Cotton yield prediction method, device and equipment based on cotton pixel segmentation and texture features
Through drone imaging technology and deep learning model, the pixel ratio and texture features in cotton images are extracted, which solves the problems of high labor intensity and environmental impact in cotton yield prediction, and achieves high-precision cotton yield prediction.
Patent Information
- Application Number
- CN202411234421.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-03
- Publication Date
- 2024-11-22
AI Technical Summary
The prior art has high labor intensity and environmental factors have a great impact on cotton yield prediction, making it difficult to achieve accurate prediction, especially in large-scale cotton planting environments.
Using a method based on cotton pixel segmentation and texture features, cotton image data is collected through drone imaging technology, cotton pixel ratio is obtained using a trained semantic segmentation model, and contrast, energy, correlation and uniformity are extracted through grayscale symbiosis matrix. Features, combined with ResNet50 feature extraction backbone network and CA attention mechanism, a cotton yield prediction model is constructed.
The accuracy and efficiency of cotton yield prediction have been improved, and accurate yield prediction in large-scale cotton planting environments have been achieved. The average absolute percentage error has dropped to 9.25%, significantly improving the accuracy of existing methods.
Smart Images

Figure CN119007048A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of cotton yield prediction, and in particular relates to a cotton yield prediction method, device and equipment based on cotton pixel segmentation and texture features. Background Art
[0002] Chinese cotton plays a pivotal role in the global cotton industry and has important economic value. Therefore, rapid and accurate prediction of cotton yield in complex backgrounds is of great significance for cotton planting management, cotton trade and policy making. At present, in order to increase cotton production, a dense planting pattern of "short, early and dense" with alternating narrow and wide rows is generally adopted. This pattern has obvious advantages in preventing natural disasters and increasing production, but the narrow rows of cotton plants are intertwined and densely grown, which brings many challenges to cotton image segmentation and cotton yield prediction.
[0003] At present, the traditional cotton growth monitoring and yield prediction are based on artificial subjective consciousness and planting experience to determine the growth status and yield of cotton. When faced with large-scale cotton growth monitoring and yield prediction, the artificial method seems to be inadequate. Compared with the traditional cotton yield prediction, remote sensing technology has certain advantages in precision agriculture. This technology uses electromagnetic bands to collect cotton canopy spectral information and converts it into physiological and biochemical information. However, the image resolution obtained by satellite remote sensing technology is not high and cannot be used for refined yield prediction operations; at the same time, this technology is greatly affected by clouds. If it encounters cloudy weather, it will cause image quality to deteriorate or fail to obtain images, which seriously affects the accuracy of yield prediction. With the continuous iteration of drone remote sensing technology, drone low-altitude remote sensing imaging technology has occupied an important position in cotton yield prediction. Compared with satellite remote sensing technology, drone remote sensing technology can make up for the shortcomings of satellite remote sensing technology to a certain extent, and can obtain ultra-high-resolution digital images. At the same time, drones can be equipped with sensors with various functions for specific image shooting.
[0004] Therefore, it is urgent to propose a cotton yield prediction method, device and equipment based on cotton pixel segmentation and texture features of UAV imaging technology. Summary of the invention
[0005] In order to solve the above technical problems, the present invention proposes a cotton yield prediction method, device and equipment based on cotton pixel segmentation and texture features to solve the problems existing in the above-mentioned prior art.
[0006] To achieve the above object, the present invention provides a cotton yield prediction method based on cotton pixel segmentation and texture features, comprising the following steps:
[0007] Collect cotton image data;
[0008] Inputting the cotton image data into a trained semantic segmentation model based on cotton image pixels, segmenting the number of cotton pixels in the cotton image from the total number of image pixels through the semantic segmentation model based on cotton image pixels, and obtaining a cotton pixel ratio;
[0009] Extracting texture features in cotton images based on gray-level co-occurrence matrix, wherein the texture features include contrast, energy, correlation and uniformity;
[0010] Based on the cotton yield prediction model, the prediction result of cotton yield is obtained according to the cotton pixel ratio and the texture feature.
