A growth monitoring device and method for field crops

Abstract

The application relates to the technical field of crop monitoring, in particular to a growth monitoring device and method for field crops, which solve the technical problem of insufficient accuracy and reliability of current local pesticide application effect monitoring. The method comprises: uniformly dividing each frame of image in a crown layer image sequence of the field crops into multiple spatial grids, determining a pesticide application center grid, a farmland boundary grid, and a new leaf germination feature day of each spatial grid; constructing a time delay topology network based on the multiple spatial grids; taking a node corresponding to the pesticide application center grid as a connection starting point and a node corresponding to the farmland boundary grid as a connection end point, executing a maximum flow algorithm to determine a network connection weight index of the time delay topology network, and extracting a time delay blocking boundary; and determining a growth diagnosis result of the field crops according to a greenness increment difference between grids inside and outside the time delay blocking boundary and the network connection weight index.

Description

Technical Field

[0001] This application relates to the field of crop monitoring technology, specifically to a growth monitoring device and method for field crops. Background Technology

[0002] In the cultivation of field crops, visual monitoring systems can be installed above the fields to analyze crop greening patterns in aerial images taken over several consecutive days. This allows for the determination of the physical boundaries of pesticide diffusion and crop recovery patterns, thereby assessing the crop growth recovery effect after soil pesticide application. However, field crops are situated in an open natural meteorological environment and are significantly affected by it. For example, a large-scale rise in temperature or widespread rainfall in early spring can cause crops throughout the field to be stimulated by moisture and temperature, resulting in the simultaneous sprouting of new leaves within a few days. This phenomenon of synchronized greening across the entire field caused by global climatic factors can easily be confused with the growth recovery after localized application of liquid pesticides.

[0003] Existing crop growth monitoring systems, when distinguishing between localized seepage and global precipitation phenomena, mostly rely on manually selecting a central grid for pesticide application and a few preset edge points on an image, making a simple judgment by comparing the time difference in greening of these discrete pixels. However, real fields often contain interference areas such as waterlogged areas, bare soil with rocks, or dead and missing seedlings. If the selected edge sampling points fall into these interference areas, calculations based on the time difference of a few fixed coordinate points will result in errors due to missing data; simple time interval measurement cannot identify pseudo-seepage phenomena that have time lags but no significant difference in crop growth caused by the slow overall rise in early spring temperatures. Therefore, existing monitoring methods are easily affected by the heterogeneity of the field surface when distinguishing between localized pesticide application effects and global climate influences, leading to insufficient monitoring accuracy and reliability. Summary of the Invention

[0004] To address the current technical problems of insufficient accuracy and reliability in monitoring the effects of localized pesticide application, this application aims to provide a method for monitoring the growth of field crops. The specific technical solution adopted is as follows: The images in the canopy image sequence of field crops collected over several consecutive days were uniformly divided into multiple spatial grids, and the injection center grid, farmland boundary grid, and new leaf germination characteristic day of each spatial grid were determined.

[0005] A time-delay topology network is constructed based on the positional relationships between multiple spatial grids and the time difference between the characteristic days of new leaf germination in adjacent spatial grids. Nodes in the time-delay topology network correspond one-to-one with multiple spatial grids, and the edge weights in the network are determined based on the time difference.

[0006] Using the nodes corresponding to the injection center grid as the connection start point and the nodes corresponding to the farmland boundary grid as the connection end point, the maximum flow algorithm is executed to determine the network connectivity weight index of the time-delay topology network, and the time-delay blocking boundary composed of directed edges whose connectivity quotas have been exhausted is extracted.

[0007] The growth diagnosis results of field crops are determined based on the difference in greenness increment between the inner and outer grids of the time delay blocking boundary and the network connectivity weight index.

[0008] In one possible implementation, determining the leaf germination characteristic day for each spatial grid includes: determining the vegetation index of each spatial grid over several consecutive days based on each frame of the canopy image sequence; determining the date on which the vegetation index of each spatial grid first reaches or exceeds the dynamic greenness reference threshold as the leaf germination characteristic day of the spatial grid; the dynamic greenness reference threshold is a threshold obtained by adding a preset increment to the arithmetic mean of the vegetation indices of all spatial grids in the canopy image of the first day over several consecutive days; if there is a spatial grid whose vegetation index is less than the dynamic greenness reference threshold over several consecutive days, then the leaf germination characteristic day of the spatial grid is set as an overtime penalty value, and the value of the overtime penalty value is greater than the value of the total number of days over several consecutive days.

[0009] In one possible implementation, the vegetation index of each spatial grid is determined daily over several consecutive days based on each frame of the canopy image sequence. This includes: determining the red, green, and blue reflectance values ​​of each pixel in the daily canopy images of each spatial grid over several consecutive days; processing the red, green, and blue reflectance values ​​of each pixel based on the absorption and reflection of visible light by vegetation to obtain the vegetation index of each pixel; and calculating the arithmetic mean of the vegetation index values ​​of all pixels within each spatial grid to obtain the daily vegetation index of each grid.

[0010] In one possible implementation, the process of determining the edge weights in the time-delay topology network includes: determining the target grid pair; the target grid pair is a grid pair consisting of any two adjacent spatial grids from multiple spatial grids; determining the leaf germination characteristic days of the two spatial grids in the target grid pair; if the leaf germination characteristic day of a spatial grid in the target grid pair is a timeout penalty value, then the edge weight of the directed edge connecting the target grid pair is set to zero; if the leaf germination characteristic day of no spatial grid in the target grid pair is a timeout penalty value, then the absolute value of the difference between the leaf germination characteristic days of the two spatial grids is calculated to obtain the time difference, and the time difference is inversely mapped to obtain the edge weight of the directed edge connecting the target grid pair.

