Wheat intelligent water-saving irrigation method

Abstract

The invention provides an intelligent water-saving irrigation method for wheat. The method comprises the steps that visible light images and hyperspectral images of all growth stages of wheat are obtained; selecting an image of one wave band from the hyperspectral image as an enhanced image; fusing the enhanced image and the visible light image to form a fused image; marking the water demand of the wheat based on the fused image of each growth stage of the wheat to form a training sample; inputting the training sample into a neural network for training to obtain a wheat water demand prediction model; and calculating the water demand of wheat in the target wheat field by using the wheat water demand prediction model, and irrigating the target wheat field according to the water demand. By acquiring the image data of the wheat in each growth stage, the water demand of the wheat can be analyzed and predicted more accurately, it is ensured that the irrigation amount meets the actual demand, waste of water resources is effectively reduced, the growth environment of crops is improved, and therefore the yield and quality of the wheat are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent irrigation, and more specifically, relates to a wheat intelligent water-saving irrigation method. Background Art

[0002] With the intensification of global climate change and the increasing shortage of water resources, agricultural irrigation is facing severe challenges. Wheat, as one of the most important food crops in the world, its production is not only directly related to food security but also closely related to the livelihood of farmers.

[0003] In recent years, with the rapid development of intelligent agricultural technologies, using advanced technical means such as sensors and the Internet of Things for irrigation has gradually become a research hotspot. For example, soil moisture sensors can monitor the soil moisture status in real time, and staff can adjust the irrigation timing and water volume according to practical experience. However, this method is affected by the subjective awareness of staff, prone to problems such as water resource waste and low irrigation efficiency, which have a negative impact on the growth and yield of wheat. Summary of the Invention

[0004] To solve the above problems, the purpose of the present invention is to provide a wheat intelligent water-saving irrigation method.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A wheat intelligent water-saving irrigation method, including:

[0007] Step 1: Obtain visible light images and hyperspectral images of wheat at each growth stage;

[0008] Step 2: Select an image of one band from the hyperspectral images as the enhanced image;

[0009] Step 3: Fuse the enhanced image and the visible light image to form a fused image;

[0010] Step 4: Label the water requirements of wheat based on the fused images of wheat at each growth stage to form training samples;

[0011] Step 5: Input the training samples into a neural network for training to obtain a wheat water requirement prediction model; Step 6: Evaluate the accuracy of the wheat water requirement prediction model. When the accuracy is not within the preset range, select an image of one band from the hyperspectral images again as the enhanced image, and return to Step 3;

[0012] Step 7: When the accuracy is within the preset range, use the wheat water requirement prediction model to calculate the water requirements of wheat in the target wheat field and irrigate the target wheat field according to the water requirements.

[0013] Preferably, in step 3: fusing the enhanced image and the visible light image to form a fused image, including:

[0014] Step 3.1: Using the non-subsampled contourlet transform to decompose the enhanced image and the visible light image respectively to obtain enhanced image transform coefficients and visible light image transform coefficients;

[0015] Step 3.2: Extracting the feature map of the enhanced image;

[0016] Step 3.3: Normalizing the feature map of the enhanced image to obtain a normalized feature map;

[0017] Step 3.4: Calculating the weight coefficient according to the normalized feature map;

[0018] Step 3.5: Using the weight coefficient to weight the enhanced image transform coefficients and the visible light image transform coefficients to form fusion coefficients;

[0019] Step 3.6: Performing an inverse non-subsampled contourlet transform on the fusion coefficients to obtain a fused image.

[0020] Preferably, in step 3.2: extracting the feature map of the enhanced image, including:

[0021] Step 3.2.1: Constructing a window with a point on the enhanced image as the center point;

[0022] Step 3.2.2: Using a non-linear filter to process the pixel points within the window, and continuously sliding the window to form the feature map of the enhanced image; wherein, the expression formula of the non-linear filter is:

[0023]

[0024] wherein, g(i,j) represents the pixel value of the feature map at the point (i,j), S(i,j) is the set of pixel points within the window formed with the point (i,j) as the center, f(k,l) represents the pixel value of the enhanced image at the point (i,j), w(i,j,k,l) represents the filter weight, σ d represents the distance standard deviation, σ r represents the pixel standard deviation.