[0011] Optionally, before inputting the cotton image data into a trained semantic segmentation model based on cotton image pixels, the method further includes: performing brightness enhancement, noise blurring, flipping, rotation and color space conversion on the cotton image.
[0012] Optionally, the construction process of the semantic segmentation model based on cotton image pixels includes: taking the semantic segmentation model as the original model, replacing the CBR downsampling feature extraction backbone network in the original model with a ResNet50 feature extraction backbone network, introducing a CA attention mechanism after the five feature layers output by the feature extraction backbone network, and replacing the ReLU activation function in the original model with a HardSwish activation function to obtain the semantic segmentation model based on cotton image pixels.
[0013] Optionally, the ResNet50 feature extraction backbone network includes a convolutional residual block and a skip residual block.
[0014] Optionally, the formula of the HardSwish activation function is:
[0015]
[0016] Optionally, the formula for the cotton pixel ratio is:
[0017]
[0018] Among them, P r is the cotton pixel ratio, P c is the number of cotton pixels in the image, P t is the total number of pixels in the image.
[0019] Optionally, the formulas for extracting contrast, energy, correlation and uniformity in the cotton image based on the gray level co-occurrence matrix are:
[0020] Contrast=∑ i ∑ j (i,j)2 P(i,j)
[0021]
[0022] Asm=∑ i ∑ j P(i,j) 2
[0023] Where i and j represent gray levels; P(i,j) represents the probability of gray levels i and j appearing simultaneously; μ x and μ y are the means of gray levels i and j respectively; σ x and σ y are the standard deviations of gray levels i and j, respectively.
[0024] Optionally, the cotton yield prediction model is:
[0025] y=ω1x1+ω2x2+ω3x3+ω4x4+ω5x5+ψ,
[0026] Among them, x1-x5 are pixel ratio, contrast, energy, correlation and uniformity respectively, w1-w5 are the weights of pixel ratio, contrast, energy, correlation and uniformity estimation factors respectively, and ψ is the error factor.
[0027] The present invention also provides a cotton yield prediction device based on cotton pixel segmentation and texture features, which is used to implement a cotton yield prediction method based on cotton pixel segmentation and texture features, comprising: an image receiving module, a first modeling module, a first processing module, a second processing module, a second modeling module and a yield determination module;
[0028] The image receiving module is used to receive the collected cotton image data;
[0029] The first modeling module is used to build a semantic segmentation model based on cotton image pixels based on the ResNet50 network, the CA attention mechanism and the HardSwish activation function;
[0030] The first processing module is used to input the cotton image data into the trained cotton image pixel-based semantic segmentation model to obtain a cotton pixel ratio;
[0031] The second processing module is used to extract texture features in the cotton image based on the gray level co-occurrence matrix, and the texture features include contrast, energy, correlation and uniformity;
[0032] The second modeling module is used to construct a cotton yield prediction model based on the cotton pixel ratio obtained by the first processing module and the contrast, energy, correlation and uniformity obtained by the second processing module;
[0033] The yield determination module is used to determine the cotton yield based on the cotton yield prediction model constructed by the second modeling module.
[0034] Optionally, the image receiving module is used to communicate with a radar of the UAV.
