Nitrogen fertilizer management system and method for planting phyllostachys praecox
Through deep learning technology, the growth status and soil condition data of thunder bamboo are integrated, and the amount of nitrogen fertilizer application is intelligently recommended, which solves the problems of excessive or insufficient in traditional fertilization methods, and achieves precise fertilization and environmental protection.
Patent Information
- Application Number
- CN202411417079.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-11
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional thunder bamboo nitrogen fertilizer fertilization methods rely on experience and fail to adjust the amount of thunder bamboo to be applied according to the actual nutritional needs and growth stage of thunder bamboo, resulting in excessive or insufficient fertilization, resulting in reduced fertilization effect, soil degradation and agricultural non-point source pollution.
The camera is used to collect hyperspectral images of the growth state of thunder bamboo and sensors to collect soil condition data in real time. Through deep learning-based image recognition technology and data analysis algorithms, the growth state and soil condition characteristics are integrated, and the amount of nitrogen fertilizer application is intelligently recommended.
Accurate fertilization has been achieved, reducing waste of nitrogen fertilizer and environmental impact, and promoting the healthy growth of thunder bamboo.
Smart Images

Figure CN118940971A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent management, and more specifically, to a nitrogen fertilizer management system and method for phyllostachys praecox cultivation. Background Art
[0002] Lei bamboo, also known as Leigong bamboo, is a plant of the bamboo subfamily unique to China, belonging to the genus Bambusa of the Poaceae family. It is a tall, fast-growing perennial plant known for its thick bamboo poles and delicious bamboo shoots, so planting Lei bamboo can bring significant economic benefits.
[0003] Since Phyllostachys praecox is an excellent bamboo species for bamboo shoots, economic benefits are generally improved by applying large amounts of fertilizers, especially nitrogen fertilizers. Specifically, nitrogen fertilizers can promote the growth and development of Phyllostachys praecox, increase the yield of bamboo shoots and increase the growth rate of bamboo poles. Reasonable application of nitrogen fertilizers can help improve the taste and nutritional value of bamboo shoots and enhance market competitiveness.
[0004] However, the traditional nitrogen fertilizer application method for thunbergii may rely mainly on experience rather than scientific data, that is, the amount of fertilizer is not adjusted according to the actual nutritional needs of thunbergii, resulting in excessive and unreasonable nitrogen fertilizer application. Secondly, thunbergii has different nitrogen requirements at different growth stages. The traditional method may not adjust the fertilization time according to the specific growth stage and soil conditions, missing the best time for fertilization, thus causing problems such as reduced fertilization effect of thunbergii forests, soil degradation and agricultural non-point source pollution.
[0005] Therefore, an optimized nitrogen fertilizer management system for P. praecox cultivation is desired. Summary of the invention
[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a nitrogen fertilizer management system and method for the cultivation of Phyllostachys praecox, which collects hyperspectral images of the growth status of Phyllostachys praecox through a camera, and collects soil condition data (pH value, nitrogen content, phosphorus content and potassium content) in real time through a sensor, and uses image recognition technology and data analysis algorithms based on deep learning to perform feature analysis of the hyperspectral image of the growth status of Phyllostachys praecox and time series correlation of the soil condition data, and fuses the features of the two, so as to comprehensively and intelligently obtain the recommended value for the application of nitrogen fertilizer. In this way, the system can quickly respond to changes in the growth environment of Phyllostachys praecox according to the actual growth status and soil conditions of Phyllostachys praecox, so as to intelligently recommend the application amount of nitrogen fertilizer, realize precise fertilization, and avoid excessive or insufficient fertilization, which helps to reduce waste and reduce potential impacts on the environment, thereby promoting the healthy growth of crops.
[0007] According to one aspect of the present application, a nitrogen fertilizer management system for planting Phyllostachys praecox is provided, comprising:
[0008] A hyperspectral image acquisition module for the growth status of Phyllostachys praecox, used for acquiring a hyperspectral image of the growth status of Phyllostachys praecox through a camera;
[0009] A soil condition data acquisition module, used to acquire soil condition data in real time through sensors to obtain a time series of soil condition data, wherein the soil condition data includes pH value, nitrogen content, phosphorus content and potassium content;
[0010] A shallow-depth feature capture module for the growth state of Phyllostachys praecox is used to capture the multi-scale features of the growth state of Phyllostachys praecox on the hyperspectral image of the growth state of Phyllostachys praecox to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox;
[0011] A semantic enhancement fusion module for deep features of the growth state of Phyllostachys thunbergii, which is used to semantically enhance the deep feature map of the growth state of Phyllostachys thunbergii to obtain a deep semantic feature map of the growth state of Phyllostachys thunbergii, and then perform feature fusion on the shallow feature map of the growth state of Phyllostachys thunbergii and the deep semantic feature map of the growth state of Phyllostachys thunbergii to obtain a multi-dimensional feature map of the growth state of Phyllostachys thunbergii;
[0012] A soil condition time series association module is used to integrate the time series of the soil condition data and input the data into a soil condition time series implicit association feature capture module to obtain a pH value time series implicit association feature vector, a nitrogen content time series implicit association feature vector, a phosphorus content time series implicit association feature vector, and a potassium content time series implicit association feature vector;
[0013] A soil condition time series feature enhancement module, for arranging the pH value time series implicit correlation feature vector, the nitrogen content time series implicit correlation feature vector, the phosphorus content time series implicit correlation feature vector and the potassium content time series implicit correlation feature vector into a soil condition time series matrix, and then performing soil condition time series enhancement to obtain a soil condition time series enhanced feature map;
[0014] A growth state-soil condition characterization module, used for fusing the multi-dimensional feature map of the growth state of Phyllostachys praecox and the time series enhanced feature map of the soil condition to obtain a comprehensive characterization feature map of the growth state of Phyllostachys praecox and soil condition, and performing pooling processing on the comprehensive characterization feature map of the growth state of Phyllostachys praecox to obtain a comprehensive characterization feature vector of the growth state of Phyllostachys praecox;
[0015] A growth state-soil condition probabilistic guidance and control module, used for performing probabilistic guidance and control on the growth state-soil condition comprehensive characterization feature vector of Phyllostachys praecox based on decoding weight data to obtain an optimized growth state-soil condition comprehensive characterization feature vector of Phyllostachys praecox;
[0016] The nitrogen fertilizer recommendation value generation module is used to obtain the recommended value for applying nitrogen fertilizer based on the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector.
[0017] In the above-mentioned nitrogen fertilizer management system for phyllostachys praecox cultivation, the phyllostachys praecox growth state deep and shallow feature capture module is used to: input the phyllostachys praecox growth state hyperspectral image into the phyllostachys praecox growth state feature capture model based on the void convolutional neural network to obtain the phyllostachys praecox growth state shallow feature map and the phyllostachys praecox growth state deep feature map.
