Crop drawing determination method and device, electronic equipment and storage medium

Abstract

The invention provides a crop drawing determination method and apparatus, an electronic device and a storage medium. The method comprises the steps of determining a crop relative height corresponding to each pixel and a crop remote sensing spectral index corresponding to each pixel in a cultivated land area image; based on the crop relative height corresponding to each pixel, determining the height dimension probability that each pixel is a target crop in the height dimension; based on the crop remote sensing spectrum index corresponding to each pixel, determining a remote sensing spectrum dimension probability that each pixel is a target crop in a remote sensing spectrum dimension; determining whether each pixel is a target crop based on the height dimension probability and the remote sensing spectrum dimension probability; determining a crop map based on the pixel corresponding to the target crop; the crop drawing is used for representing the classification condition of the target crop and the non-target crop. According to the method, the target crop is determined by using the two characteristic indexes of the height dimension and the remote sensing spectrum, so that the accuracy and precision of classification of the target crop and other crops are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a crop mapping determination method, device, electronic equipment and storage medium. Background Art

[0002] At present, the agricultural field often uses a single-dimensional characteristic index of crops to classify crops and produce crop classification maps. However, a single-dimensional characteristic index will make crop classification absolute and one-sided, resulting in misclassification and omission problems between target crops and other crops. Summary of the invention

[0003] The present invention provides a crop mapping determination method, device, electronic device and storage medium, which are used to solve the defects in the prior art that the classification of crops of different categories is low and inaccurate due to the absolutization and one-sidedness of crop classification caused by a single-dimensional characteristic index, and achieve high accuracy and high precision in crop classification.

[0004] The present invention provides a crop mapping determination method, comprising the following steps: Determine the crop relative height corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel; Based on the relative height of the crop corresponding to each pixel, the height dimension probability of each pixel being the target crop in the height dimension is determined; Based on the crop remote sensing spectral index corresponding to each pixel, the remote sensing spectral dimension probability of each pixel being the target crop in the remote sensing spectral dimension is determined; Based on the height dimension probability and remote sensing spectrum dimension probability, determine whether each pixel is the target crop; Crop mapping is determined based on the pixels corresponding to the target crop; crop mapping is used to characterize the classification of target crops and non-target crops.

[0005] A crop mapping determination method provided by the present invention determines the crop remote sensing spectral index corresponding to each pixel, including: determining the maximum value of the bimodal canopy nitrogen index corresponding to each pixel and the maximum value of the red edge position index corresponding to each pixel; normalizing the maximum value of the bimodal canopy nitrogen index to obtain the nitrogen index probability of each pixel belonging to the target crop on the bimodal canopy nitrogen index; normalizing the maximum value of the red edge position index to obtain the red edge index probability of each pixel belonging to the target crop on the red edge position index; and determining the remote sensing spectral index of each crop based on the nitrogen index probability and the red edge index probability.

[0006] According to a crop mapping determination method provided by the present invention, based on the height dimension probability and the remote sensing spectral dimension probability, whether each pixel is a target crop is determined, including: determining the height Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the height dimension of the cultivated land area image; determining the spectral Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the remote sensing spectral dimension of the cultivated land area image; based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, determining a first penalty weight for the pixel being the target crop in the remote sensing spectral dimension; based on the first penalty weight, determining a second penalty weight for the pixel being the target crop in the height dimension; based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectral dimension probability, determining whether each pixel is a target crop.

[0007] According to a crop mapping determination method provided by the present invention, the target probability of a pixel being a target crop is determined based on a first penalty weight, a second penalty weight, a height dimension probability and a remote sensing spectral dimension probability, including: determining a remote sensing spectral dimension value based on the first penalty weight and the remote sensing spectral dimension probability; determining a height dimension value based on the second penalty weight and the height dimension probability; and determining the target probability of a pixel being a target crop based on the height dimension value and the remote sensing spectral dimension value.

[0008] A crop mapping determination method provided by the present invention determines whether each pixel is a target crop based on a first penalty weight, a second penalty weight, an altitude dimension probability and a remote sensing spectral dimension probability, including: determining a target probability that the pixel is a target crop based on the first penalty weight, the second penalty weight, the altitude dimension probability and the remote sensing spectral dimension probability; determining whether each pixel is a target crop based on the target probability and a probability threshold.

[0009] According to a crop mapping determination method provided by the present invention, based on the altitude Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, a first penalty weight for a pixel being a target crop in a remote sensing spectral dimension is determined, including: based on the altitude Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, determining the sum of the Bhattacharyya coefficient weights between the altitude dimension and the remote sensing spectral dimension; based on the Bhattacharyya coefficient weight and the spectral Bhattacharyya coefficient, determining the first penalty weight.

[0010] According to a crop mapping determination method provided by the present invention, after determining the crop mapping based on pixels corresponding to the target crop, the method further includes: determining a classification accuracy evaluation index; and evaluating the accuracy of the crop mapping based on the classification accuracy evaluation index.