[0025] Preferably, in step 3.3: normalizing the feature map of the enhanced image to obtain a normalized feature map, including:

[0026] Obtaining the maximum value and the minimum value among all pixel points on the enhanced image feature map;

[0027] Normalize the feature map based on the maximum and minimum values of the pixels in the pixel map to obtain a normalized feature map; where the normalization formula is:

[0028]

[0029] In the formula, S norm (x,y) represents the normalized feature map, and S BF (x,y) represents the original feature map, and S max represents the maximum value among all the pixels on the feature map, and S min represents the minimum value among all the pixels on the feature map.

[0030] Preferably, step 3.4: Calculate the weight coefficient according to the normalized feature map, including:

[0031] Step 3.4.1: Calculate the information amount under each block according to the normalized feature map and the pixel distribution on the visible light image; where the information amount calculation formula is:

[0032]

[0033] Among them, E IR (x,y) represents the information amount of the block centered on (x,y) on the visible light image, and E ID (x,y) represents the information amount of the block centered on (x,y) on the normalized feature map, ID(x+i,y+j) represents the pixel value at the point (x+i,y+j) on the normalized feature map, IR(x+i,y+j) represents the pixel value at the point (x+i,y+j) on the visible light image, and w(i,j) represents the weight matrix;

[0034] Step 3.4.2: Calculate the information amount weight according to the information amount under each block;

[0035] Step 3.4.3: Introduce the normalized feature map to obtain the weight coefficient based on the information amount weight.

[0036] Preferably, in step 3.4.2, calculate the information amount weight according to the proportion of the information amount under each block; where the information amount weight calculation formula is:

[0037]

[0038] Among them, α 1 (x,y) represents the first information amount weight, and α 2 (x,y) represents the second information amount weight.

[0039] Preferably, step 3.4.3: Introduce the normalized feature map to obtain the weight coefficient based on the information amount weight, including:

[0040] Adopt the formula:

[0041]

[0042] Calculate the weight coefficient; where, w IR (x,y) represents the first weight coefficient, w ID (x,y) represents the second weight coefficient, α 1 ′(x,y) = S norm (x,y) + α 1 (x,y), α 1 ′(x,y) min represents the minimum value in α 1 ′(x,y), α 1 ′(x,y) max represents the maximum value in α 1 ′(x,y), α 2 ′(x,y) = 1 - S norm (x,y) + α 2 (x,y), α 2 ′(x,y) min represents the minimum value in α 2 ′(x,y), α 2 ′(x,y) max represents the maximum value in α 2 ′(x,y).

[0043] Preferably, in the step 3.5, based on the first weight coefficient and the second weight coefficient, perform weighted summation on the visible light image transformation coefficient and the enhanced image transformation coefficient respectively to form a fusion coefficient; where, the weighted summation formula is:

[0044] IF coef (x,y) = w IR (x,y)IR coef (x,y) + w ID (x,y)ID coef (x,y)

[0045] where, IR coef (x,y) represents the transformation coefficient corresponding to the (x,y) point on the visible light image, ID coef (x,y) represents the transformation coefficient corresponding to the (x,y) point on the enhanced image, IF coef (x,y) represents the fusion coefficient.

[0046] The present invention also provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored on the memory and executable on the processor. The transceiver, the memory, and the processor are connected through the bus. It is characterized in that when the computer program is executed by the processor, the steps in the above-mentioned wheat intelligent water-saving irrigation method are realized.

[0047] The present invention also provides a computer-readable storage medium, on which a computer program is stored. It is characterized in that when the computer program is executed by a processor, the steps in the above-mentioned wheat intelligent water-saving irrigation method are realized.

[0048] The beneficial effect of the wheat intelligent water-saving irrigation method provided by the present invention lies in that: compared with the prior art, by acquiring the image data of wheat at each growth stage, the present invention can analyze and predict the water requirement of wheat more accurately, not only ensuring that the irrigation amount meets the actual demand and effectively reducing the waste of water resources, but also helping to improve the growth environment of crops, thereby improving the yield and quality of wheat. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1 It is a flowchart of a wheat intelligent water-saving irrigation method provided by an embodiment of the present invention;

[0051] Figure 2 It is the relationship between the hyperspectral data of wheat at the jointing stage and the water content of wheat leaves provided by an embodiment of the present invention; the horizontal axis represents the wavelength, and the vertical axis represents the correlation coefficient;

[0052] Figure 3 It is the relationship between the hyperspectral data of wheat at the filling stage and the water content of wheat leaves provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] In order to make the technical problems, technical solutions, and beneficial effects to be solved by the present invention clearer, the following further details the present invention with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] Please refer to Figure 1 , a wheat intelligent water-saving irrigation method, including:

[0055] Step 1: Obtain visible light images and hyperspectral images of wheat at various growth stages;

[0056] Step 2: Select an image of a single band from the hyperspectral images as the enhanced image;

[0057] Step 3: Fuse the enhanced image and the visible light image to form a fused image;

[0058] Further, Step 3 includes:

[0059] Step 3.1: Use the non - subsampled contourlet transform to decompose the enhanced image and the visible light image respectively to obtain enhanced image transform coefficients and visible light image transform coefficients;

[0060] Step 3.2: Extract the feature map of the enhanced image;

[0061] Among them, Step 3.2 includes:

[0062] Step 3.2.1: Construct a window with a point on the enhanced image as the center point;

[0063] Step 3.2.2: Use a non - linear filter to process the pixel points within the window, and continuously slide the window to form the feature map of the enhanced image; among them, the expression formula of the non - linear filter is:

[0064]

[0065] Among them, g(i,j) represents the pixel value of the feature map at the point (i,j), S(i,j) is the set of pixel points within the window formed with the point (i,j) as the center, f(k,l) represents the pixel value of the enhanced image at the point (i,j), w(i,j,k,l) represents the filter weight, σ d represents the distance standard deviation, σ r represents the pixel standard deviation.

[0066] The non - linear filter of the present invention will consider the spatial position and pixel value of the pixels when processing the image. Based on this, the processing of the enhanced image can remove the noise or background interference in the enhanced image and enhance the biological characteristics of wheat, such as the texture, shape, and color changes of the plant. This has a positive effect on analyzing the health status, maturity, etc. of the plant.

[0067] Step 3.3: Perform normalization processing on the feature map of the enhanced image to obtain a normalized feature map;

[0068] In Step 3.3, the present invention can complete the normalization process based on the maximum - minimum normalization algorithm, and its specific process is as follows:

[0069] Step 3.3.1: Obtain the maximum and minimum values among all pixel points on the enhanced image feature map;

[0070] Step 3.3.2: Perform normalization processing on the feature map based on the maximum and minimum values among the pixel points to obtain a normalized feature map; where the normalization processing formula is:

[0071]

[0072] In the formula, S norm (x, y) represents the normalized feature map, and S BF (x, y) represents the original feature map, and S max represents the maximum value among all pixel points on the feature map, and S min represents the minimum value among all pixel points on the feature map.

[0073] Step 3.4: Calculate the weight coefficient according to the normalized feature map;

[0074] Further, Step 3.4 includes:

[0075] Step 3.4.1: Calculate the information amount under each block according to the normalized feature map and the pixel point distribution on the visible light image; where the information amount calculation formula is:

[0076]

[0077] Among them, E IR (x, y) represents the information amount of the block centered on (x, y) on the visible light image, and E ID (x, y) represents the information amount of the block centered on (x, y) on the normalized feature map, ID(x + i, y + j) represents the pixel value at the point (x + i, y + j) on the normalized feature map, IR(x + i, y + j) represents the pixel value at the point (x + i, y + j) on the visible light image, and w(i, j) represents the weight matrix;

[0078] Step 3.4.2: Calculate the information amount weight according to the information amount under each block;

[0079] In the said Step 3.4.2, the present invention calculates the information amount weight according to the proportion of the information amount under each block; where the information amount weight calculation formula is:

[0080]

[0081] Among them, α 1 (x, y) represents the first information amount weight, and α 2 (x, y) represents the second information amount weight.

[0082] Step 3.4.3: Introduce the normalized feature map to obtain the weight coefficients based on the information quantity weights. The calculation formula for the weight coefficients is as follows:

[0083]

[0084] where w IR (x,y) represents the first weight coefficient, w ID (x,y) represents the second weight coefficient, α 1 ′(x,y) = S norm (x,y) + α 1 (x,y), α 1 ′(x,y) min represents the minimum value in α 1 ′(x,y), α 1 ′(x,y) max represents the maximum value in α 1 ′(x,y), α 2 ′(x,y) = 1 - S norm (x,y) + α 2 (x,y), α 2 ′(x,y) min represents the minimum value in α 2 ′(x,y), α 2 ′(x,y) max represents the maximum value in α 2 ′(x,y).

