Growth irrigation data analysis method and system based on neural network
Through the neural network-based irrigation data analysis method, the problem of insufficient irrigation data analysis accuracy in the prior art was solved. Through the processing of drip irrigation image and humidity information, drip irrigation splash and soil water accumulation were identified, and combined with leaf and rhizome humidity information, the Efficient irrigation quality control for crops such as Phalaenopsis has improved the accuracy of irrigation data analysis and crop growth management.
Patent Information
- Application Number
- CN202411321935.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-23
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-23
AI Technical Summary
The existing irrigation data analysis focuses on large-scale humidity control, and fails to fully consider the subtle needs of crop growth, resulting in insufficient analysis accuracy, especially the root wetness and leaf humidity requirements of some crops such as Phalaenopsis.
Using a neural network-based growth irrigation data analysis method, the data dimensionality reduction process is carried out by collecting drip irrigation images and humidity information in crop areas, identifying drip irrigation splashes and soil water accumulation, and combining leaf and rhizome humidity information for quality analysis to obtain comprehensive drip irrigation quality and correct the information.
It improves the comprehensiveness and accuracy of irrigation data analysis, and can more accurately evaluate the wetness of leaves and rhizomes, ensure uniformity of irrigation quality, optimize irrigation strategies, and improve crop yield and quality.
Smart Images

Figure CN118887545B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of data processing, and in particular to a growth irrigation data analysis method and system based on a neural network. Background Art
[0002] Drip irrigation technology has been widely used in irrigation management of various crops. With its precise water supply and water-saving characteristics, drip irrigation is considered to be an effective way to improve water resource utilization, reduce evaporation and avoid water waste. The analysis of irrigation data plays an important role in irrigation adjustment. At present, when analyzing irrigation data, generally only whether the overall irrigation water meets the crop requirements is considered.
[0003] However, for some crops such as Phalaenopsis, during irrigation, the soil at the roots needs to remain moist but without stagnant water, while the surface humidity of the leaves should be low and without any water, otherwise there may be health problems with crop leaf diseases. Current irrigation data analysis does not take this into account.
[0004] In general, the irrigation data analysis of existing technologies focuses more on large-scale humidity analysis, fails to fully consider the subtle needs of crop growth, and has technical problems such as incomplete irrigation data analysis and insufficient analysis accuracy. Summary of the invention
[0005] The present invention aims at the technical problems that the irrigation data analysis in the prior art mostly focuses on the large-scale humidity control analysis, fails to fully consider the subtle needs of crop growth, and has the problems of imperfect irrigation data analysis and insufficient analysis accuracy.
[0006] In a first aspect, the present invention provides a growth irrigation data analysis method based on a neural network, comprising: during the crop growth irrigation process, collecting drip irrigation images of multiple crop areas, and humidity information near leaves and near rhizomes, to obtain multiple drip irrigation images, multiple leaf humidity information, and multiple rhizome humidity information;
[0007] Performing dimensionality reduction processing with data feature preservation on the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain a reduced-dimensionality leaf moisture information set and a reduced-dimensionality root and stem moisture information set, as well as obtaining distribution uniformity information, and screening the plurality of drip irrigation images according to the crop areas corresponding to the reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information to obtain a reduced-dimensionality drip irrigation image set;
[0008] Based on a neural network, drip irrigation splash recognition and soil water accumulation recognition are respectively performed on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set to obtain a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters;
[0009] Based on the neural network, the drip irrigation leaf quality analysis is performed in combination with the multiple drip irrigation splash parameters and the reduced dimension leaf humidity information set to obtain multiple drip irrigation leaf quality information; the drip irrigation root and stem quality analysis is performed in combination with the multiple soil water accumulation parameters and the reduced dimension root and stem humidity information set to obtain multiple drip irrigation root and stem quality information;
[0010] According to the multiple drip irrigation leaf quality information and the multiple drip irrigation root and stem quality information, comprehensive drip irrigation quality information and drip irrigation quality uniformity information are processed and obtained. The comprehensive drip irrigation quality information is corrected and calculated in combination with the distribution uniformity information to obtain the drip irrigation quality analysis result as the growth irrigation data analysis result.
[0011] In a second aspect, the present invention provides a growth irrigation data analysis system based on a neural network, the system comprising: an irrigation data acquisition module, for collecting drip irrigation images of multiple crop areas and humidity information near leaves and roots and stems during crop growth irrigation, to obtain multiple drip irrigation images, multiple leaf humidity information and multiple root and stem humidity information;
[0012] An irrigation data dimensionality reduction module is used to perform dimensionality reduction processing on the plurality of leaf moisture information and the plurality of root and stem moisture information with data feature preservation, obtain a reduced-dimensionality leaf moisture information set and a reduced-dimensionality root and stem moisture information set, and obtain distribution uniformity information, and screen the plurality of drip irrigation images according to the crop areas corresponding to the reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information to obtain a reduced-dimensionality drip irrigation image set;
[0013] An image recognition module, for performing drip irrigation splash recognition and soil water accumulation recognition on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set based on a neural network, and obtaining a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters;
[0014] A drip irrigation quality analysis module, for performing a drip irrigation leaf quality analysis based on a neural network, combining the multiple drip irrigation splash parameters and the reduced-dimensionality leaf humidity information set, and obtaining multiple drip irrigation leaf quality information; performing a drip irrigation root and stem quality analysis based on the multiple soil water accumulation parameters and the reduced-dimensionality root and stem humidity information set, and obtaining multiple drip irrigation root and stem quality information;
[0015] The comprehensive analysis module is used to process the comprehensive drip irrigation quality information and the drip irrigation quality uniformity information according to the multiple drip irrigation leaf quality information and the multiple drip irrigation root and stem quality information, and to correct and calculate the comprehensive drip irrigation quality information in combination with the distribution uniformity information to obtain the drip irrigation quality analysis result as the growth irrigation data analysis result.
[0016] In a third aspect, the present invention provides an electronic device comprising: a memory for storing a computer software program; and a processor for reading and executing the computer software program, thereby implementing the neural network-based growth irrigation data analysis method of the first aspect.
[0017] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer software program stored therein, and when the computer software program is executed by a processor, the neural network-based growth irrigation data analysis method of the first aspect is implemented.
[0018] The present invention obtains multiple drip irrigation images, multiple leaf moisture information and multiple root and stem moisture information by collecting drip irrigation images of multiple crop areas and moisture information near leaves and rhizomes during crop growth irrigation; performs dimensionality reduction processing on the multiple leaf moisture information and the multiple root and stem moisture information to retain data features, obtains a reduced-dimensionality leaf moisture information set and a reduced-dimensionality root and stem moisture information set, and obtains distribution uniformity information; and screens multiple drip irrigation images according to crop areas corresponding to the reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information to obtain a reduced-dimensionality drip irrigation image set, and improves the retained data features through dimensionality reduction processing to improve analysis efficiency; based on a neural network, multiple reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set are screened to obtain a reduced-dimensionality drip irrigation image set. The irrigation image is respectively subjected to drip irrigation splash recognition and soil water accumulation recognition to obtain multiple drip irrigation splash parameters and multiple soil water accumulation parameters; based on the neural network, the drip irrigation leaf quality analysis is performed in combination with multiple drip irrigation splash parameters and the reduced dimension leaf moisture information set to obtain multiple drip irrigation leaf quality information; the drip irrigation root and stem quality analysis is performed in combination with multiple soil water accumulation parameters and the reduced dimension root and stem moisture information set to obtain multiple drip irrigation root and stem quality information; based on the multiple drip irrigation leaf quality information and the multiple drip irrigation root and stem quality information, the comprehensive drip irrigation quality information and the drip irrigation quality uniformity information are processed to obtain the comprehensive drip irrigation quality information and the drip irrigation quality uniformity information, and the comprehensive drip irrigation quality information is corrected and calculated in combination with the distribution uniformity information to obtain the drip irrigation quality analysis result as the growth irrigation data analysis result. Through the above steps, the present application can analyze the leaf splash wetness and root and stem drip irrigation conditions in the drip irrigation data, and then analyze the overall drip irrigation quality, and at the same time, make corrections in combination with the distribution and drip irrigation uniformity, and consider the overall drip irrigation quality analysis, so as to achieve the technical effect of improving the comprehensiveness and accuracy of the drip irrigation data analysis. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 A schematic diagram of a flow chart of a growth irrigation data analysis method based on a neural network provided in an embodiment of the present invention;
[0020] Figure 2 A schematic diagram of the structure of a growth irrigation data analysis system based on a neural network provided in an embodiment of the present invention;
[0021] Figure 3A functional block diagram of an electronic device provided by an embodiment of the present invention;
[0022] Figure 4 A principle block diagram of the computer-readable storage medium provided by the present invention.