[0011] The present invention also provides a crop mapping determination device, comprising the following modules: The first determination module is used to determine the relative height of the crop corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel; The second determination module is used to determine the height dimension probability of each pixel being a target crop in the height dimension based on the relative height of the crop corresponding to each pixel; The third determination module is used to determine the remote sensing spectral dimension probability that each pixel is a target crop in the remote sensing spectral dimension based on the crop remote sensing spectral index corresponding to each pixel; A fourth determination module is used to determine whether each pixel is a target crop based on the height dimension probability and the remote sensing spectrum dimension probability; The fifth determination module is used to determine the crop mapping based on the pixels corresponding to the target crop; the crop mapping is used to characterize the classification of the target crop and the non-target crop.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the crop mapping determination methods described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the crop mapping determination methods described above.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the crop mapping determination method as described in any one of the above is implemented.

[0015] The crop mapping determination method, device, electronic device and storage medium provided by the present invention determine the height dimension probability that the pixel is a target crop in the height dimension by using the crop relative height corresponding to the pixel, determine the remote sensing spectral dimension probability that the pixel is a target crop in the remote sensing spectral dimension by using the crop remote sensing spectral index corresponding to the pixel, and determine whether the crop is a target crop based on the height dimension probability and the remote sensing spectral dimension probability. In this way, the two dimensional probabilities are determined by using two characteristic indices, namely the crop relative height and the crop remote sensing spectral index, and whether the crop is a target crop is comprehensively determined based on the two dimensional probabilities, thereby improving the accuracy of judging the target crop, thereby improving the accuracy and precision of classifying the target crop with other crops. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is one of the flow charts of the crop mapping determination method provided by the present invention.

[0018] Figure 2 This is the second flow chart of the crop mapping determination method provided by the present invention.

[0019] Figure 3 It is a schematic diagram of the corresponding relationship between the growth time of various crops and the bimodal canopy nitrogen index provided by the present invention.

[0020] Figure 4 It is a schematic diagram of the bimodal canopy nitrogen index of various crops provided by the present invention.

[0021] Figure 5 This is a schematic diagram of corn spatial classification provided by the present invention.

[0022] Figure 6 It is a structural schematic diagram of the crop mapping determination device provided by the present invention.

[0023] Figure 7 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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 are within the scope of protection of the present invention.

[0025] Combine the following Figure 1-Figure 5 The crop mapping determination method of the present invention is described. The determination method provided by the present invention is applicable to the classification of any crop. The execution subject of the method can be an electronic device or a determination device set in the electronic device. The determination device can be implemented by software, hardware or a combination of both.

[0026] Figure 1 is one of the flow charts of the crop mapping determination method provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 101: Determine the crop relative height corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel.

[0027] Here, the relative height of the crop corresponding to each pixel can be inferred by a height estimation model, wherein the height estimation model can be trained using the spectral characteristics of the crop and the corresponding bidirectional height difference labels. Specifically, the initial area image is input into the height estimation model, and the relative height of the crop corresponding to each pixel in the cultivated area image is directly output.

[0028] The crop remote sensing spectral index corresponding to each pixel can be comprehensively determined based on multiple indices in the remote sensing spectral image.

[0029] It should be noted that each pixel in the cultivated land area image corresponds to a crop.

[0030] Step 102: Based on the relative height of the crop corresponding to each pixel, determine the height dimension probability that each pixel is the target crop in the height dimension.

[0031] Here, the high-dimensional probability may be determined by any suitable method, for example, using a two-component Gaussian mixture model to determine the high-dimensional probability; or using a neural network model to determine the high-dimensional probability.

[0032] For example, first, a two-component Gaussian mixture model is established by programming software, and the established two-component Gaussian mixture model is defined by the following formula (1): (1)

[0033] Among them, P (GMM) is the established two-component Gaussian mixture model, α and β are the weights of the target crops and non-target crops in the Gaussian mixture model, and the sum of α and β is 1. and are the mean and variance of the Gaussian mixture model of non-target crops, and are the mean and variance of the Gaussian mixture model of the target crop.

[0034] Next, the crop relative height corresponding to each pixel is imported into the established two-component Gaussian mixture model for fitting, and the corresponding parameters in the two-component Gaussian mixture model are obtained. Finally, the height dimension probability of each pixel being the target crop in the height dimension is calculated. The calculation is as follows formula (2): ) (2)

[0036] in, is the probability density function of the non-target Gaussian mixture model, is the weight value of the non-target crop relative height dimension in the two-component Gaussian mixture model, and are the mean and variance of the non-target crop component in the height dimension of the two-component Gaussian mixture model is the Gaussian mixture model probability density function of the target crop, is the weight value of the target crop relative height dimension in the two-component Gaussian mixture model, and are the mean and variance of the target crop component in the height dimension of the two-component Gaussian mixture model, respectively.

[0037] Step 103: Based on the crop remote sensing spectral index corresponding to each pixel, determine the remote sensing spectral dimension probability that each pixel is a target crop in the remote sensing spectral dimension.

[0038] Here, the remote sensing spectrum dimension probability can be determined by any appropriate method, for example, using a two-component Gaussian mixture model to determine the remote sensing spectrum dimension probability; or using a neural network model to determine the remote sensing spectrum dimension probability.

