A remote sensing identification method and remote sensing identification system for winter wheat and land crops

Abstract

The invention discloses a remote sensing identification method and remote sensing identification system for winter wheat and land crops. The remote sensing identification method includes: using the past winter wheat distribution data of the target area to obtain the interpretation sample point collection range of winter wheat, so as to select the interpretation sample point within the range; and obtaining the multispectral image of the target area, calculating the winter wheat characteristic index parameter value corresponding to each grid pixel according to the multispectral image, and extracting the winter wheat characteristic index parameter value corresponding to the sample point; and inputting the characteristic index parameter value corresponding to each grid pixel into a pre-constructed winter wheat identification model; and the winter wheat identification model identifies each grid pixel of the multispectral image of the target area, and the identification result is that the crop attribute of the grid pixel is winter wheat, or other land objects. The invention uses past achievement data to reduce manual operations, and efficiently and accurately identifies winter wheat to determine the growth layout in the area.

Description

Technical Field

[0001] The present invention relates to the field of crop remote sensing identification, and in particular to a remote sensing identification method and a remote sensing identification system for winter wheat and land crops. Background Art

[0002] Winter wheat, as my country's main overwintering grain crop, occupies an important position in grain production. Therefore, efficient and accurate identification of winter wheat and subsequent analysis to obtain winter wheat planting area and yield data can provide scientific and effective reference opinions for decision makers to timely grasp the winter wheat planting situation, accurately estimate summer grain output, and thus scientifically judge the grain production situation. It is of great significance to ensure grain production and solve food security issues.

[0003] At present, the monitoring and identification methods for crops are mainly divided into supervised classification methods and unsupervised classification methods. The supervised classification method is a process of using known samples to train the machine to master the characteristics of various categories and classify according to the selected classification rules. Common supervised classification methods include maximum likelihood method, minimum distance method, etc. The accuracy of supervised classification results is generally high and the subsequent operation is simple, but because a sufficient number of training samples need to be collected and the training samples need to be representative, it takes a certain amount of manpower, material resources and time, and the monitoring results will be affected by subjective factors. The unsupervised classification method refers to the classification process without applying prior knowledge. The machine automatically classifies according to the spectral characteristics of the ground objects. The classification results do not have attribute information, and the attribute categories need to be subsequently compared and analyzed. Common unsupervised classification methods include K-means method, ISODATA method, etc. The advantage of the unsupervised classification method is that the early classification process is simple and easy to operate, which reduces the influence of subjective factors, but the subsequent operation process is cumbersome and the classification accuracy is lower than that of supervised classification.

[0004] In summary, the current supervised classification method and unsupervised classification method processes are too complicated, and therefore cannot solve the problem of crop monitoring and identification conveniently and quickly. Summary of the invention

[0005] In order to solve the problems of the prior art, the present invention provides a remote sensing identification method and remote sensing identification system for winter wheat and land crops. The technical solution is as follows:

[0006] In one aspect, the present invention provides a remote sensing identification method for land crops, which is used to identify winter wheat in a target area. The remote sensing identification method comprises the following steps:

[0007] Using the past winter wheat distribution data of the target area, the interpretation sample point collection range of winter wheat is obtained, so as to select a plurality of interpretation sample points within the interpretation sample point collection range; and

[0008] Acquire a multispectral image of the target area, calculate the winter wheat characteristic index parameter value corresponding to each grid pixel according to the multispectral image, and extract the winter wheat characteristic index parameter value corresponding to each interpretation sample point; and

[0009] Inputting the winter wheat characteristic index parameter value corresponding to each grid pixel into a pre-constructed winter wheat recognition model, wherein the winter wheat recognition model is related to the winter wheat characteristic index parameter value corresponding to each interpretation sample point; and

[0010] The winter wheat recognition model recognizes each grid pixel of the multispectral image of the target area, and the recognition result is that the crop attribute of the grid pixel is winter wheat, or other ground objects.

[0011] Furthermore, the winter wheat characteristic index parameter is the normalized vegetation index of winter wheat at the heading stage and / or the difference between the normalized vegetation index of winter wheat at the heading stage and the harvest stage, and the calculation formulas thereof are respectively:

[0012] Among them, NDVI is the normalized difference vegetation index, ρ nir is the reflectance value in the near-infrared band, ρ red is the reflectance value of the red light band;

[0013] ΔNDVI=NDVI 抽穗期 -NDVI 收获期 , where NDVI 抽穗期 is the normalized difference vegetation index at heading stage, NDVI 收获期 is the normalized difference vegetation index at the harvest stage, and ΔNDVI is the difference between the normalized difference vegetation index at the heading stage and the harvest stage.

[0014] Furthermore, the pre-constructed winter wheat identification model uses the NDVI of winter wheat at heading stage and the ΔNDVI between heading stage and harvest stage of winter wheat as model characteristic parameters; and

[0015] If the NDVI and ΔNDVI corresponding to the grid pixels of the multispectral image of the target area satisfy: NDVI∈T1 and ΔNDVI∈T2, where T1 is the threshold range determined according to the NDVI of the heading period of winter wheat corresponding to each interpretation sample point, and T2 is the threshold range determined according to the difference ΔNDVI of the normalized vegetation index between the heading period and the harvest period of winter wheat corresponding to each interpretation sample point, then the recognition result is that the crop attribute of the grid pixel is winter wheat, otherwise the recognition result is that the crop attribute of the grid pixel is other land objects.

[0016] Furthermore, the use of remote sensing technology to collect multispectral images of the target area and multispectral data of each grid pixel further includes:

[0017] During the heading period of winter wheat, the multispectral image of the target area and the reflectance values ​​of the near infrared band and the red light band of each raster pixel are obtained, and then the NDVI of the heading period of winter wheat corresponding to each raster pixel is calculated. 抽穗期 ;

[0018] During the winter wheat harvest period, the multispectral image of the target area and the reflectance values ​​of the near infrared band and the red light band of each raster pixel are obtained, and then the NDVI of the winter wheat harvest period corresponding to each raster pixel is calculated. 收获期 ;

[0019] Then, according to the NDVI of winter wheat heading period corresponding to each grid pixel 抽穗期 and NDVI at winter wheat harvest 收获期 Calculate the difference ΔNDVI of the normalized difference vegetation index between the heading stage and the harvest stage of winter wheat corresponding to each grid pixel.