[0023] In the accompanying drawings, the components represented by the reference numerals are listed as follows:
[0024] Irrigation data acquisition module 11, irrigation data dimension reduction module 12, image recognition module 13, drip irrigation quality analysis module 14, comprehensive analysis module 15, electronic device 100, memory 110, processor 120, computer program 111, computer readable storage medium 200, computer program 211. DETAILED DESCRIPTION
[0025] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.
[0026] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.
[0027] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details to obscure the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.
[0028] Example 1
[0029] like Figure 1 As shown, the embodiment of the present invention provides a growth irrigation data analysis method based on a neural network, comprising:
[0030] S10: During the crop growth irrigation process, drip irrigation images of multiple crop areas and humidity information near leaves and near rhizomes are collected to obtain multiple drip irrigation images, multiple leaf humidity information, and multiple rhizome humidity information;
[0031] In the embodiment of the present application, in the irrigation of crops such as Phalaenopsis, drip irrigation is often used for irrigation. The soil near the root of the crop needs to be kept moist but not waterlogged, and the leaves should not be too wet to avoid diseases such as late blight. Based on this, data from crop growth irrigation is collected for analysis, and the crop is preferably Phalaenopsis.
[0032] In order to accurately monitor and analyze the irrigation effect, firstly, drip irrigation images and humidity information are collected in multiple crop areas, such as areas with multiple crop plants.
[0033] For example, cameras and humidity sensors are arranged in each crop area to collect multiple drip irrigation images formed during the irrigation process regularly or in real time. These images record the process of the drip irrigation system delivering water to the root and stem area of the crop, reflecting the distribution of drip irrigation droplets, coverage, and water absorption effect. Drip irrigation images can provide a visual reference for irrigation management and help identify any uneven irrigation phenomena, such as splashing water droplets on leaves or water accumulation in the soil.
[0034] In addition, humidity sensors are placed near the leaves and roots of each crop area to collect environmental humidity information. The collection of leaf humidity information can help determine whether the leaves are affected by too much or too little water, which is an important indicator of crop health. The root and stem humidity information reflects the moisture level in the soil and affects the absorption of water and growth by the crop roots.
[0035] S20: performing dimensionality reduction processing with data feature preservation on the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain a reduced-dimensionality leaf moisture information set and a reduced-dimensionality root and stem moisture information set, and obtaining distribution uniformity information, and screening the plurality of drip irrigation images according to the crop areas corresponding to the reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information to obtain a reduced-dimensionality drip irrigation image set;
[0036] In the embodiment of the present application, before processing and analyzing the data collected during the drip irrigation process of crop growth, in order to improve the efficiency of data processing and analysis and ensure the accuracy of processing and analysis, it is necessary to perform dimensionality reduction processing on multiple leaf moisture information and multiple root and stem moisture information with data feature retention to obtain a reduced-dimensional leaf moisture information set and a reduced-dimensional root and stem moisture information set. In this way, a data set with data distribution characteristics is retained, the amount of data is reduced, and the efficiency of subsequent analysis is improved.
[0037] For example, dimensionality reduction can be achieved through methods such as principal component analysis (PCA) or linear discriminant analysis (LDA), which can retain the most representative humidity features while compressing information and reducing redundant data.
[0038] Among them, data feature analysis and dimension reduction are performed based on leaf moisture information and root and stem moisture information to retain the leaf moisture information and root and stem moisture information of the reduced-dimensional crop area with data features, and obtain the reduced-dimensional leaf moisture information set and the reduced-dimensional root and stem moisture information set. At the same time, according to the distribution of these reduced-dimensional crop areas in multiple crop areas, the uniformity of the distribution area of the characteristic leaf moisture information and root and stem moisture information can be analyzed to obtain the distribution uniformity information.
[0039] After the dimensionality reduction process is completed, the dimensionality reduction crop areas corresponding to the dimensionality reduction leaf wetness information and the dimensionality reduction root and stem wetness information are used to screen the drip irrigation images of the dimensionality reduction crop areas to obtain a dimensionality reduction drip irrigation image set, thereby completing the dimensionality reduction of the drip irrigation images.
[0040] Step S20 in the method provided in the embodiment of the present application includes:
[0041] Calculating and processing the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain reference leaf moisture information and reference root and stem moisture information;
[0042] Calculate the distribution probability according to the difference between other leaf humidity information and rhizome humidity information and the reference leaf humidity information and reference rhizome humidity information, and obtain reference leaf humidity probability distribution and reference rhizome humidity probability distribution;
[0043] According to the benchmark leaf moisture probability distribution and the benchmark rhizome moisture probability distribution, the plurality of leaf moisture information and the plurality of rhizome moisture information are subjected to feature-retaining optimized dimensionality reduction to obtain the reduced dimensionality leaf moisture information set and the reduced dimensionality rhizome moisture information set, and to obtain a plurality of reduced dimensionality crop regions;
[0044] Filtering the plurality of reduced-dimensionality drip irrigation images of the crop areas to obtain a reduced-dimensionality drip irrigation image set;
[0045] The distribution uniformity information is obtained by calculation according to the region coordinates of the multiple reduced-dimensionality crop regions.
[0046] In the embodiment of the present application, the processing of multiple leaf moisture information and multiple root and stem moisture information requires the calculation of baseline information first. Specifically, representative baseline leaf moisture information and baseline root and stem moisture information are extracted from the multiple leaf moisture information and multiple root and stem moisture information. This baseline information represents the overall level of the multiple leaf moisture information and multiple root and stem moisture information, for example, the mean of the multiple leaf moisture information and the multiple root and stem moisture information are calculated respectively to obtain the baseline leaf moisture information and the baseline root and stem moisture information.
[0047] Furthermore, the difference between other leaf moisture information and root moisture information and the baseline leaf moisture information and the baseline root moisture information is calculated. Based on multiple differences, the probability distribution of moisture is calculated by assigning distribution probabilities. Exemplarily, the interval of the distribution probability is (0,1), the minimum difference defines the corresponding distribution probability as 0, and the maximum difference defines the distribution probability as 1. In this way, the distribution probability corresponding to each difference is calculated, and the distribution probability of each leaf moisture information and root moisture information is obtained. Combined with the position sorting of multiple crop areas, the baseline leaf moisture probability distribution and the baseline root moisture probability distribution are constructed. The baseline leaf moisture probability distribution and the baseline root moisture probability distribution include the distribution characteristics of leaf moisture and root moisture in multiple crop areas.