[0011] In one possible implementation, the nodes corresponding to the injection center grid are used as the connection starting point, and the nodes corresponding to the farmland boundary grid are used as the connection ending point. A maximum flow algorithm is executed to determine the network connectivity weight index of the time-delay topology network, and the time-delay blocking boundary formed by directed edges whose connectivity quotas are exhausted is extracted. This includes: setting virtual source and virtual sink nodes outside the time-delay topology network; connecting the virtual source node to the node corresponding to the injection center grid; and connecting the virtual sink node to the node corresponding to the farmland boundary grid. The edge weights of the connections between the virtual source node and the node corresponding to the injection center grid, and the edge weights of the connections between the virtual sink node and the node corresponding to the farmland boundary grid, are preset values, and these preset values ​​are greater than the sum of all edge weights in the time-delay topology network. The virtual source node is used as the search... Starting from the virtual sink as the search endpoint, the system continuously searches for valid connected paths in the time-delay topology network that have not yet reached the capacity quota limit. The starting point of a valid connected path is the node corresponding to the injection center grid, and the ending point is the node corresponding to the farmland boundary grid. After finding a valid connected path, the minimum available quota that can be accommodated on the currently found valid connected path is accumulated, and the remaining available quota of all directed edges on the currently found valid connected path is deducted by the same amount. This step is iterated until no new valid connected paths can be found, and the accumulated total value is used as the network connectivity weight index. The set of directed edges in the time-delay topology network with an initial connectivity quota greater than zero and a current remaining available quota of zero is extracted as the time-delay blocking boundary.

[0012] In one possible implementation, the process of determining the difference in greenness increment between the inner and outer grids of the time-delayed blocking boundary includes: determining the inner and outer grids based on the spatial distance between the two spatial grids connected by each directed edge in the time-delayed blocking boundary and the injection center grid; the spatial distance between the inner grid and the injection center grid is less than the spatial distance between the outer grid and the injection center grid; determining the termination observation dates for the inner and outer grids; the termination observation date is after the new leaf germination characteristic day and at a preset interval of several days from the new leaf germination characteristic day; the preset number of days is less than or equal to the number of consecutive days corresponding to multiple days. If the observation termination date exceeds the total number of consecutive days, the last day of the consecutive days is taken as the observation termination date. The greenness increment of the inner grid is determined based on the difference in vegetation index between the observation termination date and the new leaf germination characteristic day. The greenness increment of the outer grid is determined based on the difference in vegetation index between the observation termination date and the new leaf germination characteristic day. The difference between the greenness increment of the inner grid and the greenness increment of the outer grid at each time delay barrier boundary is calculated, and the arithmetic mean of the greenness increment difference at each time delay barrier boundary is determined as the greenness increment difference.

[0013] In one possible implementation, the growth diagnosis result of field crops is determined based on the difference in greenness increment between the grids inside and outside the time-delay blocking boundary and the network connectivity weight index. This includes: if the network connectivity weight index is greater than the greening synchronicity extreme threshold and the difference in greenness increment is less than the greenness increment difference threshold, then a first diagnosis result is output; the first diagnosis result indicates that the growth change is caused by global climate factors; the greening synchronicity extreme threshold is determined based on an ideal zero-time-difference network model with the same grid size as the time-delay topology network; if the network connectivity weight index is less than or equal to the greening synchronicity extreme threshold and the difference in greenness increment is greater than or equal to the greenness increment difference threshold, then a second diagnosis result is output; the second diagnosis result indicates that the growth change is a growth recovery caused by local pesticide seepage; if none of the above conditions are met, then a third diagnosis result is generated, which indicates that manual verification is needed to determine the cause of the growth change.

[0014] In one possible implementation, the process of determining the extreme threshold for greening synchronicity includes: constructing an ideal zero-time-difference network model; the ideal zero-time-difference network model is consistent with the spatial grid division of the field, and the characteristic day of new leaf germination in each spatial grid is the same day, the time difference between adjacent grids is zero, and the weight of all edges is unified to the theoretical limit value; processing the ideal zero-time-difference network model based on the maximum flow algorithm to obtain the ideal network connectivity weight index; determining the extreme threshold for greening synchronicity based on the ideal network connectivity weight index and the network smoothness tolerance coefficient; the network smoothness tolerance coefficient is used to characterize the ideal smoothness ratio required to identify large-scale synchronous greening.

[0015] In one possible implementation, the process of determining the greenness increment difference threshold includes: obtaining a preset increment, and multiplying the preset increment by the growth drop manifestation coefficient as the greenness increment difference threshold; wherein, the growth drop manifestation coefficient is used to characterize the proportion of the local efficacy decline drop relative to the preset increment.

[0016] This application also provides a field crop growth monitoring device, comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the field crop growth monitoring method as described above.

[0017] This application has the following beneficial effects: By dividing a continuous multi-day canopy image sequence into a spatial grid and extracting the new leaf germination feature days of each grid, the high-dimensional continuous image time series data is reduced to a discrete spatial-temporal feature matrix; by constructing a time-delay topology network, the germination time difference between adjacent grids is transformed into the connectivity weights of the network edges, so that the lag information in the time dimension is objectively mapped into the channel-limited quota in the topology network; then, the maximum flow algorithm is used to calculate the connectivity weight index of the entire network from the injection center to the farmland boundary, and the connectivity quotas that have been exhausted are extracted. Directed edges serve as time-delayed blocking boundaries, automatically avoiding heterogeneous noise areas in the field, stably summarizing the synchronous scale of greening across the entire field, and objectively locating the geographical outer edge where pesticide penetration stagnates. Finally, by combining the difference in greenness increment inside and outside the time-delayed blocking boundary with network connectivity weight indicators for cross-analysis, the synchronous recovery of high-flux and uniform precipitation with low-flux and significant internal and external greenness cliffs with local seepage effects can be distinguished. This enables accurate differentiation between global climatic factors and local pesticide seepage-induced changes in growth, significantly improving the accuracy and reliability of field growth monitoring. Attached Figure Description

[0018] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a method for monitoring the growth of field crops provided in one embodiment of this application; Figure 2 This is a schematic diagram of the structure of a field crop growth monitoring device provided in one embodiment of this application. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method for monitoring the growth of field crops according to this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0022] The following describes in detail, with reference to the accompanying drawings, a specific scheme for a method of monitoring the growth of field crops provided in this application.

[0023] Please see Figure 1 It illustrates a flowchart of a method for monitoring the growth of field crops according to an embodiment of this application, as shown below. Figure 1 As shown, the method includes the following steps: Step 101: Divide each frame of the canopy image sequence of field crops collected over several consecutive days into multiple spatial grids.