[0020] Further optionally, the multispectral image of the target area is obtained by any of the following methods:

[0021] Download or purchase remote sensing data to obtain a multispectral image of the target area, and then obtain the reflectance value of the near-infrared band and the reflectance value of the red light band of each grid pixel thereof, wherein the remote sensing data includes satellite images; or

[0022] The target area is detected above the target area by using a multispectral imaging device carried by an unmanned aerial vehicle to obtain a multispectral image of the target area, and then the reflection value of the near-infrared band and the reflection value of the red light band of each grid pixel are obtained.

[0023] Furthermore, the winter wheat identification model is related to the winter wheat characteristic index parameter value corresponding to each interpretation sample point and includes:

[0024] According to the winter wheat characteristic index parameter values ​​corresponding to each interpreted sample point, a mathematical method is used to calculate the upper limit value and the lower limit value of the recognition threshold of the winter wheat recognition model, and the interval between the upper limit value and the lower limit value of the recognition threshold constitutes the characteristic index parameter recognition threshold range of the winter wheat recognition model, wherein the mathematical method includes a box plot analysis method or a K-means clustering algorithm or a normal distribution judgment method or a Grubbs test method;

[0025] Among them, if the winter wheat characteristic index parameter value corresponding to the grid pixel of the multispectral image of the target area is within the characteristic index parameter recognition threshold range of the winter wheat recognition model, the recognition result is that the crop attribute of the grid pixel is winter wheat.

[0026] Furthermore, the upper and lower limits of the recognition threshold of the winter wheat recognition model are calculated using a box plot analysis method, based on a sequential data set obtained by arranging the winter wheat characteristic index parameter values ​​corresponding to each interpretation sample point from small to large, and the calculation formula includes:

[0027] Q max =Min[(Q3+1.5*IQR),X max ], where Q max To identify the upper threshold value, Q3 is the upper quartile of the sequential data set, X max is the maximum value of the sequential data set, and Min[] is the minimum function;

[0028] Q min =Max[(Q1-1.5*IQR),X min ], where Q min To identify the lower threshold value, Q1 is the lower quartile of the sequential data set, X min is the minimum value of the sequential data set, and Max[] is the maximum function;

[0029] Among them, IQR is the interquartile range, and its calculation formula is IQR=Q3-Q1.

[0030] Furthermore, the obtaining of the collection range of the interpretation sample points of winter wheat by using the past winter wheat distribution data of the target area further includes:

[0031] Obtain the past winter wheat distribution raster data of the target area, and use the data conversion tool to convert its data format from raster data to vector data;

[0032] The converted vector map patches are summarized to obtain the interpreted sample point collection range of winter wheat in the target area.

[0033] Furthermore, a preset number of interpretation sample points are randomly generated within the obtained interpretation sample point collection range of winter wheat, and the generated interpretation sample points meet the following conditions:

[0034] The density of the interpreted sample points is greater than or equal to a preset per-acre point number threshold, and / or the distance between any two interpreted sample points is greater than or equal to a preset distance threshold.

[0035] Furthermore, after the winter wheat recognition model recognizes all grid pixels of the multispectral image of the target area, the recognition result is verified by the following steps:

[0036] Setting a plurality of verification points in the target area and determining the crop attributes of each verification point based on high-resolution images or actual ground survey data; and

[0037] Matching the grid pixels of the verification point on the multispectral image of the target area to determine the crop attributes of the grid pixels matched by the winter wheat recognition model;

[0038] The number of verification points correctly identified and the number of verification points incorrectly identified by the winter wheat recognition model were counted, and the Kappa coefficient was calculated using a confusion matrix.

[0039] Furthermore, if the calculated result of the Kappa coefficient is lower than the preset qualified accuracy value, the interpretation sample points are reselected within the interpretation sample point collection range and / or the number of selected interpretation sample points is adjusted to re-identify the winter wheat in the target area by remote sensing.

[0040] On the other hand, the present invention provides a remote sensing identification method for land crops, which is used to identify target crops in a target area, obtain the interpretation sample point collection range of the target crop by using the historical distribution data of the target crop in the target area in the past, and select multiple interpretation sample points within the interpretation sample point collection range; and

[0041] Acquire the multispectral image of the target area and the multispectral data of each grid pixel thereof, calculate the crop characteristic index parameter value corresponding to each grid pixel according to the multispectral data, and extract the crop characteristic index parameter value corresponding to each interpretation sample point therefrom; and

[0042] Performing data processing on the crop characteristic index parameter values ​​corresponding to the plurality of interpreted sample points according to a mathematical method to obtain a characteristic index parameter recognition threshold range of the target crop; and

[0043] Each grid pixel of the multispectral image of the target area is judged. If the crop characteristic index parameter value corresponding to the grid pixel is within the characteristic index parameter recognition threshold range of the target crop, the attribute of the grid pixel is judged to be the target crop, otherwise, its attribute is judged to be other ground objects.

[0044] Wherein, the acquisition of the interpretation sample point collection range of the target crop further includes:

[0045] Obtain the distribution grid data of the target crops in the target area, and use the data conversion tool to convert the data format from raster data to vector data;

[0046] The areas of the converted vector map patches are summarized to obtain the interpretation sample point collection range of the target crop in the target area.

[0047] Furthermore, the characteristic index parameter identification threshold range of the target crop is obtained by the following method:

[0048] The crop characteristic index parameter values ​​corresponding to multiple interpretation sample points are calculated using a box plot analysis method, a K-means clustering algorithm, a normal distribution judgment method, or a Grubbs test method to obtain an upper limit value of an identification threshold and a lower limit value of an identification threshold. The interval between the upper limit value of the identification threshold and the lower limit value of the identification threshold constitutes the characteristic index parameter identification threshold range.

[0049] Furthermore, the crop characteristic index parameters are selected according to the growth period characteristics of the target crop, and a recognition model is constructed according to the crop characteristic index parameters; and the multispectral image of the target area in the corresponding growth period and the multispectral data of each grid pixel thereof are obtained.