[0048] According to the baseline leaf moisture probability distribution and the baseline root and stem moisture probability distribution, multiple leaf moisture information and multiple root and stem moisture information are subjected to feature-retaining optimized dimensionality reduction to obtain a reduced dimensionality leaf moisture information set and a reduced dimensionality root and stem moisture information set, and the crop regions where these reduced dimensionality leaf moisture information and reduced dimensionality root and stem moisture information are located are obtained to obtain multiple reduced dimensionality crop regions.
[0049] The step of "performing feature-preserving optimization dimensionality reduction on the plurality of leaf moisture information and the plurality of root and stem moisture information according to the benchmark leaf moisture probability distribution and the benchmark root and stem moisture probability distribution" in the method provided in the embodiment of the present application includes:
[0050] According to the dimensionality reduction ratio, randomly select the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain a first dimensionality reduced leaf moisture information set and a first dimensionality reduced root and stem moisture information set;
[0051] According to the reference leaf moisture information and the reference rhizome moisture information, calculate and obtain the first reduced dimension leaf moisture probability distribution and the first reduced dimension rhizome moisture probability distribution of the first reduced dimension leaf moisture information set and the first reduced dimension rhizome moisture information set;
[0052] Calculate the similarity between the first reduced-dimensionality leaf moisture probability distribution and the reference leaf moisture probability distribution to obtain a first leaf moisture feature retention degree; calculate the similarity between the first reduced-dimensionality root and stem moisture probability distribution and the reference root and stem moisture probability distribution to obtain a first root and stem moisture feature retention degree; and combine the first leaf moisture feature retention degree to obtain a first reduced-dimensionality feature retention degree through weighted calculation;
[0053] Continue to perform iterative random dimensionality reduction until the preset number of dimensionality reduction times is reached, output the reduced dimensionality leaf moisture information set and the reduced dimensionality root and stem moisture information set with the largest dimensionality reduction feature retention, and select the crop areas corresponding to the multiple reduced dimensionality leaf moisture information as multiple reduced dimensionality crop areas.
[0054] In the embodiment of the present application, according to a preset dimensionality reduction ratio, some data are randomly selected from multiple leaf moisture information and multiple root and stem moisture information, wherein the leaf moisture information and root and stem moisture information of the same crop region are selected to obtain a first dimensionality reduced leaf moisture information set and a first dimensionality reduced root and stem moisture information set. The first dimensionality reduced leaf moisture information set and the first dimensionality reduced root and stem moisture information set include the same leaf moisture information set root and stem moisture information of multiple crop regions.
[0055] The preset dimensionality reduction ratio can be set based on those skilled in the art, for example, 40%-60%, and can be set according to the computing power of subsequent data processing and the overall number of crops.
[0056] According to the above-mentioned benchmark leaf moisture information and benchmark rhizome moisture information, the differences between the first reduced dimensionality leaf moisture information set and the first reduced dimensionality root moisture information set and the benchmark leaf moisture information and the benchmark rhizome moisture information are calculated, and then the distribution probabilities are assigned to obtain the first reduced dimensionality leaf moisture probability distribution and the first reduced dimensionality root moisture probability distribution of the first reduced dimensionality leaf moisture information set and the first reduced dimensionality root moisture information set.
[0057] Furthermore, the similarity between the first reduced dimensionality leaf moisture probability distribution and the reference leaf moisture probability distribution after random dimensionality reduction is calculated to obtain the first leaf moisture feature retention degree. Based on the same method, the similarity between the first reduced dimensionality root and stem moisture probability distribution and the reference root and stem moisture probability distribution is calculated to obtain the first root and stem moisture feature retention degree. The first reduced dimensionality feature retention degree is obtained by weighted calculation of the first leaf moisture feature retention degree and the first root and stem moisture feature retention degree. Among them, the moisture of the root and stem part is more important for crop growth, and a larger weight is set. For example, the weights of the first leaf moisture feature retention degree and the first root and stem moisture feature retention degree are 0.4 and 0.6.
[0058] For example, the similarity between the first reduced dimensionality leaf wetness probability distribution and the reference leaf wetness probability distribution can be calculated based on methods such as Kullback-Leibler divergence in the prior art to quantify the degree of feature retention of the first reduced dimensionality leaf wetness information set after dimensionality reduction.
[0059] Continue to perform multiple random dimensionality reduction processes in multiple leaf moisture information and multiple root and stem moisture information tables based on the same method and the dimensionality reduction ratio, and recalculate the dimensionality reduction feature retention after each iteration until the preset number of dimensionality reductions is reached, for example, 50 times. Through iterative dimensionality reduction optimization, the dimensionality reduction leaf moisture information set and the dimensionality reduction root and stem moisture information set with the largest dimensionality reduction feature retention are selected as the dimensionality reduction results with the best feature retention. Finally, the dimensionality reduction leaf moisture information set and the dimensionality reduction root and stem moisture information set represent the crop areas that still retain the most moisture features after data dimensionality reduction.
[0060] Furthermore, the crop areas corresponding to the multiple reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information in the reduced-dimensionality leaf moisture information set and the reduced-dimensionality root and stem moisture information set are obtained to obtain multiple reduced-dimensionality crop areas.
[0061] Among the above-mentioned multiple drip irrigation images, multiple drip irrigation images of reduced-dimensionality crop areas are screened to obtain a reduced-dimensionality drip irrigation image set.
[0062] And, the distribution uniformity information is calculated based on the regional coordinates of multiple reduced-dimensional crop regions.
[0063] Among them, ideally, the multiple reduced-dimensionality crop areas obtained by dimensionality reduction should be randomly distributed in multiple crop areas. If they are concentrated in a certain area, it means that the drip irrigation moisture distribution of the overall multiple crop areas is uneven. Therefore, based on the regional coordinates of multiple reduced-dimensionality crop areas, a distribution uniformity analysis is performed to reflect the uniformity level of the overall irrigation data for subsequent analysis.
[0064] The step of "calculating and obtaining the distribution uniformity information according to the regional coordinates of the plurality of reduced-dimensionality crop regions" in the method provided in the embodiment of the present application includes:
[0065] Constructing a crop region coordinate system in the plurality of crop regions, and calculating and obtaining reference point coordinates of the coordinates of the plurality of crop regions;
[0066] According to the crop region coordinate system, obtaining the region coordinates of the plurality of reduced-dimensionality crop regions, and calculating the reduced-dimensionality point coordinates;
[0067] The distribution uniformity information is calculated based on the deviation between the coordinates of the dimension reduction point and the coordinates of the reference point.
[0068] In the embodiment of the present application, in order to accurately evaluate the uniformity of irrigation distribution during crop growth, it is necessary to construct a crop region coordinate system in each crop region. First, the two-dimensional or three-dimensional spatial coordinates of each crop region are determined according to the physical boundaries of the crop region to form a complete coordinate system. This coordinate system provides a spatial reference framework for subsequent data analysis.
[0069] Based on the crop region coordinate system, in each crop region, the region coordinates of each crop region can be indexed, and the center values of the region coordinates of multiple crop regions, that is, the mean, can be further calculated to obtain the reference point coordinates of the coordinates of the multiple crop regions, that is, the coordinates of the centers of the multiple crop regions.
[0070] Based on the crop region coordinate system, the region coordinates of a plurality of reduced-dimensionality crop regions are obtained, and the coordinates of the center point are calculated to obtain the reduced-dimensionality point coordinates.
[0071] The deviation between the coordinates of the reduced-dimension point and the coordinates of the reference point is calculated. Ideally, the coordinates of the reduced-dimension point and the coordinates of the reference point should be the same. If there is a deviation, it means that the distribution of the reduced-dimension crop area with the humidity information of the distribution characteristics is uneven in multiple crop areas. The greater the deviation, the greater the degree of unevenness. The distribution uniformity information is further calculated based on the deviation. For example, the ratio of the deviation to the possible maximum deviation (for example, the maximum difference between the coordinates of multiple crop areas and the coordinates of the reference point) is calculated, and 1-the ratio is used as the distribution uniformity information.
