WO2024040845A1 - 一种联合收获机大数据重构方法与装置 - Google Patents

一种联合收获机大数据重构方法与装置 Download PDF

Info

Publication number
WO2024040845A1
WO2024040845A1 PCT/CN2022/144067 CN2022144067W WO2024040845A1 WO 2024040845 A1 WO2024040845 A1 WO 2024040845A1 CN 2022144067 W CN2022144067 W CN 2022144067W WO 2024040845 A1 WO2024040845 A1 WO 2024040845A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
big data
harvester
values
combine harvester
Prior art date
Application number
PCT/CN2022/144067
Other languages
English (en)
French (fr)
Inventor
石茂林
Original Assignee
江苏大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江苏大学 filed Critical 江苏大学
Publication of WO2024040845A1 publication Critical patent/WO2024040845A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2337—Non-hierarchical techniques using fuzzy logic, i.e. fuzzy clustering

Definitions

  • the invention belongs to the cross-technical field of combine harvesters, big data, and artificial intelligence, and particularly relates to a method and device for reconstructing big data of combine harvesters.
  • the combine harvester is a key piece of equipment that ensures timely harvesting of staple foods such as rice, wheat, and corn.
  • the performance of its operation process is of great significance to the safety of my country's grain production.
  • the operating performance of the combine harvester is mainly determined by the working parts composed of headers, conveyor troughs, threshing, cleaning and other devices.
  • the interaction between different devices is very complex and difficult to accurately characterize through physical derivation, numerical simulation and other means.
  • the adjustment of operating parameters of the combine harvester mostly relies on the experience of the operator, which is difficult to accurately match the real-time harvesting conditions, and the operating performance is good and bad.
  • Data reconstruction is one of the hot topics in the field of big data. It refers to estimating the true values of null values, outliers, and outliers based on the intrinsic characteristics of the data, so as to reconstruct the inconsistencies that contain null values, outliers, and outliers.
  • the data is mostly reconstructed from the perspective of data spatial distribution. For example, estimating the true values through the spatial distribution density and mean value of parameters corresponding to null values, outliers, and outliers does not take into account the correlation between data parameters, making it difficult to achieve accurate reconstruction of data.
  • one of the purposes of one aspect of the present invention is to provide a combined harvester big data reconstruction method, based on the fuzzy clustering framework, using the Kriging method to describe the parameter correlation relationship of the complete data of the combine harvester, through The correlation between data parameters, as well as data spatial distribution density, mean and other information jointly infer the true values of null values, wild values, and outliers, improve the accuracy of combine harvester big data reconstruction, and provide a basis for subsequent combine harvester big data Provide data support for modeling and analysis tasks.
  • One of the purposes of one mode of the present invention is to provide a device for realizing a combined harvester big data reconstruction method, using a data input module to obtain agricultural machinery field operation data; the data reconstruction module is used to receive a combined data input from the data input module. Harvester big data, and reconstruct the data according to the combine harvester big data reconstruction method; the data output module outputs the reconstructed data of the combine harvester big data; the data storage module is used to save the original data and reconstruction of the combine harvester big data Data and membership matrix improve the accuracy of combine harvester big data reconstruction and provide complete data support for subsequent combine harvester big data modeling and analysis tasks. .
  • the present invention achieves the above technical objectives through the following technical means.
  • a combined harvester big data reconstruction method includes the following steps:
  • Step S1 Under the fuzzy clustering framework, use the kriging method to characterize the correlation between the parameters of the complete data of the combine harvester's operating parameters that does not include null values, wild values or abnormal values, and obtain the complete data of the combine harvester. Membership;
  • Step S2 Based on the complete data of the combine harvester and its membership degree calculation prototype obtained in step S1, and then obtain the membership degree of the incomplete data of the operation parameters containing null values, wild values and or abnormal values;
  • Step S3 Estimate the true values of null values, outliers and/or outliers respectively based on the calculated class prototypes, normal values of incomplete data, and membership degrees of incomplete data to realize the reconstruction of combine harvester big data.
  • step S1 the kriging method is used to describe the correlation between data parameters, and the membership degree of the complete data is calculated through the following formula:
  • u k,i is the membership degree of the k-th complete data to the i-th subclass
  • c is the number of subcategories of clustering
  • y k is the value of the prediction target parameter of the kth complete data
  • step S1 the number c of clustering subclasses is 3, and the membership threshold ⁇ is generally taken to be 0.5.
  • step S2 the class prototype is calculated by the following formula:
  • n is the number of samples of complete data
  • v i is the class prototype of the i-th subclass
  • step S2 the membership degree of incomplete data is calculated through the following formula:
  • ⁇ k,i is the membership degree of the k-th incomplete data to the i-th subclass
  • z k,j represents the value of the j-th parameter of the k-th data
  • v i,j represents the value of the j-th parameter of the i-th class prototype
  • X M represents the set of null values, wild values, and outliers.
  • X P represents a set of normal values
  • s is the number of parameters.
  • step S3 the true values of null values, wild values, and outliers are estimated through the following formula to realize the reconstruction of the combine harvester big data:
  • x k,j ⁇ X M represents the null value, wild value or outlier value in the data
  • v i,j represents the value of the j-th parameter of the i-th class prototype
  • ⁇ k,i represents the membership degree of the k-th incomplete data to the i-th subclass.
  • a device for implementing the combined harvester big data reconstruction method including a data input module, a data reconstruction module, a data output module and a data storage module;
  • the data input module is used to obtain agricultural machinery field operation data
  • the data reconstruction module is used to receive the combine harvester big data input by the data input module, and reconstruct the data according to the combine harvester big data reconstruction method;
  • the data output and storage module is used to output the reconstructed data of the combine harvester big data and save the original data, reconstructed data and membership matrix of the combine harvester big data.
  • the data input module is connected to the combine harvester vehicle-mounted monitoring system, remote monitoring system or smart farm platform, and is used to receive the combine harvester big data with calibrated null values, wild values, and abnormal values.