[0018] In the above-mentioned nitrogen fertilizer management system for phyllostachys praecox planting, the phyllostachys praecox growth status deep feature semantic enhancement fusion module includes: a phyllostachys praecox growth status deep semantic enhancement unit, which is used to input the phyllostachys praecox growth status deep feature map into the phyllostachys praecox growth status semantic enhancement module based on the pyramid pooling module to obtain the phyllostachys praecox growth status deep semantic feature map; a phyllostachys praecox growth status shallow semantic feature fusion unit, which is used to cascade the phyllostachys praecox growth status shallow feature map and the phyllostachys praecox growth status deep semantic feature map to obtain the phyllostachys praecox growth status multi-dimensional feature map.
[0019] In the above-mentioned nitrogen fertilizer management system for planting praecox, the soil condition time series association module includes: a soil condition time series data integration unit, which is used to integrate the time series of the soil condition data based on the time dimension and the parameter sample dimension to obtain the time series of pH value, the time series of nitrogen content, the time series of phosphorus content and the time series of potassium content; a soil condition time series feature capture unit, which is used to input the time series of pH value, the time series of nitrogen content, the time series of phosphorus content and the time series of potassium content into the soil condition time series implicit association feature capture module based on the one-dimensional convolutional neural network respectively to obtain the pH value time series implicit association feature vector, the nitrogen content time series implicit association feature vector, the phosphorus content time series implicit association feature vector and the potassium content time series implicit association feature vector.
[0020] In the above-mentioned nitrogen fertilizer management system for pheasanthus praecox planting, the soil condition time series feature enhancement module is used to: arrange the pH value time series implicit correlation feature vector, the nitrogen content time series implicit correlation feature vector, the phosphorus content time series implicit correlation feature vector and the potassium content time series implicit correlation feature vector into the soil condition time series matrix and then input the matrix into the soil condition time series enhancer based on the spatial attention mechanism to obtain the soil condition time series enhanced feature map.
[0021] In the above-mentioned nitrogen fertilizer management system for the cultivation of Phyllostachys praecox, the growth state-soil condition probabilistic guidance and control module is used to: determine the decoding weight matrix of the decoder; perform matrix decoupling on the decoding weight matrix in units of row vectors to obtain a set of decoding weight vectors; calculate the Phyllostachys praecox growth state-soil condition comprehensive representation metric learning loss information item between each decoding weight vector in the set of decoding weight vectors and the Phyllostachys praecox growth state-soil condition comprehensive representation feature vector to obtain the Phyllostachys praecox growth state-soil condition comprehensive representation metric learning compensation representation vector The growth state-soil condition comprehensive representation metric learning compensation representation vector is input Activate function to obtain probabilistic Phyllostachys praecox growth state-soil condition comprehensive characterization metric learning compensation weight vector; calculate the position point multiplication of the probabilistic Phyllostachys praecox growth state-soil condition comprehensive characterization metric learning compensation weight vector and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector to obtain the point product feature vector, and calculate the weighted sum of the point product feature vector and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector to obtain the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector.
[0022] In the nitrogen fertilizer management system for the cultivation of Phyllostachys praecox, the nitrogen fertilizer recommendation value generation module is used to: input the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector into the decoder to obtain the recommended value for applying nitrogen fertilizer.
[0023] According to another aspect of the present application, a nitrogen fertilizer management method for phyllostachys praecox planting is provided, comprising:
[0024] The camera is used to collect hyperspectral images of the growth status of Phyllostachys praecox;
[0025] Collecting soil condition data in real time through sensors to obtain a time series of soil condition data, wherein the soil condition data includes pH value, nitrogen content, phosphorus content and potassium content;
[0026] Capturing multi-scale features of the growth state of Phyllostachys praecox on the hyperspectral image of the growth state of Phyllostachys praecox to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox;
[0027] After semantic enhancement is performed on the deep feature map of the growth state of Phyllostachys praecox to obtain a deep semantic feature map of the growth state of Phyllostachys praecox, feature fusion is performed on the shallow feature map of the growth state of Phyllostachys praecox and the deep semantic feature map of the growth state of Phyllostachys praecox to obtain a multi-dimensional feature map of the growth state of Phyllostachys praecox;
[0028] The soil condition data are integrated in time series and then input into a soil condition time series implicit correlation feature capture module to obtain a pH value time series implicit correlation feature vector, a nitrogen content time series implicit correlation feature vector, a phosphorus content time series implicit correlation feature vector, and a potassium content time series implicit correlation feature vector;
[0029] Arranging the pH value time series implicit correlation feature vector, the nitrogen content time series implicit correlation feature vector, the phosphorus content time series implicit correlation feature vector and the potassium content time series implicit correlation feature vector into a soil condition time series matrix, and then performing soil condition time series enhancement to obtain a soil condition time series enhancement feature map;
[0030] The multi-dimensional feature map of the growth state of Phyllostachys praecox and the time series enhanced feature map of the soil condition are integrated to obtain a feature map of the growth state of Phyllostachys praecox and soil condition, and pooling is performed on the feature map of the growth state of Phyllostachys praecox and soil condition to obtain a feature vector of the growth state of Phyllostachys praecox and soil condition;
[0031] The Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector is subjected to probabilistic guidance regulation based on decoding weight data to obtain an optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector;
[0032] Based on the optimized comprehensive characterization feature vector of Phyllostachys praecox growth state and soil conditions, the recommended value for applying nitrogen fertilizer is obtained.
[0033] In the above-mentioned nitrogen fertilizer management method for phyllostachys praecox planting, multi-scale feature capture of the phyllostachys praecox growth state is performed on the hyperspectral image of the phyllostachys praecox growth state to obtain a shallow feature map of the phyllostachys praecox growth state and a deep feature map of the phyllostachys praecox growth state, including: inputting the hyperspectral image of the phyllostachys praecox growth state into a phyllostachys praecox growth state feature capture model based on a void convolutional neural network to obtain the shallow feature map of the phyllostachys praecox growth state and the deep feature map of the phyllostachys praecox growth state.
[0034] In the above-mentioned nitrogen fertilizer management method for leucopsis thunbergii planting, after semantic enhancement is performed on the deep feature map of the growth state of leucopsis thunbergii to obtain the deep semantic feature map of the growth state of leucopsis thunbergii, the shallow feature map of the growth state of leucopsis thunbergii and the deep semantic feature map of the growth state of leucopsis thunbergii are feature fused to obtain the multi-dimensional feature map of the growth state of leucopsis thunbergii, including: inputting the deep feature map of the growth state of leucopsis thunbergii into a semantic enhancement module of the growth state of leucopsis thunbergii based on a pyramid pooling module to obtain the deep semantic feature map of the growth state of leucopsis thunbergii; cascading the shallow feature map of the growth state of leucopsis thunbergii and the deep semantic feature map of the growth state of leucopsis thunbergii to obtain the multi-dimensional feature map of the growth state of leucopsis thunbergii.