[0039] The establishment of the two-component Gaussian mixture model here can refer to the above formula (1), which will not be repeated here.

[0040] Then, the crop remote sensing spectral index corresponding to each pixel is imported into the established two-component Gaussian mixture model for fitting, and the corresponding parameters in the two-component Gaussian mixture model are obtained. Finally, the remote sensing spectral dimension probability of each pixel being the target crop in the remote sensing spectral dimension is calculated. The calculation is as follows formula (3): (3)

[0041] in, is the probability density function of the two-component Gaussian mixture model of non-target crops in the remote sensing spectral dimension, is the weight value of non-target crops in the two-component Gaussian mixture model in the remote sensing spectral dimension, and are the mean and variance of the target crop component of the two-component Gaussian mixture model in the remote sensing spectral dimension, is the probability density function of the two-component Gaussian mixture model of the target crop in the remote sensing spectral dimension, is the weight value of non-target crops in the two-component Gaussian mixture model in the remote sensing spectral dimension, and They are the mean and variance of the target crop components of the two-component Gaussian mixture model in the remote sensing spectral dimension.

[0042] Step 104: Determine whether each pixel is a target crop based on the height dimension probability and the remote sensing spectrum dimension probability.

[0043] Here, the method for determining whether each pixel is a target crop can be determined by comparing the probability that the pixel is a target crop with a probability threshold, wherein the probability that the pixel is a target crop can be determined based on the height probability and the remote sensing spectrum dimension probability; or the height dimension probability and the remote sensing spectrum dimension probability can be input into a neural network model to determine whether each pixel is a target crop. The present invention is not limited to this.

[0044] Step 105: determine crop mapping based on the pixels corresponding to the target crop; the crop mapping is used to characterize the classification of the target crop and the non-target crop.

[0045] It should be noted that crop mapping is essentially a binary image. In the cultivated land area, the value of the pixel where the target crop is located is marked as 1, and the value of the pixel where the non-target crop is located is marked as 0.

[0046] In an embodiment of the present invention, the relative height of the crop corresponding to the pixel is used to determine the height dimension probability that the pixel is the target crop in the height dimension, and the remote sensing spectral index of the crop corresponding to the pixel is used to determine the remote sensing spectral dimension probability that the pixel is the target crop in the remote sensing spectral dimension. Based on the height dimension probability and the remote sensing spectral dimension probability, it is determined whether the crop is the target crop. In this way, the two dimensional probabilities are determined using two characteristic indices, namely the relative height of the crop and the remote sensing spectral index of the crop. Based on the two dimensional probabilities, it is comprehensively determined whether the crop is the target crop, thereby improving the accuracy of judging the target crop, thereby improving the accuracy and precision of classifying the target crop with other crops.

[0047] In the embodiment of the present invention, determining the crop remote sensing spectral index corresponding to each pixel includes: Determine the maximum value of the bimodal canopy nitrogen index corresponding to each pixel and the maximum value of the red edge position index corresponding to each pixel.

[0048] Here, the bimodal canopy nitrogen index represents the nitrogen content during the peak growth period of the crop, and the red edge position index represents the chlorophyll content during the peak growth period of the crop. Among them, the maximum value of the bimodal canopy nitrogen index refers to the maximum bimodal canopy nitrogen index measured by the pixel over a period of time, and the maximum value of the red edge position index refers to the maximum red edge position index measured by the pixel over a period of time.

[0049] Maximum value of bimodal canopy nitrogen index The calculation of is as follows formula (4): (4)

[0050] in, For red edge 3 band, For the red edge 2 band, For the red edge 1 band, is the red band; It is the exponential maximum synthesis operation.

[0051] Red Edge Position Index max, REP max ) is calculated as follows: (5)

[0052] The maximum value of the bimodal canopy nitrogen index was normalized to obtain the nitrogen index probability of each pixel belonging to the target crop on the bimodal canopy nitrogen index.

[0053] Here, normalization processing refers to scaling the data so that it falls into a small specific interval, usually [0, 1]. The normalization processing includes but is not limited to minimum-maximum normalization, logarithmic normalization, inverse cotangent normalization, etc.

[0054] Specifically, the nitrogen index probability The calculation of is as follows formula (6): (6)

[0055] Among them, DCNI max is the maximum value of the bimodal canopy nitrogen index corresponding to a pixel, DCNI max，UQ For DCNI max The upper quartile of DCNI max The upper quartile refers to the value at the 75% position of the maximum bimodal canopy nitrogen index of all pixels sorted from small to large.

[0056] The maximum value of the red edge position index is normalized to obtain the red edge index probability of each pixel being the target crop at the red edge position index.

[0057] Here, the red edge exponential probability The calculation can be as follows formula (7): (7)

[0058] in, is the red edge index probability corresponding to a pixel, For REP max The lower quartile of REP max,UQ For REP max The upper quartile of For REP max The lower quartile of refers to the value at the 25% position of the maximum red edge index probability of all pixels sorted from small to large; REP max The upper quartile refers to the value at the 75% position of the maximum red edge index probability of all pixels sorted from small to large.