[0072] The embodiment of the present application reduces the dimensionality of multiple leaf moisture information, multiple root and stem moisture information, and multiple drip irrigation images of multiple crop areas, and calculates the distribution uniformity based on the distribution of the multiple reduced-dimensional crop areas after dimensionality reduction. It is possible to obtain reduced-dimensional data that retains the drip irrigation data characteristics of multiple crop areas, analyze the distribution uniformity of the overall drip irrigation data, reflect the uniformity level of the overall drip irrigation data, conduct subsequent analysis, and improve analysis efficiency and comprehensiveness.
[0073] S30: Based on a neural network, performing drip irrigation splash recognition and soil water accumulation recognition on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set, respectively, to obtain a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters;
[0074] Based on a neural network, drip irrigation splash recognition and soil water accumulation recognition are respectively performed on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set to obtain a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters.
[0075] The reduced-dimensional drip irrigation image includes the water contact conditions of the leaf area and rhizome area of the crop plant during the drip irrigation process. Based on the neural network, the reduced-dimensional drip irrigation image is respectively subjected to drip irrigation splash recognition and soil water accumulation recognition. It can analyze the water contact of the leaves and the water accumulation of the roots during the drip irrigation process, and then analyze the drip irrigation quality. Ideally, the leaves need to have as little contact with water as possible to avoid diseases, and the roots need to avoid water accumulation to improve the drip irrigation effect.
[0076] Among them, each reduced-dimensionality drip irrigation image in the reduced-dimensionality drip irrigation image set is recognized and processed through a convolutional neural network model to identify the characteristics of water drop splashing and water accumulation in the image, and obtain multiple drip irrigation splash parameters and multiple soil water accumulation parameters.
[0077] Step S30 in the method provided in the embodiment of the present application includes:
[0078] According to the historical drip irrigation data of crops, a set of sample drip irrigation images is collected. According to the proportion of drip irrigation water splashing on leaves in each sample drip irrigation image and the time of soil water accumulation after dripping, a set of sample drip irrigation splash parameters and a set of sample soil water accumulation parameters are annotated to obtain;
[0079] Based on a convolutional neural network, the sample drip irrigation image set is used as input data, and a sample drip irrigation splash parameter set and a sample soil waterlogging parameter set are used as supervision data, respectively, to train a splash recognition branch and a waterlogging recognition branch, and obtain a drip irrigation image recognizer;
[0080] The drip irrigation image recognizer is used to perform drip irrigation splash recognition and soil water accumulation recognition on the multiple reduced-dimensional drip irrigation images, respectively, to obtain multiple drip irrigation splash parameters and multiple soil water accumulation parameters.
[0081] In an embodiment of the present application, based on the historical drip irrigation data of similar crops, drip irrigation images are collected and standardized to obtain a set of sample drip irrigation images. The sample drip irrigation image includes the proportion of water splashing onto the leaves during drip irrigation, such as 10%, and the range of water accumulation on the soil near the rhizomes, for example. In order to further train the model, these sample images need to be accurately annotated. Specifically, by analyzing the proportion of drip irrigation water splashing on the leaves in each sample image, a set of sample drip irrigation splash parameters is obtained by annotation, and by annotating the range of water accumulation in the soil after dripping in the sample drip irrigation image, a set of sample soil water accumulation parameters is obtained.
[0082] Furthermore, a convolutional neural network (CNN) is used to train the sample drip irrigation image set for recognition. A convolutional neural network is a deep learning algorithm that is particularly good at processing image data. First, a set of labeled sample drip irrigation images is used as input data, and each set of sample drip irrigation splash parameters and sample soil waterlogging parameters are used as supervised data. Here, supervised data refers to the goal of neural network learning, that is, the parameters that the model is expected to predict. In this process, the model will learn the features in the image through multiple iterations, and gradually establish a splash recognition branch and a waterlogging recognition branch. These two branches are used to identify water splash and soil waterlogging in the image, respectively.
[0083] The splash recognition branch learns how to identify the splash ratio of water splashing onto leaves from drip irrigation images. Through a large number of image training, the network can extract the distribution characteristics of water droplets on leaves and convert them into a set of splash parameters. The water accumulation recognition branch learns how to identify the characteristics of water accumulation on the soil surface, such as water surface reflection, water accumulation, etc. During the training process, the network continuously updates parameters to learn how to accurately identify the area range of water accumulation.
[0084] After the network outputs the prediction results, these results are compared with the label data (i.e., the actual splash and water accumulation parameters). The error between the predicted value and the actual value is calculated through the loss function (such as mean square error or cross entropy). The loss function quantifies the quality of the model's prediction. The larger the loss, the less accurate the model prediction. Based on the error, the network parameters are adjusted until the error is reduced to meet the requirements, and the splash recognition branch and the water accumulation recognition branch are trained.
[0085] After training is completed, for example, when the accuracy of the splash recognition branch and the water accumulation recognition branch meets the requirements, such as 95%, the obtained drip irrigation image recognizer can be used. The recognizer can be applied to multiple reduced-dimensional drip irrigation images, and splash recognition and water accumulation recognition are performed on each reduced-dimensional drip irrigation image. In this process, the recognizer will extract the drip irrigation splash phenomenon in each image based on the previously learned features and generate multiple drip irrigation splash parameters. Similarly, the recognizer will detect the water accumulation range from the reduced-dimensional image and generate multiple soil water accumulation parameters.
[0086] Through the above steps, drip irrigation image recognition based on convolutional neural network can effectively improve the accuracy and precision of leaf and rhizome irrigation condition analysis in drip tubes.
[0087] S40: Based on a neural network, combining the multiple drip irrigation splash parameters and the reduced-dimensionality leaf humidity information set, performing a drip irrigation leaf quality analysis to obtain multiple drip irrigation leaf quality information, combining the multiple soil water accumulation parameters and the reduced-dimensionality root and stem humidity information set, performing a drip irrigation root and stem quality analysis to obtain multiple drip irrigation root and stem quality information;
[0088] In the embodiment of the present application, after obtaining multiple drip irrigation splash parameters and multiple soil water accumulation parameters, the reduced-dimensional leaf moisture information set and the reduced-dimensional root and stem moisture information set are respectively combined to analyze the drip irrigation quality of crop leaves and roots and stems. For example, the smaller the drip irrigation splash parameter, the smaller the soil water accumulation parameter, the more the reduced-dimensional leaf moisture information and the reduced-dimensional root and stem moisture information meet the needs of crop growth, and the higher the drip irrigation quality.
[0089] Among them, the neural network model in machine learning is used to analyze the leaf drip irrigation quality based on multiple drip irrigation splash parameters and reduced-dimensional leaf moisture information sets to identify whether drip irrigation has caused excessive moisture, which may lead to leaf diseases (such as mold or lesions), or leaf dryness caused by insufficient moisture. To obtain drip irrigation leaf quality information for drip irrigation in multiple reduced-dimensional crop areas.
[0090] In addition, through the neural network model in machine learning, the root and stem drip irrigation quality analysis is performed based on multiple soil water accumulation parameters and root and stem humidity information sets to identify whether excessive soil water accumulation will cause root suffocation, root and stem rot and other problems in the root and stem area. And identify whether the soil is kept in a suitable moist state to ensure that the crop roots and stems can grow normally. To obtain multiple drip irrigation root and stem quality information for multiple reduced-dimensional crop areas in drip irrigation.