[0035] Compared with the prior art, the present application provides a nitrogen fertilizer management system and method for the cultivation of Phyllostachys praecox, which collects hyperspectral images of the growth status of Phyllostachys praecox through a camera, and collects soil condition data (PH value, nitrogen content, phosphorus content and potassium content) in real time through a sensor, and uses image recognition technology and data analysis algorithms based on deep learning to perform feature analysis of the hyperspectral images of the growth status of Phyllostachys praecox and time series correlation of the soil condition data, and fuses the features of the two to comprehensively and intelligently obtain the recommended value for the application of nitrogen fertilizer. In this way, the system can quickly respond to changes in the growth environment of Phyllostachys praecox according to the actual growth status and soil conditions of Phyllostachys praecox, so as to intelligently recommend the application amount of nitrogen fertilizer, realize precise fertilization, and avoid excessive or insufficient fertilization, which helps to reduce waste and reduce potential impacts on the environment, thereby promoting the healthy growth of crops. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0037] Figure 1 4 is a block diagram of a nitrogen fertilizer management system for growing Phyllostachys praecox according to an embodiment of the present application.
[0038] Figure 2 The block diagram is a semantic enhancement fusion module of deep features of the growth status of Phyllostachys praecox in a nitrogen fertilizer management system for Phyllostachys praecox cultivation according to an embodiment of the present application.
[0039] Figure 3 4 is a block diagram of a soil condition time series association module in a nitrogen fertilizer management system for pheasanthus praecox planting according to an embodiment of the present application.
[0040] Figure 4 The present invention is a flow chart of a nitrogen fertilizer management method for growing Phyllostachys praecox according to an embodiment of the present application. DETAILED DESCRIPTION
[0041] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.
[0042] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.
[0043] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.
[0044] As shown in this disclosure, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular, but also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list, and the method or device may also include other steps or elements.
[0045] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.
[0046] Lei bamboo, also known as Leigong bamboo, is a plant of the bamboo subfamily unique to China, belonging to the genus Bambusa of the Poaceae family. It is a tall, fast-growing perennial plant known for its thick bamboo poles and delicious bamboo shoots, so planting Lei bamboo can bring significant economic benefits.
[0047] Since Phyllostachys praecox is an excellent bamboo species for bamboo shoots, economic benefits are generally improved by applying large amounts of fertilizers, especially nitrogen fertilizers. Specifically, nitrogen fertilizers can promote the growth and development of Phyllostachys praecox, increase the yield of bamboo shoots and increase the growth rate of bamboo poles. Reasonable application of nitrogen fertilizers can help improve the taste and nutritional value of bamboo shoots and enhance market competitiveness.
[0048] However, the traditional nitrogen fertilizer application method for thunbergii may rely mainly on experience rather than scientific data, that is, the amount of fertilizer is not adjusted according to the actual nutritional needs of thunbergii, resulting in excessive and unreasonable nitrogen fertilizer application. Secondly, thunbergii has different nitrogen requirements at different growth stages. The traditional method may not adjust the fertilization time according to the specific growth stage and soil conditions, missing the best time for fertilization, thus causing problems such as reduced fertilization effect of thunbergii forests, soil degradation and agricultural non-point source pollution.
[0049] Therefore, an optimized nitrogen fertilizer management system for the cultivation of Phyllostachys praecox is desired, which collects hyperspectral images of the growth status of Phyllostachys praecox by a camera, and collects soil condition data (PH value, nitrogen content, phosphorus content and potassium content) in real time by a sensor, and uses image recognition technology and data analysis algorithms based on deep learning to perform feature analysis of the hyperspectral images of the growth status of Phyllostachys praecox and time series correlation of the soil condition data, and fuses the features of the two to comprehensively and intelligently obtain the recommended value for the application of nitrogen fertilizer. In this way, the system can quickly respond to changes in the growth environment of Phyllostachys praecox according to its actual growth status and soil conditions, so as to intelligently recommend the application amount of nitrogen fertilizer, achieve precise fertilization, and avoid excessive or insufficient fertilization, which helps to reduce waste and reduce potential impacts on the environment, thereby promoting healthy growth of crops.
[0050] Figure 1 FIG. 1 is a block diagram of a nitrogen fertilizer management system for growing Phyllostachys praecox according to an embodiment of the present application. Figure 1As shown, according to the embodiment of the present application, the nitrogen fertilizer management system 100 for the cultivation of phragmites praecox comprises: a phragmites praecox growth state hyperspectral image acquisition module 110, which is used to acquire the phragmites praecox growth state hyperspectral image through a camera; a soil condition data acquisition module 120, which is used to acquire soil condition data in real time through a sensor to obtain a time series of soil condition data, wherein the soil condition data includes pH value, nitrogen content, phosphorus content and potassium content; a phragmites praecox growth state depth feature capture module 130, which is used to perform multi-scale feature capture of the phragmites praecox growth state hyperspectral image to obtain a shallow depth feature capture of the phragmites praecox growth state. The deep semantic feature map of the growth state of Phyllostachys thunbergii is used to perform semantic enhancement on the deep feature map of the growth state of Phyllostachys thunbergii to obtain the deep semantic feature map of the growth state of Phyllostachys thunbergii, and then fuse the shallow feature map of the growth state of Phyllostachys thunbergii and the deep semantic feature map of the growth state of Phyllostachys thunbergii to obtain the multi-dimensional feature map of the growth state of Phyllostachys thunbergii; the soil condition time series association module 150 is used to integrate the time series of the soil condition data and input it into the soil condition time series implicit association feature capture module to obtain the pH value time series implicit association feature vector, the nitrogen content time series implicit association feature vector, and the soil condition time series implicit association feature vector. The invention relates to a soil condition time series feature enhancement module 160, which is used to arrange the pH value time series implicitly associated feature vector, the nitrogen content time series implicitly associated feature vector, the phosphorus content time series implicitly associated feature vector and the potassium content time series implicitly associated feature vector into a soil condition time series matrix, and then perform soil condition time series enhancement to obtain a soil condition time series enhanced feature map; a growth state-soil condition characterization module 170, which is used to fuse the multi-dimensional feature map of the growth state of Phyllostachys praecox and the soil condition time series enhanced feature map to obtain a Phyllostachys praecox growth state. A growth state-soil condition comprehensive characterization feature map is obtained, and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature map is pooled to obtain a Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector; a growth state-soil condition probabilistic guidance and control module 180 is used to perform probabilistic guidance and control on the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector based on decoding weight data to obtain an optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector; a nitrogen fertilizer recommendation value generation module 190 is used to obtain a recommended value for applying nitrogen fertilizer based on the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector.