[0024] In some embodiments, the aforementioned consecutive days refer to a series of consecutive dates starting from the date of the first injection. The method for determining the date of the first injection includes: inserting a physicochemical probe into the rhizosphere soil of a field crop within the injection area. The physicochemical probe is used to monitor the concentration of a specific applied pesticide or the targeted conductivity. The physicochemical probe is controlled to perform periodic sampling to acquire continuously output index data, which is related to the injected pesticide (e.g., pesticide concentration). The periodic change in the index is obtained by calculating the difference between the index value of the current sampling period and the index value of the previous sampling period. This change is compared with a preset change threshold (e.g., to filter out white noise from natural fluctuations, the change threshold can be preset to 15% of the relative change amplitude). If the comparison result shows that the change is greater than the change threshold, it is determined that artificial pesticide injection has occurred in the soil, and the specific date on which the index value shows a significant change is extracted as the date of the first injection.

[0025] After establishing the date of the first pesticide application, this date is designated as day 1 of the current monitoring cycle. A continuous image acquisition command is sent to a visual camera fixed above the crop canopy. The camera continuously captures color images of the crop canopy for a preset number of observation days, following a fixed midday sunlight period (e.g., 12:00 to 13:00 daily to ensure consistent light angle). The preset number of observation days covers a complete physiological evolution cycle of the crop from absorbing water or pesticides to the emergence of new leaves (e.g., 14 days). After completing several consecutive days of shooting, these multiple independent frames are arranged chronologically to generate a sequence of crop canopy images.

[0026] Following this, for each frame in the canopy image sequence, a proportionally uniform spatial division is performed, dividing the full-frame image into an equal number of rectangular spatial grids. The total number of grids is recorded (e.g., dividing the image horizontally into 100 parts and vertically into 100 parts results in a total of 10,000 grids). Each grid in the two-dimensional plane is assigned a unique identifier, i, to determine its coordinate order, where i is an incrementing number starting from 1 and covering all grids. It should be noted that the grid side lengths should correspond to the actual ground dimensions. This is to ensure that each grid contains at least one crop.

[0027] Step 102: Determine the injection center grid, farmland boundary grid, and new leaf germination characteristic day for each spatial grid.

[0028] The process of determining the injection center grid is as follows: The three-dimensional spatial coordinates of the physicochemical probe in the actual field environment are read. Based on the visual camera's intrinsic and extrinsic parameter calibration model and coordinate transformation matrix algorithm, these three-dimensional spatial coordinates are projected and mapped onto the segmented two-dimensional image pixel coordinate system, thus determining the mapping point coordinates of the physicochemical probe in the two-dimensional image pixel coordinate system. All previously divided spatial grids are traversed, and the grid containing these mapping point coordinates is determined as the injection center grid; the injection center grid represents the geographic zero point of drug diffusion in the image.

[0029] The process of determining the farmland boundary grid is as follows: by searching by coordinate row and column numbers, all grids located at the outermost four edges of the entire two-dimensional image are extracted as the farmland boundary grid.

[0030] The process of determining the characteristic day of new leaf germination includes: determining the vegetation index of each spatial grid over several consecutive days based on each frame of the canopy image sequence; and determining the date on which the vegetation index of each spatial grid first reaches or exceeds the dynamic greenness reference threshold as the characteristic day of new leaf germination for that spatial grid. The dynamic greenness reference threshold is obtained by adding a preset increment to the arithmetic mean of the vegetation indices of all spatial grids in the canopy images of the first day over several consecutive days. If a spatial grid has a vegetation index that is less than the dynamic greenness reference threshold for several consecutive days, then the characteristic day of new leaf germination for that spatial grid is set as an overdue penalty value, and the value of the overdue penalty value is greater than the total number of days over several consecutive days.

[0031] Optionally, the above vegetation index is determined as follows: The red, green, and blue reflectance values ​​of each pixel in the daily canopy images of each spatial grid over several consecutive days are determined; based on the absorption and reflection of visible light by vegetation, the red, green, and blue reflectance values ​​of each pixel are processed to obtain the vegetation index for each pixel; the arithmetic mean of the vegetation index values ​​of all pixels within each spatial grid is calculated to obtain the daily vegetation index for each grid. It should be noted that, to ensure consistency in the calculation of the vegetation index, the reflectance values ​​of each channel are normalized to the [0,1] interval.

[0032] Specifically, for each grid in the first day's image, the reflectance values ​​of the red, green, and blue channels for all pixels within it are extracted. Utilizing the unique absorption and reflection patterns of vegetation in the visible light band, each pixel is substituted into the vegetation index formula for calculation. The optional vegetation index formula is: ,in, This represents the reflectance value of the green channel. This represents the reflectance value of the red channel. This represents the reflectance value of the blue channel. Based on the vegetation index formula, the vegetation index values ​​of all pixels within the canopy image of the grid on the first day are calculated, and their arithmetic mean is calculated to generate the average vegetation index of the grid on the first day. It should be noted that the above vegetation index formula is a technique well-known to those skilled in the art. The reason this application chooses this vegetation index is that it is only scaled as a whole under proportional changes in illumination, without altering the relative order between the grids. That is, although the ambient illumination changes, the obtained vegetation index will change synchronously and proportionally, and will not change the detection order on the timeline of the event that the vegetation index first reaches or exceeds a certain threshold in subsequent steps.

[0033] The arithmetic mean of the vegetation index of all grids on the first day is calculated to obtain the average background color of the entire map, reflecting the true background color of the current farmland vegetation. A pre-set increment is added to this average background color to generate a dynamic greenness reference threshold adapted to the specific farmland vegetation sparseness and soil background. This pre-set increment characterizes the minimum physiological range of new crop leaves that can be stably identified by the naked eye and machine (e.g., a value of 0.05). This dynamic greenness reference threshold is established as the criterion for determining whether substantial new leaf germination has occurred within a certain grid throughout the entire monitoring period. It should be noted that the method for obtaining the pre-set increment is based on the 3σ criterion. In the specific implementation of this embodiment, the pre-set increment is set to 3 times σ (i.e., the 3σ criterion) using the time-series noise standard deviation σ of the image sensor under static uniform targets to filter out false positive germination detections caused by sensor thermal noise, quantization noise, etc. The obtained value of 0.05 is a reasonable calibration value calculated based on this method under typical field conditions.

[0034] Extract the vegetation index values ​​for each day of the i-th grid from day 1 to the last day of the observation period. Concatenate these values ​​in chronological order to construct a multi-day vegetation index sequence for the i-th grid. After constructing the multi-day vegetation index sequence for each grid, iterate through the sequence along the time axis starting from day 1, performing a value comparison. During the iteration, compare the vegetation index value for each day with the dynamic greenness reference threshold. If the comparison results show that the vegetation index value for that day is greater than or equal to the dynamic greenness reference threshold, it indicates that the crop growth within that grid has crossed the germination qualification line. At this point, the search for the remaining days of the current sequence is immediately terminated, and the specific time number corresponding to the first occurrence of this condition is extracted. This extracted time number is recorded and assigned as the new leaf germination characteristic day of the i-th grid. .