[0085] Step 3.5: Use the weight coefficients to weight the enhanced image transformation coefficients and the visible light image transformation coefficients to form the fusion coefficients;

[0086] In the step 3.5, weight and sum the visible light image transformation coefficients and the enhanced image transformation coefficients respectively based on the first weight coefficient and the second weight coefficient to form the fusion coefficients; where the weighted summation formula is:

[0087] IF coef (x,y) = w IR (x,y)IR coef (x,y) + w ID (x,y)ID coef (x,y)

[0088] where IR coef (x,y) represents the transformation coefficient corresponding to the point (x,y) on the visible light image, ID coef (x,y) represents the transformation coefficient corresponding to the point (x,y) on the enhanced image, IF coef (x,y) represents the fusion coefficient.

[0089] Step 3.6: Perform inverse non-subsampled contourlet transform on the fusion coefficients to obtain the fused image.

[0090] Since the amount of information contained in each block of the image is not the same, the present invention needs to minimize the influence of background information as much as possible, and take the places with high information content as the places where the hyperspectral image needs to be fused, so as to improve the feature expression ability of the subsequent fused image in the neural network.

[0091] Step 4: Label the water requirements of wheat based on the fused images of each growth stage of wheat to form training samples;

[0092] In practical applications, the humidity values required for each growth stage of wheat can be labeled.

[0093] Step 5: Input the training samples into the neural network for training to obtain a wheat water requirement prediction model;

[0094] In Step 5, the training samples need to be divided into a training set, a validation set, and a test set. The common ratio is 70% for training, 15% for validation, and 15% for testing. Then, they are input into the CNN model for training, and the loss function is used to evaluate the training effect in real time. When the value of the loss function falls within the preset range, the training is completed to obtain a wheat water requirement prediction model.

[0095] Step 6: Evaluate the accuracy of the wheat water requirement prediction model. When the accuracy is not within the preset range, select an image of a band from the hyperspectral image as the enhanced image again, and return to Step 3;

[0096] Step 7: When the accuracy is within the preset range, use the wheat water requirement prediction model to calculate the water requirement of wheat in the target wheat field, and irrigate the target wheat field according to the water requirement.

[0097] As Figures 2-3 shown, the correlation between the leaf water content of wheat and the acquisition band of hyperspectral is a parabolic relationship. Therefore, fusing the hyperspectral image as the enhanced image into the visible light image can reflect the water supply required by wheat at each growth stage.

[0098] The visible light image and the hyperspectral image provide different types of information. The visible light image can reflect the growth state and health status of wheat, while the hyperspectral image can provide detailed information about the leaf water content of wheat. Fusing the two can obtain more comprehensive features, enabling the neural network to learn features at different levels and scales, so as to better capture the subtle differences in the water status changes of wheat at each growth stage. Through real-time analysis of this type of fused data, it can provide timely information support for agricultural management, help farmers make scientific and reasonable irrigation decisions, and optimize water resource utilization.

[0099] The present invention also provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored on the memory and executable on the processor. The transceiver, the memory, and the processor are connected through the bus. When the computer program is executed by the processor, the steps in the above-mentioned wheat intelligent water-saving irrigation method are implemented, and the same technical effects can be achieved. To avoid repetition, it will not be elaborated here.

[0100] The present invention also provides a computer-readable storage medium, on which a computer program is stored. It is characterized in that when the computer program is executed by a processor, the steps in the above-mentioned wheat intelligent water-saving irrigation method are implemented, and the same technical effects can be achieved. To avoid repetition, it will not be elaborated here.

[0101] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A smart water-saving irrigation method for wheat, characterized in that: include: Step 1: Obtain visible light images and hyperspectral images of wheat at various growth stages; Step 2: Select an image of any band from the hyperspectral image as the enhanced image; Step 3: Fusing the enhanced image and the visible light image to form a fused image; Step 4: Label the water requirement of wheat based on the fused images of wheat at various growth stages to form training samples; Step 5: Input the training samples into the neural network for training to obtain the wheat water requirement prediction model; Step 6: Evaluate the accuracy of the wheat water requirement prediction model. If the accuracy is not within the preset range, select a band of images from the hyperspectral image as the enhanced image and return to step 3. Step 7: When the accuracy is within the preset range, the water requirement of wheat in the target wheat field is calculated using the wheat water requirement prediction model, and the target wheat field is irrigated according to the water requirement.

2. A smart water-saving irrigation method for wheat according to claim 1, characterized in that: The step 3: fusing the enhanced image and the visible light image to form a fused image, comprises: Step 3.1: Decomposing the enhanced image and the visible light image by using non-subsampled contourlet transform to obtain enhanced image transformation coefficients and visible light image transformation coefficients; Step 3.2: extracting a feature map of the enhanced image; Step 3.3: normalizing the feature map of the enhanced image to obtain a normalized feature map; Step 3.4: Calculate the weight coefficient based on the normalized feature map; Step 3.5: Using the weight coefficient, weight the enhanced image transformation coefficient and the visible light image transformation coefficient to form a fusion coefficient; Step 3.6: Perform non-subsampled contourlet inverse transform on the fusion coefficients to obtain the fused image.