[0059] Based on the nitrogen index probability and the red edge index probability, the crop remote sensing spectral index corresponding to each pixel is determined.

[0060] Here, the method for determining the crop remote sensing spectral index corresponding to each pixel can be the first product, for example, the first product can be the product of the nitrogen index probability and the red edge index probability; it can also be a weighted summation, for example, the nitrogen index probability and the red edge index probability are weighted summation.

[0061] For example, the crop remote sensing spectral index The calculation of can be as follows formula (8): (8)

[0062] Among them, crop remote sensing spectral index It can also be the probability of the target crop in the remote sensing spectral dimension, is the probability that the bimodal canopy nitrogen index belongs to the target crop during the peak growth period, is the probability that the red edge position index belongs to the target crop during the peak growth period.

[0063] In an embodiment of the present invention, the maximum value of the bimodal canopy nitrogen index and the maximum value of the red edge position index are used to determine the crop remote sensing spectral index. In this way, the confusion between the target crop and the spectrally similar crops is reduced through the unique characteristics of the crop nitrogen content and chlorophyll content, thereby improving the accuracy of determining the target crop.

[0064] In the embodiment of the present invention, determining whether each pixel is a target crop based on the height dimension probability and the remote sensing spectrum dimension probability includes: Determine the height Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the height dimension of the cultivated area image.

[0065] Here, the Bhattacharyya coefficient is used to measure the similarity between two probability distributions, and is particularly suitable for Gaussian distribution. The larger the Bhattacharyya coefficient, the greater the similarity between the two probability distributions; conversely, the smaller the similarity. In fitting a two-component Gaussian mixture model using multiple sets of data, by minimizing the Bhattacharyya coefficient, the target crops and non-target crops can be effectively distinguished, thereby improving the performance of the classification model.

[0066] It should be noted that in the process of fitting the two-component Gaussian mixture model using the relative heights of multiple groups of crops, the mean of the Gaussian mixture model of each component can be obtained. and variance Then, using the mean of each component Gaussian mixture model and variance Calculate the Bhattacharyya (BC) coefficient between the probability distributions corresponding to the two components. Here, the Bhattacharyya BC coefficient is calculated as follows: (9)

[0067] Wherein, BD represents Bhattacharyya distance (BD), and BD can be calculated as follows: (10)

[0068] in, and are the mean and variance of the Gaussian mixture model of non-target crops, and are the mean and variance of the Gaussian mixture model of the target crop, and ln represents the logarithmic sign.

[0069] Determine the spectral Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the remote sensing spectral dimension of the cultivated land area image.

[0070] Here, the calculation of the spectral Barthel coefficient can refer to the above formulas (9) and (10).

[0071] It should be noted that the mean and variance of the Gaussian mixture model of each component crop in different dimensions are different, that is, the mean and variance are determined based on the imported data. If the input is the relative height of the crop, the mean and variance are determined based on the relative height of the crop; if the input is the remote sensing spectral index, the mean and variance are determined based on the remote sensing spectral index. The means and variances determined in the two dimensions may be the same or different.

[0072] Based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, the first penalty weight of the pixel for the target crop in the remote sensing spectral dimension is determined.

[0073] Here, the method for determining the first penalty weight can be the ratio of the first difference to the second difference, the first difference can be the difference between the fixed value and the remote sensing spectral dimension probability, and the second difference can be determined based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient; it can also be determined by weighted summation.

[0074] The first penalty weight indicates the importance of the remote sensing spectrum dimension, wherein the larger the first penalty weight is, the greater the importance of the remote sensing spectrum dimension is, and vice versa, the smaller the importance of the remote sensing spectrum dimension is.

[0075] Based on the first penalty weight, a second penalty weight for the pixel being the target crop in the height dimension is determined.

[0076] The second penalty weight indicates the importance of the height dimension, wherein the larger the second penalty weight is, the greater the importance of the height dimension is, and vice versa.

[0077] It should be noted that the sum of the first penalty weight and the second penalty weight is 1.

[0078] Based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectral dimension probability, it is determined whether each pixel is a target crop.

[0079] In an embodiment of the invention, the altitude Bhattacharyya coefficient in the altitude dimension and the spectral Bhattacharyya coefficient in the remote sensing spectrum dimension are used to determine the first penalty weight for the target crop in the remote sensing spectrum dimension, thereby improving the accuracy of the first penalty weight. Then, the altitude dimension probability and the remote sensing spectrum dimension probability and their corresponding penalty weights are used to determine whether the pixel is the target crop. The target crop is comprehensively judged through two dimensions, thereby improving the accuracy of judging the target crop.

[0080] In the embodiment of the present invention, based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, determining the first penalty weight of a pixel being a target crop in the remote sensing spectral dimension includes: Based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, the weight sum of the Bhattacharyya coefficients between the height dimension and the remote sensing spectral dimension is determined.

[0081] Here, the weighted sum of the Bhattacharyya coefficients may be the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, or may be the weighted sum of the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient.

[0082] Based on the Bhattacharyya coefficient weight and the spectral Bhattacharyya coefficient, a first penalty weight is determined.