[0035] Compared with the prior art, the present invention has the following advantages and technical effects:
[0036] The present invention first collects cotton image data; inputs the cotton image data into a trained semantic segmentation model based on cotton image pixels, and divides the number of cotton pixels in the cotton image from the total number of image pixels through the semantic segmentation model based on cotton image pixels to obtain the cotton pixel ratio; extracts texture features in the cotton image based on the grayscale co-occurrence matrix, and the texture features include contrast, energy, correlation and uniformity; based on the cotton yield prediction model, obtains the prediction result of cotton yield according to the cotton pixel ratio and the texture features. It can be seen that the present invention aims at the problems of high labor intensity and environmental impact in cotton yield prediction, and predicts cotton yield through a semantic segmentation model based on cotton image pixels and a texture feature method. When facing large-scale cotton yield prediction problems, accurate prediction of cotton yield is achieved, and the accuracy and efficiency of cotton yield prediction can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0038] Figure 1 It is a flow chart of a cotton yield prediction method based on cotton pixel segmentation and texture features according to an embodiment of the present invention;
[0039] Figure 2 This is a schematic diagram of the backbone feature extraction network structure of the cotton pixel segmentation model according to an embodiment of the present invention;
[0040] Figure 3 Schematic diagram of the ResNet50 structure of an embodiment of the present invention, wherein (a) is a schematic diagram of a convolutional residual block, and (b) is a schematic diagram of a skip residual block;
[0041] Figure 4 Schematic diagram of the structure of the CA attention mechanism module of an embodiment of the present invention;
[0042] Figure 5 This is a schematic diagram of comparing the cotton yield prediction accuracy according to an embodiment of the present invention;
[0043] Figure 6This is a schematic diagram of the structure of a cotton yield prediction device according to an embodiment of the present invention;
[0044] Figure 7 Schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0045] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0046] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0047] In the related art, when cotton matures, the open cotton bolls and other objects in the background, such as cotton branches and leaves, have obvious differences in color and shape, which makes the later image processing less difficult. Moreover, when cotton reaches maturity, the production characteristics of its cotton bolls are relatively stable. At this time, the present invention makes it easier to extract the characteristics of cotton itself through computer vision and deep learning technology.
[0048] Figure 1 1 is a flow chart of a cotton yield prediction method based on cotton pixel segmentation and texture features. In this embodiment, a cotton yield prediction method based on cotton pixel segmentation and texture features is provided, comprising the following steps:
[0049] S10, cotton image data collection:
[0050] Use drones to collect data on field cotton after defoliation.
[0051] S20, data preprocessing:
[0052] Includes: data enhancement and data set division;
[0053] The original image is brightness enhanced, noise blurred, flipped and rotated respectively; at the same time, the RGB color space is converted to CMYK color space.
[0054] The cotton image dataset is divided into training set, validation set and test set according to 7:2:1.
[0055] Figure 2 : is a schematic diagram of the backbone feature extraction network structure of the cotton pixel segmentation model in an embodiment of the present invention. Figure 2 As shown, the method includes:
[0056] S30, establish a cotton pixel segmentation model: including three stages: feature extraction backbone network ResNet50, feature fusion and activation function;
[0057] Figure 3 is a schematic diagram of the structure of ResNet50 in the feature extraction backbone network in an embodiment of the present invention, such as Figure 3 As shown, including:
[0058] Feature extraction backbone network stage: The original CBR downsampling backbone feature extraction network of the semantic segmentation model (UNet) is replaced with the ResNet50 feature extraction backbone network to optimize the model's local and global feature extraction capabilities and improve accuracy.
[0059] Figure 4 : is a schematic diagram of the structure of the CA attention mechanism module in an embodiment of the present invention. Figure 4 As shown, including:
[0060] Feature fusion stage: The CA attention mechanism is introduced after the five feature layers output by the feature extraction backbone network to increase the receptive field of the model and improve the segmentation accuracy of the model.
[0061] Activation function stage: Use the HardSwish activation function to replace the original ReLU activation function. Through truncation and linear scaling operations, the model is made more robust and the gradient propagation efficiency in the model training and reasoning process is improved.
[0062] Furthermore, in S30, the feature extraction backbone network is composed of a convolutional residual block and a jump residual block, wherein the convolutional residual block is used to change the network dimension, and the jump residual block is used to deepen the network.
[0063] Further, in S30, the activation function adopts the HardSwish activation function, and the calculation method of the HardSwish activation function is:
[0064]
[0065] Furthermore, in S30, the cotton pixel segmentation model uses UNet as the basic linear model.