[0035] If, after completing the traversal and retrieval of the entire observation period, all vegetation index values ​​in the sequence are strictly less than the dynamic greenness reference threshold, then it is determined that no new leaf growth has occurred in the i-th grid during the entire growth observation period (usually corresponding to waterlogged puddles, hard rocks, or completely dead plants within the grid area). In this case, the new leaf germination characteristic date of the i-th grid is recorded. The value is forcibly assigned as a timeout penalty. If the value of the timeout penalty is greater than the upper limit of the observation period, for example, it is uniformly set to the total number of days in the observation period plus 1 (when the total number of observation days is 14, the timeout penalty is 15), which means that the time taken for the region to respond to changes in growth has exceeded the limit of the observation window.

[0036] Based on the above process, the characteristic days of new leaf germination for each grid in the entire image are determined one by one, and the values ​​of all the calculated characteristic days of new leaf germination are obtained. All the values ​​of the characteristic days of new leaf germination are then rearranged according to the relative coordinates of the corresponding grid's rows and columns in the original 2D image, and assembled to generate a comprehensive image. Figure 2 The daily matrix of new leaf germination characteristics in 3D space.

[0037] In this embodiment, the canopy image sequence spanning multiple consecutive days is first divided into spatial grids, and the characteristic days of new leaf germination in each grid are extracted, transforming the originally massive high-dimensional image time-series data into a structured spatial-temporal feature matrix. This not only significantly reduces the data dimensionality for subsequent analysis but also isolates unresponsive noise areas such as dead seedlings and waterlogged areas at the source through a timeout penalty mechanism, preventing these areas from interfering with subsequent algorithms.

[0038] Step 103: Based on the positional relationships between multiple spatial grids and the time difference of the characteristic days of new leaf germination between adjacent spatial grids, a time-delay topology network is constructed. Nodes in the time-delay topology network correspond one-to-one with multiple spatial grids, and the edge weights in the time-delay topology network are determined based on the time difference.

[0039] Optionally, the process of determining the edge weights in the time-delay topology network includes: determining the target grid pair; the target grid pair is a grid pair composed of any two adjacent spatial grids from multiple spatial grids; determining the leaf germination characteristic days of the two spatial grids in the target grid pair; if the leaf germination characteristic day of a spatial grid in the target grid pair is a timeout penalty value, then the edge weight of the directed edge connecting the target grid pair is set to zero; if the leaf germination characteristic day of no spatial grid in the target grid pair is a timeout penalty value, then the absolute value of the difference between the leaf germination characteristic days of the two spatial grids is calculated to obtain the time difference, and the time difference is inversely mapped to obtain the edge weight of the directed edge connecting the target grid pair.

[0040] Specifically, the entire spatial grid covered by the new leaf germination characteristic day matrix is ​​traversed one by one. During the traversal, taking the current grid numbered as i as the base point, at most four directly adjacent grids are found in space according to the four orthogonal directions of up, down, left, and right. These are collectively referred to as the j-th adjacent grid.

[0041] Find the adjacent grid pair (i.e., grid) With grid After that, retrieve the record of the first term from the matrix respectively. New leaf germination characteristics of each grid day With the first New leaf germination characteristics of each grid day Check the extracted samples separately. and If either of the two values ​​is determined to be equal to the preset timeout penalty value in the previous step, it indicates that at least one of these two adjacent grids is a dead seedling or waterlogged area where the growth response has completely failed. In this case, the edge weight between the adjacent grid pair is directly forced to be assigned a numerical value. This zero-weight assignment operation cuts off the algorithmic connectivity path of the uncropped response region in the graph network.

[0042] If the judgment and If neither of these two day values ​​is equal to the timeout penalty value, then the reciprocal of their absolute values ​​is used to derive a weight parameter representing the smoothness of the network path. The calculation formula is as follows: in, Indicates the adjacent first The grid and the first Edge weights between grid cells; Indicates the first The characteristics of new leaf germination in each grid day; Indicates the first The characteristic days of new leaf germination for each grid are given, where e is a natural constant. The above calculation formula uses an exponential function mapping with the natural constant as the base, which can invert the time lag into the positive connectivity quota in the graph theory network: when stimulated by global climate synchronization, the time difference between adjacent grids approaches zero, and the calculated edge weight value is large, indicating that the conduction connectivity weight of the adjacent region is extremely high; when the liquid agent causes the number of greening days to differ by several days between two places due to soil barriers, the calculated edge weight value is small, indicating that the data connectivity weight of the region is severely suppressed.

[0043] After calculating the edge weights, the results are assigned to the two directed edges representing bidirectional connectivity between grid i and grid j, serving as their respective available connectivity quotas. This process of iterative calculation and quota assignment is performed on all grids in the entire graph. After traversing the entire graph, these directed edges carrying different quota values ​​are used to construct a time-delay topology network encompassing the geographic and temporal interactions across the entire region.

[0044] This embodiment constructs a time-delay topology network based on the time difference between the characteristic days of new leaf germination in adjacent grids, mapping the time-dimensional hysteresis information to the connectivity weights of network edges. In this way, if adjacent areas turn green synchronously due to rainfall, the time difference approaches zero, and the corresponding edge weights will be large; conversely, if the greening time is prolonged due to pesticide seepage, the time difference increases, and the edge weights will significantly decrease. This mapping method cleverly distinguishes between two physical processes at the topological level: global synchronous stimulation and local seepage delay.

[0045] Step 104: Using the node corresponding to the injection center grid as the connection start point and the node corresponding to the farmland boundary grid as the connection end point, execute the maximum flow algorithm to determine the network connectivity weight index of the time-delay topology network, and extract the time-delay blocking boundary formed by the directed edges whose connectivity quota has been exhausted.