[0051] In an embodiment of the present application, the hyperspectral image acquisition module 110 of the growth status of Phyllostachys praecox is used to collect hyperspectral images of the growth status of Phyllostachys praecox through a camera. It should be understood that, considering that the hyperspectral image not only includes the visible light range, but also covers the near-infrared and short-wave infrared regions, it can provide more comprehensive information than ordinary images and reflect the growth status of crops. Based on this, in the technical solution of the present application, a hyperspectral image of the growth status of Phyllostachys praecox is collected by a camera. That is, hyperspectral data is sensitive to vegetation changes, and can identify subtle changes in the growth process of Phyllostachys praecox, such as changes in leaf color, area and density, and can reveal the spectral characteristics of crops at different growth stages, help monitor and manage the growth cycle of crops, and provide information guidance for subsequent precision fertilization.
[0052] In an embodiment of the present application, the soil condition data acquisition module 120 is used to collect soil condition data in real time through sensors to obtain a time series of soil condition data, wherein the soil condition data includes pH value, nitrogen content, phosphorus content and potassium content. Accordingly, considering that pH value is an important indicator for measuring soil acidity and alkalinity, it has an important impact on crop growth and soil microbial activity, and monitoring pH value helps to maintain soil health. Nitrogen, phosphorus and potassium in the soil are the three main nutrients required for plant growth. Monitoring the content of these nutrients helps to ensure that crops obtain appropriate nutrition and avoid excess or insufficient nutrition. Therefore, in the technical solution of the present application, soil condition data is collected in real time through sensors to obtain a time series of soil condition data, wherein the soil condition data includes pH value, nitrogen content, phosphorus content and potassium content, which helps to formulate a precise fertilization plan for Leizhu and ensure that crops obtain adequate nutrition during the critical growth period.
[0053] In the embodiment of the present application, the shallow and deep feature capture module 130 of the growth state of Phyllostachys thunbergii is used to capture the multi-scale features of the growth state of Phyllostachys thunbergii on the hyperspectral image of the growth state of Phyllostachys thunbergii to obtain the shallow feature map of the growth state of Phyllostachys thunbergii and the deep feature map of the growth state of Phyllostachys thunbergii. Specifically, in the embodiment of the present application, the shallow and deep feature capture module of the growth state of Phyllostachys thunbergii is used to: input the hyperspectral image of the growth state of Phyllostachys thunbergii into the feature capture model of the growth state of Phyllostachys thunbergii based on the hole convolutional neural network to obtain the shallow feature map of the growth state of Phyllostachys thunbergii and the deep feature map of the growth state of Phyllostachys thunbergii. It should be understood that considering that the hyperspectral image of the growth state of Phyllostachys thunbergii contains features on different scales of growth information, such as the overall shape of Phyllostachys thunbergii and the color of the leaves, or the deep features of the growth cycle, physiological and biochemical characteristics of Phyllostachys thunbergii. Therefore, in the technical solution of the present application, the hyperspectral image of the growth state of Phyllostachys thunbergii is input into the feature capture model of the growth state of Phyllostachys thunbergii based on the hole convolutional neural network to obtain the shallow feature map of the growth state of Phyllostachys thunbergii and the deep feature map of the growth state of Phyllostachys thunbergii. In particular, the feature capture model of the growth state of Phyllostachys praecox based on the dilated convolutional neural network described here is a dilated convolutional neural network model. It should be understood that the dilated convolution can capture image features at different scales, which is crucial for understanding the shallow details and deep information of the growth state of Phyllostachys praecox. That is, it is possible to learn deeper feature representations from the hyperspectral image of the growth state of Phyllostachys praecox, which may be closely related to the health status and growth conditions of Phyllostachys praecox, and provide a basis and support for the subsequent recommendation of nitrogen fertilizer.
[0054] In an embodiment of the present application, the deep feature semantic enhancement fusion module 140 of the growth state of Phyllostachys thunbergii is used to semantically enhance the deep feature map of the growth state of Phyllostachys thunbergii to obtain the deep semantic feature map of the growth state of Phyllostachys thunbergii, and then perform feature fusion on the shallow feature map of the growth state of Phyllostachys thunbergii and the deep semantic feature map of the growth state of Phyllostachys thunbergii to obtain a multi-dimensional feature map of the growth state of Phyllostachys thunbergii. Figure 2 : is a block diagram of a semantic enhancement fusion module for deep feature of growth state of Phyllostachys praecox in a nitrogen fertilizer management system for Phyllostachys praecox planting according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 2 As shown, the deep feature semantic enhancement fusion module 140 of the growth state of Phyllostachys thunbergii comprises: a deep semantic enhancement unit 141 of the growth state of Phyllostachys thunbergii, which is used to input the deep feature map of the growth state of Phyllostachys thunbergii into the semantic enhancement module of the growth state of Phyllostachys thunbergii based on the pyramid pooling module to obtain the deep semantic feature map of the growth state of Phyllostachys thunbergii; and a shallow semantic feature fusion unit 142 of the growth state of Phyllostachys thunbergii, which is used to cascade the shallow feature map of the growth state of Phyllostachys thunbergii and the deep semantic feature map of the growth state of Phyllostachys thunbergii to obtain the multi-dimensional feature map of the growth state of Phyllostachys thunbergii.
[0055] Specifically, the deep semantic enhancement unit 141 of the growth state of the thallus bamboo is used to input the deep feature map of the growth state of the thallus bamboo into the semantic enhancement module of the growth state of the thallus bamboo based on the pyramid pooling module to obtain the deep semantic feature map of the growth state of the thallus bamboo. Accordingly, considering that the deep feature map of the growth state of the thallus bamboo contains deep semantic feature information about the growth of the thallus bamboo, such as the semantic information of the growth rate of the thallus bamboo, the semantic information of the growth state, etc. Based on this, in order to be able to extract and capture the deeper semantic information of the growth state of the thallus bamboo from the deep feature map of the growth state of the thallus bamboo, in the technical solution of the present application, the deep feature map of the growth state of the thallus bamboo is input into the semantic enhancement module of the growth state of the thallus bamboo based on the pyramid pooling module to obtain the deep semantic feature map of the growth state of the thallus bamboo. In particular, the essence of the semantic enhancement module of the growth state of the thallus bamboo based on the pyramid pooling module is a pyramid pooling module, that is, the pyramid pooling module uses four average pooling windows of different scales, wherein the scales of the average pooling windows of the four different scales are 1×1, 2×2, 3×3 and 6×6. That is, the pyramid pooling module can enhance the semantic information in the deep feature map of the growth state of Phyllostachys praecox, so that the obtained deep semantic feature map of the growth state of Phyllostachys praecox is richer and more expressive.
[0056] Specifically, the shallow semantic feature fusion unit 142 of the growth state of the thallus bamboo is used to cascade the shallow feature map of the growth state of the thallus bamboo and the deep semantic feature map of the growth state of the thallus bamboo to obtain the multi-dimensional feature map of the growth state of the thallus bamboo. It should be understood that, considering that the shallow feature map of the growth state of the thallus bamboo and the deep semantic feature map of the growth state of the thallus bamboo respectively represent the shallow surface relatively intuitive, local feature information detail features and deep abstract, global semantic features of the growth state of the thallus bamboo. Therefore, in order to capture the complexity and subtle differences of the growth state of the thallus bamboo, so as to comprehensively improve the description ability of the growth state of the thallus bamboo, in the technical solution of the present application, the shallow feature map of the growth state of the thallus bamboo and the deep semantic feature map of the growth state of the thallus bamboo are cascaded to obtain the multi-dimensional feature map of the growth state of the thallus bamboo.