[0091] By combining multiple drip irrigation splash parameters and multiple soil water accumulation parameters obtained from drip irrigation image recognition, as well as the reduced-dimensional leaf moisture information set and reduced-dimensional root and stem moisture information set obtained from humidity monitoring, the drip irrigation quality analysis of leaves and rhizomes can be performed. This can improve the accuracy and comprehensiveness of drip irrigation quality analysis, thereby improving the accuracy of irrigation data analysis.
[0092] Step S40 in the method provided in the embodiment of the present application includes:
[0093] According to the crop historical drip irrigation monitoring data, the sample drip irrigation splash parameter set and the sample leaf humidity information set are collected, and the sample drip irrigation leaf quality information set is obtained by annotation;
[0094] Based on a neural network, the sample drip irrigation splash parameter set, the sample leaf humidity information set and the sample drip irrigation leaf quality information set are used to train a leaf drip irrigation quality analyzer, and the drip irrigation leaf quality analysis is performed on the multiple drip irrigation splash parameters and the reduced-dimensional leaf humidity information sets to obtain multiple drip irrigation leaf quality information;
[0095] A root and rhizome drip irrigation quality analyzer is trained to perform drip irrigation root and rhizome quality analysis on the multiple soil water accumulation parameters and the reduced-dimensional root and rhizome moisture information set to obtain multiple drip irrigation root and rhizome quality information.
[0096] In an embodiment of the present application, based on the data of drip irrigation experiments or growth drip irrigation records of similar crops in history, the relevant sample drip irrigation splash parameter set and sample leaf wetness information set are processed and extracted by the same method as in the above content.
[0097] According to the sample drip irrigation splash parameter set and the sample leaf humidity information set, the sample drip irrigation leaf quality information is labeled as the sample supervision data for subsequent training. The labeling process is performed manually or by algorithm to determine the impact of the splash water ratio and humidity level on leaf growth and health, and score or classify the leaves according to their health status. For example, the sample drip irrigation leaf quality information can be scored from 0 to 10. The larger the score, the better the leaf drip irrigation quality, the smaller the corresponding sample drip irrigation splash parameter, and the more the sample leaf humidity information meets the standard humidity requirements (set according to the appropriate humidity of the leaves during crop growth). The leaves may become wet due to excessive splashing, making them susceptible to diseases; or the leaves may show signs of drying up due to insufficient humidity. Based on this, the technicians in this field perform labeling to obtain the sample drip irrigation leaf quality information set.
[0098] Next, the neural network model is trained using the sample drip irrigation splash parameter set, the sample leaf wetness information set, and the sample drip irrigation leaf quality information set. First, a leaf drip irrigation quality analyzer can be constructed based on the BP neural network. The leaf drip irrigation quality analyzer is constructed based on the BP neural network, and is trained by inputting the sample drip irrigation splash parameter set, the sample leaf wetness information set, and the sample drip irrigation leaf quality information set into the network. The training goal of the neural network is to enable the model to learn to predict the quality of the leaf from the drip irrigation splash parameters and the leaf wetness information, and to identify the leaf drip irrigation quality information under the water splash information and the leaf wetness information. Among them, the network parameters of the leaf drip irrigation quality analyzer are adjusted by the deviation between the output of the leaf drip irrigation quality analyzer and the sample drip irrigation leaf quality information, and the output deviation is reduced by iterative training until the accuracy meets the requirements.
[0099] After training, the Leaf Drip Quality Analyzer can be used to process data from actual irrigation. By inputting multiple drip irrigation splash parameters and reduced-dimensional leaf wetness information sets, the analyzer can evaluate the irrigation quality of leaves in multiple reduced-dimensional crop areas. The model combines the splash parameters and wetness data to generate multiple drip irrigation leaf quality information.
[0100] Based on the same method, a set of sample soil waterlogging parameters and a set of sample reduced-dimensional root and stem moisture information are collected, and a set of sample drip irrigation root and stem quality information is obtained by annotation. Among them, based on the technical personnel in this field, the root and stem health level is graded and annotated according to the growth conditions and disease probability of crops under the sample soil waterlogging parameters and the sample reduced-dimensional root and stem moisture information, and the sample drip irrigation root and stem quality information set is obtained.
[0101] The sample soil waterlogging parameter set and the sample reduced-dimensional root and stem moisture information set are used as input data, and the sample drip irrigation root and stem quality information set is used as supervision data. Based on the same method, a root and stem drip irrigation quality analyzer that meets the requirements is trained. The training method is the same as that of the leaf drip irrigation quality analyzer, but the specific training data is different.
[0102] After training, the root and stem drip irrigation quality analyzer can evaluate the root and stem irrigation quality in the actual drip irrigation process. By analyzing multiple soil water accumulation parameters and reduced-dimensional root and stem moisture information sets, the model will output multiple drip irrigation root and stem quality information of crops in multiple reduced-dimensional crop areas during drip irrigation.
[0103] Through the above steps, the irrigation quality of crops can be comprehensively and accurately evaluated, potential irrigation quality problems can be identified, and irrigation strategies can be optimized.
[0104] S50: Based on the multiple drip irrigation blade quality information and the multiple drip irrigation root and stem quality information, comprehensive drip irrigation quality information and drip irrigation quality uniformity information are processed and corrected and calculated on the comprehensive drip irrigation quality information in combination with the distribution uniformity information to obtain a drip irrigation quality analysis result as a growth irrigation data analysis result.
[0105] In an embodiment of the present application, based on multiple drip irrigation leaf quality information and multiple drip irrigation root and stem quality information after dimensionality reduction and image humidity data analysis, a comprehensive drip irrigation quality analysis of multiple crop areas is performed, and comprehensive drip irrigation quality information is obtained through processing. Based on the distribution uniformity of multiple drip irrigation leaf quality information and multiple drip irrigation root and stem quality information, a drip irrigation quality uniformity analysis of multiple reduced-dimensional crop areas is performed to obtain drip irrigation quality uniformity information.
[0106] Among them, the distribution uniformity information includes the distribution uniformity information of multiple reduced-dimensional crop areas with data distribution characteristics, and the drip irrigation quality uniformity information includes the distribution uniformity information of the drip irrigation quality of multiple reduced-dimensional crop areas. Good drip irrigation quality should be relatively uniform in different areas. The drip irrigation quality uniformity information can reflect whether there is an over-concentration or insufficient irrigation problem in certain areas. Therefore, the comprehensive drip irrigation quality information obtained by the current analysis is corrected and calculated in combination with the two to obtain the final drip irrigation quality analysis result, reflecting the drip irrigation quality of multiple crop areas as a whole, as the growth irrigation data analysis result. Among them, if the distribution and drip irrigation quality of multiple crop areas are relatively uniform, the corrected drip irrigation quality analysis result is larger, otherwise, it is smaller.
[0107] Through the above processing and correction, the final drip irrigation quality analysis results are generated. The results include comprehensive irrigation quality assessment of all crop areas, indicating which areas have reached the irrigation quality standard and which areas need to be adjusted. It also reflects the overall drip irrigation quality and the uniformity of drip irrigation quality. The drip irrigation quality analysis results, as the production irrigation data analysis results, provide strong decision support for crop growth management, help optimize irrigation strategies, and improve crop yields and quality.
[0108] Step S50 in the method provided in the embodiment of the present application includes:
[0109] According to the plurality of drip irrigation leaf quality information and the plurality of drip irrigation root quality information, weighted calculation is performed to obtain a plurality of drip irrigation quality information, and comprehensive drip irrigation quality information is calculated;
[0110] Calculating and obtaining drip irrigation quality uniformity information according to the plurality of drip irrigation quality information;
[0111] In combination with the drip irrigation quality uniformity information and the distribution uniformity information, a correction coefficient is set, and the comprehensive drip irrigation quality information is corrected and calculated to obtain the corrected comprehensive drip irrigation quality information as the drip irrigation quality analysis result.