[0046] Optionally, virtual source and virtual sink nodes are set outside the time-delay topology network. The virtual source node is connected to the node corresponding to the injection center grid, and the virtual sink node is connected to the node corresponding to the farmland boundary grid. The edge weights of the connections between the virtual source node and the node corresponding to the injection center grid, and the edge weights of the connections between the virtual sink node and the node corresponding to the farmland boundary grid, are preset values, which are greater than the sum of all edge weights in the time-delay topology network. Using the virtual source node as the search starting point and the virtual sink node as the search ending point, a valid connected path that has not yet reached the capacity quota limit is continuously searched in the time-delay topology network. The starting point of the path is the node corresponding to the injection center grid, and the ending point of the effective connected path is the node corresponding to the farmland boundary grid. After each effective connected path is found, the minimum available quota that can be accommodated on the currently found effective connected path is accumulated, and the remaining available quota of all directed edges on the currently found effective connected path is deducted by the same amount. This step is iterated until no new effective connected path can be found, and the accumulated total value is used as the network connectivity weight index. The set of directed edges in the time-delay topology network with an initial connectivity quota greater than zero and a current remaining available quota of zero is extracted as the time-delay blocking boundary.

[0047] Specifically, outside the currently constructed time-delay topology network, a completely independent virtual source and an equally independent virtual sink are generated. A directed edge is established starting from the virtual source and ending at the injection center grid. The initial connectivity weight of this edge is forcibly assigned a preset maximum computational upper limit value (this maximum computational upper limit is set to a value greater than the sum of the weights of all edges in the time-delay topology network, the purpose of which is to ensure that the weight supply at the entrance is absolutely sufficient, and to avoid insufficient weight affecting the subsequent algorithm operation). Subsequently, for each boundary grid in the outermost farmland boundary grid, an independent directed edge is established starting from that boundary grid and pointing to the virtual sink. The initial connectivity weight of these edges is also forcibly assigned the aforementioned maximum computational upper limit value.

[0048] After completing the virtual node mounting and maximum weight assignment operations, the maximum flow search algorithm is invoked within the closed-loop network architecture. The search is limited to the virtual source node as the starting point and the virtual sink node as the ending point, continuously searching for valid connected paths that have not yet reached their connectivity quota limits throughout the entire latency-dependent topology network.

[0049] Each time a valid connected path is found, the minimum available quota that the currently found valid connected path can accommodate is accumulated (i.e., determined by the minimum value of the current remaining available quota among all edges on that path), and the remaining available quota of all directed edges on that valid connected path is deducted by the same amount. This step is performed iteratively until no new valid connected paths can be found, and the accumulated total value is used as the network connectivity weight index. The set of directed edges in the delay topology network with an initial connectivity quota greater than zero and a current remaining available quota of zero is extracted as the delay blocking boundary.

[0050] Once the search is determined to have reached saturation, the algorithm iteration stops, and the final output is the accumulated total value scalar, which is defined as the network connectivity weight index. The larger the specific value of this index, the more abundant the quota for connectivity in the network, indicating a higher degree of smoothness in which crops in the entire field turn green synchronously due to globally indiscriminate stimuli; conversely, if the value is extremely small, it indicates that there are serious transmission faults in the network.

[0051] While outputting the network connectivity weight index, the quota consumption of each directed edge initially allocated between grids is traversed and compared. All directed edges that simultaneously satisfy the condition that the initial allocated quota is strictly greater than zero and that the remaining available quota is completely deducted to zero after multiple path accumulation consumptions in the algorithm are identified. These quota-depleted directed edges form a truncated section, objectively anchoring in real-world geographical attributes the geographical segmentation edge where, as the drug's effect penetrates outwards, the concentration continuously decreases, the time consumption drastically increases, and the time difference weight drops to its limit.

[0052] This embodiment uses the injection center grid as the starting point and the farmland boundary grid as the ending point to run the maximum flow algorithm. While calculating the network connectivity weight index, it uses the algorithm to automatically identify the characteristics of the dead seedling zero-area region and stably summarize the greening response scale of the whole field. At the same time, it extracts the directed edges where the connectivity quota is exhausted as the time delay blocking boundary, so that the physical stagnation edge of the pesticide penetration can be objectively located without the need for manual point selection.

[0053] Step 105: Determine the growth diagnosis results of field crops based on the difference in greenness increment between the inner and outer grids of the time delay blocking boundary and the network connectivity weight index.

[0054] Optionally, the process for determining the difference in greenness increment between the inner and outer grids of the time-delayed blocking boundary includes: determining the inner and outer grids based on the spatial distance between the two spatial grids connected by each directed edge in the time-delayed blocking boundary and the injection center grid; the spatial distance between the inner grid and the injection center grid is less than the spatial distance between the outer grid and the injection center grid; determining the termination observation dates for the inner and outer grids; the termination observation date is after the new leaf germination characteristic day and at a preset interval of several days from the new leaf germination characteristic day; the preset number of days is less than or equal to the number of days corresponding to multiple consecutive days; if the termination date is... If the observation termination date exceeds the total number of consecutive days, the last day of the consecutive days is taken as the termination date. The greenness increment of the inner grid is determined based on the difference in vegetation index between the termination date and the new leaf germination characteristic date. Similarly, the greenness increment of the outer grid is determined based on the difference in vegetation index between the termination date and the new leaf germination characteristic date. The difference between the greenness increments of the inner and outer grids of each time-delay barrier boundary is calculated, and the arithmetic mean of these differences is taken as the greenness increment difference. It should be noted that if the new leaf germination characteristic date of either of the two grids connected by a time-delay barrier boundary is an overdue penalty value, that boundary is removed and not included in the calculation of the greenness increment difference.

[0055] Specifically, for each independent blocking edge in the time-delay blocking boundary, a loop traversal operation is performed one by one. When entering a single traversal loop, the two specific physical grids directly connected by the blocking edge are first located by using the start and end indices of the blocking edge.

[0056] After locating these two grids, calculate the Euclidean linear spatial distance from the two-dimensional center coordinates of each grid to the previously determined injection center grid coordinates. After calculating these two linear spatial distances, define the grid with the smaller spatial distance value (i.e., the grid that is physically closer to the injection source) as the inner grid; and define the grid with the larger spatial distance value as the outer grid.

[0057] After successfully defining the inner and outer grids, to ensure fairness in the growth comparison, independent, sequential observation windows based on their respective germination dates were used for both the inner and outer grids. The specific truncation and calculation process is as follows: For the inner grid: retrieve the new leaf germination characteristic date established by the inner grid itself. Add a preset number of days to this new leaf germination characteristic date (this number of days represents the fixed extension period of development after the crop crosses the pass line, for example, it can be set to 3 days) to obtain the inner extension target number of days. After obtaining the inner extension target number of days, compare the inner extension target number of days with the total number of observation days of the system, and take the smaller value as the termination observation date of the inner grid (to prevent exceeding the observation period range).