[0057] In an embodiment of the present application, the soil condition time series association module 150 is used to integrate the time series of the soil condition data and input it into the soil condition time series implicit association feature capture module to obtain a pH value time series implicit association feature vector, a nitrogen content time series implicit association feature vector, a phosphorus content time series implicit association feature vector and a potassium content time series implicit association feature vector. Figure 3 is a block diagram of a soil condition time series association module in a nitrogen fertilizer management system for pheasanthus praecox planting according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3As shown, the soil condition time series association module 150 includes: a soil condition time series data integration unit 151, which is used to integrate the time series of the soil condition data based on the time dimension and the parameter sample dimension to obtain the time series of the pH value, the time series of the nitrogen content, the time series of the phosphorus content and the time series of the potassium content; a soil condition time series feature capture unit 152, which is used to input the time series of the pH value, the time series of the nitrogen content, the time series of the phosphorus content and the time series of the potassium content into the soil condition time series implicit association feature capture module based on the one-dimensional convolutional neural network to obtain the pH value time series implicit association feature vector, the nitrogen content time series implicit association feature vector, the phosphorus content time series implicit association feature vector and the potassium content time series implicit association feature vector.
[0058] Specifically, the soil condition time series data integration unit 151 is used to integrate the time series of the soil condition data based on the time dimension and the parameter sample dimension to obtain the time series of the pH value, the time series of the nitrogen content, the time series of the phosphorus content, and the time series of the potassium content. Accordingly, considering that each soil condition parameter in the time series of the soil condition data changes with the passage of time, and there is interrelated characteristic information between each parameter in the time dimension. Based on this, in the technical solution of the present application, the time series of the soil condition data is integrated based on the time dimension and the parameter sample dimension to obtain the time series of the pH value, the time series of the nitrogen content, the time series of the phosphorus content, and the time series of the potassium content, so as to analyze and process the time sequence of each parameter in detail.
[0059] Specifically, the soil condition time series feature capture unit 152 is used to input the time series of the pH value, the time series of the nitrogen content, the time series of the phosphorus content and the time series of the potassium content into the soil condition time series implicit correlation feature capture module based on the one-dimensional convolutional neural network to obtain the pH value time series implicit correlation feature vector, the nitrogen content time series implicit correlation feature vector, the phosphorus content time series implicit correlation feature vector and the potassium content time series implicit correlation feature vector. It should be understood that considering that the various soil condition parameters are interrelated in the local time dimension, and considering that the one-dimensional convolutional neural network is good at processing time series data and can capture the time dependence and trend in the data, that is, the local connectivity of the convolution operation enables the network to focus on the local pattern in the time series, which is very important for discovering short-term changes in soil conditions. Based on this, in the technical solution of the present application, the time series of the pH value, the time series of the nitrogen content, the time series of the phosphorus content and the time series of the potassium content are respectively input into the soil condition temporal implicit correlation feature capture module based on the one-dimensional convolutional neural network to respectively capture and mine the temporal implicit correlation feature information of each soil parameter in the local time period, thereby obtaining the pH value temporal implicit correlation feature vector, the nitrogen content temporal implicit correlation feature vector, the phosphorus content temporal implicit correlation feature vector and the potassium content temporal implicit correlation feature vector. That is, the essence of the soil condition temporal implicit correlation feature capture module based on the one-dimensional convolutional neural network described here is a one-dimensional convolutional neural network.
[0060] In an embodiment of the present application, the soil condition time series feature enhancement module 160 is used to arrange the pH value time series implicitly associated feature vector, the nitrogen content time series implicitly associated feature vector, the phosphorus content time series implicitly associated feature vector and the potassium content time series implicitly associated feature vector into a soil condition time series matrix, and then perform soil condition time series enhancement to obtain a soil condition time series enhanced feature map. Specifically, in an embodiment of the present application, the soil condition time series feature enhancement module is used to: arrange the pH value time series implicitly associated feature vector, the nitrogen content time series implicitly associated feature vector, the phosphorus content time series implicitly associated feature vector and the potassium content time series implicitly associated feature vector into the soil condition time series matrix, and then input the soil condition time series enhancer based on the spatial attention mechanism to obtain the soil condition time series enhanced feature map. Accordingly, it is considered that there is a mutual correlation influence relationship between the pH value time series implicitly associated feature vector, the nitrogen content time series implicitly associated feature vector, the phosphorus content time series implicitly associated feature vector and the potassium content time series implicitly associated feature vector on the time scale regarding soil conditions. Therefore, in order to consider the time series changes of multiple soil indicators, so as to more comprehensively understand soil fertility and health status, in the technical solution of the present application, the pH value time series implicit correlation feature vector, the nitrogen content time series implicit correlation feature vector, the phosphorus content time series implicit correlation feature vector and the potassium content time series implicit correlation feature vector are arranged into the soil condition time series matrix. Then, in order to capture and mine key time series correlation feature information from the soil condition time series matrix, in the technical solution of the present application, further, the soil condition time series matrix is input into a soil condition time series enhancer based on a spatial attention mechanism to capture key time series information using the attention mechanism, and the soil condition time series enhanced feature map is obtained.
[0061] In an embodiment of the present application, the growth state-soil condition characterization module 170 is used to fuse the multi-dimensional feature map of the growth state of Phyllostachys praecox and the time-series enhanced feature map of the soil condition to obtain a comprehensive characterization feature map of the growth state of Phyllostachys praecox and soil conditions, and to perform pooling processing on the comprehensive characterization feature map of the growth state of Phyllostachys praecox to obtain a comprehensive characterization feature vector of the growth state of Phyllostachys praecox.
[0062] Furthermore, the multi-dimensional feature map of the growth state of Phyllostachys praecox and the time-series enhanced feature map of soil conditions are integrated to enhance the model's understanding of the interaction between the growth state of Phyllostachys praecox and soil conditions, thereby obtaining a comprehensive characterization feature map of the growth state of Phyllostachys praecox and soil conditions. The comprehensive characterization feature map of the growth state of Phyllostachys praecox and soil conditions is pooled to ensure key feature information, thereby providing rich input for the recommendation of nitrogen fertilizer amount for Phyllostachys praecox.
