Oilseed rape mechanized harvesting intelligent control method and system based on online monitoring

Abstract

The invention discloses an intelligent control method for mechanical rape harvesting based on online monitoring, and relates to the field of intelligent control, and the method comprises the steps: obtaining rape growth related information of a target rape field, and predicting a harvesting time node based on the rape growth related information; acquiring rape remote sensing information at the predicted harvesting time node; dividing a plurality of grid rape fields based on rape remote sensing information; calculating the rape maturity of all the grid rape fields; screening out an oilseed rape trial cutting area from all the grid oilseed rape fields; the trial cutting loss rate of the rape trial cutting area is counted, and the total rape loss of the target rape field is analyzed; judging whether the predicted harvesting time node is an optimal harvesting node or not; and generating an optimal parameter interval of the rape harvesting equipment. The harvesting loss rate of a large-area rape field can be effectively reduced.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of intelligent control, and in particular to an intelligent control method and system for mechanized rapeseed harvesting based on online monitoring. Background Art

[0002] As an economic crop, rapeseed is the basic source of edible oil. With the advancement of agricultural mechanization, the rapeseed planting area and harvesting area have expanded year by year, and the resulting loss rate problem has become increasingly prominent. This is because in the process of harvesting rapeseed using mechanical equipment, rapeseed losses are inevitable due to factors such as the performance of the harvesting equipment, harvesting weather, and rapeseed maturity. Especially for some areas with large rapeseed planting areas, the total rapeseed loss is higher, causing serious economic losses to agricultural personnel engaged in rapeseed planting. Therefore, a method is needed to minimize the rapeseed harvesting loss rate.

[0003] The existing method to reduce the loss rate is mainly to first select the appropriate harvesting equipment and harvesting method according to the variety and area of ​​the rapeseed field, then analyze the overall maturity of the rapeseed field, and adjust the equipment parameters of the harvesting equipment according to the overall maturity, and control the harvesting equipment to harvest rapeseed according to the appropriate equipment parameters. This method can reduce the rapeseed harvesting loss rate to a certain extent. However, for a large rapeseed field, since the maturity of rapeseed in different areas may be different, the appropriate equipment parameters for areas with different maturity may also be different. At the same time, due to the large area of ​​the rapeseed field, it is difficult to make an overall assessment of the maturity of rapeseed, which is easy to cause large errors in the maturity assessment. Therefore, it may cause unreasonable selection of equipment parameters, making it difficult to further reduce the rapeseed harvesting loss rate. Summary of the invention

[0004] The embodiments of the present application provide an intelligent control method and system for mechanized rapeseed harvesting based on online monitoring, which are used to solve the problem that the existing technology is difficult to further reduce the harvesting loss rate of rapeseed fields with a large area.

[0005] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0006] In a first aspect, a method for intelligent control of mechanized rapeseed harvesting based on online monitoring is provided, the method comprising:

[0007] Acquire rapeseed growth-related information of a target rapeseed field, wherein the rapeseed growth-related information includes soil information of the target rapeseed field, meteorological information of a location of the target rapeseed field, and sowing time information and rapeseed variety information of the rapeseed planted in the target rapeseed field;

[0008] Inputting the rapeseed variety information, the meteorological information and the soil information into a pre-built rapeseed growth model to predict the rapeseed growth cycle, and determining the predicted maturity time period of the target rapeseed field according to the rapeseed growth cycle prediction result;

[0009] Determining a predicted harvesting time node of the target rapeseed field according to the sowing time information and the predicted maturity time period;

[0010] At the predicted harvest time node, collecting rapeseed remote sensing information of the target rapeseed field by a remote sensing drone;

[0011] Acquire rapeseed field size information of the target rapeseed field based on the rapeseed remote sensing information, and grid-divide the target rapeseed field according to the rapeseed field size information to obtain a plurality of grid rapeseed fields;

[0012] Performing feature analysis on all the grid rapeseed fields according to the rapeseed remote sensing information to obtain rapeseed remote sensing features, and calculating rapeseed maturity of all the grid rapeseed fields based on the rapeseed remote sensing features;

[0013] A reference grid rapeseed field is selected from all the grid rapeseed fields according to the rapeseed maturity, and adjacent grid rapeseed fields of the reference grid rapeseed field are subjected to iterative diffusion screening with the reference grid rapeseed field as the center and the maximization of the rapeseed maturity difference as the screening target, until the number of the selected rapeseed fields is greater than or equal to a preset number threshold, thereby obtaining a rapeseed trial harvesting area;

[0014] Controlling the rapeseed harvesting equipment to perform a rapeseed trial harvesting task in the rapeseed trial harvesting area, and during the execution of the rapeseed trial harvesting task, using the rapeseed harvesting equipment to count the trial harvesting loss rate of the rapeseed trial harvesting area, and analyzing the total rapeseed loss of the target rapeseed field based on the trial harvesting loss rate and using a correlation analysis method;

[0015] Determine the rapeseed maturity node of the target rapeseed field in combination with the rapeseed remote sensing characteristics and the rapeseed total loss, and judge whether the predicted harvesting time node is the optimal harvesting node according to the rapeseed maturity node;

[0016] If the predicted harvesting time node is the optimal harvesting node, the rapeseed maturity and the trial harvesting loss rate are combined and a parameter optimization algorithm is used to generate an optimal parameter range for the rapeseed harvesting equipment, and the rapeseed harvesting equipment is controlled to harvest the rapeseed in the target rapeseed field according to the optimal parameters.

[0017] Optionally, performing feature analysis on all the grid rapeseed fields according to the rapeseed remote sensing information to obtain rapeseed remote sensing features, and calculating rapeseed maturity of all the grid rapeseed fields based on the rapeseed remote sensing features comprises the following steps:

[0018] Preprocessing the rapeseed remote sensing information of all the grid rapeseed fields;

[0019] For any of the grid rapeseed fields, extracting the reflectance spectrum in the rapeseed remote sensing information that has completed preprocessing;

[0020] Calculate a spectral reflectance value according to the reflectance spectrum, perform feature extraction on the spectral reflectance value, and obtain a spectral feature of rapeseed;

[0021] Performing principal component transformation on the rapeseed remote sensing information to obtain spatial color features of the rapeseed remote sensing information;

[0022] Using a clustering algorithm to perform color classification on the spatial color features to obtain rapeseed color features of the rapeseed remote sensing information;

[0023] Integrating the rapeseed spectral characteristics and the rapeseed color characteristics into the rapeseed remote sensing characteristics of the grid rapeseed field;

[0024] A rapeseed maturity prediction model is constructed based on a neural network model, and all the rapeseed remote sensing features are input into the trained rapeseed maturity prediction model, and the rapeseed maturity prediction model is used to output the rapeseed maturity of all the grid rapeseed fields.