  • the data reconstruction module is used to use the kriging method to describe the correlation between data parameters under the fuzzy clustering framework, obtain the membership degree of the complete data, and then calculate the class based on the complete data and its membership degree. prototype, and then obtain the membership degree of the incomplete data, and finally estimate the true values of the null value, outlier value and outlier value based on the normal value and incomplete membership degree of the complete data corresponding to the class prototype, null value, outlier value and outlier value, Realize the reconstruction of combine harvester big data.
  • the data output and storage module outputs the reconstructed big data and membership matrix of the combined harvester to the cloud, and saves the original big data, reconstructed big data, and membership matrix of the combined harvester to the combined harvester vehicle.
  • Monitoring System the data output and storage module outputs the reconstructed big data and membership matrix of the combined harvester to the cloud, and saves the original big data, reconstructed big data, and membership matrix of the combined harvester to the combined harvester vehicle.
  • a combined harvester big data reconstruction method is provided.
  • the Kriging method is used to characterize the complete data of the combined harvester's operating parameters that do not include null values, outliers or outliers.
  • a device for realizing a combined harvester big data reconstruction method is provided.
  • the data input module is used to obtain agricultural machinery field operation data;
  • the data reconstruction module is used to receive the combined harvester big data input by the data input module. data, and reconstruct the data according to the combine harvester big data reconstruction method;
  • the data output module outputs the reconstructed data of the combine harvester big data;
  • the data saving module is used to save the original data, reconstructed data, and subordinate data of the combine harvester big data. degree matrix, ultimately improving the accuracy of combine harvester big data reconstruction and providing complete data support for subsequent combine harvester big data modeling and analysis tasks.
  • Figure 1 is a schematic flow diagram of an embodiment of the present invention.
  • Figure 2 is a schematic diagram of the complete data membership calculation process according to an embodiment of the present invention.
  • Figure 3 is a schematic diagram of a combine harvester big data reconstruction device according to an embodiment of the present invention.
  • Figure 4 is a partial complete data membership matrix diagram obtained by the combine harvester big data reconstruction method according to an embodiment of the present invention.
  • Figure 5 is a comparison chart of experimental results between the combine harvester big data reconstruction method and other data reconstruction methods according to one embodiment of the present invention.
  • first and second are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of indicated technical features. Therefore, features defined as “first” and “second” may explicitly or implicitly include one or more of these features.
  • “plurality” means two or more than two, unless otherwise explicitly and specifically limited.
  • connection In the present invention, unless otherwise clearly stated and limited, the terms “installation”, “connection”, “connection”, “fixing” and other terms should be understood in a broad sense. For example, it can be a fixed connection or a detachable connection. , or integrally connected; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be an internal connection between two components.
  • connection connection
  • fixing and other terms should be understood in a broad sense. For example, it can be a fixed connection or a detachable connection. , or integrally connected; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be an internal connection between two components.
  • the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
  • Figure 1 shows a combined harvester big data reconstruction method, including the following steps:
  • Step S1 Under the fuzzy clustering framework, use the kriging method to characterize the correlation between the parameters of the complete data of the combine harvester's operating parameters that does not include null values, wild values or abnormal values, and obtain the complete data of the combine harvester. Membership;
  • Step S2 Based on the complete data of the combine harvester and its membership degree calculation prototype obtained in step S1, and then obtain the membership degree of the incomplete data of the operation parameters containing null values, wild values and or abnormal values;
  • Step S3 Estimate the true values of null values, outliers and/or outliers respectively based on the calculated class prototypes, normal values of incomplete data, and membership degrees of incomplete data to realize the reconstruction of combine harvester big data.
  • the calculation of the membership degree of the complete data includes the following steps:
  • Step S1) Set the number of clustering subclasses c, the prediction target parameter y in the complete data, the membership threshold ⁇ , the maximum number of iterations g max , and the initial number of iterations g set to 1;
  • Step S2 Use the fuzzy c-means algorithm to perform initial clustering on the complete data, and obtain the membership matrix U FCM as the initial membership matrix U (0) ;
  • Step S3) For each subcategory, select the complete data with a membership value greater than ⁇ as the training data set for the subcategory;
  • Step S4) Check whether the training data set of each subclass is an empty set. If it is an empty set, randomly generate a membership matrix and return to step S3). If it is not an empty set, perform the next step;
  • Step S5) For each subcategory, use the kriging method to establish a prediction model
  • Step S6 Use the prediction model of each subcategory to obtain the prediction values of all complete data and construct the following prediction value matrix:
  • Y is the predicted value matrix
  • Step S7) Calculate the complete data membership degree using the formula:
  • u k,i represents the membership degree of the k-th complete data to the i-th subclass
  • y k is the true value of the kth complete data prediction target parameter
  • y k,t is the predicted value of the k-th complete data by the t-th subclass prediction model
  • the use of kriging method to establish a prediction model includes the following steps:
  • Step S5.1 Assuming that x and y satisfy the following relationship, the kriging method establishes a prediction model:
  • y is the prediction target parameter in the complete data
  • x is a vector composed of other parameters except the predicted target parameters
  • f j (x) is the jth basis function
  • ⁇ j is the coefficient of the jth basis function
  • p is the number of basis functions
  • Z(x) is a Gaussian process that satisfies the following conditions:
  • ⁇ 2 is the sample variance
  • R( ⁇ ,x i ,x j ) is the correlation matrix
  • ⁇ is the parameter vector of the correlation matrix
  • x i is a vector composed of the remaining parameters of the i-th data except the predicted target parameters
  • x j is a vector composed of the remaining parameters of the jth data except the predicted target parameters
