ridge direction, respectively, so that the forward reference zone and the backward reference zone have the same lateral projection range as the candidate object, and cover the ridge centerline covered by the candidate object in the lateral direction.

[0026] The front and rear directions along the ridge direction are two opposite directions determined with the candidate object as a reference. The forward reference zone is located on one side of the candidate object, and the rear reference zone is located on the other side of the candidate object.

[0027] The lateral projection ranges of both the forward and backward reference bands are the same as the lateral coverage range of the candidate object. Therefore, the row-direction centerline covered by the candidate object is simultaneously covered by both reference bands. For example, if the candidate object covers three adjacent row-direction centerlines laterally, then the forward and backward reference bands also cover those three row-direction centerlines respectively. In this way, the candidate object and the two reference bands correspond to the same set of seedling strip positions.

[0028] S105, remove pixels from the forward and backward reference bands that are outside the effective monitoring range, within other candidate objects, have invalid image data, are not located at the geometric seedling strip position corresponding to any ridge centerline, or are inconsistent with the phase of the ridge centerline covered by the current candidate object. The retained pixels are determined as valid pixels. When the number of valid pixels and the proportion of valid pixels in the corresponding reference band both reach the corresponding preset threshold, the corresponding reference band is determined to be valid.

[0029] Pixels located outside the effective monitoring range do not belong to the current monitoring area; pixels located within other candidate objects may indicate other vegetation anomalies; pixels with invalid image data cannot be used to reliably calculate vegetation characterization values; pixels not located at the geometric seedling zone position corresponding to any row-to-center line cannot characterize the vegetation status of the crop seedling zone; pixels whose phase is inconsistent with the row-to-center line covered by the current candidate object may be located in other seedling zones, but do not correspond to the same seedling zone position as the current candidate object. Therefore, the above pixels are removed from the corresponding reference zone.

[0030] The number of valid pixels represents the size of comparable pixels within the corresponding reference band, while the percentage of valid pixels represents the proportion of valid pixels among all pixels in the corresponding reference band. A reference band is considered valid only when both the number of valid pixels and the percentage of valid pixels reach their respective preset thresholds. Validity is assessed separately for the forward and backward reference bands.

[0031] S106 When both the forward and backward reference zones are valid, the same vegetation characterization value calculation method is used. The vegetation characterization value of the candidate object is determined based on the pixels in the candidate object, and the vegetation characterization values ​​of the forward and backward reference zones are determined based on the valid pixels in each reference zone. The lower of the vegetation characterization values ​​of the two reference zones is determined as the weaker reference value, and the difference between the weaker reference value and the vegetation characterization value of the candidate object is determined as the vegetation deficit.

[0032] Candidate subjects, forward reference zones, and backward reference zones are calculated using the same method to ensure comparability of their results. For reference zones, only selected valid pixels are used to determine their vegetation characterization values.

[0033] In the implementation method that uses two phases of satellite remote sensing imagery, if the first phase... The geometric range of the candidate objects is Its statistical pixel set is The effective pixel sets of the front and rear reference bands are respectively and First, calculate NDVI, ExG, and Veg for each pixel, then calculate the arithmetic mean within their respective non-empty statistical sets. For any statistical pixel set... The mean value of vegetation combination in the later stage is defined as:

[0034] The vegetation deficit is calculated by subtracting the candidate mean from the lower of the two-sided reference means.

[0035] in, for The set of pixels that meets the data validity and geometric seedling zone location conditions is defined. The geometric seedling zone location is determined by the corresponding centerline and phase tolerance, using the same rules as the reference zone. Pixels are not filtered based on the vegetation value of the candidate object itself. Vertical lines indicate the number of pixels in the set. GNDVI, SAVI, and the two NDVI periods all use a statistical approach of calculating the mean pixel by pixel first, and then calculating the regional mean. When the set is empty, the mean is not calculated, and the corresponding candidate object becomes an object to be reviewed and is not included in the vegetation deficit discrimination.

[0036] S107: When a candidate object is located between the forward reference zone and the backward reference zone and the vegetation deficit is greater than 0, the candidate object is identified as a suspected seedling gap object, and the planar vector information representing the boundary and location of the suspected seedling gap object is output.

[0037] Candidate objects are located between the forward and backward reference zones, meaning that along the ridge direction, the forward and backward reference zones are located on either side of the candidate object and correspond to the same lateral projection range as the candidate object. When the vegetation deficit is greater than 0, it indicates that the vegetation characteristic value of the candidate object is lower than the weaker reference value of the two reference zones.

[0038] When both the aforementioned spatial location conditions and vegetation deficit conditions are met, the candidate object is identified as a suspected crop gap / discontinuity object. The spatial outline of the object is converted into planar vector information, with the boundary of the planar vector representing the range of the suspected crop gap / discontinuity object, and the spatial coordinates of the planar vector representing its position in the target crop plot.

[0039] This embodiment determines the ridge direction and ridge centerline phase based on the seedling strip texture in multispectral satellite remote sensing imagery. Candidate objects extending along the ridge direction and spanning multiple ridge centerlines are extracted in local ridge coordinates, ensuring that the candidate objects correspond to the spatial distribution characteristics of the crop seedling strips. By constructing reference zones with the same lateral projection range and covering the same ridge centerline in front of and behind the candidate objects along the ridge direction, the candidate objects and reference zones correspond to the same planting location. Pixels located outside the effective monitoring range, within other candidate objects, with invalid image data, not at the geometric seedling strip position corresponding to any ridge centerline, or with a phase inconsistent with the ridge centerline covered by the current candidate object are removed. The validity of the reference zones is then assessed to reduce interference from plot edges, other abnormal areas, and seedling strip misalignment. Furthermore, using the lower vegetation characterization value from the two effective reference zones to determine vegetation deficit reduces the impact of unilateral reference zone vegetation anomalies on the identification results, thereby improving the reliability of crop seedling gap identification and obtaining the boundary and location information of suspected seedling gap objects.

[0040] Based on the above embodiments, the multispectral satellite remote sensing imagery includes pre-concurrent and post-concurrent satellite remote sensing images. These images are multispectral satellite remote sensing images of the target crop plot acquired at two different times, with the pre-concurrent images acquired earlier than the post-concurrent images. In this embodiment, a cornfield is used as an example. The pre-concurrent and post-concurrent images were acquired on June 12, 2026, and June 19, 2026, respectively, both containing blue, green, red, and near-infrared bands, with a spatial resolution of 0.25 m.

[0041] In step S101 above, multispectral satellite remote sensing images and plot vectors of the target crop plots are acquired, and the effective monitoring range is determined based on the plot vectors. Specifically, this may include the following steps.

[0042] S1011, registers early and late satellite remote sensing images to the same projection coordinate system, pixel size, and pixel grid.

[0043] Image registration refers to unifying the spatial relationship between two satellite remote sensing images, ensuring that locations with the same geographic coordinates in the two images correspond to the same ground area. The projected coordinate system represents the geographic location of each pixel in the image; the pixel size represents the ground extent corresponding to a single pixel; and the pixel grid represents the arrangement of each pixel in the projected coordinate system.

[0044] After registration, the earlier and later satellite remote sensing images adopt the same projection coordinate system and spatial resolution, and the pixel boundaries in the two images are aligned with each other. For example, a pixel in the earlier satellite remote sensing image and a pixel in the same row and column in the later satellite remote sensing image correspond to the same ground area, thus establishing a spatial correspondence between the two satellite remote sensing images.

[0045] S1012, determine the crop plot range based on the plot vector, shrink the boundary of the crop plot range according to the preset shrinkage distance, determine the area between the boundary of the crop plot range and the shrunken boundary as the plot edge buffer zone, determine the area enclosed by the shrunken boundary as the candidate range inside the plot, and exclude the areas where roads, forest belts and water bodies are located from the candidate range inside the plot to obtain the effective monitoring range.

[0046] Specifically, the area represented by the plot vector is taken as the crop plot range, and the crop plot range is made to have a consistent spatial coordinate reference with the registered previous and subsequent satellite remote sensing images.

[0047] The preset shrinkage distance is a pre-defined spatial distance used to offset the boundary of the crop plot inwards; in this embodiment, it is set to 2 meters. After shrinking the boundary of the crop plot according to this preset shrinkage distance, a buffer zone is formed between the original boundary and the shrunken boundary, enclosing the interior area of ​​the plot. The area enclosed by the shrunken boundary serves as a candidate range for the interior of the plot. For example, for crop plots with irregularly shaped boundaries, the shrunken boundary maintains the overall outline corresponding to the original boundary, but is spaced apart from the original boundary by the preset shrinkage distance.