[0025] Optionally, the method of constructing a rapeseed maturity prediction model based on a neural network model, inputting all the rapeseed remote sensing features into the trained rapeseed maturity prediction model, and outputting the rapeseed maturity of all the grid rapeseed fields through the rapeseed maturity prediction model comprises the following steps:

[0026] Construct rapeseed maturity prediction model based on BP neural network model;

[0027] Using a pre-constructed rapeseed training set to train the rapeseed maturity prediction model;

[0028] For any of the grid rapeseed fields, the rapeseed spectral characteristics and the rapeseed color characteristics of the grid rapeseed field are sequentially input into the rapeseed maturity prediction model to perform maturity prediction, and obtain a first maturity prediction value and a second maturity prediction value;

[0029] Using Kalman filtering to filter and correct the first maturity prediction value and the second maturity prediction value;

[0030] The first maturity prediction value and the second maturity prediction value after filtering and correction are fused to obtain the rapeseed maturity of all the grid rapeseed fields.

[0031] Optionally, the selecting a plurality of target grid rapeseed fields from all the grid rapeseed fields as rapeseed trial harvesting areas according to the rapeseed maturity comprises the following steps:

[0032] According to the rapeseed maturity, all the grid rapeseed fields are divided into maturity grades using a threshold method to obtain the rapeseed maturity grades of all the grid rapeseed fields;

[0033] If the rapeseed maturity level is the maximum rapeseed maturity level or the rapeseed maturity level is the minimum rapeseed maturity level, all the grid rapeseed fields corresponding to the maximum rapeseed maturity level and the minimum rapeseed maturity level are marked as trial rapeseed fields respectively;

[0034] For any of the trial-cut rapeseed fields, the trial-cut rapeseed field marking step is repeatedly performed based on the rapeseed maturity level and with the trial-cut rapeseed field as the center until the number of the trial-cut rapeseed fields is greater than or equal to a preset number threshold, and the trial-cut rapeseed field marking step is:

[0035] Calculating the maturity level difference between the adjacent rapeseed field of the trial-cut rapeseed field and the trial-cut rapeseed field, and recording the adjacent rapeseed field whose maturity level difference is greater than or equal to a preset difference threshold as the trial-cut rapeseed field;

[0036] Integrate all adjacent trial-cut rapeseed fields to obtain a plurality of initial trial-cut areas;

[0037] Constructing a target maturity level matrix of the target rapeseed field according to the rapeseed maturity levels of all the grid rapeseed fields;

[0038] For any of the initial trial cutting areas, constructing a regional maturity grade matrix of the initial trial cutting area based on the rapeseed maturity grade;

[0039] Based on a similarity calculation algorithm, the similarities between the target maturity level matrix and all the regional maturity level matrices are calculated respectively to obtain a plurality of matrix similarities;

[0040] The initial trial cutting area corresponding to the largest data similarity is used as the rapeseed trial cutting area.

[0041] Optionally, controlling the rapeseed harvesting device to perform a rapeseed trial harvesting task in the rapeseed trial harvesting area, counting the trial harvesting loss rate of the rapeseed trial harvesting area by the rapeseed harvesting device during the execution of the rapeseed trial harvesting task, and analyzing the total rapeseed loss of the target rapeseed field based on the trial harvesting loss rate and using a correlation analysis method comprises the following steps:

[0042] When the rapeseed harvesting equipment arrives at any of the grid rapeseed fields in the rapeseed trial harvesting area, the equipment parameters of the rapeseed harvesting equipment are updated based on the maturity of the rapeseed in the grid rapeseed field to obtain harvesting equipment parameters;

[0043] Controlling the rapeseed harvesting device to perform the rapeseed trial harvesting task in the rapeseed trial harvesting area according to the harvesting device parameters, and obtaining the rapeseed trial harvesting data of all the grid rapeseed fields in the rapeseed trial harvesting area;

[0044] Calculating the trial cutting loss rate of all the grid rapeseed fields in the rapeseed trial cutting area according to the rapeseed trial cutting data;

[0045] Performing a correlation analysis on the trial cutting loss rate and the rapeseed maturity to obtain a correlation analysis result;

[0046] Calculating the rapeseed loss rate of all the grid rapeseed fields except the rapeseed trial harvesting area according to the correlation analysis result;

[0047] Calculate the average rapeseed loss rate of the target rapeseed field by combining all the trial cutting loss rates and the rapeseed loss rate;

[0048] The total rapeseed loss of the target rapeseed field is calculated according to the average rapeseed loss rate and the pre-acquired estimated rapeseed yield.

[0049] Optionally, the updating of equipment parameters of the rapeseed harvesting equipment based on the rapeseed maturity of the grid rapeseed field to obtain harvesting equipment parameters comprises the following steps:

[0050] Acquiring rapeseed image information of the grid rapeseed field using an image acquisition device pre-arranged outside the rapeseed harvesting device;

[0051] Preprocessing the rapeseed image information;

[0052] The rapeseed image information that has completed the preprocessing is binarized using a threshold segmentation method to obtain a target rapeseed image;

[0053] Acquire device parameters of the image acquisition device, and calculate rapeseed height information of the rapeseed trial cutting area according to the device parameters and the target rapeseed image;

[0054] The equipment parameters of the rapeseed harvesting equipment are updated in combination with the rapeseed height information and the rapeseed maturity corresponding to the grid rapeseed field to obtain harvesting equipment parameters.

[0055] Optionally, the determining the rapeseed maturity node of the target rapeseed field in combination with the rapeseed remote sensing feature and the rapeseed total loss, and judging whether the predicted harvesting time node is the optimal harvesting node according to the rapeseed maturity node comprises the following steps:

[0056] If the total rapeseed loss is less than or equal to a preset loss threshold, determining the predicted harvesting time node as the optimal harvesting node;

[0057] If the total loss of rapeseed is greater than the loss threshold, predicting the rapeseed maturity node of the target rapeseed field by combining the rapeseed image information and the rapeseed remote sensing characteristics;

[0058] If the rapeseed maturity node is later than the predicted harvesting time node, determining that the predicted harvesting time node is not the optimal harvesting node;

[0059] If the rapeseed maturity node is not later than the predicted harvesting time node, the predicted harvesting time node is determined to be the optimal harvesting node.

[0060] Optionally, the predicting the rapeseed maturity node of the target rapeseed field by combining the rapeseed image information and the rapeseed remote sensing characteristics comprises the following steps:

[0061] Constructing a rapeseed maturity prediction model based on a neural network model and a self-attention mechanism, and using a pre-constructed training set to perform model training on the rapeseed maturity prediction model;

[0062] Using a target detection algorithm to identify the root and stem region of the rapeseed image information, and performing image segmentation on the rapeseed image information according to the root and stem region identification result to obtain a rapeseed root and stem image in the rapeseed image information;

[0063] Extracting rhizome color features of the rape rhizome image using a clustering algorithm;

[0064] Inputting the root and stem color features, the rapeseed height information and the rapeseed remote sensing features into the trained rapeseed maturity prediction model, and outputting the predicted rapeseed maturity of the target rapeseed field through the rapeseed maturity prediction model;

[0065] The predicted harvesting time node is corrected according to the predicted rapeseed maturity period to obtain the rapeseed maturity node of the target rapeseed field.