  • Step S5.2 Construct the basis function matrix F:
  • n is the number of samples in the training data
  • the basis function matrix F can be rewritten as:
  • ⁇ is a vector composed of basis function coefficients
  • Z is a vector composed of Gaussian processes of the training data:
  • Step S5.4 The output estimated value of the sample x new to be predicted is:
  • Y is a vector composed of training data prediction target parameters
  • c(x) T is the coefficient vector
  • Step S5.5) The prediction error of sample x new is as follows:
  • y(x new ) is the true output value of sample x new ;
  • Step S5.7 Establish the following optimization problem to solve the optimal parameters of the model:
  • ⁇ is the Lagrange multiplier
  • Step S5.9 The estimated value of x is obtained by the following formula:
  • Step S5.10) The output value of sample x new is as follows:
  • the step S2 uses the following formula to calculate the class prototype:
  • n is the number of samples of complete data
  • v i is the class prototype of the i-th subclass
  • the step S2 uses the following formula to calculate the membership degree of incomplete data:
  • ⁇ k,i is the membership degree of the k-th incomplete data to the i-th subclass
  • z k,j represents the value of the j-th parameter of the k-th data
  • v i,j represents the value of the j-th parameter of the i-th class prototype
  • X M represents the set of null values, wild values, and outliers
  • X P represents a set of normal values
  • s is the number of parameters.
  • the step S2 uses the following formula to estimate the true values of null values, wild values, and outliers:
  • x j,k ⁇ X M represents null values, wild values or outliers in the data
  • v i,j represents the value of the j-th parameter of the i-th class prototype
  • ⁇ k,i represents the membership degree of the k-th incomplete data to the i-th subclass.
  • the mean of the normal values of the parameters corresponding to the null values, outliers, and outliers of the combine harvester is regarded as its normal value.
  • the fuzzy c-means algorithm is used to cluster the complete data, and the corresponding parameter values of the nearest class prototypes of null values, outliers, and outliers are regarded as their true values, as shown below.
  • x is a null value, wild value or outlier value in the data
  • vi is the parameter value corresponding to the null value, wild value or outlier value in the nearest neighbor class prototype obtained through the fuzzy c-means algorithm.
  • Step (1) Set the number of clustering subclasses c, the maximum number of iterations g max , and the number of iterations g set to 1;
  • Step (2) Generate a random matrix U as the initial membership matrix
  • n is the number of complete data samples
  • c is the number of subcategories
  • ⁇ i,k represents the degree of membership of the i-th complete data to the j-th subclass, with values ranging from 0 to 1 and satisfying the following conditions:
  • Step (3) Calculate the class prototype of each subclass
  • vi is the class prototype of the i-th subclass
  • z k is the kth complete data
  • Step (4) Update the membership matrix
  • the crop harvested by the combine harvester is rice
  • the data sample capacity is 150, including 11 parameters such as header height, forward speed, fan speed, breakaway gap, and cleaning loss.
  • the combined harvester big data reconstruction method, parameter mean reconstruction method, and nearest neighbor prototype reconstruction method are used to estimate the true values of null values, outliers, and outliers in the combine harvester big data.
  • the parameters of a combined harvester big data reconstruction method provided by the present invention are set as follows:
  • the number of clusters is set to 3;
  • the prediction target parameter of the kriging method is set to the number of cleaning losses
  • the membership threshold is set to 0.5;
  • the maximum number of iterations is set to 20.
  • the parameters of the nearest neighbor prototype reconstruction method are set as follows:
  • the number of clusters is set to 3;
  • the maximum number of iterations is set to 20.
  • the average reconstruction error of the combine harvester big data reconstruction method of the present invention is 1.53%, which is significantly lower than the parameter mean reconstruction method (16.27%) and the nearest neighbor prototype reconstruction method (7.85%).
  • the reconstruction errors of the combine harvester big data reconstruction method of the present invention are all below 3.00%, the errors of the parameter mean reconstruction method are all higher than 10.00%, and the partial data reconstruction errors of the nearest neighbor prototype reconstruction method are lower than 5.00%. But it is still higher than the combine harvester big data reconstruction method of the present invention. It can be seen that the combined harvester big data reconstruction method of the present invention improves the accuracy of the reconstruction of big data null values, outliers, and outliers, and can achieve more accurate combine harvester big data reconstruction.
  • this embodiment provides a device for implementing the combine harvester big data reconstruction method described in Embodiment 1, and therefore has the beneficial effects of Embodiment 1, which will not be described again here.
  • the device of the combine harvester big data reconstruction method includes a data input module, a data reconstruction module, a data output and a data storage module.
  • the device can be integrated into, but is not limited to, combine harvester on-board monitoring, remote monitoring, smart farm and other platforms.
  • the data input module is connected to platforms such as combine harvester vehicle-mounted monitoring, remote monitoring, and smart farms, and receives combined harvester big data that has been calibrated for null values, wild values, and abnormal values;
  • the data reconstruction module reconstructs the combine harvester big data according to the combined harvester big data reconstruction method of the present invention
  • the data output and data storage module outputs the reconstructed data of the combine harvester big data, and saves the original data, reconstructed data and membership matrix of the combine harvester big data.
  • the device for implementing the combined harvester big data reconstruction method of the present invention is integrated into the combined harvester vehicle-mounted monitoring system, and is mainly used to obtain the combined harvester data before transmitting the data to the cloud. Big data is reconstructed.
  • the combined harvester on-board monitoring system uses the data acquisition devices of each subsystem to obtain the combine harvester big data and summarizes it through the CAN bus.
  • the combined harvester on-board monitoring system calibrates null values, outliers and outliers in the data through integrated null value judgment, outlier identification, anomaly identification and other algorithms.
  • the combined harvester big data reconstruction method provided by the present invention is used to reconstruct the input combine harvester big data.
  • the combined harvester's reconstructed big data and membership matrix are output to the cloud, and the combined harvester's reconstructed big data and membership matrix are output to the cloud.
  • the original big data, reconstructed big data, and membership matrix are saved to the combined harvester on-board monitoring system.