[0050] Furthermore, the target crop is winter wheat, and the corresponding crop characteristic index parameter is any one or a combination of multiple of the normalized vegetation index of winter wheat at heading stage, the difference between the normalized vegetation index of winter wheat at heading stage and harvest stage, the blue band of winter wheat at heading stage, and the green band of winter wheat at heading stage; or,

[0051] The target crop is rice, and the corresponding crop characteristic index parameter is the normalized vegetation index at the rice heading stage and / or the water index at the rice field flooding stage; or,

[0052] The target crop is corn, and the corresponding crop characteristic index parameter is any one or a combination of multiple ones of the normalized vegetation index at the heading stage of corn, the red edge parameter at the heading stage of corn, the blue band at the heading stage of corn, and the green band at the heading stage of corn.

[0053] Furthermore, the target crop is winter wheat, and the crop characteristic index parameter is the normalized vegetation index of winter wheat at heading stage and / or the difference between the normalized vegetation index of winter wheat at heading stage and harvest stage, and the calculation formulas thereof are respectively:

[0054] Among them, NDVI is the normalized difference vegetation index, ρ nir is the reflectance value in the near-infrared band, ρ red is the reflectance value of the red light band;

[0055] ΔNDVI=NDVI 抽穗期 -NDVI 收获期 , where NDVI 抽穗期 is the normalized difference vegetation index at heading stage, NDVI 收获期 is the normalized difference vegetation index at the harvest stage, and ΔNDVI is the difference between the normalized difference vegetation index at the heading stage and the harvest stage.

[0056] In another aspect, the present invention provides a remote sensing identification system for land crops, comprising the following modules:

[0057] Remote sensing data acquisition module, used to obtain multispectral images of the target area and multispectral data of each grid pixel;

[0058] A historical distribution data storage module is used to store the distribution grid data of the target crops in the target area in the past;

[0059] An interpretation sample point selection module is used to convert the distribution grid data of the target crop in the target area into vector patches, and randomly generate a preset number of interpretation sample points within the interpretation sample point collection range obtained by summarizing the converted vector patches;

[0060] A crop characteristic index parameter value calculation module, used to calculate the crop characteristic index parameter value corresponding to each grid pixel according to the multispectral data acquired by the remote sensing data acquisition module;

[0061] A model recognition threshold calculation module is used to extract the crop characteristic index parameter value corresponding to each interpretation sample point from the calculation result of the crop characteristic index parameter value calculation module, and use a mathematical method to calculate the upper limit value and the lower limit value of the recognition threshold of the crop recognition model, wherein the interval between the upper limit value and the lower limit value of the recognition threshold constitutes the characteristic index parameter recognition threshold range of the winter wheat recognition model, wherein the mathematical method includes a box plot analysis method or a K-means clustering algorithm or a normal distribution judgment method or a Grubbs test method;

[0062] The crop recognition model module recognizes each grid pixel of the multispectral image of the target area according to the characteristic index parameter recognition threshold range calculated by the model recognition threshold calculation module, and the recognition result is that the crop attribute of the grid pixel is the target crop, or other ground objects.

[0063] The present invention proposes a remote sensing identification method for winter wheat. According to the growth and development characteristics of winter wheat, the optimal characteristic parameters are selected to construct a model. At the same time, the model threshold is quickly obtained by using past achievements and statistical principles, thereby realizing efficient and accurate identification of winter wheat. The technical solution provided by the present invention has the following beneficial effects:

[0064] a. Based on the past achievements, winter wheat interpretation sample points are obtained. At the same time, the box plot algorithm can be used to eliminate outliers and screen out the correct numerical range of the characteristic parameters of the sample points as the model threshold, thereby realizing rapid identification and monitoring of winter wheat;

[0065] b. Reduce manual intervention in the selection of interpretation sample points, reduce the complex workload caused by collecting a large number of interpretation sample points, and improve monitoring accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0067] Figure 1 A schematic flow chart of a remote sensing identification method for winter wheat provided as an exemplary embodiment of the present disclosure;

[0068] Figure 2 A schematic diagram of a discrimination flow of a winter wheat monitoring model provided for an exemplary embodiment of the present disclosure;

[0069] Figure 3 A schematic diagram of the distribution grid data of winter wheat in Wen County in 2019 provided for an exemplary embodiment of the present disclosure;

[0070] Figure 4 An exemplary embodiment of the present disclosure provides a Figure 3 Schematic diagram of the data collection range of winter wheat interpretation sample points determined by the distribution grid data;

[0071] Figure 5 An exemplary embodiment of the present disclosure provides a Figure 4 A distribution diagram of interpretation sample points selected based on the interpretation sample point collection range;

[0072] Figure 6 An exemplary embodiment of the present disclosure provides Figure 5 Schematic diagram of NDVI value scatter plots of sample points in the interpretation;

[0073] Figure 7 An exemplary embodiment of the present disclosure provides Figure 5 Schematic diagram of scatter plots of ΔNDVI values ​​of sample points in the interpretation;

[0074] Figure 8 A schematic diagram of the distribution grid data of winter wheat in Wen County in 2020 provided for an exemplary embodiment of the present disclosure;

[0075] Fig. 9 A schematic module block diagram of a remote sensing identification system for land crops provided as an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0076] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0077] It should be noted that the terms "comprises" and "has" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or equipment comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or equipment.

[0078] The inventive concept of the present invention is as follows: as crops grow and develop, the spectral characteristics change with certain regularity, and different crops have different growth periods, and the spectral characteristics change law is also different. The present invention can effectively distinguish the target crops by using the spectral characteristics of crops and the regularity and difference of spectral changes during the growth and development process. Therefore, a model is proposed based on the growth and development characteristics of winter wheat, combined with the planting system, and the NDVI of the heading period of winter wheat and the difference ΔNDVI of the heading period and the harvest period are selected to construct a model. At the same time, sample points are obtained by using past achievements, and the normal value range of the sample points is obtained by using statistical principles as the model threshold to realize efficient and accurate identification of winter wheat. Specifically, the present invention proposes an identification method for winter wheat using a decision tree classification method combined with statistical principles. Make full use of the growth and development characteristics of winter wheat, select the NDVI of the heading period of winter wheat and the difference ΔNDVI of the heading period and the harvest period to construct an identification model. At the same time, using past achievement data, winter wheat sample points are obtained by random sampling, and the normal value range of the characteristic parameters of the winter wheat sample points is screened out by using the box plot algorithm as the model threshold, thereby realizing efficient and accurate identification of winter wheat. While ensuring the recognition accuracy, the recognition method reduces the manual intervention in the selection process of interpretation sample points, reduces the complex workload caused by collecting a large number of interpretation sample points, and improves the recognition accuracy and efficiency.