[0066] Optionally, the combining the rapeseed maturity and the trial harvesting loss rate and using a parameter optimization algorithm to generate an optimal parameter range for the rapeseed harvesting equipment comprises the following steps:

[0067] All the grid rapeseed fields in the rapeseed trial harvesting area where the rapeseed loss rate is less than the loss rate threshold are used as standard rapeseed fields;

[0068] Acquire the harvesting equipment parameters corresponding to the rapeseed harvesting equipment in all the standard rapeseed fields;

[0069] Calculating the parameter interval of the rapeseed harvesting equipment according to all the harvesting equipment parameters to obtain a plurality of different types of initial parameter intervals;

[0070] Generate multiple random parameters according to all the initial parameter intervals, and encode all the random parameters to obtain multiple initial parameter populations;

[0071] Based on the genetic algorithm, all the initial parameter populations are optimized respectively until a maximum number of parameter optimization times is reached, and a plurality of target device parameters are output;

[0072] Performing information exchange on all the target device parameters, and calculating a target parameter range based on all the target device parameters that have completed the information exchange;

[0073] The update iteration step is repeated for the target parameter interval until a preset maximum number of iterations is reached, thereby obtaining the optimal parameter interval of the rapeseed harvesting equipment.

[0074] In a second aspect, the present application provides an intelligent control device for mechanized rapeseed harvesting based on online monitoring, characterized in that it includes:

[0075] a memory configured to store instructions; and

[0076] A processor is configured to call the instructions from the memory and to implement the method for intelligent control of mechanized rapeseed harvesting based on online monitoring according to any one of the first aspects when executing the instructions.

[0077] Through the above technical scheme, the predicted maturity time period of the target rapeseed field is predicted through the rapeseed growth-related information of the target rapeseed field and the pre-constructed rapeseed growth model, and the predicted harvesting time node is determined according to the predicted maturity time period. A time suitable for rapeseed harvesting is determined by the model prediction method, which can avoid the increase of rapeseed loss rate due to immature or over-mature rapeseed. Then, a trial harvesting is carried out at the predicted harvesting time node, and the optimal harvesting time node can be further determined. At the same time, the optimal parameter adjustment interval can be determined for the subsequent formal harvesting according to the adjustment status of the rapeseed harvesting equipment during the trial harvesting process, so as to ensure that the rapeseed is at the optimal harvesting time node during the formal harvesting, and the rapeseed harvesting equipment can also maintain the optimal equipment parameters for harvesting during the harvesting process. In summary, the present application provides a method that can further reduce the rapeseed harvesting loss rate, which can reduce the rapeseed harvesting loss rate to the greatest extent and increase the rapeseed yield.

[0078] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 A schematic diagram of a process flow of an intelligent control method for mechanized rapeseed harvesting based on online monitoring provided in an embodiment of the present application;

[0080] Figure 2 A schematic diagram of the rapeseed maturity grade distribution structure of the target rapeseed field provided in an embodiment of the present application. DETAILED DESCRIPTION

[0081] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0082] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0083] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0084] Figure 1 The following schematically shows a flow chart of a method for intelligent control of mechanized rapeseed harvesting based on online monitoring according to an embodiment of the present application. Figure 1 As shown, the embodiment of the present application provides a method for intelligent control of mechanized rapeseed harvesting based on online monitoring, which may include the following steps:

[0085] S101. Acquire rapeseed growth-related information of a target rapeseed field, where the rapeseed growth-related information includes soil information of the target rapeseed field, weather information of a location where the target rapeseed field is located, and sowing time information and rapeseed variety information of the rapeseed planted in the target rapeseed field.

[0086] In this embodiment, soil information includes soil pH value, organic matter content and other information of the target rapeseed field. Meteorological information refers to the precipitation, temperature and other information of the target rapeseed field during the rapeseed planting period. Sowing time information refers to the time of sowing rapeseed in the target rapeseed field, such as sowing on March 25. Rapeseed variety information refers to the type of rapeseed sown in the target rapeseed field, such as Zhongyouza No. 19, Fengyou 737, Qinyou No. 10, etc. Rapeseed growth-related information is an important factor affecting the maturity time of rapeseed. For example, winter rapeseed sown in winter takes 160-290 days from sowing to maturity, and spring rapeseed sown during the Spring Festival takes 85-130 days from sowing to maturity. The growth cycle of Zhongyouza No. 19 is about 230 days, and the growth cycle of Qinyou No. 10 is between 230 days and 248 days.

[0087] S102: Input rapeseed variety information, meteorological information and soil information into a pre-built rapeseed growth model to predict the rapeseed growth cycle, and determine the predicted maturity time period of the target rapeseed field according to the rapeseed growth cycle prediction result.

[0088] In this embodiment, the predicted maturity time period refers to the time period from sowing to maturity of the rapeseed in the target rapeseed field, which is also the growth cycle of rapeseed. For example, the predicted maturity time period is 201 days. A rapeseed growth model is constructed based on a neural network model. Taking a multi-layer perceptron (MLP) as an example, historical rapeseed information of the area where the target rapeseed field is located is collected. The historical rapeseed information includes the growth cycle of different varieties of rapeseed under different weather and soil conditions. The historical rapeseed information is randomly divided into a model training set and a model verification set of the rapeseed growth model after data cleaning and information annotation. The rapeseed growth model includes an input layer for receiving information, a hidden layer for performing nonlinear transformation and feature extraction on the input information, and an output layer for generating the final prediction result. The initial number of nodes is set for the input layer, the hidden layer, and the output layer according to the characteristic dimension of the historical rapeseed information, and a suitable loss function and optimizer are selected. For example, the mean square error (MSE) can be selected as the loss function, and the optimizer can select SGD (stochastic gradient descent). The loss function is used to measure the difference between the prediction result of the model and the true label, and to measure the performance of the model. The core function of the optimizer is to update the parameters of the model according to the gradient of the loss function. It can be used to select an appropriate learning rate or dynamically adjust the learning rate during the training process, thereby accelerating the convergence speed and improving the model training speed. After multiple iterations of training the rapeseed growth model using the model training set, the model parameters are continuously optimized, and the model learning rate, batch size and other parameters are adjusted until the maximum number of iterations is reached. The validation set is then used to evaluate the model performance of the trained rapeseed growth model. The parameters used for evaluation include accuracy, recall rate, etc. When each parameter reaches the preset parameter threshold, the rapeseed growth model training is completed. The rapeseed variety information, meteorological information and soil information are input into the trained rapeseed growth model, and the predicted maturity period of the target rapeseed field is output through the rapeseed growth model.

[0089] S103: Determine a predicted harvesting time node of the target rapeseed field according to the sowing time information and the predicted maturity time period.

[0090] In this embodiment, the sowing time information is used as the initial node, and the initial node plus the predicted maturity time period can obtain the predicted harvesting time node of the target rapeseed field. For example, if the rapeseed planting time information is March 21, 2021, and the predicted maturity time period is 100 days, the predicted harvesting time node is June 29, 2021. The predicted harvesting time node is the rapeseed maturity time node of the target rapeseed field predicted by the rapeseed growth model, and it is also the most suitable time node for harvesting rapeseed. Harvesting at the time node when the rapeseed in the target rapeseed field is just mature can effectively reduce the rapeseed harvesting loss rate. This is because if the rapeseed is harvested at the time node when it is over-mature or immature, the rapeseed harvesting loss rate will increase, so it is necessary to choose the right time for harvesting.

[0091] S104. At the predicted harvest time node, remote sensing information of rapeseed in the target rapeseed field is collected by remote sensing drones.