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Automation & Control Theory (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明提供了一种联合收获机大数据重构方法与装置，该方法基于模糊聚类框架，利用克里金法刻画联合收获机完整数据的参数关联关系，通过数据参数之间的关联关系，以及数据空间分布密度、均值等信息共同推断空值、野值、异常值的真实数值。该装置，采用数据输入模块用于获取农机现场作业数据；数据重构模块用于接收数据输入模块输入的联合收获机大数据，并根据联合收获机大数据重构方法重构数据；数据输出模块输出联合收获机大数据的重构数据；数据保存模块用于保存联合收获机大数据的原始数据、重构数据、隶属度矩阵，提高了联合收获机大数据重构的精准度，为后续联合收获机大数据建模与分析任务提供数据支撑。

Description

一种联合收获机大数据重构方法与装置 技术领域
本发明属于联合收获机、大数据、人工智能的交叉技术领域，特别涉及一种联合收获机大数据重构方法与装置。
背景技术
联合收获机是保障水稻、小麦、玉米等主粮及时收获的关键装备，其作业过程的性能表现对于我国粮食生产安全具有重要意义。联合收获机的作业性能主要由割台、输送槽、脱粒、清选等装置组成的工作部件所决定，不同装置之间的相互作用十分复杂，难以通过物理推导、数值仿真等手段准确刻画。在实际作业过程中，联合收获机的作业参数调整多依赖机手经验，难以准确匹配实时收获工况，作业性能时好时坏。
随着联合收获机工作部件运行监测的日益完善，来源于作业现场的实测数据不断累积。这些数据记录下了联合收获机工作部件内部的运行过程，也隐含了联合收获机工作部件与作物的相互作用机理。通过挖掘这些数据的内部信息，可以有效地指导联合收获机的运行作业，提升联合收获机的作业性能表现。
但是联合收获机在田间作业，部分传感器安装于脱粒、清选等装置内部，服役环境温度高、湿度大、粉尘多；田间地域广阔，难以提供稳定、良好的通讯条件。以上因素共同导致实测数据存在大量空值、野值与异常值，数据不完备。因此，就需要对这些空值、野值、异常值进行数据重构，以形成完备的数据集，为后续数据建模与分析任务提供数据支撑。
数据重构是大数据领域的热点话题之一，其是指根据数据内在特征，对空值、野值、异常值的真实数值进行估计，从而将内含空值、野值、异常值的不完备数据转变为完备数据的过程。
目前，对于联合收获机大数据的重构问题，多是从数据空间分布角度重构数据。例如，通过空值、野值、异常值对应参数的空间分布密度、均值等估计其真实数值，并没有考虑数据参数之间的关联关系，难以实现数据的精准重构。
发明内容
针对上述技术问题，本发明的一个方式的目的之一是提供一种联合收获机大数据重构方法，基于模糊聚类框架，利用克里金法刻画联合收获机完整数据的参数关联关系，通过数据参数之间的关联关系，以及数据空间分布密度、均值等信息共同推断空值、野值、异常值的 真实数值，提高联合收获机大数据重构的精准度，为后续联合收获机大数据建模与分析任务提供数据支撑。
本发明的一个方式的目的之一是提供一种实现联合收获机大数据重构方法的装置，采用数据输入模块用于获取农机现场作业数据；数据重构模块用于接收数据输入模块输入的联合收获机大数据，并根据联合收获机大数据重构方法重构数据；数据输出模块输出联合收获机大数据的重构数据；数据保存模块用于保存联合收获机大数据的原始数据、重构数据、隶属度矩阵，提高联合收获机大数据重构的精准度，为后续联合收获机大数据建模与分析任务提供完备的数据支撑。。
注意，这些目的的记载并不妨碍其他目的的存在。本发明的一个方式并不需要实现所有上述目的。可以从说明书、附图、权利要求书的记载中抽取上述目的以外的目的。
本发明是通过以下技术手段实现上述技术目的的。
实施例1
一种联合收获机大数据重构方法，包括如下步骤：
步骤S1、在模糊聚类框架下，利用克里金法刻画联合收获机不包括空值、野值或异常值的作业参数的完整数据的参数之间的关联关系，获取联合收获机完整数据的隶属度；
步骤S2、基于步骤S1获得的联合收获机的完整数据及其隶属度计算类原型，继而获得包含空值、野值和或异常值的作业参数的不完整数据的隶属度；
步骤S3、根据计算的到的类原型、不完整数据的正常数值、不完整数据的隶属度分别估计空值、野值和或异常值的真实数值，实现联合收获机大数据的重构。
上述方案中，所述步骤S1中，采用克里金法刻画数据参数之间的关联关系，通过如下公式计算完整数据的隶属度：
Figure PCTCN2022144067-appb-000001
u k,i是第k个完整数据对第i个子类的隶属度；
c是聚类的子类个数；
y k是第k个完整数据的预测目标参数的数值；
Figure PCTCN2022144067-appb-000002
是第i个子类的克里金模型给出第k个完整数据的预测目标参数的预测值；
Figure PCTCN2022144067-appb-000003
是第t个子类的克里金模型给出第k个完整数据的预测目标参数的预测值。
进一步的，所述步骤S1中，聚类的子类个数c为3，隶属度阈值θ，一般取为为0.5。
上述方案中，所述步骤S2中，通过如下公式计算的类原型：
Figure PCTCN2022144067-appb-000004
n是完整数据的样本个数，
v i是第i个子类的类原型，
Figure PCTCN2022144067-appb-000005
是第k个完整数据，
Figure PCTCN2022144067-appb-000006
Figure PCTCN2022144067-appb-000007
是第k个完整数据的预测目标参数的数值，
Figure PCTCN2022144067-appb-000008
是第k个完整数据的除预测目标参数之外其余参数数值组成的向量。
上述方案中，所述步骤S2中，通过如下公式计算不完整数据的隶属度：
Figure PCTCN2022144067-appb-000009
Figure PCTCN2022144067-appb-000010
μ k,i是第k个不完整数据对第i个子类的隶属度，
z k,j表示第k个数据的第j个参数的数值，
v i,j表示第i个类原型的第j个参数的数值，
X M表示空值、野值、异常值组成的集合，
X P表示正常数值组成的集合，
s是参数的个数。
进一步的，所述步骤S3中，通过如下公式估计空值、野值、异常值的真实值，实现联合收获机大数据的重构：
Figure PCTCN2022144067-appb-000011
x k,j∈X M，表示数据中的空值、野值或异常值，
v i,j表示第i个类原型的第j个参数的数值，
μ k,i表示第k个不完整数据对第i个子类的隶属度。
一种实现所述联合收获机大数据重构方法的装置，包括数据输入模块、数据重构模块、数据输出模块和数据存储模块；
所述数据输入模块用于获取农机现场作业数据；
所述数据重构模块用于接收数据输入模块输入的联合收获机大数据，并根据联合收获机大数据重构方法重构数据；
所述数据输出和保存模块用于输出联合收获机大数据的重构数据，保存联合收获机大数据的原始数据、重构数据、隶属度矩阵。
上述方案中，所述数据输入模块与联合收获机车载监测系统、远程监控系统或智慧农场平台连接，用于接收标定好空值、野值、异常值的联合收获机大数据。
上述方案中，所述数据重构模块用于在模糊聚类框架下，利用克里金法刻画数据参数之间的关联关系，获取完整数据的隶属度，再基于完整数据及其隶属度计算类原型，继而获得不完整数据的隶属度，最后根据类原型、空值、野值和异常值对应的完整数据的正常数值、不完整隶属度，估计空值、野值和异常值的真实数值，实现联合收获机大数据的重构。
上述方案中，所述数据输出和保存模块将联合收获机的重构大数据与隶属度矩阵输出至云端，将联合收获机原始大数据、重构大数据、隶属度矩阵保存至联合收获机车载监测系统。
与现有技术相比，本发明的有益效果是：