[0079] In one embodiment of the present invention, a remote sensing identification method for winter wheat is provided, which is used to identify winter wheat in a target area. Figure 1 , the remote sensing identification method comprises the following steps:

[0080] Using the past winter wheat distribution data of the target area, the interpretation sample point collection range of winter wheat is obtained, so as to select a plurality of interpretation sample points within the interpretation sample point collection range; and

[0081] Acquire a multispectral image of the target area, calculate the winter wheat characteristic index parameter value corresponding to each grid pixel according to the multispectral image, and extract the winter wheat characteristic index parameter value corresponding to each interpretation sample point; and

[0082] Inputting the winter wheat characteristic index parameter value corresponding to each grid pixel into a pre-constructed winter wheat recognition model, wherein the winter wheat recognition model is related to the winter wheat characteristic index parameter value corresponding to each interpretation sample point; and

[0083] The winter wheat recognition model recognizes each grid pixel of the multispectral image of the target area, and the recognition result is that the crop attribute of the grid pixel is winter wheat, or other ground objects.

[0084] The present invention does not strictly limit the sequence of the above steps. The sequence of the steps can be adaptively adjusted or changed within a reasonable range. The following is an example of a feasible sequence:

[0085] Wen County, Jiaozuo City, Henan Province was selected as the research area. Based on sentinel-210m resolution remote sensing images, the decision tree classification method was used, and the existing 2019 winter wheat distribution pattern data of Wen County was used to efficiently and accurately obtain the 2020 winter wheat distribution pattern data of Wen County according to this scheme.

[0086] Step 1: Use past results to obtain the collection range of winter wheat sample points. Wen County, Jiaozuo City, Henan Province was selected as the target area for research, and the distribution pattern data of winter wheat in Wen County in 2019 was obtained. Figure 3 , use the data conversion tool to convert the raster data into vector patches. Calculate the vector area S for each patch 矢量 , filter out all S 矢量 For vector patches of ≥1 mu, all patches are merged to form the collection range of winter wheat interpretation sample points in the target area. Figure 4 .

[0087] Specifically, the existing ArcGIS 10.4 software tool can be used for data conversion and calculation of the vector area of ​​the map patch, and both the raster data before conversion and the vector map patch after conversion have position coordinate information.

[0088] Among them, by excluding S 矢量 Vector map patches of less than 1 mu can effectively prevent areas with erroneous sample points from being included in the collection range of interpretation sample points, so as to ensure the accuracy of the collection range of interpretation sample points, and thus improve the precision and accuracy of selecting interpretation sample points in the next step.

[0089] However, the present invention does not limit the critical value of the vector area to 1 mu, and the specific value can be adjusted / set according to different situations, for example, it can also be 0.1 mu-0.5 mu, or 1.1 mu-1.5 mu. The values ​​here and the parts involving values ​​below are all examples and should not be used as the basis for limiting the scope of protection claimed by the present invention.

[0090] Step 2: Obtain interpretation sample points within the collection range of winter wheat sample points. Obtain winter wheat interpretation sample points by random sampling within the collection range of winter wheat interpretation sample points in Wen County. Figure 5 The specific number of interpretation sample points N is set according to the vector area of ​​the winter wheat sample point collection range to ensure the interpretation sample point density P 密度 ≥500 mu / point, and the minimum distance between interpretation sample points is limited to 100 meters.

[0091] Likewise, the sample point density and the set threshold for the minimum distance between interpreted sample points can be adjusted.

[0092] Among them, the interpretation sample point density P 密度 ≥500 mu / point, that is, at least one interpretation sample point needs to be set up within every 500 mu. Setting a minimum limit can ensure that the sample points are not selected too loosely within the collection range of the interpretation sample points, and at the same time avoid too few sample points resulting in unrepresentative samples and affecting the accuracy; limiting the minimum distance between two adjacent interpretation sample points can avoid subsequent analysis errors caused by too dense sample points. For example, when the maturity of winter wheat in different areas within the collection range of the interpretation sample points is different, the corresponding winter wheat characteristic index parameter values ​​will be different. If the sampling points are collected too densely in a small area, it will obviously bring large errors to the determination of the winter wheat characteristic index parameter values ​​within the overall collection range.

[0093] In one embodiment of the present invention, after selecting a plurality of interpretation sample points within the interpretation sample point collection range, the selected interpretation sample points are screened according to the terrain characteristic parameters of the winter wheat planting area, including:

[0094] The slope data at each selected interpretation sample point is obtained and compared with the preset slope threshold. If the slope exceeds the preset slope threshold, the corresponding sample point is screened out. Specifically, according to agronomic prior knowledge, winter wheat consumes a lot of water during its growth and development and needs irrigation. In order to avoid soil erosion, winter wheat is basically planted in areas with gentle slopes and flat terrain. Using the above characteristics, based on terrain data, the slope data P of the interpretation sample point is obtained from the terrain data using an extraction tool. slope , filter out P slope For interpretation sample points with an angle of >6°, continue to perform the following steps with the remaining interpretation sample points.

[0095] Step 3: Construct a monitoring model based on the growth and development characteristics of the target crop. According to agronomic prior knowledge, winter wheat is in the heading stage in April, and its growth condition is higher than that of most landforms in the same period. In June, the vegetation index of winter wheat decreases significantly due to harvest, while other landforms do not change significantly during the same period. Using the above characteristics, the normalized vegetation index NDVI of winter wheat in the heading stage and the difference between the normalized vegetation index in the heading stage and the harvest stage ΔNDVI are selected as characteristic parameters to construct the recognition model. The specific steps are as follows:

[0096] Condition ①: NDVI∈T1, if the condition is true, then proceed to step ②, otherwise the pixel is judged as other ground objects;

[0097] Condition ②: ΔNDVI∈T2, if the condition is true, the pixel is identified as winter wheat, otherwise the pixel is identified as other ground objects.