[0092] In this embodiment, at the predicted harvest time node, a remote sensing drone equipped with remote sensing equipment is controlled to collect rapeseed remote sensing information of the target rapeseed field. Common remote sensing equipment includes hyperspectral imagers, thermal infrared sensors, optical cameras, infrared cameras, etc. At the predicted harvest time node, the remote sensing drone is controlled to shoot the target rapeseed field in all directions and in different regions according to the pre-set monitoring method and monitoring path. After the shooting is completed, the remote sensing images captured are spliced ​​according to the actual scene of the target rapeseed field according to the real-time positioning information of the remote sensing drone to obtain complete rapeseed remote sensing information of the target rapeseed field. Rapeseed remote sensing information refers to optical remote sensing images, which are collected by multispectral cameras and hyperspectral imagers and other equipment set inside the remote sensing drone.

[0093] S105 . Acquire rapeseed field size information of the target rapeseed field based on the rapeseed remote sensing information, and divide the target rapeseed field into grids according to the rapeseed field size information to obtain a plurality of grid rapeseed fields.

[0094] In this embodiment, the rapeseed remote sensing information is first subjected to geometric correction, image denoising, and radiation calibration, wherein geometric correction refers to the process of eliminating or correcting the geometric errors of remote sensing images, and these geometric errors refer to image distortion caused by factors such as objective lens distortion and atmospheric refraction. The main purpose of image denoising is to remove noise in rapeseed remote sensing information and improve the accuracy of rapeseed remote sensing information. Radiation calibration is the process of converting the brightness grayscale value of rapeseed remote sensing information into an absolute radiation brightness value. Through the above steps, the image quality of rapeseed remote sensing information can be improved, image distortion can be reduced, and the error of subsequent analysis of rapeseed maturity can be reduced. After completing the above preprocessing steps, the rapeseed field size information of the target rapeseed field is calculated according to the rapeseed remote sensing information. Since the edge area of ​​the rapeseed field may be connected to non-rapeseed field areas such as water bodies, roads, ridges, and buildings when the remote sensing drone collects the target rapeseed field, in order to more accurately calculate the rapeseed field size information of the target rapeseed field, the non-rapeseed field area can be eliminated according to the rapeseed remote sensing information.

[0095] The specific method is as follows: the red edge normalized index of the target rapeseed field is calculated based on the reflectance of multiple bands of different wavelengths in the rapeseed remote sensing information, and the area with a red edge normalized index greater than zero is eliminated. The calculation formula is: N NDVI 705 ＝( ρ 750 -ρ 705 ) / (ρ 750 +ρ 705 ), where ρ 750 Represents the reflectivity of the band with a central wavelength of 750nm, ρ705 The red edge normalization index is the reflectance of the band with a central wavelength of 705nm. By calculating the red edge normalization index, non-rapeseed field areas such as water bodies can be eliminated, and then according to the rapeseed remote sensing information, non-rapeseed field areas such as roads, ridges, and buildings whose reflectance of the green light band (550nm) is not less than that of the red light band (680nm) can be eliminated, and finally a complete target rapeseed field can be obtained.

[0096] After obtaining the complete target rapeseed field, the scale information in the rapeseed remote sensing information can be directly obtained by using geographic information system (GIS) software or other remote sensing data processing software. The scale information of the rapeseed remote sensing information can also be determined according to the flight altitude of the UAV and the parameters of the remote sensing equipment. After obtaining the scale information, the image size information of the target rapeseed field in the rapeseed remote sensing information can be directly calculated by using GIS software. Finally, the actual size information of the target rapeseed field, that is, the rapeseed field size information, can be calculated according to the scale information and the image size information.

[0097] After obtaining the rapeseed field size information, the target rapeseed field is gridded according to the rapeseed field size information, and the target rapeseed field can be divided into multiple grid rapeseed fields of equal area according to the rapeseed field size information. The purpose of gridding the target rapeseed field is to more accurately select a rapeseed trial harvesting area in the target rapeseed field that can represent the entire target rapeseed field. Since the target rapeseed field is large in area, the rapeseed maturity in different areas of the target rapeseed field may not be the same, and there may be certain errors in the predicted harvesting time node of the target rapeseed field. Therefore, it is necessary to select an area in the target rapeseed field where the rapeseed maturity difference is large and the distribution of rapeseed maturity is similar to that of the target rapeseed field as the rapeseed trial harvesting area. First, a trial harvest is performed in the rapeseed trial harvesting area, and the predicted harvesting time node is judged to be the optimal harvesting node based on the trial harvesting results.

[0098] S106. Performing feature analysis on all grid rapeseed fields according to the rapeseed remote sensing information to obtain rapeseed remote sensing features, and calculating rapeseed maturity of all grid rapeseed fields based on the rapeseed remote sensing features.

[0099] In this embodiment, the rapeseed remote sensing feature includes the rapeseed spectral feature and the rapeseed color feature. The step of extracting the rapeseed spectral feature includes: using the geographic information system software (ArcGIS), remote sensing image processing platform (ENVI) and other remote sensing image processing software to extract the reflection values ​​of the red edge band, blue band and near infrared band in the rapeseed remote sensing information, that is, the spectral reflection value, and then using the spectral reflection value to calculate the normalized vegetation index (NDVI), enhanced vegetation index (EVI) and water index (WBI) of each grid rapeseed field in the target rapeseed field, and integrating them to obtain the rapeseed spectral index. The rapeseed spectral index can reflect the maturity of the rapeseed. For example, the mature rapeseed leaves will gradually turn yellow, so its NDVI value will decrease. After the rapeseed matures, the moisture content will also decrease, so the WBI value will decrease accordingly. After the rapeseed spectral index is normalized, the rapeseed spectral index that has completed the normalization is characterized by linear discriminant analysis, random forest, weighted average and principal component analysis to obtain the rapeseed spectral feature.

[0100] The extraction steps of rapeseed color features include: forming a matrix of R, G, and B values ​​of image pixels in rapeseed remote sensing information to obtain an RGB matrix, then calculating the covariance matrix of the RGB matrix, calculating its eigenvalues ​​and eigenvectors according to the calculated covariance matrix, sorting the eigenvalues ​​from large to small, forming a matrix of their corresponding eigenvectors, completing the principal component transformation, obtaining the spatial color features, and then clustering the spatial color features through a clustering algorithm to obtain the cluster center, that is, the most representative color. The most representative color is the rapeseed color feature. The principle of using rapeseed color features to analyze rapeseed maturity is that the color of rapeseed will gradually change from tender green to yellow during the gradual maturity process, and the rapeseed pods will also gradually change to yellow-brown. Therefore, the rapeseed maturity of each grid rapeseed field can be determined according to the color change of rapeseed.