[0083] Here, the first penalty weight can be the ratio of the first difference to the second difference. The first difference can be the difference between the first fixed constant and the spectral Bhattacharyya coefficient, or the difference between the first fixed constant and the weighted sum of the spectral Bhattacharyya coefficient; the second difference can be the difference between the second fixed constant and the weighted sum of the Bhattacharyya coefficients, or the difference between the second fixed constant and the weighted sum of the Bhattacharyya coefficients. Generally, the first fixed constant is 1 and the second fixed constant is 2.

[0084] Exemplarily, the first penalty weight The calculation of is as follows formula (11): (11)

[0085] in, is the spectral Barthel coefficient, It is the height Bhattacharyya coefficient.

[0086] In an embodiment of the present invention, the height Bhattacharyya coefficient in the height dimension and the spectral Bhattacharyya coefficient in the remote sensing spectral dimension are used to determine the first penalty weight for the target crop in the remote sensing spectral dimension, thereby reducing spectral confusion between the target crop and highly similar crops such as non-target crops in the spectral dimension, improving the ability to distinguish the target crop from crops with significant height differences such as non-target crops in the height dimension, and improving the accuracy of the first penalty weight.

[0087] In the embodiment of the present invention, determining whether each pixel is a target crop based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectrum dimension probability includes: Based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectral dimension probability, the target probability of the pixel being the target crop is determined.

[0088] It should be noted that the greater the target probability, the greater the possibility that the pixel is the target crop; conversely, the smaller the possibility that the pixel is the target crop.

[0089] Based on the target probability and the probability threshold, it is determined whether each pixel is the target crop.

[0090] Here, the probability threshold may be any appropriate value, such as 0.4 or 0.5.

[0091] Exemplarily, the crops corresponding to pixels with a target probability greater than or equal to 0.5 are regarded as target crops, and the crops corresponding to pixels with a target probability less than 0.5 are regarded as non-target crops.

[0092] In the embodiment of the present invention, the target probability that a pixel is a target crop is comprehensively determined by utilizing the height dimension probability and the remote sensing spectral dimension probability and their corresponding penalty weights, thereby improving the accuracy of the target probability. Moreover, based on the comparison between the target probability and the threshold, it is determined whether the pixel is a target crop, thereby improving the accuracy of determining the target crop.

[0093] In the embodiment of the present invention, determining the target probability that a pixel is a target crop based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectrum dimension probability includes: Based on the first penalty weight and the remote sensing spectrum dimension probability, a remote sensing spectrum dimension value is determined.

[0094] Here, the high remote sensing spectrum dimension value may be a first product, wherein the first product may be a product of a first penalty weight and a remote sensing spectrum dimension probability, or may be a weighted product of the first penalty weight and a weighted product of the remote sensing spectrum dimension probability.

[0095] Based on the second penalty weight and the height dimension probability, a height dimension value is determined.

[0096] Here, the height dimension value may be the second product, wherein the second product may be the product of the second penalty weight and the height dimension probability, or may be the weighted product of the second penalty weight and the weighted product of the remote sensing spectrum dimension probability.

[0097] Based on the height dimension value and the remote sensing spectral dimension value, the target probability of the pixel being the target crop is determined.

[0098] Here, the target probability may be a third product, wherein the third product may be a product of a height dimension value and a remote sensing spectrum dimension value, or may be a weighted product of a height dimension value and a weighted product of a remote sensing spectrum dimension value.

[0099] For example, the target probability of a pixel being a target crop is The calculation of is as follows formula (12): (12)

[0100] in, is the probability that each pixel is the target crop, is the second penalty weight of the remote sensing spectral dimension, is the second penalty weight of the height dimension.

[0101] In the embodiment of the present invention, the target probability that a pixel is a target crop is comprehensively determined by utilizing the height dimension probability and the remote sensing spectrum dimension probability and their corresponding penalty weights, thereby improving the accuracy of the target probability and further improving the accuracy of determining the target crop.

[0102] In the embodiment of the present invention, after determining the crop mapping based on the pixels corresponding to the target crop, the method further includes: Determine the classification accuracy evaluation index; It should be noted that the classification map is a binary image. In the cultivated land area, the value of the pixel where the target crop is located is marked as 1, and the value of the pixel where the non-target crop is located is marked as 0.

[0103] The accuracy of crop mapping is evaluated based on the classification accuracy evaluation index.

[0104] Here, the classification accuracy evaluation indicators include but are not limited to Overall Accuracy (OA), User's Accuracy (UA), Producer's Accuracy (PA) and Cohen's Kappa (Kappa).

[0105] Specifically, the overall accuracy OA represents the ratio of the number of correctly classified samples to the total number of samples, and is calculated as follows: (13)

[0106] Among them, TP is the true positive (TP), that is, the number of samples correctly classified as the positive category, TN is the true negative (TN), that is, the number of samples correctly classified as the negative category, FP is the false positive (FP), that is, the number of samples that are actually negative but mistakenly classified as positive, and FN is the false negative (FN), that is, the number of samples that are actually positive but mistakenly classified as negative.