[0066] Furthermore, the performance of the cotton pixel segmentation model after optimization in S30 was compared with that before optimization, and the three models on the cotton independent test set were analyzed by TP, TN, FP and FN. TP means that the true category of the sample is cotton, and the model correctly identifies it; TN means that the true category of the sample is background, and the model correctly identifies it; FP means that the true category of the sample is background, but the model misidentifies it; FN means that the true category of the sample is cotton, but the model misidentifies it.
[0067] S40, segmentation model training:
[0068] The Adam optimizer is used in model training; the initial learning rate is set to 10 -4 , the minimum learning rate is set to 10 -6 The learning rate decrease method is cosine decrease. The training rounds are set to 100, of which 50 rounds are frozen training; the batch size is 64 when the model is frozen, and the batch size is 32 when the model is unfrozen.
[0069] S50, cotton image texture feature extraction:
[0070] GLCM is used to extract four texture features of cotton images, namely contrast, energy, correlation and uniformity, as yield prediction factors of the cotton yield prediction model.
[0071] Further, in S50, the gray level co-occurrence matrix (GLCM) is used to extract four texture features of contrast, energy, correlation, and uniformity in the cotton image as yield prediction factors of the cotton yield prediction model. The calculation method of contrast, energy, correlation, and uniformity is:
[0072]
[0073] Asm=∑ i ∑ j P(i,j) 2
[0074] Where i and j represent gray levels; P(i,j) represents the probability of gray levels i and j appearing simultaneously; μ x and μ y are the means of gray levels i and j respectively; σ x and σ y are the standard deviations of gray levels i and j, respectively.
[0075] S60, establish cotton yield prediction model:
[0076] A multi-factor yield prediction model was constructed based on the cotton pixel ratio extracted by the cotton pixel segmentation model and the five eigenvalues of cotton image contrast, energy, correlation and uniformity extracted by GLCM with the help of ridge regression analysis.
[0077] Further, in S60, the cotton pixel ratio, contrast, energy, correlation and uniformity are used as yield estimation factors to establish a cotton yield prediction model using ridge regression analysis, wherein the cotton pixel ratio is calculated as follows:
[0078]
[0079] Where P r is the cotton pixel ratio, P c is the number of cotton pixels in the image, Pt is the total number of pixels in the image.
[0080] Furthermore, in S60, the cotton yield prediction model is:
[0081] y=ω1x1+ω2x2+ω3x3+ω4x4+ω5x5+ψ
[0082] Among them, x1-x5 are pixel ratio, contrast, energy, correlation and uniformity respectively, w1-w5 are the weights of pixel ratio, contrast, energy, correlation and uniformity estimation factors respectively, and ψ is the error factor.
[0083] Figure 5 A schematic diagram for comparing the cotton yield prediction accuracy provided by an embodiment of the present invention is shown in FIG. Figure 5 As shown, Figure 5 The middle grid column is the actual yield, and the oblique line column is the predicted yield obtained by the cotton yield prediction method provided by the embodiment of the present invention. It can be observed from the bar graph that the average absolute percentage error of the cotton yield prediction method proposed by the present invention is 9.25%. It shows that the cotton yield prediction method provided by the present invention can significantly improve the cotton yield prediction accuracy.
[0084] In summary, the solution of the embodiment of the present invention achieved an average pixel accuracy of 96.22% and a precision of 96.30% on an independent test data set for cotton yield prediction, and the average absolute percentage error between the predicted yield and the actual yield was 9.254%, which is better than the existing method. The model proposed by the present invention has more flexible sampling capabilities and excellent robustness for the characteristics of cotton images, and can accurately predict the yield of field cotton.
[0085] The present invention aims to solve the problems of high labor intensity and environmental impact in cotton yield prediction, and to predict cotton yield by using a cotton pixel segmentation model and a texture feature method. When facing large-scale cotton yield prediction problems, accurate cotton yield prediction can be achieved, and the accuracy and efficiency of cotton yield prediction can be improved.
[0086] Figure 6 A schematic diagram of the structure of a cotton yield prediction device provided by an embodiment of the present invention is shown in FIG. Figure 6 As shown, the cotton yield prediction device comprises:
[0087] The receiving module 601 is used to receive the field cotton image data of the target area from the client corresponding to each target field cotton area.