[0101] After obtaining the spectral characteristics and color characteristics of rapeseed, they are sequentially input into the rapeseed maturity prediction model. The rapeseed maturity prediction model can be constructed based on a multi-layer feedforward neural network model (BP neural network model). The BP neural network is a forward multi-layer network that uses an error back propagation algorithm to train the network. It usually includes an input layer, a hidden layer, and an output layer. The number of neurons in the input layer can be set to two, one of which is the spectral characteristics of rapeseed, and the other is the color characteristics of rapeseed. The output layer can also be set to multiple neurons, for example, four, and a four-bit binary number is used to represent the maturity of rapeseed. The rapeseed maturity prediction model is trained using a pre-constructed training set, which includes historical rapeseed remote sensing characteristics of various maturity levels. The rapeseed maturity prediction model that has been trained is used to predict the rapeseed maturity of each grid rapeseed field.

[0102] S107. Select a benchmark grid rapeseed field from all grid rapeseed fields according to rapeseed maturity. Taking the benchmark grid rapeseed field as the center and maximizing the rapeseed maturity difference as the screening target, perform iterative diffusion screening on adjacent grid rapeseed fields of the benchmark grid rapeseed field until the number of screened rapeseed fields is greater than or equal to a preset number threshold, thereby obtaining a rapeseed trial harvesting area.

[0103] In this embodiment, all grid rapeseed fields are first divided into maturity levels according to rapeseed maturity and using a threshold method to obtain rapeseed maturity levels of all grid rapeseed fields. The grid rapeseed fields with the largest rapeseed maturity level or the smallest rapeseed maturity level among all grid rapeseed fields are sequentially marked as trial-cut rapeseed fields. For any trial-cut rapeseed field, the trial-cut rapeseed field is taken as the center, and the maturity level difference between the adjacent rapeseed field of the trial-cut rapeseed field and the trial-cut rapeseed field is calculated. The adjacent rapeseed fields with a maturity level difference greater than or equal to a preset difference threshold are marked as trial-cut rapeseed fields. After the adjacent rapeseed fields are marked as trial-cut rapeseed fields, the marked adjacent rapeseed field is taken as the center, and the above trial-cut rapeseed field marking step is repeated. After the trial-cut rapeseed field marking step is repeated multiple times, the number of trial-cut rapeseed fields gradually increases until the number of trial-cut rapeseed fields is greater than or equal to the preset number threshold, so as to prevent the purpose of trial cutting from being unable to be achieved due to too many or too few trial-cut rapeseed fields.

[0104] After completing the marking step, the adjacent trial-cut rapeseed fields are integrated to obtain multiple initial trial-cut areas. In order to further enable the final selected trial-cut areas to more accurately represent the target rapeseed fields, it is necessary to perform a secondary screening on all initial trial-cut areas. First, the target maturity level matrix of the target rapeseed field is constructed according to the rapeseed maturity levels of different grid rapeseed fields in the target rapeseed field. Similarly, a regional maturity level matrix is ​​constructed for all initial trial-cut areas based on the rapeseed maturity level. Then, the similarity between each regional maturity level matrix and the target maturity level matrix is ​​calculated using a similarity calculation algorithm to obtain multiple matrix similarities. The matrix similarity is also the similarity of the rapeseed maturity distribution characteristics. The initial trial-cut area of ​​the matrix similarity is used as the rapeseed trial-cut area.

[0105] S108, controlling the rapeseed harvesting equipment to perform a rapeseed trial harvesting task in the rapeseed trial harvesting area, and during the execution of the rapeseed trial harvesting task, using the rapeseed harvesting equipment to count the trial harvesting loss rate in the rapeseed trial harvesting area, and analyzing the total rapeseed loss in the target rapeseed field based on the trial harvesting loss rate and using a correlation analysis method.

[0106] In this embodiment, when the rapeseed harvesting equipment is controlled to reach different grid rapeseed fields in the rapeseed trial harvesting area, the rapeseed image information of the grid rapeseed fields is collected by using an image acquisition device (which can be a high-definition camera) arranged outside the rapeseed harvesting equipment, and the rapeseed height information of the different grid rapeseed fields is calculated according to the rapeseed image information and the parameters of the image acquisition device. The equipment parameters of the rapeseed harvesting equipment are set in combination with the rapeseed height information and the rapeseed maturity, and the harvesting equipment parameters of the different grid rapeseed fields are obtained, and the harvesting equipment parameters are stored. While storing the harvesting equipment parameters, the rapeseed trial harvesting data of the current time node is also stored and matched with the corresponding grid rapeseed field coordinate range. The rapeseed trial harvesting data includes the rapeseed harvesting weight and the rapeseed falling amount. The rapeseed harvesting weight is obtained by a weight sensor arranged inside the rapeseed harvester, which refers to the weight of the rapeseed harvested in the rapeseed harvester at the current time node. The rapeseed falling amount is obtained by an image acquisition device arranged at the tail of the rapeseed harvester. The image of the rapeseed left on the ground after the harvest is completed is collected by the image acquisition device (which can be a high-definition camera), and the rapeseed density in the rapeseed image is calculated using image recognition technology. Then, the rapeseed falling amount in each grid rapeseed field is calculated based on the rapeseed density in the rapeseed image. The trial cutting loss rate of all grid rapeseed fields in the rapeseed trial cutting area is calculated according to the rapeseed harvesting weight and the rapeseed falling amount. The trial cutting loss rate of each grid rapeseed field and the corresponding rapeseed maturity are correlated to obtain the correlation analysis results. The correlation analysis can be performed by calculating the Pearson correlation coefficient, the Spearman rank correlation coefficient, the Kendall rank correlation coefficient, etc., and then the rapeseed loss rate of all grid rapeseed fields except the rapeseed trial cutting area is calculated according to the correlation analysis results. The average rapeseed loss rate of the target rapeseed field is calculated by combining the total trial cutting loss rate and the rapeseed loss rate. The total rapeseed loss of the target rapeseed field is calculated according to the average rapeseed loss rate and the pre-acquired estimated rapeseed yield. The estimated rapeseed yield is calculated based on the historical rapeseed yield of the target rapeseed field. The rapeseed loss rate refers to the rapeseed falling during the rapeseed harvesting process due to reasons such as unripe or overripe rapeseed, insufficient performance of rapeseed harvesting equipment, etc., resulting in rapeseed harvesting losses. Therefore, the rapeseed loss rate is also the rapeseed drop rate. By estimating the rapeseed loss rate and adjusting the equipment parameters of the rapeseed harvesting equipment according to the rapeseed loss rate, the rapeseed loss rate can be reduced, the rapeseed yield can be increased, and the economic benefits for rapeseed growers can be increased.

[0107] S109. Determine the rapeseed maturity node of the target rapeseed field in combination with the rapeseed remote sensing characteristics and the rapeseed total loss, and judge whether the predicted harvesting time node is the optimal harvesting node based on the rapeseed maturity node.