[0058] After determining the termination date, extract the vegetation index value for that day from the multi-day vegetation index sequence corresponding to the inner grid, and simultaneously extract the vegetation index value for the day of new leaf germination characteristic. Subtract the value for the day of new leaf germination characteristic from the value for the termination date to obtain the greenness increment of the inner grid within the follow-up period (if the new leaf germination characteristic day of the grid in the previous step is an overtime penalty value, then the greenness increment of the inner grid is forcibly assigned to the value 0).

[0059] For the outer grid: the same processing method as the inner grid is adopted, retrieving the new leaf germination characteristic date of the outer grid itself. The outer grid's observation termination date is determined by adding a preset number of days to this new leaf germination characteristic date and within the limit of the total number of observation days in the system. The vegetation index values ​​for these two corresponding dates are extracted from the multi-day vegetation index sequence of the outer grid, and the difference is calculated to obtain the greenness increment of the outer grid (similarly, if the new leaf germination characteristic date of the grid is an overtime penalty value, its greenness increment is assigned to 0).

[0060] After obtaining the greenness increments of the inner and outer grids, the difference in greenness increment at the time-delayed blocking boundary is calculated using this set of paired data. : in, Indicates the current number The time delay blocks the difference in green density increment inside and outside the boundary; This represents the calculated increase in greenness of the inner grid. This represents the calculated increase in greenness of the outer grid. The larger the difference calculated by this formula, the more objectively it proves that the soil pesticide concentration outside the blocking edge has significantly declined and can no longer support normal crop growth.

[0061] Calculate the current delay blocking boundary. Then, the loop continues until all time-delay blocking boundaries in the set have been traversed. After traversal, the differences obtained from all time-delay blocking boundaries are summed, and the arithmetic mean is calculated using a formula to obtain the difference in greenness increment. : in, Indicates the difference in greenness increment; This indicates the total number of network edges contained in the delay blocking boundary; Indicates the first The difference in greenness increment along the strip, the summation subscript of this formula. Based on this formula, the difference in greenness increment, representing the global micro-level difference in drug efficacy, was calculated, quantifying the concentration cliff phenomenon at the drug efficacy boundary.

[0062] Optionally, in determining the growth diagnosis results of field crops, if the network connectivity weight index is greater than the greening synchronicity extreme threshold and the difference in greenness increment is less than the greenness increment difference threshold, then the first diagnosis result is output; the first diagnosis result is used to indicate that the growth change is caused by global climate factors; the greening synchronicity extreme threshold is determined based on an ideal zero-time-difference network model with the same grid size as the time-delay topology network; if the network connectivity weight index is less than or equal to the greening synchronicity extreme threshold and the difference in greenness increment is greater than or equal to the greenness increment difference threshold, then the second diagnosis result is output; the second diagnosis result is used to indicate that the growth change is caused by local pesticide seepage and growth recovery; if none of the above conditions are met, then the third diagnosis result is generated, and the third diagnosis result is used to indicate that manual verification is needed to determine the cause of the growth change.

[0063] The process of determining the extreme threshold for greening synchronicity includes: constructing an ideal zero-time-difference network model; the ideal zero-time-difference network model is consistent with the spatial grid division of the field, and the characteristic day of new leaf germination in each spatial grid is the same day, the time difference between adjacent grids is zero, and the weight of all edges is unified to the theoretical limit value; processing the ideal zero-time-difference network model based on the maximum flow algorithm to obtain the ideal network connectivity weight index; determining the extreme threshold for greening synchronicity based on the ideal network connectivity weight index and the network smoothness tolerance coefficient; the network smoothness tolerance coefficient is used to characterize the ideal smoothness ratio required to identify large-scale synchronous greening.

[0064] Specifically, a virtual ideal zero-time-difference network model is dynamically generated, perfectly matching the scale of the current field grid matrix. In this virtual ideal zero-time-difference network model, it is assumed that all grids in the entire map synchronously sprout new leaves on the same day (i.e., a perfect rainfall climate with an absolute time difference of 0). Based on this assumption, the capacity quota of all directed edges between grids within this virtual ideal zero-time-difference network model is uniformly and forcibly assigned to the ideal zero-time-difference full-load quota (i.e., the constant limit value in the formula when the time difference is 0). ).

[0065] After assigning the virtual full-load quota, the maximum flow algorithm, identical to that in step 104, is run in the virtual ideal zero-time-difference network model, still initiating the search from the same central location until saturation. After the algorithm finishes running, the ideal network connectivity weight index under the theoretical extreme condition is obtained. This ideal network connectivity weight index is then multiplied by a pre-set network throughput tolerance coefficient (for example, a tolerance coefficient of 0.7 means that only 70% of the ideal throughput is needed to be considered large-scale synchronization). This product is then used as the extreme threshold for greening synchronization applicable to the current grid partitioning density. This extreme threshold becomes the objective criterion for determining whether the current measured network connectivity weight index belongs to high-flow concurrent throughput.

[0066] The process of determining the threshold for the difference in greenness increment includes: obtaining a preset increment, and multiplying the preset increment by the growth drop manifestation coefficient as the threshold for the difference in greenness increment; wherein, the growth drop manifestation coefficient is used to characterize the proportion of the local efficacy decline drop relative to the preset increment.

[0067] Specifically, a preset increment is retrieved and multiplied by a pre-defined growth drop manifestation coefficient (this coefficient represents the proportion of a precipitous drop that can be clearly identified by the naked eye or instruments to the passing line; for example, the coefficient is set to 0.2). This product is then used as the greenness increment difference threshold adapted to the current specific farmland vegetation background color. After generating this greening synchronicity extreme value threshold and the greenness increment difference threshold, these two dynamically generated scales are saved as the core judgment criteria for subsequent logical comparison stages.