3. A smart water-saving irrigation method for wheat as claimed in claim 2, characterized in that: The step 3.2: extracting the feature map of the enhanced image comprises: Step 3.2.1: Construct a window with a point on the enhanced image as the center point; Step 3.2.2: Use a nonlinear filter to process the pixels in the window, and continuously slide the window to form a feature map of the enhanced image; wherein the expression formula of the nonlinear filter is: Among them, g(i,j) represents the pixel value of the feature map at point (i,j), S(i,j) is the set of pixel points in the window formed with point (i,j) as the center, f(k,l) represents the pixel value of the enhanced image at point (i,j), w(i,j,k,l) ​​represents the filter weight, σ d represents the distance standard deviation, σ r Represents the pixel standard deviation.

4. A smart water-saving irrigation method for wheat as claimed in claim 3, characterized in that: The step 3.3: normalizing the feature map of the enhanced image to obtain a normalized feature map, comprises: Get the maximum and minimum values ​​of all pixels on the enhanced image feature map; The feature map is normalized based on the maximum and minimum values ​​of the pixels to obtain a normalized feature map; the normalization formula is: In the formula, S norm (x,y) represents the normalized feature map, S BF (x,y) represents the original feature map, S max Represents the maximum value of all pixels on the feature map, S min Represents the minimum value of all pixels on the feature map.

5. A smart water-saving irrigation method for wheat as claimed in claim 4, characterized in that: The step 3.4: calculating the weight coefficient according to the normalized feature map includes: Step 3.4.1: Calculate the amount of information in each block based on the normalized feature map and the pixel distribution on the visible light image; the information amount calculation formula is: Among them, E IR (x, y) represents the amount of information of the block centered at (x, y) on the visible light image, E ID (x,y) represents the amount of information of the block centered at (x,y) on the normalized feature map, ID(x+i,y+j) represents the pixel value of the normalized feature map at point (x+i,y+j), IR(x+i,y+j) represents the pixel value of the visible light image at point (x+i,y+j), and w(i,j) represents the weight matrix; Step 3.4.2: Calculate the information weight according to the information amount of each block; Step 3.4.3: Introduce the normalized feature map to obtain the weight coefficient based on the information weight.

6. A smart water-saving irrigation method for wheat as claimed in claim 5, characterized in that: In step 3.4.2, the information weight is calculated according to the proportion of the information in each block; wherein the information weight calculation formula is: Among them, α1(x, y) represents the first information weight, and α2(x, y) represents the second information weight.

7. A smart water-saving irrigation method for wheat according to claim 6, characterized in that: The step 3.4.3: introducing the normalized feature map to obtain the weight coefficient based on the information weight, includes: Using the formula: Calculate the weight coefficient; where w IR (x,y) represents the first weight coefficient, w ID (x, y) represents the second weight coefficient, α1′(x, y)=S norm (x,y)+α1(x,y),α1′(x,y) min represents the minimum value in α1′(x,y), α1′(x,y) max represents the maximum value in α1′(x,y), α2′(x,y)=1-S norm (x,y)+α2(x,y),α2′(x,y) min represents the minimum value in α2′(x,y), α2′(x,y) max Represents the maximum value in α2′(x,y).

8. A smart water-saving irrigation method for wheat as claimed in claim 7, characterized in that: In step 3.5, based on the first weight coefficient and the second weight coefficient, weighted summation is performed on the visible light image transformation coefficient and the enhanced image transformation coefficient to form a fusion coefficient; wherein the weighted summation formula is: IF coef (x,y)=w IR (x,y)IR coef (x,y)+w ID (x,y)ID coef (x,y) Among them, IR coef (x, y) represents the transformation coefficient corresponding to the point (x, y) on the visible light image, ID coef (x, y) represents the transformation coefficient corresponding to the point (x, y) on the enhanced image. coef (x,y) represents the fusion coefficient.

9. An electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, wherein: When the computer program is executed by the processor, the steps in the intelligent water-saving irrigation method for wheat as described in any one of claims 1-8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps in the intelligent water-saving irrigation method for wheat as described in any one of claims 1-8 are implemented.

Images