[0108] In this embodiment, firstly, it is judged whether the predicted harvesting time node is the optimal harvesting node according to the total loss of rapeseed. If the predicted harvesting time node is not the optimal harvesting node, the rapeseed maturity node is predicted according to the rapeseed remote sensing characteristics and rapeseed image information. Specifically, a rapeseed maturity prediction model is constructed based on a neural network model and a self-attention mechanism is introduced. The rapeseed maturity prediction model is iteratively trained using a training set with different rhizome colors, rapeseed heights and remote sensing characteristics and pre-annotated. After reaching the maximum number of training times, the rhizome color characteristics, rapeseed height information and rapeseed remote sensing characteristics are input into the trained rapeseed maturity prediction model, and the predicted rapeseed maturity of the target rapeseed field is output through the rapeseed maturity prediction model. The rapeseed maturity is added to the current time node (predicted harvesting time node) to obtain the rapeseed maturity node of the target rapeseed field. If the rapeseed maturity node is later than the predicted harvesting time node, it is determined that the predicted harvesting time node is not the optimal harvesting node; if the rapeseed maturity node is not later than the predicted harvesting time node, it is determined that the predicted harvesting time node is the optimal harvesting node. During the rapeseed harvesting process, the rapeseed loss rate is closely related to the maturity of the rapeseed. Over-maturity or under-maturity of the rapeseed will lead to an increase in the rapeseed loss rate. In order to minimize the rapeseed loss rate, it is necessary to accurately determine the time when the rapeseed is moderately mature, that is, the optimal harvesting node. Harvesting rapeseed at the optimal harvesting node can effectively reduce the rapeseed loss rate. At the same time, adjusting the harvesting parameters of the rapeseed harvesting equipment according to the maturity of the rapeseed can further reduce the rapeseed harvesting loss rate. A two-pronged approach of selecting the optimal harvesting node and adjusting the optimal parameters can minimize the rapeseed harvesting loss rate and increase the yield of the target rapeseed field.

[0109] S110. If the predicted harvesting time node is the optimal harvesting node, the rapeseed maturity and the trial harvesting loss rate are combined and the parameter optimization algorithm is used to generate the optimal parameter range of the rapeseed harvesting equipment, and the rapeseed harvesting equipment is controlled to harvest the rapeseed in the target rapeseed field according to the optimal parameters.

[0110] In this embodiment, if the predicted harvesting time node is the optimal harvesting node, all grid rapeseed fields in the rapeseed trial harvesting area with a rapeseed loss rate less than the loss rate threshold are first screened out to obtain multiple standard rapeseed fields, and the harvesting equipment parameters of the rapeseed harvesting equipment within the corresponding standard rapeseed field range are obtained according to the coordinate range of each standard rapeseed field. Multiple random parameters are first generated according to all initial parameter intervals, and the generated multiple random parameters are all located in the corresponding initial parameter interval. Multiple random parameters are generated multiple times to ensure that the initial population has sufficient diversity to avoid the algorithm from converging to the local optimal solution too early. Real number coding can be used to construct multiple initial parameter populations. The initial parameter population is optimized by using a genetic algorithm until the maximum number of parameter optimization times is reached, and multiple target device parameters are output. The target parameter interval is constructed according to the multiple target device parameters. The above update iteration steps are repeated until the preset maximum number of iterations is reached to obtain the optimal parameter interval of the rapeseed harvesting equipment. The genetic algorithm is a computational model of the biological evolution process that simulates the natural selection and genetic mechanism of Darwin's theory of biological evolution. It is a method for searching for the optimal solution by simulating the natural evolution process. The algorithm uses a mathematical method and computer simulation to convert the problem-solving process into a process similar to the crossover and mutation of chromosome genes in biological evolution. In the subsequent formal harvesting process, the rapeseed harvesting equipment is controlled to make small parameter adjustments within the optimal parameter range according to the maturity and height of the rapeseed in the target rapeseed field, ensuring that the harvesting equipment parameters of the rapeseed harvesting equipment can continue to remain within the optimal parameter range, thereby reducing the rapeseed loss rate.

[0111] In one embodiment, the rapeseed fields of all grids are analyzed for characteristics according to the rapeseed remote sensing information to obtain the rapeseed remote sensing characteristics, and the rapeseed maturity of all grid rapeseed fields is calculated based on the rapeseed remote sensing characteristics, including the following steps:

[0112] Pre-processing the rapeseed remote sensing information of all grid rapeseed fields;

[0113] For any grid rapeseed field, the reflectance spectrum of the rapeseed remote sensing information that has completed preprocessing is extracted;

[0114] The spectral reflectance value is obtained according to the reflectance spectrum calculation, and the spectral reflectance value is subjected to feature extraction to obtain the spectral feature of rapeseed;

[0115] The principal component transformation of rapeseed remote sensing information is performed to obtain the spatial color characteristics of rapeseed remote sensing information;

[0116] The clustering algorithm is used to classify the spatial color features and obtain the rapeseed color features of rapeseed remote sensing information;

[0117] The spectral characteristics and color characteristics of rapeseed are integrated into the remote sensing characteristics of rapeseed in grid rapeseed fields;

[0118] A rapeseed maturity prediction model was constructed based on the neural network model, and all rapeseed remote sensing features were input into the trained rapeseed maturity prediction model. The rapeseed maturity prediction model was used to output the rapeseed maturity of all grid rapeseed fields.

[0119] In this embodiment, the preprocessing steps of rapeseed remote sensing information include: geometric correction, image denoising, and radiation calibration, wherein geometric correction refers to the process of eliminating or correcting the geometric errors of remote sensing images, and these geometric errors refer to image distortion caused by factors such as objective lens distortion and atmospheric refraction. The main purpose of image denoising is to remove noise in rapeseed remote sensing information and improve the accuracy of rapeseed remote sensing information. Radiation calibration is the process of converting the brightness grayscale value of rapeseed remote sensing information into an absolute radiation brightness value. Through the above steps, the image quality of rapeseed remote sensing information can be improved and image distortion can be reduced.

[0120] The remote sensing characteristics of rapeseed include rapeseed spectral characteristics and rapeseed color characteristics. The steps of extracting the spectral characteristics of rapeseed include: using remote sensing image processing software such as geographic information system software (ArcGIS) and remote sensing image processing platform (ENVI) to extract the reflectance values ​​of the red edge band, blue band and near infrared band in the rapeseed remote sensing information, that is, the spectral reflectance values, and then using the spectral reflectance values ​​to calculate the normalized vegetation index (NDVI), enhanced vegetation index (EVI) and water content index (WBI) of each grid rapeseed field in the target rapeseed field.

[0121] The above index calculation formula includes: NDVI = (NR) / (N + R), where N is the reflectance of the near infrared band and R is the reflectance of the red light band; EVI = 2.5 * ((NR) / (N + 6R-7.5B + ​​1)), where B represents the reflectance of the blue light band, and WBI = (NS) / (N + S), where S is the reflectance of the short-wave infrared band. The rapeseed spectral index can reflect the maturity of rapeseed. For example, mature rapeseed leaves will gradually turn yellow, so its NDVI value will decrease. When rapeseed matures, its moisture content will also decrease, so the WBI value will decrease accordingly. After the rapeseed spectral index is normalized, the rapeseed spectral index that has been normalized is fused using linear discriminant analysis, random forest, weighted average and principal component analysis to obtain the rapeseed spectral characteristics. Taking principal component analysis as an example, the covariance matrix, eigenvalues ​​and eigenvectors between the indexes are calculated, and then the principal components are selected for fusion to obtain the rapeseed spectral characteristics.