[0098] Alternatively, the judgment order of the above conditions ① and ② can be swapped. In short, if both conditions are met, the pixel is judged as winter wheat, otherwise it is judged as other land features.

[0099] Among them, the calculation formulas for the normalized difference vegetation index NDVI at the heading stage of winter wheat and the difference ΔNDVI between the heading stage and the harvest stage are:

[0100] Among them, NDVI is the normalized difference vegetation index, ρ nir is the reflectance value in the near-infrared band, ρ red is the reflectance value of the red light band;

[0101] ΔNDVI=NDVI 抽穗期 -NDVI 收获期 , where NDVI 抽穗期 is the normalized difference vegetation index at heading stage, NDVI 收获期 is the normalized difference vegetation index at the harvest stage, and ΔNDVI is the difference between the normalized difference vegetation index at the heading stage and the harvest stage.

[0102] Among them, T1 and T2 are threshold ranges. This step is to build a monitoring model. The next step is to determine the threshold ranges of T1 and T2 based on the interpretation sample points obtained in step 2.

[0103] Obviously, this step can be performed before step 1.

[0104] Step 4: Obtain multispectral images of the target area, and then set thresholds to identify winter wheat based on the analysis results of the interpreted sample points. According to the growth characteristics and image conditions of winter wheat in the target area, obtain images of the heading and harvest periods of winter wheat that meet the quality requirements of the target area, and obtain the NDVI and difference ΔNDVI of the heading period of winter wheat by superposition operation. This step can be performed before step 1 or step 3. Use the extraction tool to obtain the NDVI value and ΔNDVI value of the interpreted sample points, see Figure 6 and Figure 7 The upper edge Q of the NDVI and ΔNDVI of the sample points were calculated using the box plot analysis method. max and lower edge Q min The numerical value and the calculated result are used as the model threshold to identify winter wheat.

[0105] Specifically, the multispectral image of the target area is obtained by any of the following methods:

[0106] Download or purchase remote sensing data to obtain a multispectral image of the target area, and then obtain the reflectance value of the near-infrared band and the reflectance value of the red light band of each grid pixel thereof, wherein the remote sensing data includes satellite images; or

[0107] The target area is detected above the target area by using a multispectral imaging device carried by an unmanned aerial vehicle to obtain a multispectral image of the target area, and then the reflection value of the near-infrared band and the reflection value of the red light band of each grid pixel are obtained.

[0108] First, the upper edge Q of the model threshold T1 of the NDVI value is used max and lower edge Q min The calculation method of the numerical value is explained below:

[0109] Based on Figure 6 The NDVI of winter wheat corresponding to each interpreted sample point is shown 抽穗期 The sequential data set arranged from small to large is used to calculate the upper edge Q of T1 using the following formula max and lower edge Q min :

[0110] Q max =Min[(Q3+1.5*IQR),X max ], where Q max To identify the upper threshold value, Q3 is the upper quartile of the sequential data set, X max is the maximum value of the sequential data set, and Min[] is the minimum function;

[0111] Q min =Max[(Q1-1.5*IQR),X min ], where Qmin To identify the lower threshold value, Q1 is the lower quartile of the sequential data set, X min is the minimum value of the sequential data set, and Max[] is the maximum function;

[0112] Among them, IQR is the interquartile range, and its calculation formula is IQR=Q3-Q1, Q3 and Q1 are the upper quartile and lower quartile respectively, which are the i-th and j-th values ​​after all the values ​​in the sample are arranged from small to large.

[0113] Where i = 3*(n+1) / 4, j = (n+1) / 4, and n is NDVI 抽穗期 The number of sample points of the sequential data set arranged from small to large. From the above formula, we can see that i and j may be integers (the decimal part is 0) or non-integers.

[0114] Q3=X i +(X i+1 -X i )*i fractional part, where X i is the i-th position of the sequential data set, X i+1 is the i+1th position of the sequential data set;

[0115] Q1=X j +(X j+1 -X j )*j fractional part, where X j is the jth bit of the sequential data set, X j+1 It is the j+1th bit of the sequential data set.

[0116] The upper edge Q of the model threshold T2 of the ΔNDVI value max and lower edge Q min The calculation method of the numerical value is the same as above, that is, Figure 7 The sequence data set obtained by arranging the normalized vegetation index difference ΔNDVI of winter wheat heading period and harvest period corresponding to each interpreted sample point from small to large is shown in the figure. The upper edge Q of T2 is calculated using the same formula as above. max and lower edge Q min .

[0117] In one embodiment of the present invention, the NDVI of the sample point is interpreted 抽穗期 The results of the ΔNDVI box plot analysis are shown in Table 1:

[0118] Table 1 Analysis results of NDVI and difference ΔNDVI box plot method at heading stage of interpreted sample points

[0119]

[0120] The data in Table 1 above,

[0121] T1∈[069,0.90],

[0122] T2∈[0.48,0.77],

[0123] The flowchart of using the winter wheat recognition model / monitoring model to identify each raster pixel of the multispectral image of the target area is shown in Figure 2 The distribution diagram of winter wheat identified is as follows: Figure 8 shown.

[0124] The mechanism of this step is that the sample points obtained through past results are generally of reference significance, but there are also factors of change. For example, an area that planted winter wheat last year may plant other crops next year. In this case, the interpreted sample points calibrated in the area will not have reference significance. Therefore, the NDVI calculated from the corresponding raster data needs to be used. 抽穗期 and / or ΔNDVI from the upper edge Q max and lower edge Q min However, the present invention is not limited to using the box plot analysis method, a mathematical method, to delete some abnormal data that obviously do not meet the general characteristics from a pile of data.

[0125] For example, common mathematical methods can also use K-means clustering algorithm or normal distribution judgment method or Grubbs test method to determine the upper and lower critical values ​​of T1 and T2 (Q max , Q min ). Taking the K-means clustering algorithm as an example, we can use SPSS software to use K-means clustering analysis to cluster NDVI 抽穗期 The sample points are divided into two categories, among which NDVI 抽穗期 The NDVI value range of the category with large values ​​is used as the model threshold T1; the ΔNDVI sample points are divided into two categories using K-means cluster analysis, among which the ΔNDVI value range of the category with large ΔNDVI values ​​is used as the model threshold T2.