[0048] Furthermore, the areas containing roads, forest belts, and water bodies within the candidate areas of the land parcel are identified, and these areas are excluded from the candidate areas within the land parcel. Specifically, road areas include areas accessible to vehicles or pedestrians but not planted with crops; forest belt areas include areas formed by a continuous distribution of trees; and water body areas include ditches, ponds, and other areas with water features within or crossing the land parcel.

[0049] The candidate area within the plot, excluding areas containing roads, forest belts, and water bodies, is determined as the effective monitoring area. If... Indicates the candidate range within the land parcel, in order to , and The effective monitoring range is defined by the areas where roads, forest belts, and water bodies are located, respectively. It can be represented as:

[0050] Therefore, the effective monitoring range used in subsequent processing is located within the crop plot and does not include the plot edge buffer zone, roads, forest belts and water bodies.

[0051] In step S102 above, the ridge direction and ridge centerline phase are determined based on the seedling strip texture in the multispectral satellite remote sensing image, the ridge spacing is obtained, and a set of ridge centerlines is generated based on the ridge direction, ridge centerline phase and ridge spacing. Specifically, this may include the following steps.

[0052] S1021, based on the internal boundaries of the plot, the continuity of the seedling strip, and the local directional differences, the target crop plot is divided into multiple planting zones with consistent ridge orientation.

[0053] Among them, the internal boundary of the plot refers to the spatial boundary within the target crop plot used to distinguish different planting areas; the continuity of the seedling strip is used to characterize whether the seedling strip maintains a continuous extension in space; and the local directional difference is used to characterize whether the extension direction of the seedling strip in different areas within the plot is consistent.

[0054] Based on the above information, areas within the target crop plot where the seedling strips are relatively consistent in direction and continuously distributed are identified, and each area is defined as a ridge-oriented consistent planting zone. For example, if different areas of the same target crop plot have different seedling strip extension directions due to different sowing directions, ridge-oriented consistent planting zones can be defined according to the location where the seedling strip direction changes. The ridge direction and ridge centerline phase are then determined for each ridge-oriented consistent planting zone.

[0055] The local coordinates along the ridge direction use the same projection coordinate reference as the two phases of imagery, with a fixed reference origin for each zone. The ridge direction is defined as the longitudinal direction, and the direction rotated 90 degrees counterclockwise is defined as the lateral direction. The ridge direction angle is defined with the eastward direction of the projection coordinates as zero and the counterclockwise direction as positive, and the phase is defined relative to the same origin. The preceding and following reference zones are only translated longitudinally, without re-estimating the lateral position. When an object crosses a zone boundary, it is first divided into zones and then processed separately.

[0056] Candidate objects are compared with the forward and backward reference bands within the same ridge-direction consistent planting zone.

[0057] S1022, for each ridge-oriented consistent planting zone, obtain the ridge spacing of the ridge-oriented consistent planting zone, set a local analysis window within the ridge-oriented consistent planting zone, and perform a two-dimensional discrete Fourier transform on the spatial distribution of vegetation combination quantity within the spatial frequency range corresponding to the ridge spacing, and determine the power spectrum based on the obtained frequency domain transform result.

[0058] Row spacing is the horizontal distance between adjacent seedling strips. Vegetation composition is a combination of values ​​determined from multiple spectral bands in multispectral satellite remote sensing imagery, used to characterize the spatial distribution of crop vegetation. The local analysis window is a local image region within a row-oriented planting zone used to analyze seedling strip texture; each pixel in the window has a corresponding vegetation composition.

[0059] In the calculation of vegetation combination, the two images are converted to reflectance with consistent dimensions and subjected to radiometric consistency processing; the normalized vegetation index and supergreen index are first calculated based on the reflectance of the near-infrared, red, green and blue bands of the same period, and then the two are weighted and combined:

[0060]

[0061]

[0062] in, Let 1 and 2 represent the early and late stages, respectively; the subscripts of each reflectance symbol , , , Corresponding to the near-infrared, red, green, and blue bands respectively, reflectance uses dimensionless small values, and numerical stability constants. A uniform value of 10 to the power of -6 is used for denominator protection in calculations of various indices and brightness normalization. Vegetation combination amounts are calculated for both periods; the ridge orientation and phase use time phases with higher texture quality, and their time phase numbers are denoted as follows. Later Used for candidate extraction and vegetation deficiency detection.

[0063] Among a set of implementation parameters, row spacing Based on farm planting records and confirmed by on-site measurements, the spacing is 1.10m, with a local analysis window of 64m × 64m; when the image pixel size is 0.25m, this window is 256 × 256 pixels. The spatial frequency modulus range corresponding to the ridge spacing is 0.72–1.08 Hz / m, covering the dominant frequency of 0.9091 Hz / m corresponding to a ridge spacing of 1.10m. This represents the set of discrete physical frequency vectors whose magnitudes fall within this range after spectral centering and coordinate transformation.

[0064] Both image phases used identical local analysis windows within each ridge-oriented planting zone, and calculations were performed using the following formula; the results used for subsequent spatial construction were taken from the selected texture phase. Indicates partition number, This indicates the window number, with M and M cells in the two grid directions respectively. The vegetation combination amount after removing the mean is denoted as The two-dimensional discrete Fourier transform is expressed as:

[0065] in, subscript and These represent integer cell indices for the two raster directions, respectively. and For discrete frequency indexing, The unit is imaginary. After spectral centering, the frequency index is converted to the physical frequency in projected coordinates based on the pixel size and image affine transformation; angle calculations use the projected coordinate direction to avoid mistakenly assigning north as the direction of increasing image row number. (The rest of the text is omitted.) Represents the physical frequency vector Corresponding discrete frequency index The same power spectrum value at that location; and The only difference is the coordinate representation method; the power spectrum is not defined separately.

[0066] The power spectrum is determined by the square of the magnitude of the corresponding frequency domain transform result:

[0067] When seedling strips are periodically distributed according to corresponding row spacing, they form a frequency response in the power spectrum corresponding to that row spacing. Therefore, by analyzing the power spectrum within the spatial frequency range corresponding to the row spacing, the directional energy distribution corresponding to the periodic texture of the seedling strips can be determined.

[0068]

[0069]

[0070]

[0071]

[0072] Both phases use the same window position, frequency range, and angle division. Texture quality is evaluated by the ratio of peak energy to average energy in each window direction. Then, the median of the evaluation values ​​of the common effective windows within the partition is compared, and the phase corresponding to the higher median is selected as the time phase. If they are the same, the later one will be selected. Windows with zero average directional energy will not be included in the comparison. If the common valid windows are empty, the comparison will be pending.

[0073] in, Indicates the first Within the first ridge-oriented planting zone A local analysis window is located in the center of the sector. The directional energy at a given location is a set of frequencies. The sum of the power spectrum values ​​corresponding to each physical frequency vector located within the sector in that direction; Indicates direction angle and by The minimum angular distance with periodicity. and The range of values ​​is ; For physical frequency vectors, This represents the set of frequencies whose modulus falls within the frequency range corresponding to the row spacing; α is the direction of the sector center. The sector half-width is used, and the direction angle is normalized to a range of radians from zero to less than π. The formula consistently uses radians; degrees in text are converted to radians before being substituted into the formula. In one implementation, the sector center direction is sampled in 1-degree increments, and the half-width is 0.5 degrees; when equal peak values ​​exist, the smaller sector direction is used. The peak value of the window spectrum represents the seedling normal direction, obtained by a fixed 90-degree rotation. ; For the corresponding peak energy, The directional angle of the partition is obtained by averaging the peak energy of each effective window using a double-angle circular method. The first parameter of atan2 is the sine weighted sum, and the second parameter is the cosine weighted sum; when the composite vector is zero or there is no effective directional window, the partition is to be reviewed.

[0074] S1023, determine the directional energy peak within the spatial frequency range based on the power spectrum, and rotate the direction corresponding to the directional energy peak by 90° to obtain the ridge angle of the ridge-oriented planting zone.