[0112] In an embodiment of the present application, weighted calculation is performed on multiple drip irrigation blade quality information and multiple drip irrigation root and stem quality information to obtain multiple drip irrigation quality information, wherein weighted calculation is performed on the drip irrigation leaf quality information and the drip irrigation root and stem quality information of each reduced-dimensionality crop area, the weight of the drip irrigation quality of the root and stem area is relatively large, for example, 0.6, and multiple drip irrigation quality information is obtained by calculation.
[0113] Furthermore, the mean of multiple drip irrigation quality information is calculated as the comprehensive drip irrigation quality information of multiple crop areas. Weights can also be set according to the growth of multiple dimensionality-reduced crop areas to perform weighted calculation on multiple drip irrigation quality information. The better the growth, the greater the weight, which can be set based on technical personnel in this field.
[0114] According to the multiple drip irrigation quality information, the drip irrigation quality uniformity information is calculated, for example, the variance of the multiple drip irrigation quality information is calculated, and the inverse of the variance is used as the drip irrigation quality uniformity information. The larger the variance, the larger the drip irrigation quality uniformity information. Optionally, a drip irrigation quality uniformity information mapping table can also be constructed, and the corresponding drip irrigation quality uniformity information is obtained according to the variance mapping, and the variance in the drip irrigation quality uniformity information mapping table and the drip irrigation quality uniformity information meet a negative correlation.
[0115] In combination with the drip irrigation quality uniformity information and the distribution uniformity information, a correction coefficient is set. For example, the drip irrigation quality uniformity information and the distribution uniformity information are multiplied as the correction coefficient. The corresponding correction coefficient can also be set according to the size of the drip irrigation quality uniformity information and the distribution uniformity information, as long as the correction coefficient is positively correlated with the size of the drip irrigation quality uniformity information and the distribution uniformity information.
[0116] The correction coefficient is used to perform correction calculation on the above-mentioned comprehensive drip irrigation quality information to obtain the corrected comprehensive drip irrigation quality information as the drip irrigation quality analysis result.
[0117] By correcting the comprehensive drip irrigation quality information based on the uniformity of the distribution of multiple reduced-dimensional crop areas and the uniformity of the overall drip irrigation quality information, an evaluation standard for the uniformity of overall drip irrigation quality is introduced to improve the comprehensiveness and visibility of the overall irrigation data quality analysis and provide a decision-making reference for irrigation management.
[0118] The embodiments of the present application achieve the following technical effects through the above specific implementation methods:
[0119] The present invention obtains multiple drip irrigation images, multiple leaf moisture information and multiple root and stem moisture information by collecting drip irrigation images of multiple crop areas and moisture information near leaves and rhizomes during crop growth irrigation; performs dimensionality reduction processing on the multiple leaf moisture information and the multiple root and stem moisture information to retain data features, obtains a reduced-dimensionality leaf moisture information set and a reduced-dimensionality root and stem moisture information set, and obtains distribution uniformity information; and screens multiple drip irrigation images according to crop areas corresponding to the reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information to obtain a reduced-dimensionality drip irrigation image set, and improves the retained data features through dimensionality reduction processing to improve analysis efficiency; based on a neural network, multiple reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set are screened to obtain a reduced-dimensionality drip irrigation image set. The irrigation image is respectively subjected to drip irrigation splash recognition and soil water accumulation recognition to obtain multiple drip irrigation splash parameters and multiple soil water accumulation parameters; based on the neural network, the drip irrigation leaf quality analysis is performed in combination with multiple drip irrigation splash parameters and the reduced dimension leaf moisture information set to obtain multiple drip irrigation leaf quality information; the drip irrigation root and stem quality analysis is performed in combination with multiple soil water accumulation parameters and the reduced dimension root and stem moisture information set to obtain multiple drip irrigation root and stem quality information; based on the multiple drip irrigation leaf quality information and the multiple drip irrigation root and stem quality information, the comprehensive drip irrigation quality information and the drip irrigation quality uniformity information are processed to obtain the comprehensive drip irrigation quality information and the drip irrigation quality uniformity information, and the comprehensive drip irrigation quality information is corrected and calculated in combination with the distribution uniformity information to obtain the drip irrigation quality analysis result as the growth irrigation data analysis result. Through the above steps, the present application can analyze the leaf splash wetting condition and the root and stem drip irrigation condition in the drip irrigation data, and then analyze the overall drip irrigation quality, and at the same time, make corrections in combination with the distribution and drip irrigation uniformity, and analyze and process considering the overall drip irrigation quality uniformity, which can achieve the technical effect of improving the comprehensiveness and accuracy of the drip irrigation data analysis.
[0120] Example 2
[0121] like Figure 2 As shown, based on the same inventive concept as the neural network-based growth irrigation data analysis method provided in Example 1, an embodiment of the present invention provides a neural network-based growth irrigation data analysis system, the system comprising:
[0122] The irrigation data acquisition module 11 is used to collect drip irrigation images of multiple crop areas and humidity information near leaves and roots during crop growth irrigation, and obtain multiple drip irrigation images, multiple leaf humidity information and multiple root humidity information;
[0123] The irrigation data dimension reduction module 12 is used to perform dimension reduction processing on the plurality of leaf moisture information and the plurality of root and stem moisture information with data feature preservation, obtain a reduced dimension leaf moisture information set and a reduced dimension root and stem moisture information set, and obtain distribution uniformity information, and screen the plurality of drip irrigation images according to the crop areas corresponding to the reduced dimension leaf moisture information and the reduced dimension root and stem moisture information to obtain a reduced dimension drip irrigation image set;
[0124] An image recognition module 13 is used to perform drip irrigation splash recognition and soil water accumulation recognition on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set based on a neural network, and obtain a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters;
[0125] The drip irrigation quality analysis module 14 is used to perform drip irrigation leaf quality analysis based on a neural network, in combination with the multiple drip irrigation splash parameters and the reduced-dimensionality leaf humidity information set, to obtain multiple drip irrigation leaf quality information, and to perform drip irrigation root and stem quality analysis in combination with the multiple soil water accumulation parameters and the reduced-dimensionality root and stem humidity information set, to obtain multiple drip irrigation root and stem quality information;
[0126] The comprehensive analysis module 15 is used to process the comprehensive drip irrigation quality information and the drip irrigation quality uniformity information according to the multiple drip irrigation leaf quality information and the multiple drip irrigation root and stem quality information, and to correct and calculate the comprehensive drip irrigation quality information in combination with the distribution uniformity information to obtain the drip irrigation quality analysis result as the growth irrigation data analysis result.
[0127] Furthermore, the irrigation data dimension reduction module 12 is also used to: calculate and process the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain reference leaf moisture information and reference root and stem moisture information;
[0128] Calculate the distribution probability according to the difference between other leaf humidity information and rhizome humidity information and the reference leaf humidity information and reference rhizome humidity information, and obtain reference leaf humidity probability distribution and reference rhizome humidity probability distribution;
[0129] According to the benchmark leaf moisture probability distribution and the benchmark rhizome moisture probability distribution, the plurality of leaf moisture information and the plurality of rhizome moisture information are subjected to feature-retaining optimized dimensionality reduction to obtain the reduced dimensionality leaf moisture information set and the reduced dimensionality rhizome moisture information set, and to obtain a plurality of reduced dimensionality crop regions;
[0130] Filtering the plurality of reduced-dimensionality drip irrigation images of the crop areas to obtain a reduced-dimensionality drip irrigation image set;
[0131] The distribution uniformity information is obtained by calculation according to the region coordinates of the multiple reduced-dimensionality crop regions.