根据本发明的一个方式，提供一种联合收获机大数据重构方法，基于模糊聚类框架，利用克里金法刻画联合收获机不包括空值、野值或异常值的作业参数的完整数据的参数之间的关联关系，通过数据参数之间的关联关系，以及数据空间分布密度、均值等信息共同推断空值、野值、异常值的真实数值，提高联合收获机大数据重构的精准度，为后续联合收获机大数据建模与分析任务提供完备的数据支撑。
根据本发明的一个方式，提供一种实现联合收获机大数据重构方法的装置，采用数据输入模块用于获取农机现场作业数据；数据重构模块用于接收数据输入模块输入的联合收获机大数据，并根据联合收获机大数据重构方法重构数据；数据输出模块输出联合收获机大数据的重构数据；数据保存模块用于保存联合收获机大数据的原始数据、重构数据、隶属度矩阵，最终实现提高联合收获机大数据重构的精准度，为后续联合收获机大数据建模与分析任务提供完备的数据支撑。
注意，这些效果的记载不妨碍其他效果的存在。本发明的一个方式并不一定必须具有所有上述效果。可以从说明书、附图、权利要求书等的记载显而易见地看出并抽出上述以外的效果。
附图说明
图1是本发明一实施方式的流程示意图。
图2是本发明一实施方式的完整数据隶属度计算流程示意图。
图3是本发明一实施方式的联合收获机大数据重构装置的示意图。
图4是本发明一实施方式的联合收获机大数据重构方法获取的部分完整数据隶属度矩阵 图。
图5是本发明一实施方式的联合收获机大数据重构方法与其他数据重构方法的实验结果对比图。
具体实施方式
下面详细描述本发明的实施例，所述实施例的示例在附图中示出，其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施例是示例性的，旨在用于解释本发明，而不能理解为对本发明的限制。
在本发明的描述中，需要理解的是，术语“中心”、“纵向”、“横向”、“长度”、“宽度”、“厚度”、“前”、“后”、“左”、“右”、“上”、“下”、“轴向”、“径向”、“竖直”、“水平”、“内”、“外”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本发明和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明的限制。此外，术语“第一”、“第二”仅用于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者更多个该特征。在本发明的描述中，“多个”的含义是两个或两个以上，除非另有明确具体的限定。
在本发明中，除非另有明确的规定和限定，术语“安装”、“相连”、“连接”、“固定”等术语应做广义理解，例如，可以是固定连接，也可以是可拆卸连接，或一体地连接；可以是机械连接，也可以是电连接；可以是直接相连，也可以通过中间媒介间接相连，可以是两个元件内部的连通。对于本领域的普通技术人员而言，可以根据具体情况理解上述术语在本发明中的具体含义。
图1所示一种联合收获机大数据重构方法，包括如下步骤：
步骤S1、在模糊聚类框架下，利用克里金法刻画联合收获机不包括空值、野值或异常值的作业参数的完整数据的参数之间的关联关系，获取联合收获机完整数据的隶属度；
步骤S2、基于步骤S1获得的联合收获机的完整数据及其隶属度计算类原型，继而获得包含空值、野值和或异常值的作业参数的不完整数据的隶属度；
步骤S3、根据计算的到的类原型、不完整数据的正常数值、不完整数据的隶属度分别估计空值、野值和或异常值的真实数值，实现联合收获机大数据的重构。
如图2所示，所述完整数据的隶属度计算包括如下步骤：
步骤S1)设定聚类的子类个数c，完整数据中的预测目标参数y，隶属度阈值θ，最大迭代次数g max，初始迭代次数g设定为1；
步骤S2)使用模糊c均值算法对完整数据进行初始聚类，获得隶属度矩阵U FCM，作 为初始隶属度矩阵U (0)；
步骤S3)针对每个子类，选用隶属度数值大于θ的完整数据作为该子类的训练数据集；
步骤S4)检查每个子类的训练数据集是否为空集，如果为空集，随机生成一个隶属度矩阵，返回步骤S3)，不为空集，执行下一步骤；
步骤S5)针对每个子类，利用克里金法建立预测模型；
步骤S6)利用每个子类的预测模型，获取对所有完整数据的预测值，构建如下预测值矩阵：
Figure PCTCN2022144067-appb-000012
式中，Y是预测值矩阵，
Figure PCTCN2022144067-appb-000013
是第i个子类预测模型对第k个完整数据的预测值；
步骤S7)利用公式计算完整数据隶属度：
Figure PCTCN2022144067-appb-000014
式中，u k,i表示第k个完整数据对第i个子类的隶属度；
y k是第k个完整数据预测目标参数的真实数值；
Figure PCTCN2022144067-appb-000015
是第i个子类预测模型对第k个完整数据的预测值；
y k,t是第t个子类预测模型对第k个完整数据的预测值；
步骤S8)对于任意i,j，如果满足
Figure PCTCN2022144067-appb-000016
或g≥g max，ε为预设阈值，将当前隶属度矩阵视为最终隶属度矩阵，否则，返回步骤S3)，同时，g＝g+1。
所述利用克里金法建立预测模型包括如下步骤：
步骤S5.1)假设x和y满足如下关系,克里金法建立预测模型：
Figure PCTCN2022144067-appb-000017
式中，y是完整数据中的预测目标参数；
x是除预测目标参数之外其余参数组成的向量；
f j(x)为第j个基函数；
β j为第j个基函数的系数；
p是基函数的个数；
Z(x)为高斯过程，满足如下条件：
E(Z(x))＝0
E(Z(x i)Z(x j))＝σ 2R(θ,x i,x j)
σ 2为样本方差；
R(θ,x i,x j)为相关矩阵；
θ为相关矩阵的参数向量；
x i是第i个数据除预测目标参数之外的其余参数组成的向量；
x j是第j个数据除预测目标参数之外的其余参数组成的向量；
步骤S5.2)构建基函数矩阵F：
F＝(f(x 1),f(x 2),…,f(x n)) T
其中n是训练数据的样本数；
f(x i)(i＝1,2,…,n)表示第i个数据通过基函数得到的数值，如下所示：
f(x i)＝(f 1(x i),f 2(x i),..,f p(x i))
基函数矩阵F可改写为：
Figure PCTCN2022144067-appb-000018
步骤S5.3)输入向量表示如下：
Y＝Fβ+Z
β是基函数系数组成的向量；
Z是训练数据的高斯过程组成的向量：
Z＝(Z(x 1),Z(x 2),…,Z(x n)) T
步骤S5.4)待预测样本x new的输出估计值为：
Figure PCTCN2022144067-appb-000019
Y是训练数据预测目标参数组成的向量；
c(x) T是的系数向量；
步骤S5.5)样本x new的预测误差如下：
Figure PCTCN2022144067-appb-000020
y(x new)是样本x new的真实输出值；
由线性无偏条件，得：
Figure PCTCN2022144067-appb-000021
F Tc(x)-f(x)＝0
步骤S5.6)计算回归均方差
Figure PCTCN2022144067-appb-000022
Figure PCTCN2022144067-appb-000023
σ 2＝D(Z)＝E(Z 2)-(E(Z)) 2＝E(Z 2)
其中
Figure PCTCN2022144067-appb-000024
是样本x new与训练数据输入通过相关矩阵得到的对应向量；
步骤S5.7)建立如下优化问题求解模型最优参数：
Figure PCTCN2022144067-appb-000025
s.t.F Tc(x)-f(x)＝0
s.t.表示约束条件，
步骤S5.8)构建朗格朗日函数：
Figure PCTCN2022144067-appb-000026
其中λ是拉格朗日乘子；
求偏导，得：
Figure PCTCN2022144067-appb-000027
Figure PCTCN2022144067-appb-000028