[0075] The direction corresponding to the directional energy peak represents the lateral frequency change direction of the seedling strip periodic texture and is perpendicular to the seedling strip extension direction. Therefore, in this embodiment, rotating this direction by 90° and normalizing the resulting angle to the range of 0° (inclusive) to 180° (exclusive) yields the ridge angle corresponding to the seedling strip extension direction.

[0076] The ridge angle is used to represent the extension angle of the seedling strip in a ridge-oriented planting zone relative to a preset coordinate direction. When multiple local analysis windows are set for the same ridge-oriented planting zone, the ridge angle of the ridge-oriented planting zone can be determined jointly based on the directional energy peak values ​​corresponding to each local analysis window.

[0077] S1024. Based on the ridge angle and ridge spacing, generate parallel candidate ridge center lines. Within a phase range with a length equal to the ridge spacing, sequentially adjust the phase of the candidate ridge center lines. Determine the average vegetation combination amount of the pixels covered by the candidate ridge center lines under each phase. Then, determine the phase corresponding to the largest average vegetation combination amount as the ridge center line phase of the ridge-oriented consistent planting zone.

[0078] Specifically, the extension direction of the candidate ridge centerline is determined according to the ridge angle, and the lateral spacing between adjacent candidate ridge centerlines is determined according to the ridge spacing, thereby generating a set of mutually parallel candidate ridge centerlines.

[0079] The phase of the candidate ridge centerline represents the overall lateral position of that group of candidate ridge centerlines. Keeping the ridge angle and ridge spacing constant, the lateral position of the candidate ridge centerlines is sequentially changed within a phase range equal to the ridge spacing. After a phase change of one ridge spacing, the relative position between the candidate ridge centerline and the seedling strip begins to repeat; therefore, the phase adjustment range is set to one ridge spacing.

[0080] For each candidate phase, the average vegetation combination at the sampling location of the candidate centerline is calculated for the selected texture phase, and the centerline position is determined by the phase with the largest average value.

[0081] Wherein, CL represents the set of centerline sampling positions generated based on the partitioned row angle, candidate phase, and fixed row spacing. This represents the row spacing. Each candidate phase is defined relative to the same projection coordinate origin, traversed within a row spacing range according to a preset step size, and uses the same sampling interval and interpolation rules along the line; when the phase mean is the same, the smaller phase is taken.

[0082] For example, when the row spacing is 1.10 m, candidate phases are traversed from 0 m (inclusive) to 1.10 m (exclusive) with a step size of 0.05 m. Sampling is performed along each candidate centerline at approximately 50 m intervals, with at least two sampling points taken for each centerline. Bilinear interpolation is used to obtain sampled values ​​on the vegetation combination image of the selected texture phase. The sampled mean values ​​of each candidate phase are compared to determine the centerline position corresponding to the vegetation distribution of the seedling strip. The phase step size can be set to a value smaller than the image pixel size; the resulting sub-pixel phase is not equivalent to the corresponding decimal place accuracy in the field.

[0083] S1025 generates a set of ridge centerlines based on the ridge angle, ridge centerline phase, and ridge spacing.

[0084] Among them, the ridge angle is used to define the extension direction of each ridge centerline, the ridge centerline phase is used to define the starting position of each ridge centerline in the lateral direction, and the ridge spacing is used to define the interval between adjacent ridge centerlines. According to the above three parameters, multiple ridge centerlines that are parallel to each other and correspond to the spatial position of the seedling strip are generated in the corresponding ridge-oriented consistent planting zone, thus obtaining the set of ridge centerlines of the ridge-oriented consistent planting zone.

[0085] By processing each ridge-oriented planting zone in the target crop plot separately, a set of ridge centerlines corresponding to different local planting directions can be obtained.

[0086] Based on the above embodiments, step S103, which involves extracting candidate objects from the effective monitoring range, may specifically include the following steps.

[0087] S1031. Calculate the vegetation combination amount or vegetation index of each pixel based on multispectral satellite remote sensing imagery to obtain the vegetation characterization value of each pixel. Pixels with vegetation characterization values ​​lower than the corresponding local reference values ​​are identified as candidate pixels. The local reference values ​​are determined based on the vegetation characterization values ​​of pixels in the local area surrounding the corresponding pixel.

[0088] The vegetation combination amount can be adopted as defined in step S1022. Vegetation indices are vegetation characterization parameters determined based on the differences in reflectance between different spectral bands. Either vegetation combination amount or vegetation index can be selected as the vegetation characterization value, and all pixels within the effective monitoring range can be processed using the same calculation caliber.

[0089] For any pixel within the effective monitoring range, a local area is defined around that pixel, and a local reference value is determined based on the vegetation characterization values ​​of each pixel within that local area. The local reference value reflects the vegetation status of the area surrounding the pixel. For example, a local sliding window can be established centered on the pixel to be assessed, removing pixels that are missing, saturated, or located outside the effective monitoring range, and the median of the vegetation characterization values ​​of the remaining pixels within the window is determined as the local reference value. The size of the local sliding window is set according to the row spacing and image spatial resolution, with a horizontal dimension greater than one row spacing to cover the seedling strip where the pixel to be assessed is located and adjacent seedling strips; its specific determination method can be set according to the vegetation characterization values ​​and image data characteristics used.

[0090] If we take Represents a pixel The vegetation characterization value, in Represents a pixel The corresponding local reference value, when satisfying the following formula, will be the pixel. Identified as a candidate pixel:

[0091] For example, when a pixel is located in an area with relatively low vegetation characterization values, while the overall vegetation characterization values ​​of its surrounding local areas are relatively high, this pixel can be identified as a candidate pixel. The candidate pixels obtained in this way reflect the pixel locations that show a decline in vegetation status relative to the surrounding local vegetation.

[0092] S1032, perform connected component aggregation on candidate pixels, and perform morphological opening and closing operations, merging of adjacent regions and boundary reconstruction on the aggregated connected regions to obtain a planar object.

[0093] Connected region aggregation refers to grouping interconnected candidate cells into the same connected region based on their spatial adjacency. A single connected region consists of one or more spatially contiguous candidate cells and is used to represent an initial low-vegetation area.

[0094] Morphological opening and closing operations are performed on connected regions. Morphological opening is used to handle small protrusions at the edges of connected regions and isolated cells, while morphological closing is used to connect closely spaced regions and handle small discontinuities within connected regions. After morphological opening and closing operations, the overall outline of the connected regions corresponds to the spatial distribution of candidate cells.

[0095] For connected regions that are spatially adjacent and belong to the same continuous low-value area, adjacent regions are merged to form a complete region. Then, boundary reconstruction is performed based on the outer contour of the merged region, transforming the region composed of discrete pixels into a planar object with closed boundaries. Each planar object corresponds to a low-value vegetation area to be morphologically screened.

[0096] S1033, determine the planar objects that have an angle between the main axis direction and the ridge direction that is not greater than a preset angle threshold, a ridge length that is not less than a preset length threshold, and a lateral width that is not less than a preset width threshold as candidate objects.

[0097] The principal axis direction of the areal object represents the direction in which the areal object extends the most. The angle between the principal axis direction and the ridge direction of the corresponding planting zone is determined, and this angle is compared with a preset angle threshold. If the angle is not greater than the preset angle threshold, it indicates that the extension direction of the areal object is the same as or approximately the same as the ridge direction.

[0098] In the local coordinate system along the ridge, the projected length of the planar object along the ridge direction is defined as the ridge length, and the projected width of the planar object in the lateral direction is defined as the lateral width. The ridge length is used to characterize the continuous range of the planar object along the extension direction of the seedling strip, and the lateral width is used to characterize the coverage range of the planar object perpendicular to the direction of the seedling strip.

[0099] The length along the ridge and the width laterally of the areal object are compared with the corresponding preset length threshold and preset width threshold, respectively. An areal object is identified as a candidate object when it simultaneously meets the following conditions:

[0100] in, Let be the minimum undirected angle between the principal axis of the object and the longitudinal direction. The included angle threshold, and These are the length along the ridge and its threshold, respectively. and These are the horizontal width and its threshold, respectively. Candidate objects must also cover at least two center lines in the horizontal direction.

[0101] Therefore, objects with relatively low vegetation characterization values ​​compared to the surrounding local areas, extending along the ridge direction and having corresponding ridge length and lateral width, are retained from the areal objects as candidate objects for subsequent construction of forward and backward reference zones.