[0122] The extraction step of rapeseed color features includes: forming a matrix of R, G, and B values ​​of image pixels in rapeseed remote sensing information to obtain an RGB matrix, calculating the average value of each variable in the RGB matrix, calculating the covariance of each variable in the RGB matrix according to the calculated average value, constructing the covariance matrix of the RGB matrix according to the calculated covariance, calculating its eigenvalues ​​and eigenvectors using the covariance matrix calculated by the characteristic polynomial, sorting the eigenvalues ​​from large to small, forming a matrix of the corresponding eigenvectors to obtain a PCA transformation matrix, i.e., a spatial color feature, and then clustering the spatial color feature through a clustering algorithm to obtain a cluster center, i.e., the most representative color, and the most representative color is the rapeseed color feature. Commonly used clustering algorithms include hierarchical clustering algorithms, spectral clustering algorithms, and K-means algorithms. Taking the K-means algorithm as an example, first select k initial cluster centers from the spatial color feature, assign each pixel point to the nearest cluster center, recalculate the center point of each cluster, repeat the assignment and update steps until the cluster center no longer changes or reaches a predetermined number of iterations, the iteration ends, and the rapeseed color feature is obtained. The principle of using rapeseed color characteristics to analyze rapeseed maturity is that the color of rapeseed will gradually change from light green to yellow during the gradual maturity process. In addition, the rapeseed pods will gradually turn into yellow-brown. Therefore, the rapeseed maturity of each grid rapeseed field can be determined based on the color changes of rapeseed.

[0123] After obtaining the spectral characteristics and color characteristics of rapeseed, they are sequentially input into the rapeseed maturity prediction model. The rapeseed maturity prediction model can be constructed based on a multi-layer feedforward neural network model (BP neural network model). The BP neural network is a forward multi-layer network that uses an error back propagation algorithm to train the network. It usually includes an input layer, a hidden layer, and an output layer. The number of neurons in the input layer can be set to two, one of which is the spectral characteristics of rapeseed, and the other is the color characteristics of rapeseed. The output layer can also be set to multiple neurons, for example, four, and a four-bit binary number is used to represent the maturity of rapeseed. The rapeseed maturity prediction model is trained using a pre-constructed training set, which includes historical rapeseed remote sensing characteristics of various maturity levels. The rapeseed maturity prediction model that has been trained is used to predict the rapeseed maturity of each grid rapeseed field.

[0124] In one embodiment, a rapeseed maturity prediction model is constructed based on a neural network model, and all rapeseed remote sensing features are input into the trained rapeseed maturity prediction model. Outputting the rapeseed maturity of all grid rapeseed fields through the rapeseed maturity prediction model includes the following steps:

[0125] Construct rapeseed maturity prediction model based on BP neural network model;

[0126] The rapeseed maturity prediction model was trained using the pre-built rapeseed training set;

[0127] For any grid rapeseed field, the rapeseed spectral characteristics and rapeseed color characteristics of the grid rapeseed field are sequentially input into the rapeseed maturity prediction model to perform maturity prediction, and a first maturity prediction value and a second maturity prediction value are obtained;

[0128] Using Kalman filtering to filter and correct the first maturity prediction value and the second maturity prediction value;

[0129] The first maturity prediction value and the second maturity prediction value after filtering and correction are fused to obtain the rapeseed maturity of all grid rapeseed fields.

[0130] In this embodiment, a rapeseed maturity prediction model is constructed based on a BP neural network model. The BP neural network is a multi-layer feedforward neural network, which is characterized by forward transmission of signals and back propagation of errors. It usually includes an input layer, a hidden layer, and an output layer. The input layer is used to receive rapeseed spectral features and rapeseed color features. Two neurons can be set to receive the above two features respectively. The hidden layer is the core part of the neural network, which is responsible for feature extraction and nonlinear transformation. The input data is mapped to the output layer through the connection and activation function between the multi-layer neurons. Softmax (normalized exponential function) can be selected as the activation function of the hidden layer. The hidden layer can be set to two layers. The neurons of the output layer are calculated by weighting and activation functions to obtain output results. The neurons of the output layer can be set to four or five, and four or five binary numbers can be used to represent rapeseed maturity. The pre-marked training samples are divided into a rapeseed training set and a rapeseed verification set. The rapeseed training set is used to train the rapeseed maturity prediction model, and the learning rate, batch size and other parameters are continuously and dynamically adjusted until the maximum number of iterations is reached. Then, the rapeseed validation set is used to evaluate the model performance of the trained rapeseed maturity prediction model. The parameters used for evaluation include accuracy, recall rate, etc. When each parameter reaches the preset parameter threshold, the rapeseed maturity prediction model training is completed. The rapeseed spectral characteristics and rapeseed color characteristics of each grid rapeseed field are input into the rapeseed maturity prediction model in sequence for maturity prediction. The rapeseed maturity prediction model predicts the first maturity prediction value and the second maturity prediction value based on the rapeseed spectral characteristics and rapeseed color characteristics.

[0131] Then, the first maturity prediction value and the second maturity prediction value are filtered and corrected by using Kalman filtering. Specifically, a secondary prediction is performed based on the first maturity prediction value and the second maturity prediction value to obtain two secondary maturity prediction values. The first maturity prediction value and the second maturity prediction value are corrected by using the secondary maturity prediction value, and then the Kalman gain is calculated by using the Kalman filtering equation. The Kalman gain is a weight coefficient between 0 and 1. The Kalman gain is used to perform data fusion on the first maturity prediction value and the second maturity prediction value after the correction to obtain a more accurate rapeseed maturity. The Kalman filtering algorithm is an optimal recursive filtering method under a discrete system. It can reduce data errors based on data prediction and data update. The Kalman filtering can improve the accuracy of the maturity prediction value and obtain a more accurate rapeseed maturity, so that when the harvesting parameters of the rapeseed harvesting equipment are adjusted according to the rapeseed maturity, the adjusted harvesting equipment parameters can be more reasonable, thereby reducing the rapeseed loss rate.

[0132] In one embodiment, selecting a plurality of target grid rapeseed fields from all grid rapeseed fields as rapeseed trial harvesting areas according to rapeseed maturity includes the following steps:

[0133] According to the maturity of rapeseed, the maturity level of all grid rapeseed fields is divided by threshold method to obtain the maturity level of rapeseed in all grid rapeseed fields;

[0134] If the rapeseed maturity level is the maximum rapeseed maturity level or the rapeseed maturity level is the minimum rapeseed maturity level, all grid rapeseed fields corresponding to the maximum rapeseed maturity level and the minimum rapeseed maturity level are marked as trial rapeseed fields respectively;

[0135] For any trial rapeseed field, the trial rapeseed field marking step is repeatedly performed based on the rapeseed maturity level and with the trial rapeseed field as the center until the number of trial rapeseed fields is greater than or equal to a preset number threshold. The trial rapeseed field marking step is:

[0136] Calculating the maturity grade difference between the adjacent rapeseed field of the trial-cut rapeseed field and the trial-cut rapeseed field, and recording the adjacent rapeseed field whose maturity grade difference is greater than or equal to a preset difference threshold as the trial-cut rapeseed field;

[0137] Integrate all adjacent trial-harvesting rapeseed fields to obtain multiple initial trial-harvesting areas;

[0138] According to the rapeseed maturity levels of all grid rapeseed fields, a target maturity level matrix of the target rapeseed field is constructed;

[0139] For any initial trial cutting area, a regional maturity grade matrix of the initial trial cutting area is constructed based on the rapeseed maturity grade;

[0140] Based on the similarity calculation algorithm, the similarities between the target maturity level matrix and the maturity level matrices of all regions are calculated respectively to obtain multiple matrix similarities;

[0141] The initial trial cutting area corresponding to the maximum data similarity is used as the rapeseed trial cutting area.