得
Figure PCTCN2022144067-appb-000029
步骤S5.9)x的估计值通过下式获得：
Figure PCTCN2022144067-appb-000030
由
Figure PCTCN2022144067-appb-000031
得β的最优估计：
Figure PCTCN2022144067-appb-000032
步骤S5.10)样本x new的输出值如下所示：
Figure PCTCN2022144067-appb-000033
所述步骤S2利用如下公式计算类原型：
Figure PCTCN2022144067-appb-000034
n是完整数据的样本个数；
v i是第i个子类的类原型；
Figure PCTCN2022144067-appb-000035
是第k个完整数据，
Figure PCTCN2022144067-appb-000036
Figure PCTCN2022144067-appb-000037
是第k个完整数据的预测目标参数的数值；
Figure PCTCN2022144067-appb-000038
是第k个完整数据的除预测目标参数之外其余参数数值组成的向量。
所述步骤S2利用如下公式计算不完整数据的隶属度：
Figure PCTCN2022144067-appb-000039
Figure PCTCN2022144067-appb-000040
其中μ k,i是第k个不完整数据对第i个子类的隶属度，
z k,j表示第k个数据的第j个参数的数值，
是第个不完整数据对第个子类的隶属度；表示第个数据的第个参数的数值；
v i,j表示第i个类原型的第j个参数的数值；
X M表示空值、野值、异常值组成的集合；
X P表示正常数值组成的集合；
s是参数的个数。
所述步骤S2利用如下公式估计空值、野值、异常值的真实值：
Figure PCTCN2022144067-appb-000041
其中x j,k∈X M，表示数据中的空值、野值或异常值；
v i,j表示第i个类原型的第j个参数的数值；
μ k,i表示第k个不完整数据对第i个子类的隶属度。
为了更好地说明本发明所述联合收获机大数据重构方法的优势，引入以下2种常用的数据重构方法作为对比。
1.参数均值重构方法
在参数均值重构方法中，将联合收获机的空值、野值、异常值对应参数正常值的均值视为其正常数值。
2.最近邻原型重构方法
在最近邻原型重构方法中，利用模糊c均值算法对完整数据聚类，并将空值、野值、异常值的最近类原型的对应参数值视为其真实数值，如下所示。
x＝v i；
式中，x是数据中的空值、野值或异常值；
v i是通过模糊c均值算法获取的最近邻类原型中，空值、野值或异常值对应的参数值。
利用模糊c均值算法对完整数据的聚类过程如下：
步骤(1)设定聚类的子类数目c，最大迭代次数g max，迭代次数g设定为1；
步骤(2)生成一个随机矩阵U作为初始隶属度矩阵；
Figure PCTCN2022144067-appb-000042
式中，n是完整数据样本个数；
c是子类个数；
μ i,k表示第i个完整数据对第j个子类的隶属程度，取值从0到1，并满足如下条件：
Figure PCTCN2022144067-appb-000043
步骤(3)计算每个子类的类原型；
Figure PCTCN2022144067-appb-000044
式中，v i是第i个子类的类原型；
z k是第k个完整数据；
步骤(4)更新隶属度矩阵；
Figure PCTCN2022144067-appb-000045
步骤(5)对于任意i,j，如果满足
Figure PCTCN2022144067-appb-000046
或g≥g max，将当前类原型视 为最终类原型；否则，返回步骤(3)，g＝g+1。
根据本实施例，优选的，联合收获机收获作物为水稻，数据样本容量150，包括割台高度、前进速度、风机转速、脱离间隙、清选损失等11个参数。
随机选取10％的数据中作为不完整数据，随机抽取其中的一个数值作为待重构数值，剩余数据作为完整数据。
利用所述联合收获机大数据重构方法、参数均值重构方法、最近邻原型重构方法估计联合收获机大数据中的空值、野值和异常值的真实值。
优选的，本发明提供的一种联合收获机大数据重构方法的参数设定如下：
聚类数设定为3；
克里金法的预测目标参数设定为清选损失数；
隶属度阈值设定为0.5；
最大迭代次数设定为20。
最近邻原型重构方法参数设定如下：
聚类数设定为3；
最大迭代次数设定为20。
进一步详细说明所述联合收获机大数据重构方法的数据重构过程：
1.计算完整数据的隶属度。利用克里金法刻画数据参数之间的关联关系，获取完整数据的隶属度，其中部分数据的隶属度如附图3所示。可以看出，数据对于不同子类的隶属程度有所不同。以第一个数据为例，隶属度分别为0.91，0.08，0.01，其属于第一个子类的可能性最大，为0.91，因此其归为第一类。
2.计算完整数据的类原型。基于完整数据及其隶属度，计算完整数据的类原型。
3.计算不完整数据的隶属度。基于不完整数据和类原型计算其隶属度。
4.计算空值、野值、异常值的真实数值。基于空值、野值、异常值对应的完整数据的参数值，结合获取的不完整数据隶属度，计算空值、野值、异常值的真实数值，部分不完整数据隶属结果如附图4所示。
所述联合收获机大数据重构方法、参数均值重构方法、最近邻原型重构方法获取的数据重构结果如表1所示。
表1数据重构结果
Figure PCTCN2022144067-appb-000047
Figure PCTCN2022144067-appb-000048
本发明所述联合收获机大数据重构方法的平均重构误差为1.53％，明显低于参数均值重构方法(16.27％)和最近邻原型重构方法(7.85％)。
本发明所述联合收获机大数据重构方法的重构误差均在3.00％以下，参数均值重构方法误差均高于10.00％，最近邻原型重构方法部分数据重构误差低于5.00％，但仍高于本发明所述联合收获机大数据重构方法。由此可见，本发明所述联合收获机大数据重构方法提高了大数据空值、野值、异常值重构的精准度，能够实现更为精准的联合收获机大数据重构。
实施例2
如图5所示，本实施例提供一种实现实施例1所述联合收获机大数据重构方法的装置，因此具有实施例1的有益效果，此处不再赘述。
所述联合收获机大数据重构方法的装置包括数据输入模块、数据重构模块、数据输出和数据存储模块。所述装置可集成于但不限于联合收获机车载监测、远程监控、智慧农场等平台。
所述数据输入模块对接联合收获机车载监测、远程监控、智慧农场等平台，接收已经标定好空值、野值、异常值的联合收获机大数据；
所述数据重构模块根据本发明所述联合收获机大数据重构方法，重构、联合收获机大数据；
所述数据输出和数据存模块输出联合收获机大数据的重构数据，保存联合收获机大数据的原始数据、重构数据、隶属度矩阵。
在本实施例中，优选的，将本发明实现所述联合收获机大数据重构方法的装置集成于联合收获机车载监测系统，主要用于在数据传输至云端前，将获取的联合收获机大数据进行重构。
联合收获机车载监测系统利用各子系统数据采集装置获取联合收获机大数据，通过CAN总线汇总。
联合收获机车载监测系统通过集成的空值判断、野值识别、异常识别等算法标定数据中的空值、野值与异常值。
将标定的联合收获机大数据输入到本发明实现所述联合收获机大数据重构方法的装置中的数据输入模块；
利用本发明提供的一种联合收获机大数据重构方法，重构输入的联合收获机大数据。
通过本发明提供的一种实现所述联合收获机大数据重构方法的装置的数据输出模块和数据存储模块，将联合收获机的重构大数据与隶属度矩阵输出至云端，将联合收获机原始大数据、重构大数据、隶属度矩阵保存至联合收获机车载监测系统。
以上公开的本发明优选实施例只是用于帮助阐述本发明。优选实施例并没有详尽叙述所有的细节，也不限制该发明的具体实施方式。本说明书选取并具体描述实施例，是为了更好地解释本发明的原理和实际应用，从而使所属技术领域技术人员能很好地理解和利用本发明。本发明仅受权利要求书及其全部范围和等效物的限制。