[0126] This embodiment further evaluates the accuracy of the recognition results of the present invention according to national standards. Referring to the national standard "Quality Inspection and Acceptance of Digital Surveying and Mapping Results", 300 verification points are randomly generated within the scope of Wen County, and the true type of the verification points is determined based on high-resolution images or actual ground survey data, and compared with the remote sensing classification results, and the number of verification points with correct classification, wrong classification and missing classification is counted, see Table 2, and the confusion matrix is ​​used to calculate the Kappa coefficient, mapping accuracy and user accuracy respectively.

[0127] Table 2 Accuracy test table of winter wheat monitoring results in Wen County in 2020

[0128]

[0129] The calculation formula of Kappa coefficient is as follows:

[0130] Among them, P o It is the sum of the number of correctly classified samples in each category divided by the total number of samples. Taking the data in Table 2 as an example, P o =(113+169) / 300;

[0131] where a1, a2, ..., a n is the number of real samples of each class, b1,b2,...,b n is the number of monitoring samples of each category, N is the total number of samples, taking the data in Table 2 as an example, P e =(126*113+174*169) / (300*300).

[0132] The Kappa coefficient calculated in this embodiment exceeds 0.88, indicating that the remote sensing identification method for winter wheat provided in this embodiment has high accuracy and great practical significance.

[0133] If the calculated result of the Kappa coefficient is lower than the preset qualified accuracy value, it means that compared with the winter wheat distribution raster data in 2019, the winter wheat planting situation in 2020 may have changed significantly. For example, the area where winter wheat was planted in 2019 was changed to other landforms in 2020, or there was an error in the determination of the collection range of the interpretation sample points, or there was an error in the selection of the interpretation sample points, or the number of interpretation sample points was insufficient, excessive, too dense or too sparse, and it is necessary to redefine the collection range of the interpretation sample points, or reselect the interpretation sample points within the collection range of the interpretation sample points and / or adjust the number of selected interpretation sample points, or change the generation density or adjacent spacing of the interpretation sample points, and re-identify the winter wheat in the target area by remote sensing.

[0134] In the above embodiment, based on the growth and development characteristics of winter wheat, the normalized vegetation index NDVI of winter wheat at the heading period and the difference ΔNDVI of the normalized vegetation index between the heading period and the harvest period are used as characteristic parameters of the monitoring model to eliminate interference from other ground objects and accurately identify winter wheat. However, the present invention is not limited to the above as the only monitoring model. For example, using only the normalized vegetation index NDVI of winter wheat at the heading period, or only the difference ΔNDVI of the normalized vegetation index between the heading period and the harvest period of winter wheat as the characteristic parameters of the monitoring model is also a feasible implementation plan of the monitoring model.

[0135] The present invention does not limit the implementation scheme of constructing the recognition model with other crop characteristic index parameters, nor does it exclude the above normalized vegetation index as a characteristic parameter of the monitoring model of crops other than winter wheat, nor does it limit the monitoring model of crops other than winter wheat to construct the recognition model with other crop characteristic index parameters. In one embodiment of the present invention, a remote sensing recognition method for land crops is provided for identifying target crops in a target area:

[0136] Using the historical distribution data of the target crop in the target area in the past, the interpretation sample point collection range of the target crop is obtained, so as to select a plurality of interpretation sample points within the interpretation sample point collection range; and

[0137] Acquire the multispectral image of the target area and the multispectral data of each grid pixel thereof, calculate the crop characteristic index parameter value corresponding to each grid pixel according to the multispectral data, and extract the crop characteristic index parameter value corresponding to each interpretation sample point therefrom; and

[0138] Performing data processing on the crop characteristic index parameter values ​​corresponding to the plurality of interpreted sample points according to a mathematical method to obtain a characteristic index parameter recognition threshold range of the target crop; and

[0139] Each grid pixel of the multispectral image of the target area is judged. If the crop characteristic index parameter value corresponding to the grid pixel is within the characteristic index parameter recognition threshold range of the target crop, the attribute of the grid pixel is judged to be the target crop, otherwise, its attribute is judged to be other ground objects.

[0140] The above mathematical method can specifically obtain the characteristic index parameter identification threshold range of the target crop in the following way:

[0141] The crop characteristic index parameter values ​​corresponding to multiple interpretation sample points are calculated using a box plot analysis method, a K-means clustering algorithm, a normal distribution judgment method, or a Grubbs test method to obtain an upper limit value of an identification threshold and a lower limit value of an identification threshold. The interval between the upper limit value of the identification threshold and the lower limit value of the identification threshold constitutes the characteristic index parameter identification threshold range.

[0142] Among them, the crop characteristic index parameters are selected according to the growth period characteristics of the target crop, and the recognition model is constructed according to the crop characteristic index parameters; and the multispectral image of the target area in the corresponding growth period and the multispectral data of each raster pixel thereof are obtained. For example: if the target crop is winter wheat, the normalized vegetation index at the heading period and / or the difference between the normalized vegetation index at the heading period and the harvest period can be selected as the corresponding crop characteristic index parameters to construct the recognition model according to its growth period characteristics, and the blue band or green band at the heading period can also be selected as the crop characteristic index parameter; for example, if the crop is rice, the water body index at the flooding period and / or the normalized vegetation index at the heading period can be selected as the crop characteristic index parameter; for example, if the crop is corn, the normalized vegetation index at the heading period and / or the red edge parameter at the heading period and / or the blue band at the heading period and / or the green band at the heading period can be selected as the crop characteristic index parameter.