[0063] In an embodiment of the present application, the growth state-soil condition probabilistic guidance and control module 180 is used to perform probabilistic guidance and control on the growth state-soil condition comprehensive characterization feature vector of the thunbergii based on the decoding weight data to obtain an optimized thunbergii growth state-soil condition comprehensive characterization feature vector. In particular, in the technical solution of the present application, it is considered that the hyperspectral image and soil condition data may have quality differences due to different acquisition equipment, environmental conditions, time points and other factors, resulting in inconsistency of the original data. The hollow neural network and pyramid pooling module may not be able to fully capture all the key features of the growth state of thunbergii, especially in a complex and changeable natural environment. The time series analysis of soil condition data may be affected by a variety of dynamic factors, such as seasonal changes, weather conditions, etc., making it difficult for the temporal implicit correlation feature capture module to accurately model. When fusing the multi-dimensional feature map of the growth state of thunbergii and the temporal enhanced feature map of soil conditions, if the fusion strategy is inappropriate, it may lead to the loss of important information or the amplification of outliers. Global pooling based on the feature matrix may lose local feature information, especially when the feature map has non-uniformity in spatial distribution, and then the thunbergii growth state-soil condition comprehensive characterization feature vector has local abnormal distribution. If there is a local abnormal distribution of the eigenvector of the comprehensive characterization of the growth state of Phyllostachys praecox-soil conditions, the decoder may not be able to accurately predict the recommended value for the application of nitrogen fertilizer during decoding regression, resulting in category outlier shift. Category outlier shift will affect the prediction accuracy and stability of the decoder, causing the recommended value to deviate from the actual optimal amount of fertilizer, affecting the healthy growth of Phyllostachys praecox and soil management. In order to solve this problem, in the technical solution of the present application, the eigenvector of the comprehensive characterization of the growth state of Phyllostachys praecox-soil conditions is probabilistically guided and regulated based on the decoding weight data to obtain an optimized eigenvector of the comprehensive characterization of the growth state of Phyllostachys praecox-soil conditions.
[0064] Specifically, the growth state-soil condition probabilistic guidance and control module is used to: determine the decoding weight matrix of the decoder; perform matrix decoupling on the decoding weight matrix in units of row vectors to obtain a set of decoding weight vectors; calculate the growth state-soil condition comprehensive representation metric learning loss information item of the thunbergii between each decoding weight vector in the set of decoding weight vectors and the thunbergii growth state-soil condition comprehensive representation feature vector to obtain the thunbergii growth state-soil condition comprehensive representation metric learning compensation representation vector The growth state-soil condition comprehensive representation metric learning compensation representation vector is input Activate function to obtain probabilistic Phyllostachys praecox growth state-soil condition comprehensive characterization metric learning compensation weight vector; calculate the position point multiplication of the probabilistic Phyllostachys praecox growth state-soil condition comprehensive characterization metric learning compensation weight vector and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector to obtain the point product feature vector, and calculate the weighted sum of the point product feature vector and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector to obtain the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector.
[0065] In the embodiment of the present application, specifically, the following loss information calculation formula is used to calculate the loss information item of the growth state of Phyllostachys praecox-soil condition comprehensive characterization metric learning between each decoding weight vector in the set of decoding weight vectors and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector to obtain the growth state of Phyllostachys praecox-soil condition comprehensive characterization metric learning compensation representation vector; wherein, the loss information calculation formula is:
[0066]
[0067] in, represents the first in the set of decoding weight vectors The decoding weight vector The eigenvalues at the positions, The first characterization vector representing the comprehensive characterization of the growth state and soil conditions of Phyllostachys praecox is represented by The eigenvalues at the positions, represents a natural constant, represents the length of the feature vector, The first one represents the comprehensive representation metric learning compensation representation vector of the growth state-soil condition of Phyllostachys praecox. The feature value at each position.
[0068] That is, the pheasanthus growth state-soil condition comprehensive characterization feature vector is probabilistically guided and regulated based on the decoding weight data, which first determines the decoding weight matrix of the decoder to understand the decoder distinguishing different features. Subsequently, the decoding weight matrix is row-vector decoupled to obtain a set of decoding weight vectors to allow the model to independently analyze the weight of each category and provide more refined control for feature compensation. Next, the pheasanthus growth state-soil condition comprehensive characterization metric learning loss information item is calculated. This process involves quantitative analysis of the interaction between each weight vector and the pheasanthus growth state-soil condition comprehensive characterization feature vector. Through the calculation of a specific loss function, the inconsistency between the feature vector and the decoding is captured, providing a quantitative indicator for compensation. Next, the pheasanthus growth state-soil condition comprehensive characterization metric learning compensation representation vector is input into the activation function, nonlinearly transformed, and converted into a probabilistic form. The application of the activation function introduces nonlinearity, enhances the model's ability to interpret the compensation weight, and makes the compensation weight more consistent with the actual distribution characteristics. Then, the compensation representation vector of the comprehensive characterization metric learning of the growth state of Phyllostachys thunbergii-soil condition processed by the activation function is used to calculate the probabilistic compensation weight vector of the comprehensive characterization metric learning of the growth state of Phyllostachys thunbergii-soil condition. This step provides probabilistic guidance for the adjustment of the feature vector in a probabilistic form, making the optimization process more in line with statistical characteristics, thereby enhancing the accuracy of the model's feature adjustment. Finally, by calculating the dot product of the probabilistic comprehensive characterization metric learning compensation weight vector of the growth state of Phyllostachys thunbergii-soil condition and the comprehensive characterization feature vector of the growth state of Phyllostachys thunbergii-soil condition, and then calculating the weighted sum with the original comprehensive characterization feature vector of the growth state of Phyllostachys thunbergii-soil condition, the optimized comprehensive characterization feature vector of the growth state of Phyllostachys thunbergii-soil condition is obtained. This step combines the original feature information and the compensation information to generate an adjusted feature vector, reducing the impact of abnormal distribution and improving the accuracy of decoding.
[0069] In an embodiment of the present application, the nitrogen fertilizer recommendation value generation module 190 is used to obtain the recommended value for applying nitrogen fertilizer based on the optimized phenotypic bamboo growth state-soil condition comprehensive characterization feature vector. Specifically, in an embodiment of the present application, the nitrogen fertilizer recommendation value generation module is used to: input the optimized phenotypic bamboo growth state-soil condition comprehensive characterization feature vector into the decoder to obtain the recommended value for applying nitrogen fertilizer. That is, the optimized phenotypic bamboo growth state-soil condition comprehensive characterization feature vector is classified and processed to comprehensively and intelligently obtain the recommended value for applying nitrogen fertilizer. In this way, the system can quickly respond to changes in the growth environment of phenotypic bamboo according to the actual growth state and soil conditions of phenotypic bamboo, so as to intelligently recommend the application amount of nitrogen fertilizer, realize precise fertilization, and avoid excessive or insufficient fertilization, which helps to reduce waste and reduce potential impacts on the environment, thereby promoting healthy growth of crops.