Claims (10)

  1. 一种联合收获机大数据重构方法，其特征在于，包括如下步骤：
    步骤S1、在模糊聚类框架下，利用克里金法刻画联合收获机不包括空值、野值或异常值的作业参数的完整数据的参数之间的关联关系，获取联合收获机完整数据的隶属度；
    步骤S2、基于步骤S1获得的联合收获机的完整数据及其隶属度计算类原型，继而获得包括空值、野值和或异常值的作业参数的不完整数据的隶属度；
    步骤S3、根据计算的到的类原型、不完整数据的正常数值、不完整数据的隶属度分别估计空值、野值和或异常值的真实数值，实现联合收获机大数据的重构。
  2. 根据权利要求1所述的联合收获机大数据重构方法，其特征在于，所述步骤S1中，采用克里金法刻画数据参数之间的关联关系，通过如下公式计算完整数据的隶属度：
    Figure PCTCN2022144067-appb-100001
    u k,i是第k个完整数据对第i个子类的隶属度；
    c是聚类的子类个数；
    y k是第k个完整数据的预测目标参数的数值；
    Figure PCTCN2022144067-appb-100002
    是第i个子类的克里金模型给出第k个完整数据的预测目标参数的预测值；
    Figure PCTCN2022144067-appb-100003
    是第t个子类的克里金模型给出第k个完整数据的预测目标参数的预测值。
  3. 根据权利要求2所述的联合收获机大数据重构方法，其特征在于，所述步骤S1中，聚类的子类个数c为3，隶属度阈值θ，一般取为为0.5。
  4. 根据权利要求1所述的联合收获机大数据重构方法，其特征在于，所述步骤S2中，通过如下公式计算的类原型：
    Figure PCTCN2022144067-appb-100004
    n是完整数据的样本个数，
    v i是第i个子类的类原型，
    Figure PCTCN2022144067-appb-100005
    是第k个完整数据，
    Figure PCTCN2022144067-appb-100006
    Figure PCTCN2022144067-appb-100007
    是第k个完整数据的预测目标参数的数值，
    Figure PCTCN2022144067-appb-100008
    是第k个完整数据的除预测目标参数之外其余参数数值组成的向量。
  5. 根据权利要求1所述的联合收获机大数据重构方法，其特征在于，所述步骤S2中，通过如下公式计算不完整数据的隶属度：
    Figure PCTCN2022144067-appb-100009
    Figure PCTCN2022144067-appb-100010
    μ k,i是第k个不完整数据对第i个子类的隶属度，
    z k,j表示第k个数据的第j个参数的数值，
    v i,j表示第i个类原型的第j个参数的数值，
    X M表示空值、野值、异常值组成的集合，
    X P表示正常数值组成的集合，
    s是参数的个数。
  6. 根据权利要求5所述的联合收获机大数据重构方法，其特征在于，所述步骤S3中，通过如下公式估计空值、野值、异常值的真实值，实现联合收获机大数据的重构：
    Figure PCTCN2022144067-appb-100011
    x k,j∈X M，表示数据中的空值、野值或异常值，
    v i,j表示第i个类原型的第j个参数的数值，
    μ k,i表示第k个不完整数据对第i个子类的隶属度。
  7. 一种实现权利要求1-6任意一项所述联合收获机大数据重构方法的装置，其特征在于，包括数据输入模块、数据重构模块、数据输出模块和数据存储模块；
    所述数据输入模块用于获取农机现场作业数据；
    所述数据重构模块用于接收数据输入模块输入的联合收获机大数据，并根据联合收获机大数据重构方法重构数据；
    所述数据输出和保存模块用于输出联合收获机大数据的重构数据，保存联合收获机大数据的原始数据、重构数据、隶属度矩阵。
  8. 根据权利要求7所述的联合收获机大数据重构方法的装置，其特征在于，所述数据输入模块与联合收获机车载监测系统、远程监控系统或智慧农场平台连接，用于接收标定好空值、野值、异常值的联合收获机大数据。
  9. 根据权利要求8所述的联合收获机大数据重构方法的装置，其特征在于，所述数据重构模块用于在模糊聚类框架下，利用克里金法刻画数据参数之间的关联关系，获取完整数据 的隶属度，再基于完整数据及其隶属度计算类原型，继而获得不完整数据的隶属度，最后根据类原型、空值、野值和异常值对应的完整数据的正常数值、不完整隶属度估计空值、野值和异常值的真实数值，实现联合收获机大数据的重构。
  10. 根据权利要求8所述的联合收获机大数据重构方法的装置，其特征在于，所述数据输出和保存模块将联合收获机的重构大数据与隶属度矩阵输出至云端，将联合收获机原始大数据、重构大数据、隶属度矩阵保存至联合收获机车载监测系统。
PCT/CN2022/144067 2022-08-22 2022-12-30 一种联合收获机大数据重构方法与装置 WO2024040845A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211007850.5 2022-08-22
CN202211007850.5A CN115423005B (zh) 2022-08-22 2022-08-22 一种联合收获机大数据重构方法与装置