[0143] In one embodiment of the present invention, a remote sensing identification system for land crops is provided. Fig. 9 , the remote sensing identification system includes the following modules:

[0144] Remote sensing data acquisition module, used to obtain multispectral images of the target area and multispectral data of each grid pixel;

[0145] A historical distribution data storage module is used to store the distribution grid data of the target crops in the target area in the past;

[0146] An interpretation sample point selection module is used to convert the distribution grid data of the target crop in the target area into vector patches, and randomly generate a preset number of interpretation sample points within the interpretation sample point collection range obtained by summarizing the converted vector patches;

[0147] A crop characteristic index parameter value calculation module is used to calculate the crop characteristic index parameter value corresponding to each grid pixel according to the multispectral data acquired by the remote sensing data acquisition module, including the crop characteristic index parameter value corresponding to the interpretation sample point generated by the corresponding interpretation sample point selection module;

[0148] A model recognition threshold calculation module is used to extract the crop characteristic index parameter value corresponding to each interpretation sample point from the calculation result of the crop characteristic index parameter value calculation module, and use a mathematical method to calculate the upper limit value and the lower limit value of the recognition threshold of the crop recognition model, wherein the interval between the upper limit value and the lower limit value of the recognition threshold constitutes the characteristic index parameter recognition threshold range of the winter wheat recognition model, wherein the mathematical method includes a box plot analysis method or a K-means clustering algorithm or a normal distribution judgment method or a Grubbs test method;

[0149] The crop recognition model module recognizes each grid pixel of the multispectral image of the target area according to the characteristic index parameter recognition threshold range calculated by the model recognition threshold calculation module, and the recognition result is that the crop attribute of the grid pixel is the target crop, or other ground objects.

[0150] It should be noted that: the remote sensing identification system for land crops provided in the above embodiment only uses the division of the above functional modules as an example when performing remote sensing identification of crops. In actual applications, the above functional distribution can be completed by different functional modules, that is, it can be divided into different functional modules to complete all or part of the functions described above. In addition, the remote sensing identification method embodiment of land crops and the remote sensing identification system embodiment of land crops of the present invention belong to the same inventive concept as the above-mentioned remote sensing identification method embodiment of winter wheat. The entire content of the remote sensing identification method embodiment of winter wheat is incorporated into the remote sensing identification method embodiment of land crops and the remote sensing identification system embodiment of land crops by reference, and will not be repeated.

[0151] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0152] The above is only a specific implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A remote sensing identification method for winter wheat, characterized in that: For identifying winter wheat in a target area, the remote sensing identification method comprises the following steps: Using the past winter wheat distribution data of the target area, obtaining the interpretation sample point collection range of winter wheat, and selecting a plurality of interpretation sample points within the interpretation sample point collection range; and Acquire a multispectral image of the target area, calculate the winter wheat characteristic index parameter value corresponding to each grid pixel according to the multispectral image, and extract the winter wheat characteristic index parameter value corresponding to each interpretation sample point; and The winter wheat characteristic index parameter value corresponding to each grid pixel is input into a pre-constructed winter wheat recognition model, and the winter wheat recognition model is related to the winter wheat characteristic index parameter value corresponding to each interpretation sample point, including: according to the winter wheat characteristic index parameter value corresponding to each interpretation sample point, the recognition threshold upper limit value and the recognition threshold lower limit value of the winter wheat recognition model are calculated by a mathematical method, and the interval between the recognition threshold upper limit value and the recognition threshold lower limit value constitutes the characteristic index parameter recognition threshold range of the winter wheat recognition model; wherein, if the winter wheat characteristic index parameter value corresponding to the grid pixel of the multispectral image of the target area is within the characteristic index parameter recognition threshold range of the winter wheat recognition model, the recognition result is that the crop attribute of the grid pixel is winter wheat; wherein, the recognition threshold upper limit value and the recognition threshold lower limit value of the winter wheat recognition model are calculated by a box plot analysis method, and the sequential data set obtained by arranging the winter wheat characteristic index parameter values ​​corresponding to each interpretation sample point from small to large is calculated by a calculation formula including: Q max =Min[(Q3+1.5*IQR),X max ], where Q max To identify the upper threshold value, Q3 is the upper quartile of the sequential data set, X max is the maximum value of the sequential data set, Min[] is the minimum function; Q min =Max[(Q1-1.5*IQR),X min ], where Q min To identify the lower threshold value, Q1 is the lower quartile of the sequential data set, X min is the minimum value of the sequential data set, Max[] is the maximum function; wherein IQR is the interquartile range, and its calculation formula is IQR=Q3-Q1; and The winter wheat recognition model recognizes each grid pixel of the multispectral image of the target area, and the recognition result is that the crop attribute of the grid pixel is winter wheat, or other ground objects.

2. The remote sensing identification method of winter wheat according to claim 1, characterized in that: The winter wheat characteristic index parameter is the normalized vegetation index of winter wheat at the heading stage and / or the difference between the normalized vegetation index of winter wheat at the heading stage and the harvest stage, and the calculation formulas thereof are: Among them, NDVI is the normalized difference vegetation index, ρ nir is the reflectance value in the near-infrared band, ρ red is the reflectance value of the red light band; ΔNDVI=NDVI 抽穗期 -NDVI 收获期 , where NDVI 抽穗期 is the normalized difference vegetation index at heading stage, NDVI 收获期 is the normalized difference vegetation index at the harvest stage, and ΔNDVI is the difference between the normalized difference vegetation index at the heading stage and the harvest stage.

3. The remote sensing identification method of winter wheat according to claim 2, characterized in that: The pre-constructed winter wheat identification model uses the NDVI of winter wheat at heading stage and the ΔNDVI of winter wheat between heading stage and harvest stage as model characteristic parameters; and If the NDVI and ΔNDVI corresponding to the grid pixels of the multispectral image of the target area satisfy: NDVI∈T1 and ΔNDVI∈T2, where T1 is the threshold range determined according to the NDVI of the heading period of winter wheat corresponding to each interpretation sample point, and T2 is the threshold range determined according to the difference ΔNDVI of the normalized vegetation index between the heading period and the harvest period of winter wheat corresponding to each interpretation sample point, then the recognition result is that the crop attribute of the grid pixel is winter wheat, otherwise the recognition result is that the crop attribute of the grid pixel is other land objects.

4. The remote sensing identification method of winter wheat according to claim 3, characterized in that: Also includes: During the heading period of winter wheat, the multispectral image of the target area and the reflectance values ​​of the near infrared band and the red light band of each raster pixel are obtained, and then the NDVI of the heading period of winter wheat corresponding to each raster pixel is calculated. 抽穗期 ; During the winter wheat harvest period, the multispectral image of the target area and the reflectance values ​​of the near infrared band and the red light band of each raster pixel are obtained, and then the NDVI of the winter wheat harvest period corresponding to each raster pixel is calculated. 收获期 ; Then, according to the NDVI of winter wheat heading period corresponding to each grid pixel 抽穗期 and NDVI at winter wheat harvest 收获期 Calculate the difference ΔNDVI of the normalized difference vegetation index between the heading stage and the harvest stage of winter wheat corresponding to each grid pixel.