[0102] Based on the above embodiments, in step S104, the lengths along the ridge of the forward reference band and the backward reference band are 0.8 to 1.5 times the length along the ridge of the candidate object, and the forward reference band and the backward reference band are spaced at least one pixel apart from the candidate object along the ridge direction.

[0103] Specifically, if the length of the candidate object along the ridge is The length of the forward reference band along the ridge is The length of the backward reference band along the ridge is Then the following condition is met:

[0104] The lengths along the ridges of the forward and backward reference zones can be the same, or they can be determined separately within the ranges mentioned above. The lateral projection range of the reference zone is the same as the lateral coverage range of the candidate object. Therefore, by limiting the length along the ridges of the reference zone, the reference zone and the candidate object can have similar spatial scales.

[0105] For example, when the ridge length of a candidate object corresponds to 10 pixels, the ridge lengths of the forward and backward reference bands can correspond to 8 to 15 pixels, respectively. The specific ridge lengths of the two reference bands can be determined based on the available areas in front of and behind the candidate object along the ridge direction.

[0106] A forward reference band is positioned in front of the candidate object along the ridge direction, and a backward reference band is positioned behind the candidate object along the ridge direction, with each reference band separated from the candidate object by at least one pixel. If... and These represent the ridge spacing between the forward reference band, the backward reference band, and the candidate object, respectively. Let the distance of a cell along the ridge direction be represented, then the following condition is met:

[0107] Therefore, neither the forward reference band nor the backward reference band overlaps with the candidate object, and the candidate object is located between the two reference bands.

[0108] Furthermore, in step S105, the lateral distance between each valid pixel and the nearest ridge centerline covered by the candidate object is not greater than a preset phase tolerance.

[0109] Specifically, the ridge-direction centerline covered by the candidate object is determined based on the lateral coverage of the candidate object. For any pixel to be screened in the forward reference band and the backward reference band, the lateral distance between the pixel and each of the above-mentioned ridge-direction centerlines is determined, and the minimum value among them is taken as the lateral distance between the pixel and the nearest ridge-direction centerline.

[0110] by Represents the set of centerlines of the rows covered by the candidate objects, in order to Number the center line; Represents a pixel With the center line The lateral distance is: The lateral distance between a pixel and the nearest center line is:

[0111] Compare the nearest lateral distance with the preset phase tolerance. When the following conditions are met:

[0112] At that time, like a pixel It corresponds horizontally to the center line of the ridge covered by the candidate object. Among them, This indicates the preset phase tolerance, used to limit the allowable lateral positional deviation of a pixel in the reference band relative to the corresponding centerline.

[0113] For example, if a candidate object covers three ridge-direction centerlines laterally, the lateral distance between each pixel in the reference band and these three ridge-direction centerlines is determined. A pixel satisfies the corresponding ridge-direction centerline phase requirement as long as its lateral distance to the nearest ridge-direction centerline is not greater than a preset phase tolerance. Simultaneously satisfying the steps... When considering criteria such as same partition, data validity, and exclusion of other candidate objects, the cell is determined as a valid cell.

[0114] To standardize the above screening criteria, the first The set of valid pixels in the reference band on one side of a candidate object can be represented as:

[0115] in, Indicates the first The candidate is the first The geometric range of the side reference zone, Take respectively , The corresponding reference valid pixel sets are respectively and ; Indicates the partition to which the candidate object belongs. A mask is used to determine the geometric seedling strip location corresponding to any ridge-to-center line, based on both the geometric position and the validity of the image data. The location is defined by the ridge-to-center line and phase tolerance; invalid observations such as missing or saturated data are assigned zeros. Seedling strip locations are not re-screened using vegetation height thresholds. The apostrophe number indicates other candidate objects, which are excluded based on their corresponding pixels within the strip. Object set. The same geometric location and data validity rules are used. The cell locations and geometric extents in the set conditions correspond to a unified image grid.

[0116] The effective pixel percentage of the corresponding reference band is the ratio of the number of effective pixels in that reference band to the total number of pixels in that reference band. If we take... Indicates the number of valid pixels in the corresponding reference band, in Indicates the total number of pixels in the corresponding reference band, in If we represent the percentage of effective pixels, then:

[0117] The total number of reference pixels is determined on a unified image grid based on the reference zone boundary before removal. When the total number of pixels is zero, the reference zone is invalid and its proportion is not calculated; when the candidate object statistical set is empty, it is marked as pending review and is not included in the vegetation deficiency discrimination.

[0118] The preset threshold for the number of effective pixels is the preset minimum number of effective pixels, and the preset threshold for the proportion of effective pixels is the preset minimum proportion of effective pixels. These parameters are preset based on spatial resolution, row spacing, candidate object scale, and available pixel proportion. Indicates the preset minimum number of valid pixels, in order to This indicates the preset minimum effective pixel ratio. The corresponding reference band is determined to be valid only when the following two conditions are met:

[0119] In a set of embodiment parameters with 0.25 m spatial resolution and 1.10 m row spacing, the phase tolerance... Set the value to 0.35m and preset the minimum number of effective pixels. Select 8, preset minimum effective pixel ratio Take 50%.

[0120] For example, if a reference band has a high percentage of valid pixels, but the total number of valid pixels does not reach the preset minimum number of valid pixels, then that reference band is invalid. Similarly, if another reference band has the preset minimum number of valid pixels, but the percentage of valid pixels does not reach the preset minimum percentage of valid pixels, then that reference band is also invalid. A reference band is only determined to be valid when both the number of valid pixels and the percentage of valid pixels reach the corresponding preset thresholds.

[0121] The joint valid state of the two reference bands can be represented as:

[0122] in, The logical "AND" is valid only when both the forward reference band and the backward reference band are valid. ; Reference band valid When invalid If any reference band is invalid, then The corresponding candidate objects are marked as objects to be reviewed or objects with low confidence, and are not proceeded to step S106 for determining the two-sided reference vegetation deficit; the other side reference band is not used for unilateral compensation, nor is a simple average of the valid and invalid reference bands performed. Step S106 is only executed when both reference bands are valid.

[0123] Based on the above embodiments, in step S107, the candidate object is identified as a suspected missing seedling or broken strip object, which may specifically include the following steps.

[0124] S1071, determine the differences in green light normalized vegetation index and soil-regulated vegetation index between the candidate object and the reference area, wherein the reference area consists of effective pixels in the forward reference zone and the backward reference zone.

[0125] The effective pixel sets of the front and rear reference bands are respectively and After merging, we get the first Spectral auxiliary reference set of candidate objects :

[0126] In the candidate object statistical set respectively and auxiliary reference set Within this step, the arithmetic mean of GNDVI and SAVI calculated pixel by pixel is taken, and then the difference between the reference mean and the object mean is calculated. The premise for this step is that both sides of the reference are valid and the vegetation deficit is positive.

[0127] The green light normalized vegetation index (RTV) is used to characterize the difference in reflectance between the near-infrared band and the green light band. For the first... The green normalized vegetation index (GRI) of any pixel in the image is expressed as:

[0128] in, and They represent the first Near-infrared and green band reflectance of the image. The value obtained in step S1022 is used.

[0129] The soil-modified vegetation index is used to characterize the difference in reflectance between the near-infrared and red light bands, and the soil background influence is adjusted by the soil moderating coefficient. For the first... The soil-modified vegetation index for any pixel in the image is expressed as:

[0130] The reflectivity of the red light band follows the definition mentioned above. The soil adjustment coefficient is set to 0.5 in this implementation method, and is related to the length of the ridge along which the soil is shaped. Use them differently. Reflectance must be a dimensionless decimal value that matches the dimensions of the soil conditioning system, and should not be directly substituted into the original DN or displayed as an elongation value.

[0131] Determine the green light normalized vegetation index of the candidate objects respectively and the green light normalized vegetation index of the reference area The difference between the normalized green vegetation index (NVR) of the reference area and the candidate areas was defined as the NVR difference.

[0132] Soil-modified vegetation index determined for each candidate. Soil-modified vegetation index and reference area The difference between the soil moderating vegetation index of the reference area and the candidate areas was defined as the soil moderating vegetation index difference.

[0133] A difference in the normalized green vegetation index (NVR) greater than 0 indicates that the NVR of the candidate area is lower than that of the reference area; a difference in the soil-regulated vegetation index (SVR) greater than 0 indicates that the SVR of the candidate area is lower than that of the reference area.