[0088] The processing module 602 is used to train a cotton pixel segmentation model, extract cotton image texture features and construct a cotton yield prediction model based on the field cotton image data of each target area.
[0089] The determination module 603 predicts the cotton yield of the target area based on the cotton pixel segmentation model, the image texture features and the cotton yield prediction model.
[0090] As a specific embodiment, the embodiment of the present invention also provides a cotton yield prediction device based on cotton pixel segmentation and texture features, which is used to implement a cotton yield prediction method based on cotton pixel segmentation and texture features, comprising: an image receiving module, a first modeling module, a first processing module, a second processing module, a second modeling module and a yield determination module;
[0091] The image receiving module is used to receive the collected cotton image data;
[0092] The first modeling module is used to build a semantic segmentation model based on cotton image pixels based on the ResNet50 network, the CA attention mechanism and the HardSwish activation function;
[0093] The first processing module is used to input the cotton image data into the trained cotton image pixel-based semantic segmentation model to obtain a cotton pixel ratio;
[0094] The second processing module is used to extract texture features in the cotton image based on the gray level co-occurrence matrix, and the texture features include contrast, energy, correlation and uniformity;
[0095] The second modeling module is used to construct a cotton yield prediction model based on the cotton pixel ratio obtained by the first processing module and the contrast, energy, correlation and uniformity obtained by the second processing module;
[0096] The yield determination module is used to determine the cotton yield based on the cotton yield prediction model constructed by the second modeling module.
[0097] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7As shown, the electronic device may include: a processor (Processor) 701, a communication interface (Communication Interface) 702, a memory (memory) 703 and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 communicate with each other through the communication bus 704. The processor 701 may call the logic instructions in the memory 703 to execute the cotton yield prediction method, which includes: receiving the field cotton image data of each target field cotton area from the client corresponding to the target field cotton area; based on the field cotton image data of each target area, training the cotton pixel segmentation model, extracting the cotton image texture features and constructing the cotton yield prediction model; predicting the cotton yield of the target area based on the cotton pixel segmentation model, the image texture features and the cotton yield prediction model.
[0098] In addition, the logic instructions in the above-mentioned memory 703 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 such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and 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 the embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, ReadOnly Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0099] On the other hand, the present invention also provides a computer program product, which includes a computer program, which 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 cotton yield prediction method provided by the above-mentioned methods, and the method includes: receiving field cotton image data of each target field cotton area from a client corresponding to the target field cotton area; based on the field cotton image data of each target area, training a cotton pixel segmentation model, extracting cotton image texture features and constructing a cotton yield prediction model; predicting the cotton yield of the target area based on the cotton pixel segmentation model, image texture features and cotton yield prediction model.
[0100] 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 cotton yield prediction method provided by the above-mentioned methods, the method comprising: receiving field cotton image data of each target field cotton area from a client corresponding to the target field cotton area; training a cotton pixel segmentation model, extracting cotton image texture features and constructing a cotton yield prediction model based on the field cotton image data of each target area; and predicting the cotton yield of the target area based on the cotton pixel segmentation model, image texture features and cotton yield prediction model.
[0101] The device examples described above are merely illustrative, wherein the units described as separated and not illustrated may or may not be physically separated, and the components shown 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. Those of ordinary skill in the art may understand and implement it without creative work.
[0102] 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, it can also be implemented by hardware. Based on the above 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, CD, etc., including a number of instructions to enable 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.
[0103] 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 cotton yield prediction method based on cotton pixel segmentation and texture features, It is characterized in that The following steps are involved: Collect cotton image data; Inputting the cotton image data into a trained semantic segmentation model based on cotton image pixels, segmenting the number of cotton pixels in the cotton image from the total number of image pixels through the semantic segmentation model based on cotton image pixels, and obtaining a cotton pixel ratio; Extracting texture features in cotton images based on gray-level co-occurrence matrix, wherein the texture features include contrast, energy, correlation and uniformity; Based on the cotton yield prediction model, the prediction result of cotton yield is obtained according to the cotton pixel ratio and the texture feature.