[0132] According to the reference leaf moisture probability distribution and the reference rhizome moisture probability distribution, the plurality of leaf moisture information and the plurality of rhizome moisture information are subjected to feature-preserving optimization dimensionality reduction, including:
[0133] According to the dimensionality reduction ratio, randomly select the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain a first dimensionality reduced leaf moisture information set and a first dimensionality reduced root and stem moisture information set;
[0134] According to the reference leaf moisture information and the reference rhizome moisture information, calculate and obtain the first reduced dimension leaf moisture probability distribution and the first reduced dimension rhizome moisture probability distribution of the first reduced dimension leaf moisture information set and the first reduced dimension rhizome moisture information set;
[0135] Calculate the similarity between the first reduced-dimensionality leaf moisture probability distribution and the reference leaf moisture probability distribution to obtain a first leaf moisture feature retention degree; calculate the similarity between the first reduced-dimensionality root and stem moisture probability distribution and the reference root and stem moisture probability distribution to obtain a first root and stem moisture feature retention degree; and combine the first leaf moisture feature retention degree to obtain a first reduced-dimensionality feature retention degree through weighted calculation;
[0136] Continue to perform iterative random dimensionality reduction until the preset number of dimensionality reduction times is reached, output the reduced dimensionality leaf moisture information set and the reduced dimensionality root and stem moisture information set with the largest dimensionality reduction feature retention, and select the crop areas corresponding to the multiple reduced dimensionality leaf moisture information as multiple reduced dimensionality crop areas.
[0137] The step of calculating and obtaining the distribution uniformity information according to the region coordinates of the plurality of reduced-dimensionality crop regions includes:
[0138] Constructing a crop region coordinate system in the plurality of crop regions, and calculating and obtaining reference point coordinates of the coordinates of the plurality of crop regions;
[0139] According to the crop region coordinate system, obtaining the region coordinates of the plurality of reduced-dimensionality crop regions, and calculating the reduced-dimensionality point coordinates;
[0140] The distribution uniformity information is calculated based on the deviation between the coordinates of the dimension reduction point and the coordinates of the reference point.
[0141] Furthermore, the image recognition module 13 is also used for:
[0142] According to the historical drip irrigation data of crops, a set of sample drip irrigation images is collected. According to the proportion of drip irrigation water splashing on leaves in each sample drip irrigation image and the range of water accumulation after soil dripping, a set of sample drip irrigation splash parameters and a set of sample soil water accumulation parameters are annotated to obtain;
[0143] Based on a convolutional neural network, the sample drip irrigation image set is used as input data, and a sample drip irrigation splash parameter set and a sample soil waterlogging parameter set are used as supervision data, respectively, to train a splash recognition branch and a waterlogging recognition branch, and obtain a drip irrigation image recognizer;
[0144] The drip irrigation image recognizer is used to perform drip irrigation splash recognition and soil water accumulation recognition on the multiple reduced-dimensional drip irrigation images, respectively, to obtain multiple drip irrigation splash parameters and multiple soil water accumulation parameters.
[0145] Furthermore, the drip irrigation quality analysis module 14 is also used for:
[0146] According to the crop historical drip irrigation monitoring data, the sample drip irrigation splash parameter set and the sample leaf humidity information set are collected, and the sample drip irrigation leaf quality information set is obtained by annotation;
[0147] Based on a neural network, the sample drip irrigation splash parameter set, the sample leaf humidity information set and the sample drip irrigation leaf quality information set are used to train a leaf drip irrigation quality analyzer, and the drip irrigation leaf quality analysis is performed on the multiple drip irrigation splash parameters and the reduced-dimensional leaf humidity information sets to obtain multiple drip irrigation leaf quality information;
[0148] A root and rhizome drip irrigation quality analyzer is trained to perform drip irrigation root and rhizome quality analysis on the multiple soil water accumulation parameters and the reduced-dimensional root and rhizome moisture information set to obtain multiple drip irrigation root and rhizome quality information.
[0149] Furthermore, the comprehensive analysis module 15 is also used for:
[0150] According to the plurality of drip irrigation leaf quality information and the plurality of drip irrigation root quality information, weighted calculation is performed to obtain a plurality of drip irrigation quality information, and comprehensive drip irrigation quality information is calculated;
[0151] Calculating and obtaining drip irrigation quality uniformity information according to the plurality of drip irrigation quality information;
[0152] In combination with the drip irrigation quality uniformity information and the distribution uniformity information, a correction coefficient is set, and the comprehensive drip irrigation quality information is corrected and calculated to obtain the corrected comprehensive drip irrigation quality information as the drip irrigation quality analysis result.
[0153] See also Figure 3 , Figure 3 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, an embodiment of the present invention provides an electronic device 100, including a memory 110, a processor 120, and a computer program 111 stored in the memory 110 and executable on the processor 120. When the processor 120 executes the computer program 111, the steps of the method in Example 1 are implemented.
[0154] See also Figure 4 , Figure 4 A schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present invention. Figure 4 As shown, this embodiment provides a computer-readable storage medium 200 on which a computer program 211 is stored. When the computer program 211 is executed by a processor, the steps of the method in Embodiment 1 are implemented.
[0155] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0156] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0157] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0158] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0159] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0160] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0161] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A growth irrigation data analysis method based on a neural network, characterized in that: The method comprises: During the crop growth irrigation process, drip irrigation images of multiple crop areas and humidity information near leaves and roots are collected to obtain multiple drip irrigation images, multiple leaf humidity information and multiple root humidity information; Performing dimensionality reduction processing with data feature preservation on the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain a reduced-dimensionality leaf moisture information set and a reduced-dimensionality root and stem moisture information set, as well as obtaining distribution uniformity information, and screening the plurality of drip irrigation images according to the crop areas corresponding to the reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information to obtain a reduced-dimensionality drip irrigation image set; Based on a neural network, drip irrigation splash recognition and soil water accumulation recognition are respectively performed on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set to obtain a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters; Based on the neural network, the drip irrigation leaf quality analysis is performed in combination with the multiple drip irrigation splash parameters and the reduced dimension leaf humidity information set to obtain multiple drip irrigation leaf quality information; the drip irrigation root and stem quality analysis is performed in combination with the multiple soil water accumulation parameters and the reduced dimension root and stem humidity information set to obtain multiple drip irrigation root and stem quality information; According to the plurality of drip irrigation leaf quality information and the plurality of drip irrigation root quality information, comprehensive drip irrigation quality information and drip irrigation quality uniformity information are obtained by processing, and the comprehensive drip irrigation quality information is corrected and calculated in combination with the distribution uniformity information to obtain a drip irrigation quality analysis result as a growth irrigation data analysis result; The dimensionality reduction processing for retaining data features of the plurality of leaf moisture information and the plurality of root and stem moisture information comprises: Calculating and processing the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain reference leaf moisture information and reference root and stem moisture information; Calculate the distribution probability according to the difference between other leaf humidity information and rhizome humidity information and the reference leaf humidity information and reference rhizome humidity information, and obtain reference leaf humidity probability distribution and reference rhizome humidity probability distribution; According to the benchmark leaf moisture probability distribution and the benchmark rhizome moisture probability distribution, the plurality of leaf moisture information and the plurality of rhizome moisture information are subjected to feature-retaining optimized dimensionality reduction to obtain the reduced dimensionality leaf moisture information set and the reduced dimensionality rhizome moisture information set, and to obtain a plurality of reduced dimensionality crop regions; Filtering the plurality of reduced-dimensionality drip irrigation images of the crop areas to obtain a reduced-dimensionality drip irrigation image set; Calculating and obtaining the distribution uniformity information according to the regional coordinates of the plurality of reduced-dimensionality crop regions; The step of calculating and obtaining the distribution uniformity information according to the region coordinates of the plurality of reduced-dimensionality crop regions includes: Constructing a crop region coordinate system in a plurality of crop regions, and calculating reference point coordinates of the coordinates of the plurality of crop regions; According to the crop region coordinate system, obtaining the region coordinates of the plurality of reduced-dimensionality crop regions, and calculating the reduced-dimensionality point coordinates; Calculating and obtaining distribution uniformity information according to the deviation between the coordinates of the dimension reduction point and the coordinates of the reference point; According to the plurality of drip irrigation blade quality information and the plurality of drip irrigation root quality information, comprehensive drip irrigation quality information and drip irrigation quality uniformity information are processed and obtained, and the comprehensive drip irrigation quality information is corrected and calculated in combination with the distribution uniformity information to obtain the drip irrigation quality analysis result, including: According to the plurality of drip irrigation leaf quality information and the plurality of drip irrigation root quality information, weighted calculation is performed to obtain a plurality of drip irrigation quality information, and comprehensive drip irrigation quality information is calculated; Calculating and obtaining drip irrigation quality uniformity information according to the plurality of drip irrigation quality information; In combination with the drip irrigation quality uniformity information and the distribution uniformity information, a correction coefficient is set, and the comprehensive drip irrigation quality information is corrected and calculated to obtain the corrected comprehensive drip irrigation quality information as the drip irrigation quality analysis result.