[0134] S1072, determine the degree of edge interference, water-wet interference, shadow interference, and bright spot interference of the candidate object, wherein the degree of shadow interference and bright spot interference are determined based on the locally normalized visible light brightness.

[0135] Edge interference level is used to characterize the degree to which candidate objects are affected by the edge area of ​​the plot; water moisture interference level is used to characterize the degree to which candidate objects are affected by abnormal water or soil moisture; shadow interference level is used to characterize the degree to which candidate objects exhibit shadow features due to relatively low visible light brightness; bright spot interference level is used to characterize the degree to which candidate objects exhibit bright spot features due to relatively high visible light brightness.

[0136] Visible light brightness is a brightness index determined based on the visible light band in multispectral satellite remote sensing imagery. Local normalization of visible light brightness refers to scaling the visible light brightness of a candidate object to that of its surrounding local area, using the brightness difference of the candidate object relative to its surrounding local area to represent the local brightness characteristic.

[0137] When the locally normalized visible light intensity is relatively low compared to the surrounding local area, the degree of shadow interference of the candidate object is determined accordingly; when the locally normalized visible light intensity is relatively high compared to the surrounding local area, the degree of bright spot interference of the candidate object is determined accordingly. The locally normalized visible light intensity is used to determine the degree of shadow interference and bright spot interference, but it is not used as a substitute indicator for the difference in green light normalized vegetation index or soil-regulated vegetation index.

[0138] In one implementation, the visible light brightness is first calculated based on the blue, green, and red band reflectance of the later image, and then normalized within a 64 m × 64 m local window centered on the centroid of the candidate object.

[0139]

[0140] in, Represents pixels in later images The visible light brightness is the arithmetic mean of the reflectance of the blue, green, and red bands of that pixel; Represents a cell The visible light intensity is locally normalized and truncated to The dimensionless brightness value obtained after the interval; and These are the 2% and 98th percentiles of the effective pixel brightness within the local window, respectively. The effective pixels in the window are those located within the effective monitoring range and effective in all four bands of the subsequent image. Brightness is calculated using dimensionless reflectance. Using the same meaning and values ​​as in step S1022, this is only used to prevent numerical instability caused by a brightness quantile difference of zero or too small, and does not constitute an independent brightness discrimination threshold. It is only applicable when the number of effective pixels within the window is not less than 500 and the high / low quantile difference is greater than... Normalization is performed only when the object's brightness is within acceptable limits; otherwise, the object's brightness discrimination information is invalid, and it is transferred to a pending review stage. A default normalization value is not assigned, and it is not considered interference-free based on this. Shadow masks are formed when the normalized brightness is less than 0.20 or greater than 0.80. and bright spot mask .

[0141] Water-wet mask Based on pixels with a later normalized water index greater than 0.00, in Internal generation indicates localized water quality anomalies that persist even after prior exclusion of clearly identified water bodies; it is not a duplicate masking of water areas such as ditches and ponds excluded in step S1012, nor is it used for quantitative inversion of soil moisture content. The normalized water index is:

[0142] Edge checking uses the distance from the centroid of the candidate object to the original parcel boundary to avoid repeatedly calculating the overlap ratio of excluded edge buffers. This represents the original land parcel extent determined by the land parcel vector. This indicates the distance. The edge influence distance is greater than the pre-recession distance; in this embodiment, it is set to 8 m. The edge interference level is set to 1 to indicate that the area is within this influence distance, and 0 to indicate that the area is outside of it.

[0143]

[0144] in, It only represents the geometric extent of the candidate object; This refers to the set of pixels located within this area according to a unified image grid that meet the conditions for validity of subsequent image data, without being filtered based on geometric seedbed location or vegetation value, as mentioned above. Use them separately. , and Unified as a binary indicator function, for cells within the set The function value is 1 when the corresponding interference condition is met, and 0 otherwise; invalid observations such as missing measurements and saturation are not included. The three types of interference levels are calculated as follows:

[0145]

[0146]

[0147] The levels of interference from water, shadows, and bright spots represent the percentage of pixels in the corresponding mask within the candidate object. The vertical lines in the denominator only indicate the percentage of pixels in the mask. The number of pixels within the set; if the set is empty, it will be transferred to the review stage and the proportion will not be calculated. The edge interference level is zero or one, and the other three items are between zero and one, with larger values ​​indicating stronger interference. The comprehensive interference level can be calculated using the following formula:

[0148] Visible light brightness is only used to form shadow masks and bright spot masks and to calculate the corresponding interference level; it does not trigger the output of suspected missing seedlings or broken bands on its own.

[0149] S1073, candidates with differences in green light normalized vegetation index and soil-regulated vegetation index that are both greater than 0, and with edge interference level of 0, water-wet interference level, shadow interference level and bright spot interference level that are all less than their respective interference exclusion thresholds are identified as suspected seedling gaps and broken belts.

[0150] The four levels of interference are retained. , , , The thresholds for excluding interference from water, shadows, and bright spots are respectively... , , All values ​​are greater than zero and not greater than one; in this embodiment, all values ​​are 0.20. Objects with an edge interference level of 1 are directly excluded, and only objects with an edge interference level of 0 are included in the subsequent judgment; no additional edge interference exclusion threshold is set. Under the premise that both sides of the reference are valid, the object is located in between, and the vegetation deficit is positive, the additional screening condition is:

[0151] The difference between two vegetation indices is used to determine whether the candidate object shows a decline in vegetation status relative to the reference area, and the degree of interference is used to determine whether the candidate object is affected by factors such as plot edges, waterlogging, shading, or bright spots. Based on the aforementioned two-sided validity and vegetation deficit conditions, the candidate object is identified as a suspected crop gap object only when the difference between both vegetation indices is greater than 0, the edge interference degree is 0, and the other three types of interference degrees are all less than their respective interference exclusion thresholds. If any waterlogging, shading, or bright spot interference degree is greater than or equal to its corresponding threshold, the object is not output as a suspected crop gap object and can be transferred to the pending review stage.

[0152] Based on the above embodiments, multispectral satellite remote sensing imagery includes earlier satellite remote sensing imagery and later satellite remote sensing imagery. The earlier satellite remote sensing imagery was acquired earlier than the later satellite remote sensing imagery, and the two sets of satellite remote sensing imagery, after registration, have a corresponding spatial relationship.

[0153] Both early and late images are used in the seedling texture quality comparison in step S1022, and the selected time phases The images are used to determine the ridge orientation and centerline phase of the zoning; later images are used for candidate object extraction and basic vegetation deficit discrimination, and the two images are also used together for vegetation change ranking as described later.

[0154] Post-conversion satellite remote sensing imagery is used to extract candidate objects, construct forward and backward reference bands, and determine vegetation deficit. Specifically, in post-conversion satellite remote sensing imagery, areal objects with vegetation characterization values ​​lower than their surrounding local areas are identified, and two reference bands are constructed based on the ridge length and lateral coverage of these areal objects. Furthermore, the vegetation deficit is determined based on the vegetation characterization values ​​of the candidate objects and the two effective reference bands in the post-conversion satellite remote sensing imagery.

[0155] In step S107, after identifying suspected seedling gaps and broken vegetation zones, the suspected seedling gaps and broken vegetation zones can be sorted according to the changes in the normalized vegetation index in the two phases of satellite remote sensing images. Specifically, the steps are as follows.

[0156] S1074, Based on the pixels in the candidate objects and the effective pixels in each reference zone, determine the normalized vegetation index changes of the candidate objects, the forward reference zone, and the backward reference zone, respectively.

[0157] The normalized vegetation index is defined in step S1022. Calculate the caliber; here, let... and This allows us to obtain the normalized vegetation index of the corresponding pixels in the early and later images.

[0158] After converting both images to the same dimension of surface reflectance and completing geometric registration and radiometric consistency processing, common valid pixel locations from both periods are selected within the candidate object and each reference zone statistical set. If the number of common valid pixels in each statistical set is no less than 8, the corresponding NDVI mean values ​​for the preceding and following periods are calculated. The same sampling locations are used for both periods of the same area, without resampling due to changes in vegetation values. If the image data used for comparison in either period is still the original DN, exhibits saturation, or cannot be unified to the same reflectance scale, it is considered that the radiometric conditions are incomparable. If the number of common valid pixels in any statistical set is less than 8, or the radiometric conditions of the two periods are incomparable, the basic discrimination result is retained but marked as insufficient temporal evidence, and it is not included in the temporal ranking.