[0107] User accuracy UA represents the probability of being classified as a certain category among samples that are actually of a certain category. The calculation formula is as follows: (14)

[0108] in, is the user precision of category i, is a true example of category i, is a false negative example of class i.

[0109] Mapping accuracy PA represents the probability that a sample predicted by the model to be a certain category actually belongs to that category. The calculation formula is as follows (15): (15)

[0110] in, is the cartographic accuracy of category i, is a true example of category i, is a false positive example of class i.

[0111] The Kappa number is a statistic that measures the performance of a classifier. It takes into account the randomness of the classification and is calculated using the following formula (16): (16)

[0112] in, The overall accuracy obtained by random classification is, where, The calculation formula is as follows: Formula (17): (17)

[0113] In the embodiment of the present invention, the precision of the classification map of target crops and non-target crops is comprehensively estimated by using multi-dimensional evaluation indicators, thereby improving the accuracy of the classification map.

[0114] The following is an application scenario of the crop mapping determination method provided by an embodiment of the present invention.

[0115] It should be noted that this application scenario uses the crop mapping determination method provided by the present invention to determine corn and other crops in the cultivated land area. Therefore, the target crop in this application scenario is corn.

[0116] Figure 2 FIG. 2 is a flow chart of the crop mapping determination method provided by the present invention. Figure 2 As shown, the method includes: Step 201: Determine the relative height of crops at each pixel in the cultivated land area using a height estimation model.

[0117] Step 202: Using a two-component Gaussian mixture model, based on the relative heights of the crops, determine the height-dimensional probability that each pixel is corn in the height dimension.

[0118] Step 203: Determine the crop remote sensing spectral index corresponding to each pixel based on the bimodal canopy nitrogen index and the red edge position index.

[0119] Figure 3 Schematic diagram of the corresponding relationship between the growth time of various crops and the bimodal canopy nitrogen index provided by the present invention, such as Figure 3 As shown, the horizontal axis is the crop generation time, and the vertical axis is the bimodal canopy nitrogen index of different crops. The different horizontal line segments in the coordinate axis represent the bimodal canopy nitrogen index of different crops at different growth times. For example, segment A represents the correspondence between the growth time of corn and the bimodal canopy nitrogen index, segment B represents the correspondence between the growth time of potato and the bimodal canopy nitrogen index, segment C represents the correspondence between the growth time of wheat and the bimodal canopy nitrogen index, segments D, E, F, and G represent the correspondence between the growth time of sunflower and the bimodal canopy nitrogen index, respectively, and segment E represents the correspondence between the growth time of sorghum, wheat, soybean and rice and the bimodal canopy nitrogen index.

[0120] Depend on Figure 3 It can be seen that during the same growth period, the bimodal canopy nitrogen index of corn is the largest.

[0121] Figure 4 Schematic diagram of the bimodal canopy nitrogen index of various crops provided by the present invention, such as Figure 4 As shown in the figure, the horizontal axis is the bimodal canopy nitrogen index, and the vertical axis is different crops, such as wheat, potato, sorghum, sunflower, rice, soybean and corn. It can be seen that the bimodal canopy nitrogen index of corn is the largest, and the bimodal canopy nitrogen index of rice is the smallest.

[0122] Step 204: using a two-component Gaussian mixture model and based on the crop remote sensing spectral index, determine the remote sensing spectral dimension probability that each pixel is corn in the remote sensing spectral dimension.

[0123] Step 205: using a two-component Gaussian mixture model, determine the height Bhattacharyya coefficient between the probability distribution corresponding to corn and the probability distribution corresponding to non-corn in the height dimension of the cultivated land area image.

[0124] Step 206: using a two-component Gaussian mixture model, determine the spectral Bhattacharyya coefficient between the probability distribution corresponding to corn and the probability distribution corresponding to non-corn in the remote sensing spectral dimension of the cultivated land area image.

[0125] Step 207: Based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, determine the first penalty weight of the pixel being corn in the remote sensing spectral dimension.

[0126] Step 208: Based on the first penalty weight, determine a second penalty weight for the pixel being corn in the height dimension.

[0127] Step 209: Perform weighted summation on the second penalty weight in the height dimension and the probability in the height dimension, and the first penalty weight in the remote sensing spectrum dimension and the probability in the remote sensing spectrum dimension to obtain the target probability of corn.

[0128] Step 210: Create a classification map of corn and non-corn on the cultivated land area image according to the target probability.

[0129] Figure 5 This is a schematic diagram of corn spatial classification provided by the present invention, such as Figure 5 As shown, the black area represents the corn area, and the white area represents the non-corn area.

[0130] Step 211: Evaluate the accuracy of the classification graph based on the classification accuracy evaluation index.

[0131] The present invention uses the maximum value synthesis and probability combination of the bimodal canopy nitrogen index and the red edge position index to determine the DRI index for corn mapping, inputs the crop relative height and the DRI index as independent variables into the height-spectral Gaussian mixture model, and uses the Bhattacharyya coefficient as the penalty weight for height-spectral probability purification, thereby achieving accurate corn mapping on a large regional scale; the present invention integrates the unique characteristics of corn physiological height, plant nitrogen content, and chlorophyll content, reduces the height confusion between corn and spectrally similar crops, and achieves spectral separation of corn and highly similar crops. In addition, the present invention reduces the dependence of traditional sample-driven mapping methods on samples, solves the bottleneck problem that traditional sample-driven methods and single feature index methods cannot improve the spatiotemporal mobility of corn mapping, and has broad potential for large-region corn mapping in the country and even the world.