2. The cotton yield prediction method based on cotton pixel segmentation and texture features according to claim 1, It is characterized in that Before the cotton image data is input into the trained semantic segmentation model based on cotton image pixels, the method also includes: performing brightness enhancement, noise blurring, flipping, rotation and color space conversion on the cotton image.
3. The cotton yield prediction method based on cotton pixel segmentation and texture features according to claim 1, It is characterized in that The construction process of the semantic segmentation model based on cotton image pixels includes: taking the semantic segmentation model as the original model, replacing the CBR downsampling feature extraction backbone network in the original model with the ResNet50 feature extraction backbone network, introducing the CA attention mechanism after the five feature layers output by the feature extraction backbone network, and replacing the ReLU activation function in the original model with the HardSwish activation function to obtain the semantic segmentation model based on cotton image pixels.
4. The cotton yield prediction method based on cotton pixel segmentation and texture features according to claim 3, It is characterized in that The ResNet50 feature extraction backbone network includes a convolutional residual block and a skip residual block.
5. The cotton yield prediction method based on cotton pixel segmentation and texture features according to claim 3, It is characterized in that The formula of the HardSwish activation function is:
6. The cotton yield prediction method based on cotton pixel segmentation and texture features according to claim 1, It is characterized in that The formula for the cotton pixel ratio is: Among them, P r is the cotton pixel ratio, P c is the number of cotton pixels in the image, P t is the total number of pixels in the image.
7. The cotton yield prediction method based on cotton pixel segmentation and texture features according to claim 1, It is characterized in that The formulas for extracting contrast, energy, correlation and uniformity in cotton images based on gray-level co-occurrence matrix are: Contrast=∑ i ∑ j (i,j) 2 P(i,j) Asm=∑ i ∑ j P(i,j) 2 Where i and j represent gray levels; P(i,j) represents the probability of gray levels i and j appearing simultaneously; μ x and μ y are the means of gray levels i and j respectively; σ x and σ y are the standard deviations of gray levels i and j, respectively.
8. The cotton yield prediction method based on cotton pixel segmentation and texture features according to claim 1, It is characterized in that The cotton yield prediction model is: y=ω 1 x 1 +oh 2 x 2 +oh 3 x 3 +oh 4 x 4 +oh 5 x 5 +ψ, Among them, x 1 -x 5 They are pixel ratio, contrast, energy, correlation and uniformity respectively, and w 1 -w 5 are the weights of pixel ratio, contrast, energy, correlation and uniformity estimation factors respectively, and ψ is the error factor.
9. A cotton yield prediction device based on cotton pixel segmentation and texture features, It is characterized in that The method for predicting cotton yield based on cotton pixel segmentation and texture features according to any one of claims 1 to 8 comprises: an image receiving module, a first modeling module, a first processing module, a second processing module, a second modeling module and a yield determination module; The image receiving module is used to receive the collected cotton image data; The first modeling module is used to build a semantic segmentation model based on cotton image pixels based on the ResNet50 network, the CA attention mechanism and the HardSwish activation function; The first processing module is used to input the cotton image data into the trained cotton image pixel-based semantic segmentation model to obtain a cotton pixel ratio; The second processing module is used to extract texture features in the cotton image based on the gray level co-occurrence matrix, and the texture features include contrast, energy, correlation and uniformity; The second modeling module is used to construct a cotton yield prediction model based on the cotton pixel ratio obtained by the first processing module and the contrast, energy, correlation and uniformity obtained by the second processing module; The yield determination module is used to determine the cotton yield based on the cotton yield prediction model constructed by the second modeling module.
10. The cotton yield prediction device based on cotton pixel segmentation and texture features according to claim 9, It is characterized in that The image receiving module is used for communication connection with the radar of the UAV.
Citation Information
Cited By
Cucumber yield prediction method and device, electronic equipment and storage medium
CN120181336A