[0159] If the normalized vegetation index of the candidate object in the early and later satellite remote sensing images are respectively and The change in the normalized vegetation index of the candidate objects for:

[0160] For the forward reference band, the normalized vegetation index (NDI) of the effective pixels within the reference band is determined in both earlier and later satellite remote sensing images. The change in the NDI of the forward reference band is also considered. for:

[0161] For the backward reference zone, the effective pixels within this zone are also used to determine their normalized vegetation index (NDI) in the two phases of satellite remote sensing imagery. The change in NDI of the backward reference zone is also considered. for:

[0162] The changes in the three normalized vegetation indices were all determined by subtracting the normalized vegetation index from the normalized vegetation index in the previous satellite remote sensing image from the normalized vegetation index in the later satellite remote sensing image, in order to maintain the consistency of the temporal direction of the changes.

[0163] S1075, sort the suspected seedling gaps and broken seedlings based on the differences in changes between the two periods.

[0164] The difference in changes between the two periods is obtained by subtracting the change in the normalized vegetation index (NVI) of the candidate object from the lower of the changes in the forward and backward reference zones. To represent the difference in changes between two periods, then:

[0165] The lower of the normalized vegetation index (NDI) changes in the forward and backward reference bands is used to characterize the relatively weaker vegetation changes in the two reference bands. The NDI change of the candidate object is used to characterize the vegetation change of the candidate object between the two satellite remote sensing images. The difference between the two periods of change is used to represent the difference between the vegetation change of the candidate object and the weaker vegetation change in the two reference bands.

[0166] For suspected candidates with valid temporal evidence, they are sorted from largest to smallest difference in change between the two periods. The larger the value, the less the candidate's NDVI change is relative to the weaker bilateral references, and the higher the priority for verification. This value is not equivalent to the probability of missing seedlings; the difference in change between the two periods is only used for sorting and is not used as an independent trigger or rejection condition for suspected missing seedlings or broken bands.

[0167] S1076 outputs attribute information associated with the area vector information.

[0168] When outputting planar vector information representing the boundary and location of suspected seedling gaps and breaks, corresponding attribute information is established for each suspected seedling gap and break, and this attribute information is associated with the corresponding planar vector object.

[0169] The attribute information includes object number, plot number, object area, effective pixel count of each of the forward and backward reference zones, vegetation deficit, and ranking information. Specifically, the object number distinguishes different suspected crop loss zones within the same plot; the plot number indicates the target crop plot to which the suspected crop loss zone belongs; the object area is the ground area covered by the corresponding areal vector object; the two effective pixel counts correspond to the forward and backward reference zones, respectively; the vegetation deficit is the difference between the candidate object determined based on subsequent satellite remote sensing imagery and the weaker reference value; and the ranking information indicates the position of the suspected crop loss zone in the ranking results.

[0170] To illustrate the output results for suspected seedling gaps or broken strips, UAV imagery can be used to compare the image representations of different seedling conditions. For example... Figure 2 As shown, areas with obvious seedling deficiency are characterized by the absence of local seedlings in continuous seedling strips, while areas with weak seedlings or low growth are characterized by reduced vegetation cover in the seedling strips compared to the surrounding areas, and normal seedling strips maintain a relatively continuous strip-like distribution.

[0171] like Figure 3 As shown, the left side is a later satellite remote sensing image, and the right side is a UAV image of the same location. The closed boundary indicates the area of ​​suspected objects to be verified. UAV imagery is used to verify the continuity of seedling belts and the location of anomalies; objects adjacent to forest belts, roads, wetlands, or shaded areas still need to be verified according to the aforementioned interference conditions. This comparison image is used to illustrate the result verification method. UAV imagery is not involved in the candidate extraction, two-sided reference construction, effective pixel screening, vegetation deficiency calculation, or interference threshold determination of this method.

[0172] Another embodiment of the present invention relates to a satellite remote sensing monitoring system for crop seedling gaps and crop belt breaks, such as... Figure 4 As shown, it includes: The monitoring range determination module 40 is used to acquire multispectral satellite remote sensing images and plot vectors of the target crop plots, and determine the effective monitoring range based on the plot vectors. The ridge direction spatial construction module 41 is used to determine the ridge direction and ridge direction centerline phase based on the seedling strip texture in the multispectral satellite remote sensing image, obtain the ridge spacing, generate a set of ridge direction centerlines based on the ridge direction, the ridge direction centerline phase and the ridge spacing, and establish ridge direction local coordinates including ridge direction and transverse direction, wherein the transverse direction is perpendicular to the ridge direction. The object processing module 42 is used to extract candidate objects from the effective monitoring range. The candidate objects are planar objects whose vegetation characterization values ​​are lower than the vegetation characterization values ​​of the local area surrounding the candidate object, extend along the ridge direction and cross multiple ridge direction center lines, and determine the ridge length and lateral coverage of each candidate object. The reference strip processing module 43 is used to construct a forward reference strip and a backward reference strip in front of and behind the candidate object along the ridge direction, respectively, based on the ridge length and lateral coverage of the candidate object, so that the forward reference strip and the backward reference strip have the same lateral projection range as the candidate object, and cover the ridge centerline covered by the candidate object in the lateral direction. The validity determination module 44 is used to remove pixels from the forward reference band and the backward reference band that are located outside the effective monitoring range, located within other candidate objects, have invalid image data, are not located at the geometric seedling strip position corresponding to any of the ridge center lines, or are inconsistent with the phase of the ridge center line covered by the current candidate object. The retained pixels are determined as valid pixels. When the number of valid pixels and the proportion of valid pixels in the corresponding reference band both reach the corresponding preset threshold, the corresponding reference band is determined to be valid. The vegetation deficit determination module 45 is used to determine the vegetation deficit amount by using the same vegetation characterization value calculation method when both the forward reference zone and the backward reference zone are valid, based on the pixels in the candidate object, the vegetation characterization value of the candidate object is determined, and the vegetation characterization value of the forward reference zone and the backward reference zone is determined based on the valid pixels in each reference zone, the lower of the vegetation characterization values ​​of the two reference zones is determined as the weaker reference value, and the difference between the weaker reference value and the vegetation characterization value of the candidate object is determined as the vegetation deficit amount. Output module 46 is used to determine the candidate object as a suspected seedling gap object when the candidate object is located between the forward reference zone and the backward reference zone and the vegetation deficit is greater than 0, and output planar vector information representing the boundary and position of the suspected seedling gap object.

[0173] Based on the above embodiments, the multispectral satellite remote sensing imagery includes pre-satellite remote sensing imagery and post-satellite remote sensing imagery. Both images contain at least blue, green, red, and near-infrared bands, and their spatial resolution is better than 0.5 m. The pre-satellite imagery was acquired earlier than the post-satellite imagery. The monitoring range determination module 40 is also used to register the pre-satellite remote sensing imagery and the post-satellite remote sensing imagery to the same projection coordinate system, pixel size, and pixel grid. The crop plot range is determined based on the plot vector. The boundary of the crop plot range is shrunk by a preset shrinkage distance. The area between the boundary of the crop plot range and the shrunk boundary is determined as the plot edge buffer zone. The area enclosed by the shrunk boundary is determined as the candidate range inside the plot. The areas where roads, forest belts, and water bodies are located are excluded from the candidate range inside the plot to obtain the effective monitoring range.

[0174] In some embodiments, the ridge-direction spatial construction module 41 is specifically used to divide the target crop plot into multiple ridge-direction consistent planting zones based on the plot's internal boundary, seedling strip continuity, and local directional differences; for each ridge-direction consistent planting zone, the ridge spacing of the ridge-direction consistent planting zone is obtained; a local analysis window is set within the ridge-direction consistent planting zone; and a two-dimensional discrete Fourier transform is performed on the spatial distribution of vegetation combination quantity within the local analysis window within the spatial frequency range corresponding to the ridge spacing; the power spectrum is determined based on the obtained frequency domain transform result; the directional energy peak value within the spatial frequency range is determined based on the power spectrum; and the direction corresponding to the directional energy peak value is rotated by 90° to obtain the ridge angle of the ridge-direction consistent planting zone; parallel candidate ridge-direction center lines are generated based on the ridge angle and ridge spacing; the phase of the candidate ridge-direction center lines is adjusted sequentially within a phase range with a length equal to the ridge spacing; the average vegetation combination quantity of the pixels covered by the candidate ridge-direction center lines is determined for each phase; and the phase corresponding to the largest average vegetation combination quantity is determined as the ridge-direction center line phase of the ridge-direction consistent planting zone; and a set of ridge-direction center lines is generated based on the ridge angle, ridge-direction center line phase, and ridge spacing.