[0068] The system receives network connectivity weight indicators, differences in greenness increments, extreme thresholds for greening synchronicity, and thresholds for greenness increment differences. After obtaining these four parameters, a two-dimensional cross-logic comparison operation is performed. First, a numerical comparison is performed on the macroscopic network characteristics, i.e., comparing the measured network connectivity weight indicators with the extreme thresholds for greening synchronicity. Based on this, a numerical comparison is performed on the microscopic difference characteristics, i.e., comparing the measured differences in greenness increments with the thresholds for greenness increment differences. Based on the combination of these two sets of numerical comparison results, the system proceeds to determine the following: Scenario 1: Crop growth recovery triggered by global meteorological factors. If the comparison results show that the network connectivity weight index is greater than the extreme threshold for greening synchronicity and simultaneously satisfies the condition that the difference in greenness increment is less than the threshold for greenness increment difference, this indicates that the farmland grid is almost completely connected across the time axis (characterized by high throughput), and the local space does not exhibit a stepped decline in greenness within the cutoff time window (characterized by low drop). This combination of data characteristics highly conforms to the pattern of indiscriminate coverage of farmland by large-area meteorological precipitation or early spring warming. When this criterion is met, a classification status report is generated that clearly indicates that the crop growth change across the entire map is caused by global meteorological factors.

[0069] Scenario 2: Growth recovery triggered by localized pesticide seepage. If the comparison results show that the network connectivity weight index is less than or equal to the extreme threshold of greening synchronicity and simultaneously satisfies the condition that the difference in greenness increment is greater than or equal to the threshold of greenness increment difference, this indicates, in terms of data characterization, that the farmland topology network has a regional time delay fault (manifested as a low flux characteristic) that severely restricts the total connectivity weight, and that significant greenness change cliffs (manifested as high drop characteristics) are present on both sides of this fault. This combination of data characteristics highly conforms to the inherent law of increasing seepage time and gradient decline of concentration of locally injected liquid media when overcoming soil pore resistance to diffuse outward. When this judgment condition is met, a classification status report is generated that clearly indicates that localized pesticide seepage has occurred and triggered a sequential recovery of growth.

[0070] Scenario 3: Feature conflict caused by unknown external interference. When this condition is met, a report indicating complex noise interference in a clearly defined area is generated, triggering an alarm to the user terminal and prompting for manual on-site verification. In other words, if the network connectivity weight index is less than or equal to the extreme threshold for greening synchronicity and the difference in greening increment is less than the threshold for greening increment difference, or if the network connectivity weight index is greater than the extreme threshold for greening synchronicity and the difference in greening increment is greater than or equal to the threshold for greening increment difference, it indicates a contradictory state where the grid exhibits macroscopic transmission obstruction but microscopic no drop, or macroscopic unobstructed flow but local abnormal drops, indicating a feature conflict caused by unknown external interference. It should be noted that if the network connectivity weight index is less than or equal to the extreme threshold for greening synchronicity and the difference in greening increment is less than the threshold for greening increment difference (i.e., low throughput, low drop), it may indicate that the farmland has suffered from large-scale uniform stress such as drought or disease, leading to stagnation of crop growth throughout the field. If the network connectivity weight index exceeds the extreme threshold for greening synchronicity, and the difference in greening increment is greater than or equal to the threshold for greening increment difference (i.e., high throughput, high drop), it may indicate abnormal data collection or interference from non-natural factors such as local replanting. Both of these conflict characteristics are included in the third diagnostic result and trigger the manual review mechanism.

[0071] This application cross-compares the aforementioned network connectivity weight index with the difference in green intensity increment inside and outside the time-delay blocking boundary: when the network connectivity weight index is large and the difference in green intensity increment is small, it indicates that the entire field is synchronously greening due to global climate influence; conversely, when the network connectivity weight index is small and the difference in green intensity increment is large, it indicates that local pesticide seepage has triggered a tiered growth recovery. Through this two-dimensional cross-analysis, the pseudo-synchronous greening caused by rainfall and the true growth recovery after local pesticide application are successfully separated from the data structure, and a fallback check branch is used to handle contradictory states, ensuring the robustness of the diagnostic conclusions.

[0072] like Figure 2 As shown in the diagram, this application also provides a structural schematic of a field crop growth monitoring device, including: one or more processors 202; and a memory 201 for storing one or more programs. When one or more programs are executed by one or more processors, the one or more processors 202 implement the field crop growth monitoring method described in the above embodiments. The memory 202 can be a memory in the field crop growth monitoring device, and the memory can include volatile memory, such as random access memory; the memory can also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; the memory can also include combinations of the above types of memory.

[0073] Optionally, the field crop growth monitoring device also includes a communication interface 203 and a bus 204. The communication interface 203 supports communication between the field crop growth monitoring device and other network entities. The memory 202 stores the program code and data of the field crop growth monitoring device. The bus 204 can be an Extended Industry Standard Architecture (EISA) bus, etc. The bus 204 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0074] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0075] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for monitoring the growth of field crops, characterized in that, include: The images in the canopy image sequence of field crops collected over several consecutive days are uniformly divided into multiple spatial grids, and the injection center grid, farmland boundary grid, and new leaf germination characteristic day of each spatial grid are determined. Based on the positional relationships between the multiple spatial grids and the time difference of the characteristic days of new leaf germination between adjacent spatial grids, a time-delay topology network is constructed; the nodes in the time-delay topology network correspond one-to-one with the multiple spatial grids, and the edge weights in the time-delay topology network are determined based on the time difference; Using the nodes corresponding to the injection center grid as the connection start point and the nodes corresponding to the farmland boundary grid as the connection end point, the maximum flow algorithm is executed to determine the network connectivity weight index of the time-delay topology network, and the time-delay blocking boundary composed of directed edges whose connectivity quotas are exhausted is extracted. The growth diagnosis results of field crops are determined based on the difference in greenness increment between the inner and outer grids of the time delay blocking boundary and the network connectivity weight index.

2. The method for monitoring the growth of field crops according to claim 1, characterized in that, Determine the characteristic day of new leaf germination for each spatial grid, including: Based on each frame of the canopy image sequence, the vegetation index for each spatial grid is determined daily over the consecutive days. The date on which the vegetation index of each spatial grid first reaches or exceeds the dynamic greenness reference threshold is determined as the new leaf germination characteristic day of the spatial grid; the dynamic greenness reference threshold is a threshold obtained by adding a preset increment to the arithmetic mean of the vegetation index of all spatial grids in the canopy image of the first day in the consecutive days. If a spatial grid has a vegetation index that is less than the dynamic greenness reference threshold for several consecutive days, then the new leaf germination characteristic day of the spatial grid is set as the timeout penalty value, and the value of the timeout penalty value is greater than the total number of days in the several consecutive days.