Publications (1)

Publication Number Publication Date
WO2024040845A1 true WO2024040845A1 (zh) 2024-02-29

Family

ID=84198624

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/144067 WO2024040845A1 (zh) 2022-08-22 2022-12-30 一种联合收获机大数据重构方法与装置

Country Status (2)

Country Link
CN (1) CN115423005B (zh)
WO (1) WO2024040845A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115423005B (zh) * 2022-08-22 2023-10-31 江苏大学 一种联合收获机大数据重构方法与装置

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001032679A (ja) * 1999-07-16 2001-02-06 Ohbayashi Corp 切羽前方亀裂分布予測方法
CN107729943A (zh) * 2017-10-23 2018-02-23 辽宁大学 信息反馈极限学习机优化估值的缺失数据模糊聚类算法及其应用
CN115423005A (zh) * 2022-08-22 2022-12-02 江苏大学 一种联合收获机大数据重构方法与装置

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101633360B1 (ko) * 2015-01-12 2016-06-27 한국과학기술원 구조물 상태 평가를 위한 크리깅 모델 기반 순차적 샘플링 방법
CN107452059B (zh) * 2017-08-09 2019-08-20 中国地质大学（武汉） 一种三维插值建模方法、设备及其存储设备
WO2019218263A1 (zh) * 2018-05-16 2019-11-21 深圳大学 基于极限学习机的极限ts模糊推理方法及系统
WO2021007744A1 (zh) * 2019-07-15 2021-01-21 广东工业大学 一种集成空间约束的核模糊c均值快速聚类算法
CN111340069A (zh) * 2020-02-11 2020-06-26 大连理工大学 基于交替学习的不完整数据精细建模及缺失值填补方法
CN111738412A (zh) * 2020-05-28 2020-10-02 江门职业技术学院 一种不完全网络的大数据异常挖掘方法、系统及存储介质
CN112229403B (zh) * 2020-08-31 2024-02-20 中国空间技术研究院 基于大地水准面三维修正原理提高海洋重力重构精度方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001032679A (ja) * 1999-07-16 2001-02-06 Ohbayashi Corp 切羽前方亀裂分布予測方法
CN107729943A (zh) * 2017-10-23 2018-02-23 辽宁大学 信息反馈极限学习机优化估值的缺失数据模糊聚类算法及其应用
CN115423005A (zh) * 2022-08-22 2022-12-02 江苏大学 一种联合收获机大数据重构方法与装置

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
"Doctoral Dissertation", 1 March 2018, DALIAN UNIVERSITY OF TECHNOLOGY, CN, article ZHANG, LIYONG: "Research on Fuzzy Clustering Method Based on Information Granules", pages: 1 - 170, XP009552843 *
"Master's Thesis", 1 April 2016, LIAONING UNIVERSITY, CN, article WANG, LU: "Research on Interval Supervised Fuzzy Clustering Algorithm for Improved Ant Colony Optimization", pages: 1 - 75, XP009552846 *
SHI, MAOLIN ET AL.: "A robust prediction method based on Kriging method and fuzzy c-means algorithm with application to a combine harvester", STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION, vol. 65, no. 9, 30 September 2022 (2022-09-30), XP037932055, DOI: 10.1007/s00158-022-03364-0 *

Also Published As

Publication number Publication date
CN115423005B (zh) 2023-10-31
CN115423005A (zh) 2022-12-02

Similar Documents

Publication Publication Date Title
WO2021098472A1 (zh) 一种深度时空特征联合学习的农作物产量估测方法
WO2024040845A1 (zh) 一种联合收获机大数据重构方法与装置
CN105739575B (zh) 一种设施蔬菜环境参数的数据融合方法
WO2016118686A1 (en) Modeling of crop growth for desired moisture content of targeted livestock feedstuff for determination of harvest windows using field-level diagnosis and forecasting of weather conditions and observations and user input of harvest condition states
US20160217231A1 (en) Modeling of crop growth for desired moisture content of bovine feedstuff and determination of harvest windows for high-moisture corn using field-level diagnosis and forecasting of weather conditions and observations and user input of harvest condition states
EP3816880A1 (en) A yield estimation method for arable crops and grasslands, coping with extreme weather conditions and with limited reference data requirements
CN116307190B (zh) 一种基于蓝牙mesh网络的果园环境对产量的预测方法
CN117391482B (zh) 一种基于大数据监测的大棚温度智能预警方法及系统
US20160217228A1 (en) Modeling of crop growth for desired moisture content of targeted livestock feedstuff for determination of harvest windows using field-level diagnosis and forecasting of weather conditions and observations and user input of harvest condition states
CN114167922A (zh) 一种基于多传感器数据采集的农牧智能分析的方法及系统
CN114638536A (zh) 基于人工智能的畜牧健康监测方法与系统
US20160217229A1 (en) Modeling of crop growth for desired moisture content of bovine feedstuff and determination of harvest windows for corn silage using field-level diagnosis and forecasting of weather conditions and field observations
US10180998B2 (en) Modeling of crop growth for desired moisture content of bovine feedstuff and determination of harvest windows for corn earlage using field-level diagnosis and forecasting of weather conditions and field observations
CN114611804A (zh) 基于tso-grnn组合模型的玉米产量预测方法
Pal Identification of paddy leaf diseases using a supervised neural network
Sun et al. Design of feed rate monitoring system and estimation method for yield distribution information on combine harvester
Sudduth et al. AI down on the farm
CN110414859A (zh) 一种基于物联网技术的优质稻谷收储作业5t管理系统及其评价方法
CN109886721A (zh) 一种猪肉价格预测系统算法
CN115423148A (zh) 一种基于克里金法和决策树的农机作业性能预测方法与装置
Jesi et al. IoT Enabled Smart Irrigation and Cultivation Recommendation System for Precision Agriculture
Salvi et al. Hydroiot: An iot and edge computing based multi-level hydroponics system
Sreemathy et al. Crop Recommendation with BiLSTM-MERNN Algorithm for Precision Agriculture
CN116720635B (zh) 一种基于实测数据的广西油茶估产方法
Yu et al. Fusion Learning of Functional Linear Regression with Application to Genotype-by-Environment Interaction Studies

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22956376

Country of ref document: EP

Kind code of ref document: A1