[0132] The crop mapping determination device provided by the present invention is described below. The determination device described below and the determination method described above can be referenced to each other.

[0133] Figure 6is a schematic diagram of the structure of the crop mapping determination device provided by the present invention, referring to Figure 6 As shown, the determination device 600 includes: a first determination module 601, a second determination module 602, a third determination module 603, a fourth determination module 604 and a fifth determination module 605, wherein: The first determination module 601 is used to determine the relative height of the crop corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel; A second determination module 602 is used to determine the height dimension probability of each pixel being a target crop in the height dimension based on the relative height of the crop corresponding to each pixel; The third determination module 603 is used to determine the remote sensing spectral dimension probability that each pixel is the target crop in the remote sensing spectral dimension based on the crop remote sensing spectral index corresponding to each pixel; A fourth determination module 604 is used to determine whether each pixel is the target crop based on the height dimension probability and the remote sensing spectrum dimension probability; The fifth determination module 605 is used to determine the crop map based on the pixels corresponding to the target crop; the crop map is used to characterize the classification of the target crop and the non-target crop.

[0134] In an exemplary embodiment, the first determination module 601 is specifically used to determine the maximum value of the bimodal canopy nitrogen index corresponding to each of the pixels and the maximum value of the red edge position index corresponding to each of the pixels; normalize the maximum value of the bimodal canopy nitrogen index to obtain the nitrogen index probability that each of the pixels belongs to the target crop on the bimodal canopy nitrogen index; normalize the maximum value of the red edge position index to obtain the red edge index probability that each of the pixels belongs to the target crop on the red edge position index; and determine the crop remote sensing spectral index corresponding to each of the pixels based on the nitrogen index probability and the red edge index probability.

[0135] In an exemplary embodiment, the fourth determination module 604 is specifically used to: determine the height Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the height dimension of the cultivated land area image; determine the spectral Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the remote sensing spectral dimension of the cultivated land area image; based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, determine the first penalty weight for the pixel to be the target crop in the remote sensing spectral dimension; based on the first penalty weight, determine the second penalty weight for the pixel to be the target crop in the height dimension; based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectral dimension probability, determine whether each of the pixels is the target crop.

[0136] In an exemplary embodiment, the fourth determination module 604 is further specifically used to: determine the sum of the Bhattacharyya coefficient weights between the altitude dimension and the remote sensing spectral dimension based on the altitude Bhattacharyya coefficient and the spectral Bhattacharyya coefficient; determine the first penalty weight based on the Bhattacharyya coefficient weight and the spectral Bhattacharyya coefficient.

[0137] In an exemplary embodiment, the fourth determination module 604 is further specifically used to: determine the target probability that the pixel is the target crop based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectral dimension probability; and determine whether each of the pixels is a target crop based on the target probability and a probability threshold.

[0138] In an exemplary embodiment, the fourth determination module 604 is further specifically used to: determine a remote sensing spectral dimension value based on the first penalty weight and the remote sensing spectral dimension probability; determine a height dimension value based on the second penalty weight and the height dimension probability; and determine the target probability that the pixel is the target crop based on the height dimension value and the remote sensing spectral dimension value.

[0139] In an exemplary embodiment, the crop mapping determination device 600 further includes an evaluation module, which is specifically used to: determine a classification accuracy evaluation index; and evaluate the accuracy of the crop mapping based on the classification accuracy evaluation index.

[0140] Figure 7 is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the determination method, which includes: determining the relative height of the crop corresponding to each pixel in the cultivated land area image and the crop remote sensing spectrum index corresponding to each pixel; determining the height dimension probability of each pixel being the target crop in the height dimension based on the relative height of the crop corresponding to each pixel; determining the remote sensing spectrum dimension probability of each pixel being the target crop in the remote sensing spectrum dimension based on the crop remote sensing spectrum index corresponding to each pixel; determining whether each pixel is the target crop based on the height dimension probability and the remote sensing spectrum dimension probability; determining crop mapping based on the pixels corresponding to the target crop; and crop mapping is used to characterize the classification of target crops and non-target crops.

[0141] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0142] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the determination method provided by the above-mentioned methods, which includes: determining the relative height of the crop corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel; based on the relative height of the crop corresponding to each pixel, determining the height dimension probability that each pixel is a target crop in the height dimension; based on the crop remote sensing spectral index corresponding to each pixel, determining the remote sensing spectral dimension probability that each pixel is a target crop in the remote sensing spectral dimension; based on the height dimension probability and the remote sensing spectral dimension probability, determining whether each pixel is a target crop; based on the pixels corresponding to the target crop, determining crop mapping; crop mapping is used to characterize the classification of target crops and non-target crops.