3. The method for monitoring the growth of field crops according to claim 2, characterized in that, Based on each frame of the canopy image sequence, the vegetation index for each spatial grid is determined daily over the consecutive days, including: Determine the red, green, and blue reflectance values ​​of each pixel in the canopy images of each spatial grid over the consecutive days. Based on the absorption and reflection of visible light by vegetation, the red, green and blue reflectance values ​​of each pixel are processed to obtain the vegetation index of each pixel. Calculate the arithmetic mean of the vegetation index values ​​of all pixels within each spatial grid to obtain the daily vegetation index for each grid.

4. The method for monitoring the growth of field crops according to claim 2, characterized in that, The process of determining the edge weights in the time-delay topology network includes: Determine the target grid pair; the target grid pair is a grid pair consisting of any two adjacent spatial grids among the plurality of spatial grids; Determine the leaf germination characteristic days of the two spatial grids in the target grid pair; If the new leaf germination characteristic day of a spatial grid in the target grid pair is the timeout penalty value, then the edge weight of the directed edge connecting the target grid pair is set to zero. If the leaf germination characteristic day of the target grid pair does not exist in the spatial grid is the timeout penalty value, then the absolute value of the difference between the leaf germination characteristic days of the two spatial grids is calculated to obtain the time difference. The time difference is then mapped inversely to obtain the edge weight of the directed edge connecting the target grid pair.

5. The method for monitoring the growth of field crops according to claim 1, characterized in that, Using the nodes corresponding to the injection center grid as the connection start point and the nodes corresponding to the farmland boundary grid as the connection end point, the maximum flow algorithm is executed to determine the network connectivity weight index of the time-delay topology network, and the time-delay blocking boundary formed by directed edges whose connectivity quotas have been exhausted is extracted, including: Virtual source points and virtual sink points are set outside the time-delay topology network. The virtual source points are connected to the nodes corresponding to the pesticide injection center grid, and the virtual sink points are connected to the nodes corresponding to the farmland boundary grid. The edge weights of the connection edges between the virtual source points and the nodes corresponding to the pesticide injection center grid, and the edge weights of the connection edges between the virtual sink points and the nodes corresponding to the farmland boundary grid, are all preset values. The preset values ​​are greater than the sum of all edge weights in the time-delay topology network. Using the virtual source point as the search starting point and the virtual sink point as the search ending point, the system continuously searches for valid connected paths in the time-delay topology network that have not yet reached the capacity quota limit; the starting point of the valid connected path is the node corresponding to the injection center grid, and the ending point of the valid connected path is the node corresponding to the farmland boundary grid. After each valid connected path is found, the minimum available quota that can be accommodated on the currently found valid connected path is accumulated, and the remaining available quota of all directed edges on the currently found valid connected path is deducted by the same amount; until no new valid connected path can be found, the total accumulated value is used as the network connectivity weight index. Extract the set of directed edges in the time-delay topology network whose initial connectivity quota is greater than zero and whose current remaining available quota is zero, and use them as the time-delay blocking boundary.

6. The method for monitoring the growth of field crops according to claim 1, characterized in that, The process for determining the difference in greenness increment between the inner and outer grids of the time-delay blocking boundary includes: Based on the spatial distance between the two spatial grids connected by each directed edge in the time delay blocking boundary and the drug injection center grid, the inner grid and the outer grid are determined; the spatial distance between the inner grid and the drug injection center grid is less than the spatial distance between the outer grid and the drug injection center grid; The termination observation dates for the inner grid and the outer grid are determined; the termination observation date is after the characteristic day of new leaf germination and is spaced apart from the characteristic day of new leaf germination by a preset number of days; the preset number of days is less than or equal to the number of days corresponding to the consecutive multi-day period; if the termination observation date exceeds the total number of days of the consecutive multi-day period, then the last day of the consecutive multi-day period is taken as the termination observation date. The greenness increment of the inner grid is determined based on the difference in vegetation index between the end of observation date and the characteristic date of new leaf germination. The greenness increment of the outer grid is determined based on the difference in vegetation index between the observation termination date and the new leaf germination characteristic date. Calculate the difference in greenness increment between the inner grid and the outer grid of each of the time delay blocking boundaries, and determine the greenness increment difference as the arithmetic mean of the differences in greenness increment between each of the time delay blocking boundaries.

7. The method for monitoring the growth of field crops according to claim 1, characterized in that, Based on the difference in greenness increment between the inner and outer grids of the time-delay blocking boundary and the network connectivity weight index, the growth diagnosis results of field crops are determined, including: If the network connectivity weight index is greater than the greening synchronicity extreme threshold and the greening increment difference is less than the greening increment difference threshold, then the first diagnostic result is output; the first diagnostic result is used to indicate that the growth change is caused by global climate factors; the greening synchronicity extreme threshold is determined based on an ideal zero-time-difference network model with the same grid size as the time-delay topology network. If the network connectivity weight index is less than or equal to the extreme threshold of greening synchronicity, and the difference in greenness increment is greater than or equal to the threshold of greenness increment difference, then a second diagnostic result is output; the second diagnostic result is used to indicate that the growth change is a growth recovery caused by local pesticide seepage; If none of the above conditions are met, a third diagnostic result is generated, which is used to indicate that manual review is required to determine the cause of the growth change.

8. The method for monitoring the growth of field crops according to claim 7, characterized in that, The process of determining the extreme threshold of greening synchronicity includes: Construct an ideal zero-time-difference network model; the ideal zero-time-difference network model is consistent with the spatial grid division of the field, and the characteristic day of new leaf germination of each spatial grid is the same day, the time difference between adjacent grids is zero, and the weight of all edges is unified to the theoretical limit value; The ideal zero-time-difference network model is processed using the maximum flow algorithm to obtain the ideal network connectivity weight index. The greening synchronization extreme threshold is determined based on the ideal network connectivity weight index and the network smoothness tolerance coefficient; the network smoothness tolerance coefficient is used to characterize the ideal smoothness ratio required to identify large-scale synchronous greening.

9. The method for monitoring the growth of field crops according to claim 7, characterized in that, The process of determining the threshold for the difference in greenness increment includes: A preset increment is obtained, and the product of the preset increment and the growth drop manifestation coefficient is used as the threshold of the greenness increment difference; wherein, the growth drop manifestation coefficient is used to characterize the proportion of the local drug efficacy decline drop to the preset increment.

10. A growth monitoring device for field crops, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for monitoring the growth of field crops as described in any one of claims 1-9.

Images