[0070] In summary, the nitrogen fertilizer management system 100 for the cultivation of Phyllostachys praecox based on the embodiment of the present application is explained, which collects the hyperspectral image of the growth state of Phyllostachys praecox through a camera, and collects soil condition data (PH value, nitrogen content, phosphorus content and potassium content) in real time through a sensor, and uses image recognition technology and data analysis algorithm based on deep learning to perform feature analysis of the hyperspectral image of the growth state of Phyllostachys praecox and time series association of the soil condition data, and fuses the features of the two, so as to comprehensively and intelligently obtain the recommended value for the application of nitrogen fertilizer. In this way, the system can quickly respond to changes in the growth environment of Phyllostachys praecox according to the actual growth state and soil conditions of Phyllostachys praecox, so as to intelligently recommend the application amount of nitrogen fertilizer, realize precise fertilization, and avoid excessive or insufficient fertilization, which helps to reduce waste and reduce potential impacts on the environment, thereby promoting the healthy growth of crops.
[0071] Figure 4 Flow chart of the nitrogen fertilizer management method for the cultivation of Phyllostachys praecox according to an embodiment of the present application. Figure 4As shown, according to the embodiment of the present application, the nitrogen fertilizer management method for the planting of Phyllostachys praecox comprises: S110, collecting a hyperspectral image of the growth state of Phyllostachys praecox by a camera; S120, collecting soil condition data in real time by a sensor to obtain a time series of soil condition data, wherein the soil condition data comprises pH value, nitrogen content, phosphorus content and potassium content; S130, performing multi-scale feature capture of the growth state of Phyllostachys praecox on the hyperspectral image of the growth state of Phyllostachys praecox to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox; S140, after performing semantic enhancement on the deep feature map of the growth state of Phyllostachys praecox to obtain a deep semantic feature map of the growth state of Phyllostachys praecox, the shallow feature map of the growth state of Phyllostachys praecox and the deep semantic feature map of the growth state of Phyllostachys praecox are feature fused to obtain a multi-dimensional feature map of the growth state of Phyllostachys praecox; S150, performing data integration on the time series of the soil condition data and inputting it into a soil condition temporal implicit correlation feature capture module to obtain a pH value temporal implicit correlation feature vector, a nitrogen content temporal implicit correlation feature vector, a phosphorus content temporal implicit correlation feature vector, and a phosphorus content temporal implicit correlation feature vector. S160, arranging the pH value time series implicit association feature vector, the nitrogen content time series implicit association feature vector, the phosphorus content time series implicit association feature vector and the potassium content time series implicit association feature vector into a soil condition time series matrix, and then performing soil condition time series enhancement to obtain a soil condition time series enhanced feature map; S170, fusing the multi-dimensional feature map of the growth state of Phyllostachys praecox and the soil condition time series enhanced feature map to obtain a Phyllostachys praecox growth state-soil condition comprehensive characterization feature map, and performing pooling processing on the Phyllostachys praecox growth state-soil condition comprehensive characterization feature map to obtain a Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector; S180, performing probabilistic guidance and regulation on the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector based on decoding weight data to obtain an optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector; S190, based on the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector, obtaining a recommended value for applying nitrogen fertilizer.
[0072] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned nitrogen fertilizer management method for the cultivation of Phyllostachys praecox have been described in the above reference. Figures 1 to 3 It has been described in detail in the description of the nitrogen fertilizer management system for pheasanthus praecox cultivation, and therefore, its repeated description will be omitted.
[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit of the technical solution of the present invention.
Claims
1. A nitrogen fertilizer management system for growing Phyllostachys praecox, characterized in that: include: A hyperspectral image acquisition module for the growth status of Phyllostachys praecox, used for acquiring a hyperspectral image of the growth status of Phyllostachys praecox through a camera; A soil condition data acquisition module, used to acquire soil condition data in real time through sensors to obtain a time series of soil condition data, wherein the soil condition data includes pH value, nitrogen content, phosphorus content and potassium content; A shallow-depth feature capture module for the growth state of Phyllostachys praecox is used to capture the multi-scale features of the growth state of Phyllostachys praecox on the hyperspectral image of the growth state of Phyllostachys praecox to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox; A semantic enhancement fusion module for deep features of the growth state of Phyllostachys thunbergii, which is used to semantically enhance the deep feature map of the growth state of Phyllostachys thunbergii to obtain a deep semantic feature map of the growth state of Phyllostachys thunbergii, and then perform feature fusion on the shallow feature map of the growth state of Phyllostachys thunbergii and the deep semantic feature map of the growth state of Phyllostachys thunbergii to obtain a multi-dimensional feature map of the growth state of Phyllostachys thunbergii; A soil condition time series association module is used to integrate the time series of the soil condition data and input the data into a soil condition time series implicit association feature capture module to obtain a pH value time series implicit association feature vector, a nitrogen content time series implicit association feature vector, a phosphorus content time series implicit association feature vector, and a potassium content time series implicit association feature vector; A soil condition time series feature enhancement module, for arranging the pH value time series implicit correlation feature vector, the nitrogen content time series implicit correlation feature vector, the phosphorus content time series implicit correlation feature vector and the potassium content time series implicit correlation feature vector into a soil condition time series matrix, and then performing soil condition time series enhancement to obtain a soil condition time series enhanced feature map; A growth state-soil condition characterization module, used for fusing the multi-dimensional feature map of the growth state of Phyllostachys praecox and the time series enhanced feature map of the soil condition to obtain a comprehensive characterization feature map of the growth state of Phyllostachys praecox and soil condition, and performing pooling processing on the comprehensive characterization feature map of the growth state of Phyllostachys praecox to obtain a comprehensive characterization feature vector of the growth state of Phyllostachys praecox; A growth state-soil condition probabilistic guidance and control module, used for performing probabilistic guidance and control on the growth state-soil condition comprehensive characterization feature vector of Phyllostachys praecox based on decoding weight data to obtain an optimized growth state-soil condition comprehensive characterization feature vector of Phyllostachys praecox; The nitrogen fertilizer recommendation value generation module is used to obtain the recommended value for applying nitrogen fertilizer based on the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector.
2. The nitrogen fertilizer management system for planting Phyllostachys praecox according to claim 1, characterized in that: The module for capturing the depth and shallowness characteristics of the growth state of Phyllostachys praecox is used for: The hyperspectral image of the growth state of Phyllostachys praecox is input into a Phyllostachys praecox growth state feature capture model based on a hole convolutional neural network to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox.
3. The nitrogen fertilizer management system for planting Phyllostachys praecox according to claim 2, characterized in that: The deep feature semantic enhancement fusion module of the growth state of Phyllostachys praecox includes: A deep semantic enhancement unit for the growth state of Phyllostachys praecox is used to input the deep feature map of the growth state of Phyllostachys praecox into a semantic enhancement module for the growth state of Phyllostachys praecox based on a pyramid pooling module to obtain the deep semantic feature map of the growth state of Phyllostachys praecox; The shallow semantic feature fusion unit of the growth state of Phyllostachys praecox is used to cascade the shallow feature map of the growth state of Phyllostachys praecox and the deep semantic feature map of the growth state of Phyllostachys praecox to obtain the multi-dimensional feature map of the growth state of Phyllostachys praecox.