[0175] In the above-described implementation of ridge-oriented consistent planting partitions, the reference strip processing module 43 is also used to construct forward reference strips and backward reference strips within the ridge-oriented consistent planting partition to which the candidate object belongs.

[0176] In some embodiments, the object processing module 42 is specifically used to calculate the vegetation combination amount or vegetation index of each pixel based on multispectral satellite remote sensing imagery, obtain the vegetation characterization value of each pixel, and determine the pixels whose vegetation characterization value is lower than the corresponding local reference value as candidate pixels. The local reference value is determined based on the vegetation characterization value of pixels in the local area surrounding the corresponding pixel. Connected component aggregation is performed on the candidate pixels, and morphological opening and closing operations, adjacent region merging and boundary reconstruction are performed on the aggregated connected regions to obtain a planar object. Planar objects in the planar object whose angle between the main axis direction and the ridge direction is not greater than a preset angle threshold, whose ridge length is not less than a preset length threshold and whose lateral width is not less than a preset width threshold are determined as candidate objects.

[0177] In some embodiments, the lengths along the ridge of the forward reference band and the backward reference band are 0.8 to 1.5 times the length along the ridge of the candidate object, respectively. The reference band processing module 43 is further configured to space the forward reference band and the backward reference band from the candidate object by at least one cell along the ridge direction.

[0178] The validity determination module 44 is also used to determine the lateral distance between each pixel to be screened and the nearest ridge centerline covered by the candidate object, and to determine the pixel to be screened as a valid pixel when the lateral distance is not greater than the preset phase tolerance and the pixel to be screened meets other pixel screening conditions; to determine the percentage of valid pixels according to the ratio of the number of valid pixels in the corresponding reference band to the total number of pixels, and to determine the corresponding reference band as valid when the number of valid pixels reaches the preset minimum number of valid pixels and the percentage of valid pixels reaches the preset minimum percentage of valid pixels.

[0179] In some embodiments, the output module 46 is further configured to determine the difference in green light normalized vegetation index and soil-regulated vegetation index between the candidate object and the reference area. The reference area consists of effective pixels in the forward reference zone and the backward reference zone. The difference in green light normalized vegetation index is the difference between the green light normalized vegetation index of the reference area and the candidate object. The difference in soil-regulated vegetation index is the difference between the soil-regulated vegetation index of the reference area and the candidate object. The output module 46 is also configured to determine the edge interference degree, water-wet interference degree, shadow interference degree and bright spot interference degree of the candidate object. The edge interference degree is set to 1 when the distance from the centroid of the candidate object to the boundary of the original plot is not greater than the preset edge influence distance, and otherwise it is set to 0. The shadow interference degree and bright spot interference degree are determined based on the locally normalized visible light brightness. The output module 46 is configured to identify the candidate object with both the difference in green light normalized vegetation index and the difference in soil-regulated vegetation index as greater than 0, and the edge interference degree is 0, while the water-wet interference degree, shadow interference degree and bright spot interference degree are all less than their respective interference exclusion thresholds as suspected seedling gap objects.

[0180] In some embodiments, the ridge-direction spatial construction module 41 is also used to compare the seedling strip texture quality of the early satellite remote sensing images and the later satellite remote sensing images, and select the early satellite remote sensing images with higher seedling strip texture quality to determine the ridge direction and the phase of the ridge-direction centerline; the object processing module 42 is used to extract candidate objects using the later satellite remote sensing images; the reference strip processing module 43 is used to construct forward reference strips and backward reference strips using the later satellite remote sensing images; and the vegetation deficit determination module 45 is used to determine the vegetation deficit using the later satellite remote sensing images.

[0181] The output module 46 is also used to determine the changes in the normalized vegetation index (NDI) of the candidate objects, the forward reference zone, and the backward reference zone based on the pixels in the candidate objects and the effective pixels in each reference zone, and to sort the suspected missing vegetation zone objects from largest to smallest according to the difference between the changes in the two periods. Specifically, the change in each NDI is the difference between the NDI of the corresponding object or reference zone in the later satellite remote sensing image and the NDI in the earlier satellite remote sensing image. The difference between the changes in the two periods is the lower of the NDI changes in the forward reference zone and the backward reference zone, minus the NDI change of the candidate object.

[0182] The output module 46 is also used to output attribute information associated with the areal vector information. The attribute information includes object number, plot number, object area, effective number of pixels in the forward reference zone and the backward reference zone, vegetation deficit and sorting information.

[0183] Another embodiment of the present invention relates to an electronic device, such as Figure 5 As shown, it includes: at least one processor 501; and a memory 502 communicatively connected to the at least one processor 501; wherein the memory 502 stores instructions executable by the at least one processor 501, the instructions being executed by the at least one processor 501 to enable the at least one processor 501 to perform the satellite remote sensing monitoring method for crop seedling loss and crop belt breakage in the above embodiments.

[0184] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges. The bus connects various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein.

[0185] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0186] Another embodiment of the present invention relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0187] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0188] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0189] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0190] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A satellite remote sensing monitoring method for crop seedling gaps and crop strip breaks, characterized in that, include: Acquire multispectral satellite remote sensing images and plot vectors of the target crop plots, and determine the effective monitoring range based on the plot vectors; Based on the seedling strip texture in the multispectral satellite remote sensing image, determine the ridge direction and ridge centerline phase, obtain the ridge spacing, generate a set of ridge centerlines based on the ridge direction, the ridge centerline phase, and the ridge spacing, and establish local ridge coordinates including those along the ridge direction and the transverse direction, wherein the transverse direction is perpendicular to the ridge direction. Candidate objects are extracted from the effective monitoring range. The candidate objects are planar objects whose vegetation characterization values ​​are lower than the vegetation characterization values ​​of the local area surrounding the candidate object, which extend along the ridge direction and cross multiple ridge center lines. The ridge length and lateral coverage of each candidate object are determined. For each candidate object, based on the ridge length and lateral coverage of the candidate object, a forward reference band and a backward reference band are constructed in front of and behind the candidate object along the ridge direction, respectively, so that the forward reference band and the backward reference band have the same lateral projection range as the candidate object, and cover the ridge centerline covered by the candidate object in the lateral direction. Pixels located outside the effective monitoring range, located within other candidate objects, with invalid image data, not located at the geometric seedling strip position corresponding to any of the ridge center lines, or with a phase inconsistent with the ridge center line covered by the current candidate object are removed from the forward reference band and the backward reference band, respectively. The retained pixels are determined as valid pixels. When the number of valid pixels and the proportion of valid pixels in the corresponding reference band both reach the corresponding preset threshold, the corresponding reference band is determined to be valid. When both the forward reference zone and the backward reference zone are valid, the same vegetation characterization value calculation method is used. The vegetation characterization value of the candidate object is determined based on the pixels in the candidate object, and the vegetation characterization values ​​of the forward reference zone and the backward reference zone are determined based on the valid pixels in each reference zone. The lower of the vegetation characterization values ​​of the two reference zones is determined as the weaker reference value, and the difference between the weaker reference value and the vegetation characterization value of the candidate object is determined as the vegetation deficit. When the candidate object is located between the forward reference zone and the backward reference zone and the vegetation deficit is greater than 0, the candidate object is identified as a suspected seedling gap object, and the planar vector information representing the boundary and position of the suspected seedling gap object is output.

2. The method according to claim 1, characterized in that, The multispectral satellite remote sensing imagery includes early satellite remote sensing imagery and later satellite remote sensing imagery. Both types of imagery contain at least blue, green, red, and near-infrared bands, and have a spatial resolution better than 0.5 m. The early satellite remote sensing imagery was acquired earlier than the later satellite remote sensing imagery. The acquisition of multispectral satellite remote sensing images and plot vectors of the target crop plots, and the determination of the effective monitoring range based on the plot vectors, includes: The earlier and later satellite remote sensing images are registered to the same projection coordinate system, pixel size, and pixel grid. The crop plot range is determined based on the plot vector. The boundary of the crop plot range is shrunk by a preset shrinkage distance. The area between the boundary of the crop plot range and the shrunk boundary is determined as the plot edge buffer zone. The area enclosed by the shrunk boundary is determined as the candidate range inside the plot. The areas where roads, forest belts and water bodies are located are excluded from the candidate range inside the plot to obtain the effective monitoring range.