2. The neural network-based growth irrigation data analysis method according to claim 1, characterized in that: According to the reference leaf moisture probability distribution and the reference rhizome moisture probability distribution, the plurality of leaf moisture information and the plurality of rhizome moisture information are subjected to feature-preserving optimization dimensionality reduction, including: According to the dimensionality reduction ratio, randomly select the plurality of leaf moisture information and the plurality of root and stem moisture information to obtain a first dimensionality reduced leaf moisture information set and a first dimensionality reduced root and stem moisture information set; According to the reference leaf moisture information and the reference rhizome moisture information, calculate and obtain the first reduced dimension leaf moisture probability distribution and the first reduced dimension rhizome moisture probability distribution of the first reduced dimension leaf moisture information set and the first reduced dimension rhizome moisture information set; Calculate the similarity between the first reduced-dimensionality leaf moisture probability distribution and the reference leaf moisture probability distribution to obtain a first leaf moisture feature retention degree; calculate the similarity between the first reduced-dimensionality root and stem moisture probability distribution and the reference root and stem moisture probability distribution to obtain a first root and stem moisture feature retention degree; and combine the first leaf moisture feature retention degree to obtain a first reduced-dimensionality feature retention degree through weighted calculation; Continue to perform iterative random dimensionality reduction until the preset number of dimensionality reduction times is reached, output the reduced dimensionality leaf moisture information set and the reduced dimensionality root and stem moisture information set with the largest dimensionality reduction feature retention, and select the crop areas corresponding to the multiple reduced dimensionality leaf moisture information as multiple reduced dimensionality crop areas.
3. The neural network-based growth irrigation data analysis method according to claim 1, characterized in that: Based on a neural network, drip irrigation splash recognition and soil water accumulation recognition are performed on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set, and a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters are obtained, including: According to the historical drip irrigation data of crops, a set of sample drip irrigation images is collected. According to the proportion of drip irrigation water splashing on leaves in each sample drip irrigation image and the range of water accumulation after soil dripping, a set of sample drip irrigation splash parameters and a set of sample soil water accumulation parameters are annotated to obtain; Based on a convolutional neural network, the sample drip irrigation image set is used as input data, and a sample drip irrigation splash parameter set and a sample soil waterlogging parameter set are used as supervision data, respectively, to train a splash recognition branch and a waterlogging recognition branch, and obtain a drip irrigation image recognizer; The drip irrigation image recognizer is used to perform drip irrigation splash recognition and soil water accumulation recognition on the multiple reduced-dimensional drip irrigation images, respectively, to obtain multiple drip irrigation splash parameters and multiple soil water accumulation parameters.
4. The neural network-based growth irrigation data analysis method according to claim 3, characterized in that: Based on the neural network, the drip irrigation leaf quality analysis is performed in combination with the multiple drip irrigation splash parameters and the reduced dimension leaf humidity information set to obtain multiple drip irrigation leaf quality information; the drip irrigation root and stem quality analysis is performed in combination with the multiple soil water accumulation parameters and the reduced dimension root and stem humidity information set to obtain multiple drip irrigation root and stem quality information, including: According to the crop historical drip irrigation monitoring data, the sample drip irrigation splash parameter set and the sample leaf humidity information set are collected, and the sample drip irrigation leaf quality information set is obtained by annotation; Based on a neural network, the sample drip irrigation splash parameter set, the sample leaf humidity information set and the sample drip irrigation leaf quality information set are used to train a leaf drip irrigation quality analyzer, and the drip irrigation leaf quality analysis is performed on the multiple drip irrigation splash parameters and the reduced-dimensional leaf humidity information sets to obtain multiple drip irrigation leaf quality information; A root and rhizome drip irrigation quality analyzer is trained to perform drip irrigation root and rhizome quality analysis on the multiple soil water accumulation parameters and the reduced-dimensional root and rhizome moisture information set to obtain multiple drip irrigation root and rhizome quality information.
5. A growth irrigation data analysis system based on neural network, characterized in that: The system is used to execute the method according to any one of claims 1 to 4, and the system comprises: The irrigation data acquisition module is used to collect drip irrigation images of multiple crop areas and humidity information near leaves and roots during the crop growth irrigation process, and obtain multiple drip irrigation images, multiple leaf humidity information and multiple root humidity information; An irrigation data dimensionality reduction module is used to perform dimensionality reduction processing on the plurality of leaf moisture information and the plurality of root and stem moisture information with data feature preservation, obtain a reduced-dimensionality leaf moisture information set and a reduced-dimensionality root and stem moisture information set, and obtain distribution uniformity information, and screen the plurality of drip irrigation images according to the crop areas corresponding to the reduced-dimensionality leaf moisture information and the reduced-dimensionality root and stem moisture information to obtain a reduced-dimensionality drip irrigation image set; An image recognition module, for performing drip irrigation splash recognition and soil water accumulation recognition on a plurality of reduced-dimensionality drip irrigation images in the reduced-dimensionality drip irrigation image set based on a neural network, and obtaining a plurality of drip irrigation splash parameters and a plurality of soil water accumulation parameters; A drip irrigation quality analysis module, for performing a drip irrigation leaf quality analysis based on a neural network, combining the multiple drip irrigation splash parameters and the reduced-dimensionality leaf humidity information set, and obtaining multiple drip irrigation leaf quality information; performing a drip irrigation root and stem quality analysis based on the multiple soil water accumulation parameters and the reduced-dimensionality root and stem humidity information set, and obtaining multiple drip irrigation root and stem quality information; The comprehensive analysis module is used to process the comprehensive drip irrigation quality information and the drip irrigation quality uniformity information according to the multiple drip irrigation leaf quality information and the multiple drip irrigation root and stem quality information, and to correct and calculate the comprehensive drip irrigation quality information in combination with the distribution uniformity information to obtain the drip irrigation quality analysis result as the growth irrigation data analysis result.
6. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the neural network-based growth irrigation data analysis method described in any one of claims 1-4.
7. A computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, the neural network-based growth irrigation data analysis method as described in any one of claims 1 to 4 is implemented.