4. The nitrogen fertilizer management system for planting Phyllostachys praecox according to claim 3, characterized in that: The soil condition time series association module includes: A soil condition time series data integration unit, used for integrating the time series of the soil condition data based on the time dimension and the parameter sample dimension to obtain a time series of pH value, a time series of nitrogen content, a time series of phosphorus content and a time series of potassium content; The soil condition time series feature capture unit is used to input the time series of the pH value, the time series of the nitrogen content, the time series of the phosphorus content and the time series of the potassium content into a soil condition time series implicit association feature capture module based on a one-dimensional convolutional neural network to obtain the pH value time series implicit association feature vector, the nitrogen content time series implicit association feature vector, the phosphorus content time series implicit association feature vector and the potassium content time series implicit association feature vector.
5. The nitrogen fertilizer management system for planting Phyllostachys praecox according to claim 4, characterized in that: The soil condition time series feature enhancement module is used to: The pH value time series implicit association feature vector, the nitrogen content time series implicit association feature vector, the phosphorus content time series implicit association feature vector and the potassium content time series implicit association feature vector are arranged into the soil condition time series matrix and then input into the soil condition time series enhancer based on the spatial attention mechanism to obtain the soil condition time series enhanced feature map.
6. The nitrogen fertilizer management system for planting Phyllostachys praecox according to claim 5, characterized in that: The growth state-soil condition probabilistic guidance and control module is used to: Determine a decoding weight matrix of a decoder; Decoupling the decoding weight matrix in units of row vectors to obtain a set of decoding weight vectors; Calculate the loss information item of the comprehensive characterization metric learning of the growth state of Phyllostachys thunbergii-soil condition between each decoding weight vector in the set of decoding weight vectors and the comprehensive characterization feature vector of the growth state of Phyllostachys thunbergii-soil condition to obtain the comprehensive characterization metric learning compensation representation vector of the growth state of Phyllostachys thunbergii-soil condition The growth state-soil condition comprehensive representation metric learning compensation representation vector is input The activation function is used to obtain a probabilistic comprehensive representation metric of the growth state of Phyllostachys praecox and soil conditions to learn the compensation weight vector; Calculate the positional point multiplication of the probabilistic Phyllostachys praecox growth state-soil condition comprehensive characterization metric learning compensation weight vector and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector to obtain a point product feature vector, and calculate the weighted sum of the point product feature vector and the Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector to obtain the optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector.
7. The nitrogen fertilizer management system for planting Phyllostachys praecox according to claim 6, characterized in that: The nitrogen fertilizer recommendation value generation module is used for: The optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector is input into the decoder to obtain the recommended value for applying nitrogen fertilizer.
8. A nitrogen fertilizer management method for Phyllostachys praecox planting, characterized in that: include: The camera is used to collect hyperspectral images of the growth status of Phyllostachys praecox; Collecting soil condition data in real time through sensors to obtain a time series of soil condition data, wherein the soil condition data includes pH value, nitrogen content, phosphorus content and potassium content; Capturing multi-scale features of the growth state of Phyllostachys praecox on the hyperspectral image of the growth state of Phyllostachys praecox to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox; After semantic enhancement is performed on the deep feature map of the growth state of Phyllostachys praecox to obtain a deep semantic feature map of the growth state of Phyllostachys praecox, feature fusion is performed on the shallow feature map of the growth state of Phyllostachys praecox and the deep semantic feature map of the growth state of Phyllostachys praecox to obtain a multi-dimensional feature map of the growth state of Phyllostachys praecox; The soil condition data are integrated in time series and then input into a soil condition time series implicit correlation feature capture module to obtain a pH value time series implicit correlation feature vector, a nitrogen content time series implicit correlation feature vector, a phosphorus content time series implicit correlation feature vector, and a potassium content time series implicit correlation feature vector; Arranging the pH value time series implicit correlation feature vector, the nitrogen content time series implicit correlation feature vector, the phosphorus content time series implicit correlation feature vector and the potassium content time series implicit correlation feature vector into a soil condition time series matrix, and then performing soil condition time series enhancement to obtain a soil condition time series enhancement feature map; The multi-dimensional feature map of the growth state of Phyllostachys praecox and the time series enhanced feature map of the soil condition are integrated to obtain a feature map of the growth state of Phyllostachys praecox and soil condition, and pooling is performed on the feature map of the growth state of Phyllostachys praecox and soil condition to obtain a feature vector of the growth state of Phyllostachys praecox and soil condition; The Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector is subjected to probabilistic guidance regulation based on decoding weight data to obtain an optimized Phyllostachys praecox growth state-soil condition comprehensive characterization feature vector; Based on the optimized comprehensive characterization feature vector of Phyllostachys praecox growth state and soil conditions, the recommended value for applying nitrogen fertilizer is obtained.
9. The nitrogen fertilizer management method for planting Phyllostachys praecox according to claim 8, characterized in that: Capturing the multi-scale features of the growth state of Phyllostachys praecox on the hyperspectral image of the growth state of Phyllostachys praecox to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox, including: The hyperspectral image of the growth state of Phyllostachys praecox is input into a Phyllostachys praecox growth state feature capture model based on a hole convolutional neural network to obtain a shallow feature map of the growth state of Phyllostachys praecox and a deep feature map of the growth state of Phyllostachys praecox.
10. The nitrogen fertilizer management method for Phyllostachys praecox planting according to claim 9, characterized in that: After semantic enhancement is performed on the deep feature map of the growth state of Phyllostachys praecox to obtain the deep semantic feature map of the growth state of Phyllostachys praecox, feature fusion is performed on the shallow feature map of the growth state of Phyllostachys praecox and the deep semantic feature map of the growth state of Phyllostachys praecox to obtain the multi-dimensional feature map of the growth state of Phyllostachys praecox, including: Inputting the deep feature map of the growth state of Phyllostachys praecox into a semantic enhancement module of the growth state of Phyllostachys praecox based on a pyramid pooling module to obtain the deep semantic feature map of the growth state of Phyllostachys praecox; The shallow feature map of the growth state of Phyllostachys praecox and the deep semantic feature map of the growth state of Phyllostachys praecox are cascaded to obtain a multi-dimensional feature map of the growth state of Phyllostachys praecox.
Citation Information
Patent Citations
Corn high-yield cultivation method
CN116612435A
Crop planting management method
CN116721389A
Intelligent water, fertilizer and pesticide integrated device for sweet potatoes and control method
CN117730669A
Method for improving soil environment of hillside cultivated land
CN117743975A
Method for improving nitrogen fertilizer utilization rate in corn planting
CN117764762A
Cited By
Intelligent fire-fighting patrol alarm system and method based on Internet of Things technology
CN118767398A
Steel structure building construction quality control system and method
CN118840019A