3. The method according to claim 1, characterized in that, The step of determining the ridge direction and ridge centerline phase based on the seedling strip texture in the multispectral satellite remote sensing image, obtaining the ridge spacing, and generating a set of ridge centerlines based on the ridge direction, the ridge centerline phase, and the ridge spacing includes: Based on the internal boundaries of the plot, the continuity of the seedling strip, and the differences in local direction, the target crop plot is divided into multiple planting zones with consistent ridge orientation; For each ridge-oriented planting zone, the ridge spacing of the ridge-oriented planting zone is obtained. A local analysis window is set within the ridge-oriented planting zone. A two-dimensional discrete Fourier transform is performed on the spatial distribution of vegetation combination within the local analysis window within the spatial frequency range corresponding to the ridge spacing. The power spectrum is determined based on the obtained frequency domain transform result. The directional energy peak value within the spatial frequency range is determined based on the power spectrum, and the direction corresponding to the directional energy peak value is rotated by 90° to obtain the ridge angle of the ridge-oriented consistent planting zone; Based on the ridge angle and the ridge spacing, parallel candidate ridge center lines are generated. Within a phase range with a length equal to the ridge spacing, the phase of the candidate ridge center lines is adjusted sequentially. The average vegetation combination amount of the pixels covered by the candidate ridge center lines under each phase is determined. The phase corresponding to the largest average vegetation combination amount is determined as the ridge center line phase of the ridge-oriented consistent planting zone. Based on the ridge angle, the phase of the ridge centerline, and the ridge spacing, the set of ridge centerlines is generated; the candidate object and its forward and backward reference bands are located within the same ridge-aligned planting zone.

4. The method according to claim 1, characterized in that, The step of extracting candidate objects from the effective monitoring range includes: The vegetation combination amount or vegetation index of each pixel is calculated based on the multispectral satellite remote sensing image to obtain the vegetation characterization value of each pixel. Pixels with vegetation characterization values ​​lower than the corresponding local reference value are identified as candidate pixels. The local reference value is determined based on the vegetation characterization value of pixels in the local area surrounding the corresponding pixel. Connected component aggregation is performed on the candidate pixels, and morphological opening and closing operations, adjacent region merging and boundary reconstruction are performed on the aggregated connected regions to obtain a planar object. The planar objects whose angle between the main axis direction and the ridge direction is not greater than a preset angle threshold, whose ridge length is not less than a preset length threshold, and whose lateral width is not less than a preset width threshold are identified as candidate objects.

5. The method according to claim 1, characterized in that, The lengths along the ridge of the forward reference band and the backward reference band are 0.8 to 1.5 times the length along the ridge of the candidate object, respectively, and the forward reference band and the backward reference band are spaced at least one pixel apart from the candidate object along the ridge direction. The lateral distance between each effective pixel and the nearest ridge centerline covered by the candidate object is not greater than a preset phase tolerance; The effective pixel ratio is the ratio of the number of effective pixels in the corresponding reference band to the total number of pixels in the corresponding reference band. The preset threshold corresponding to the number of effective pixels is the preset minimum number of effective pixels, and the preset threshold corresponding to the effective pixel ratio is the preset minimum effective pixel proportion.

6. The method according to claim 1, characterized in that, The step of identifying the candidate objects as suspected seedling gaps or interrupted seedlings includes: The differences in green normalized vegetation index (GNV) and soil-regulated vegetation index (SRI) between the candidate object and the reference area are determined. The reference area is composed of effective pixels in the forward reference zone and the backward reference zone. The GNV difference is the difference between the GNV of the reference area and the candidate object, and the SRI difference is the difference between the SRI of the reference area and the candidate object. The edge interference degree, water-wet interference degree, shadow interference degree, and bright spot interference degree of the candidate object are determined. The edge interference degree is set to 1 when the distance from the centroid of the candidate object to the original plot boundary represented by the plot vector is not greater than a preset edge influence distance, and otherwise it is set to 0. The shadow interference degree and the bright spot interference degree are determined based on the locally normalized visible light brightness. Candidates with differences in the normalized green vegetation index and the soil-regulated vegetation index both greater than 0, and with an edge interference level of 0, and with water-wet interference, shadow interference, and bright spot interference all less than their respective interference exclusion thresholds, are identified as suspected seedling gap / segmentation objects.

7. The method according to claim 2, characterized in that, Compare the seedling strip texture quality of the early-stage satellite remote sensing images and the later-stage satellite remote sensing images, and select the early-stage satellite remote sensing images with higher seedling strip texture quality to determine the ridge direction and the phase of the ridge direction centerline; the later-stage satellite remote sensing images are used to extract the candidate objects, construct the forward reference zone and the backward reference zone, and determine the vegetation deficit. Based on the pixels in the candidate objects and the effective pixels in each of the reference bands, the normalized vegetation index changes of the candidate objects, the forward reference band, and the backward reference band are determined respectively, and the suspected missing seedlings and broken belt objects are sorted from largest to smallest according to the difference between the two periods. Wherein, the change in normalized vegetation index of each of the above is the difference between the normalized vegetation index of the corresponding object or reference zone in the later satellite remote sensing image and the normalized vegetation index in the earlier satellite remote sensing image, and the difference between the two periods is the difference obtained by subtracting the change in normalized vegetation index of the candidate object from the lower of the change in normalized vegetation index of the forward reference zone and the backward reference zone. The output also includes attribute information associated with the areal vector information, including object number, plot number, object area, effective number of each of the forward reference zone and the backward reference zone, vegetation deficit, and sorting information.

8. A satellite remote sensing monitoring system for crop seedling gaps and crop belt breaks, characterized in that, include: The monitoring range determination module is used to acquire multispectral satellite remote sensing images and plot vectors of the target crop plots, and determine the effective monitoring range based on the plot vectors. The ridge direction spatial construction module is used to determine the ridge direction and ridge direction centerline phase based on the seedling strip texture in the multispectral satellite remote sensing image, obtain the ridge spacing, generate a set of ridge direction centerlines based on the ridge direction, the ridge direction centerline phase and the ridge spacing, and establish ridge direction local coordinates including ridge direction and transverse direction, wherein the transverse direction is perpendicular to the ridge direction. The object processing module is used to extract candidate objects from the effective monitoring range. The candidate objects are planar objects whose vegetation characterization values ​​are lower than the vegetation characterization values ​​of the local area surrounding the candidate object, extend along the ridge direction and cross multiple ridge direction center lines, and determine the ridge length and lateral coverage of each candidate object. The reference strip processing module is used to construct a forward reference strip and a backward reference strip in front of and behind the candidate object along the ridge direction, respectively, based on the ridge length and lateral coverage of the candidate object, so that the forward reference strip and the backward reference strip have the same lateral projection range as the candidate object, and cover the ridge centerline covered by the candidate object in the lateral direction. The validity determination module is used to remove pixels from the forward reference band and the backward reference band that are located outside the effective monitoring range, located within other candidate objects, have invalid image data, are not located at the geometric seedling strip position corresponding to any of the ridge center lines, or are inconsistent with the phase of the ridge center line covered by the current candidate object. The retained pixels are determined as valid pixels. When the number of valid pixels and the proportion of valid pixels in the corresponding reference band both reach the corresponding preset threshold, the corresponding reference band is determined to be valid. The vegetation deficit determination module is used to determine the vegetation deficit amount by using the same vegetation characterization value calculation method when both the forward reference zone and the backward reference zone are valid. It determines the vegetation characterization value of the candidate object based on the pixels in the candidate object, and determines the vegetation characterization value of the forward reference zone and the backward reference zone based on the valid pixels in each reference zone. The lower of the vegetation characterization values ​​of the two reference zones is determined as the weaker reference value, and the difference between the weaker reference value and the vegetation characterization value of the candidate object is determined as the vegetation deficit amount. The output module is used to determine the candidate object as a suspected seedling gap object when the candidate object is located between the forward reference zone and the backward reference zone and the vegetation deficit is greater than 0, and output the planar vector information characterizing the boundary and position of the suspected seedling gap object.

9. An electronic device, characterized in that, The device includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 7.

Images