[0143] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the determination method provided by the above-mentioned methods, the method comprising: determining the relative height of the crop corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel; determining the height dimension probability that each pixel is a target crop in the height dimension based on the relative height of the crop corresponding to each pixel; determining the remote sensing spectral dimension probability that each pixel is a target crop in the remote sensing spectral dimension based on the crop remote sensing spectral index corresponding to each pixel; determining whether each pixel is a target crop based on the height dimension probability and the remote sensing spectral dimension probability; determining crop mapping based on the pixels corresponding to the target crop; crop mapping is used to characterize the classification of target crops and non-target crops.

[0144] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0145] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A crop mapping determination method, characterized in that: include: Determine the relative height of the crop corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel; Based on the crop height and crop relative height corresponding to each pixel, determining the height dimension probability of each pixel being the target crop in the height dimension; Based on the crop remote sensing spectral index corresponding to each pixel, determining the remote sensing spectral dimension probability that each pixel is the target crop in the remote sensing spectral dimension; Determining whether each of the pixels is the target crop based on the height dimension probability and the remote sensing spectrum dimension probability; The crop mapping is determined based on the pixels corresponding to the target crop; the crop mapping is used to characterize the classification of the target crop and non-target crops.

2. The crop mapping determination method according to claim 1, characterized in that: Determining the crop remote sensing spectral index corresponding to each of the pixels comprises: Determine the maximum value of the bimodal canopy nitrogen index corresponding to each pixel and the maximum value of the red edge position index corresponding to each pixel; Normalizing the maximum value of the bimodal canopy nitrogen index to obtain the nitrogen index probability of each pixel belonging to the target crop on the bimodal canopy nitrogen index; Normalizing the maximum value of the red edge position index to obtain a red edge index probability of each pixel being the target crop at the red edge position index; Based on the nitrogen index probability and the red edge index probability, the crop remote sensing spectral index corresponding to each of the pixels is determined.

3. The crop mapping determination method according to claim 1 or 2, characterized in that: The determining whether each pixel is the target crop based on the height dimension probability and the remote sensing spectrum dimension probability includes: Determine the height Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the height dimension of the cultivated land area image; Determine the spectral Bhattacharyya coefficient between the probability distribution corresponding to the target crop and the probability distribution corresponding to the non-target crop in the remote sensing spectral dimension of the cultivated land area image; Based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, determining a first penalty weight of the pixel for the target crop in the remote sensing spectral dimension; Based on the first penalty weight, determining a second penalty weight for the pixel being the target crop in a height dimension; Based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectrum dimension probability, it is determined whether each of the pixels is the target crop.

4. The crop mapping determination method according to claim 3, characterized in that: The determining, based on the height Bhattacharyya coefficient and the spectral Bhattacharyya coefficient, a first penalty weight of the pixel for the target crop in the remote sensing spectral dimension comprises: Determine a weight sum of the Bhattacharyya coefficients between the altitude dimension and the remote sensing spectral dimension based on the altitude Bhattacharyya coefficient and the spectral Bhattacharyya coefficient; The first penalty weight is determined based on the Bhattacharyya coefficient weight and the spectral Bhattacharyya coefficient.

5. The crop mapping determination method according to claim 3, characterized in that: The determining whether each pixel is the target crop based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectrum dimension probability comprises: Determining a target probability that the pixel is the target crop based on the first penalty weight, the second penalty weight, the height dimension probability, and the remote sensing spectrum dimension probability; Based on the target probability and the probability threshold, it is determined whether each pixel is a target crop.

6. The crop mapping determination method according to claim 4, characterized in that: The determining the target probability that the pixel is the target crop based on the first penalty weight, the second penalty weight, the height dimension probability and the remote sensing spectrum dimension probability includes: Determining a remote sensing spectrum dimension value based on the first penalty weight and the remote sensing spectrum dimension probability; Determining a height dimension value based on the second penalty weight and the height dimension probability; Based on the height dimension value and the remote sensing spectrum dimension value, the target probability of the pixel being the target crop is determined.

7. The crop mapping determination method according to claim 1 or 2, characterized in that: After determining the crop mapping based on the pixels corresponding to the target crop, the method further includes: Determine the classification accuracy evaluation index; Based on the classification accuracy evaluation index, the accuracy of the crop mapping is evaluated.

8. A crop mapping determination device, characterized in that: include: The first determination module is used to determine the relative height of the crop corresponding to each pixel in the cultivated land area image and the crop remote sensing spectral index corresponding to each pixel; A second determination module is used to determine the height dimension probability of each pixel being a target crop in the height dimension based on the relative height of the crop corresponding to each pixel; A third determination module is used to determine the remote sensing spectral dimension probability that each pixel is the target crop in the remote sensing spectral dimension based on the crop remote sensing spectral index corresponding to each pixel; A fourth determination module, used for determining whether each pixel is the target crop based on the height dimension probability and the remote sensing spectrum dimension probability; a fifth determination module, configured to determine the crop mapping based on pixels corresponding to the target crop; The crop mapping is used to characterize the classification of the target crops and non-target crops.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the crop mapping determination method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the crop mapping determination method according to any one of claims 1 to 7 is implemented.

Images