5. The remote sensing identification method of winter wheat according to claim 1, characterized in that: The method of using the past winter wheat distribution data of the target area to obtain the collection range of the interpretation sample points of winter wheat further includes: Obtain the past winter wheat distribution raster data of the target area, and use the data conversion tool to convert the data format from raster data to vector patches; summarize the area of ​​the converted vector patches to obtain the interpretation sample point collection range of the winter wheat in the target area; and A preset number of interpretation sample points are randomly generated within the obtained interpretation sample point collection range of winter wheat, and the generated interpretation sample points meet the following conditions: the density of the interpretation sample points is greater than or equal to the preset per-acre point number threshold, and / or the distance between any two interpretation sample points is greater than or equal to the preset distance threshold.

6. A remote sensing identification method for land crops, used to identify target crops in a target area, characterized in that: Using the historical distribution data of the target crop in the target area in the past, the interpretation sample point collection range of the target crop is obtained, so as to select a plurality of interpretation sample points within the interpretation sample point collection range; as well as Acquire the multispectral image of the target area and the multispectral data of each grid pixel thereof, calculate the crop characteristic index parameter value corresponding to each grid pixel according to the multispectral data, and extract the crop characteristic index parameter value corresponding to each interpretation sample point therefrom; as well as The crop characteristic index parameter values ​​corresponding to the plurality of interpretation sample points are processed according to a mathematical method to obtain a characteristic index parameter recognition threshold range of the target crop, including: calculating an upper limit value of the recognition threshold and a lower limit value of the recognition threshold, wherein the interval between the upper limit value of the recognition threshold and the lower limit value of the recognition threshold constitutes a characteristic index parameter recognition threshold range corresponding to the target crop, wherein the mathematical method is a box plot analysis method: based on a sequential data set obtained by arranging the crop characteristic index parameter values ​​corresponding to each interpretation sample point from small to large, the calculation formula includes: Q max =Min[(Q3+1.5*IQR),X max ], where Q max To identify the upper threshold value, Q3 is the upper quartile of the sequential data set, X max is the maximum value of the sequential data set, Min[] is the minimum function; Q min =Max[(Q1-1.5*IQR),X min ], where Q min To identify the lower threshold value, Q1 is the lower quartile of the sequential data set, X min is the minimum value of the sequential data set, Max[] is the maximum function; wherein IQR is the interquartile range, and its calculation formula is IQR=Q3-Q1; and Each grid pixel of the multispectral image of the target area is judged. If the crop characteristic index parameter value corresponding to the grid pixel is within the characteristic index parameter recognition threshold range of the target crop, the attribute of the grid pixel is judged to be the target crop, otherwise, its attribute is judged to be other ground objects.

7. The remote sensing identification method of land crops according to claim 6, characterized in that: The crop characteristic index parameters are selected according to the growth period characteristics of the target crop, and a recognition model is constructed according to the crop characteristic index parameters; and a multispectral image of the target area during the corresponding growth period and multispectral data of each grid pixel thereof are obtained.

8. The remote sensing identification method of land crops according to claim 6, characterized in that: The target crop is winter wheat, and the corresponding crop characteristic index parameter is any one or a combination of multiple of the normalized vegetation index of winter wheat at heading stage, the difference between the normalized vegetation index of winter wheat at heading stage and harvest stage, the blue band of winter wheat at heading stage, and the green band of winter wheat at heading stage; or, The target crop is rice, and the corresponding crop characteristic index parameter is the normalized vegetation index at the rice heading stage and / or the water index at the rice field flooding stage; or, The target crop is corn, and the corresponding crop characteristic index parameter is any one or a combination of multiple ones of the normalized vegetation index at the heading stage of corn, the red edge parameter at the heading stage of corn, the blue band at the heading stage of corn, and the green band at the heading stage of corn.

9. A remote sensing identification system for land crops, characterized in that: Includes the following modules: Remote sensing data acquisition module, used to obtain multispectral images of the target area and multispectral data of each grid pixel; A historical distribution data storage module is used to store the distribution grid data of the target crops in the target area in the past; An interpretation sample point selection module is used to convert the distribution grid data of the target crop in the target area into vector patches, and randomly generate a preset number of interpretation sample points within the interpretation sample point collection range obtained by summarizing the converted vector patches; A crop characteristic index parameter value calculation module, used to calculate the crop characteristic index parameter value corresponding to each grid pixel according to the multispectral data acquired by the remote sensing data acquisition module; The model recognition threshold calculation module is used to extract the crop characteristic index parameter value corresponding to each interpretation sample point from the calculation result of the crop characteristic index parameter value calculation module, and use a mathematical method to calculate the upper limit value and the lower limit value of the recognition threshold of the crop recognition model, and the interval between the upper limit value and the lower limit value of the recognition threshold constitutes the characteristic index parameter recognition threshold range of the crop recognition model, wherein the mathematical method is a box plot analysis method: based on the sequential data set obtained by arranging the crop characteristic index parameter values ​​corresponding to each interpretation sample point from small to large, the calculation formula includes: Q max =Min[(Q3+1.5*IQR),X max ], where Q max To identify the upper threshold value, Q3 is the upper quartile of the sequential data set, X max is the maximum value of the sequential data set, Min[] is the minimum function; Q min =Max[(Q1-1.5*IQR),X min ], where Q min To identify the lower threshold value, Q1 is the lower quartile of the sequential data set, X min is the minimum value of the sequential data set, and Max[] is the maximum function; wherein IQR is the interquartile range, and its calculation formula is IQR=Q3-Q1; The crop recognition model module recognizes each grid pixel of the multispectral image of the target area according to the characteristic index parameter recognition threshold range calculated by the model recognition threshold calculation module, and the recognition result is that the crop attribute of the grid pixel is the target crop, or other ground objects.

Images