[0142] In this embodiment, refer to Figure 2 First, all grid rapeseed fields are divided into maturity levels according to rapeseed maturity and using the threshold method. For example, when the rapeseed maturity is 1-3, it is divided into the first maturity level, when the rapeseed maturity is 4-6, it is divided into the second maturity level, when the rapeseed maturity is 7-9, it is divided into the third maturity level, when the rapeseed maturity is 10-13, it is divided into the fourth maturity level, and when the rapeseed maturity is 14-16, it is the fifth maturity level. And so on, the rapeseed maturity level of all grid rapeseed fields is obtained. The grid rapeseed fields with the largest rapeseed maturity level or the smallest rapeseed maturity level in all grid rapeseed fields are marked as trial rapeseed fields in turn. For example, if the maximum rapeseed maturity level in all grid rapeseed fields is the fifth maturity level and the smallest rapeseed maturity level is the second maturity level, then all grid rapeseed fields corresponding to the second maturity level and the fifth maturity level are marked as trial rapeseed fields.

[0143] In addition, in some cases, if the distance between a certain grid rapeseed field with the largest or smallest rapeseed maturity level and other grid rapeseed fields with the largest or smallest rapeseed maturity level is less than the preset distance threshold, that is, the distance between the two earliest marked trial rapeseed fields is less than the preset distance threshold, at this time, the straight-line distance between the center points of the two and the edge of the target rapeseed field is calculated, and the trial rapeseed field with a smaller straight-line distance is selected for marking, and the mark of the grid rapeseed field with a larger distance is removed. If the distances between the two and the edge of the target rapeseed field are equal, the mark of one trial rapeseed field is randomly removed to ensure that the distances between the earliest marked trial rapeseed fields are greater than the distance threshold, so as to prevent large-scale overlap between the two primary trial rapeseed areas in the future, thereby saving computing power.

[0144] For any trial rapeseed field, the trial rapeseed field is taken as the center, and the maturity level difference between the adjacent rapeseed field of the trial rapeseed field and the trial rapeseed field is calculated, and the adjacent rapeseed field whose maturity level difference is greater than or equal to the preset difference threshold is marked as a trial rapeseed field, wherein the adjacent rapeseed field refers to a grid rapeseed field with overlapping edges with the trial rapeseed field. After the adjacent rapeseed field is marked as a trial rapeseed field, the marked adjacent rapeseed field is taken as the center, and the trial rapeseed field marking step is repeated. After repeating the trial rapeseed field marking step multiple times, the number of trial rapeseed fields gradually increases until the number of trial rapeseed fields is greater than or equal to the preset number threshold. The number threshold is set to prevent the purpose of trial cutting from being unable to be achieved due to too many or too few trial rapeseed fields. The number threshold is determined based on the rapeseed field size information of the target rapeseed field and the total number of grid rapeseed fields. In addition, the reason why the rapeseed field marking step is stopped when the number is greater than or equal to the quantity threshold is that in some cases, the number of marked trial-cut rapeseed fields may not be exactly equal to the quantity threshold. For example, when the number of marked trial-cut rapeseed fields is less than the quantity threshold, and the above trial-cut rapeseed field marking steps are repeated, there are exactly two trial-cut rapeseed fields that meet the marking conditions, and therefore the number of trial-cut rapeseed fields finally marked is greater than the quantity threshold.

[0145] After completing the marking step, the adjacent trial rapeseed fields are integrated. Since the target rapeseed field is large in area and the total number of grid rapeseed fields is large, multiple initial trial rapeseed areas may be obtained after the adjacent trial rapeseed fields are integrated. The above steps are to ensure that the rapeseed maturity level difference between different grid rapeseed fields in the final selected trial rapeseed area is the largest. This is because if the rapeseed maturity level difference between adjacent grid rapeseed fields in the trial rapeseed area is large, then the difference between the rapeseed trial rapeseed data of different grid rapeseed fields will also be greater, and the change of rapeseed harvesting equipment parameters in different grid rapeseed fields will also be greater, so that it is easier to analyze the relationship between rapeseed maturity level and rapeseed trial rapeseed data, as well as the relationship between rapeseed maturity and harvesting equipment parameters, so as to achieve the purpose of overall evaluation of the target rapeseed field according to the execution results of the rapeseed trial rapeseed task.

[0146] In addition, in order to further make the final selected trial cutting area more accurately represent the target rapeseed field, it is necessary to perform a secondary screening on all the initial trial cutting areas. First, the target maturity level matrix of the target rapeseed field is constructed according to the rapeseed maturity level of different grid rapeseed fields in the target rapeseed field. Similarly, the regional maturity level matrix is ​​constructed for all the initial trial cutting areas based on the rapeseed maturity level. Then, the similarity between each regional maturity level matrix and the target maturity level matrix is ​​calculated using the similarity calculation algorithm to obtain multiple matrix similarities, and the initial trial cutting area with the largest matrix similarity is used as the rapeseed trial cutting area. Commonly used similarity calculation algorithms include Frobenius norm, cosine similarity, Pearson correlation coefficient, and relative entropy (Kullback-Leibler divergence). Taking the Frobenius norm as an example, the regional maturity level matrix is ​​first filled with data to ensure that the filled regional maturity level matrix and the target maturity level matrix have the same size. Then the difference matrix between the regional maturity level matrix and the target maturity level matrix is ​​calculated, and the square root of the sum of squares of each element in the difference matrix is ​​calculated to obtain the Frobenius norm between the regional maturity level matrix and the target maturity level matrix. The Frobenius norm indicates the degree of difference between the two matrices. The smaller the Frobenius norm, the more similar the two matrices are. Therefore, the Frobenius norm between the target maturity level matrix and all regional maturity level matrices is calculated.

[0147] Then, the grid rapeseed field located at the edge of the target rapeseed field is taken as the edge rapeseed field, and the distance between the center point of each initial trial cutting area and the center point of the edge rapeseed field in the horizontal and vertical directions is calculated, and the shortest distance is selected as the edge center distance of the initial trial cutting area. Weights are assigned to the Frobenius norm and the edge center distance respectively, for example, the Frobenius norm weight is 0.8, and the edge center distance weight is 0.2, and the initial trial cutting area with a smaller Frobenius norm and a smaller edge center distance is selected as the trial cutting area. The purpose of the secondary screening is to select the initial trial cutting area that is most similar to the rapeseed maturity grade distribution of the target rapeseed field and closest to the edge of the target rapeseed field as the trial cutting area. The matrix similarity is calculated to make the screened trial cutting area more representative. The edge center distance is calculated to facilitate the rapeseed harvesting equipment to enter the trial cutting area. Because the rapeseed harvesting equipment will pre-harvest the rapeseed in a part of the area before entering the trial cutting area to pave the way for entering the trial cutting area later, if the edge center distance is too large, the area of ​​the pre-harvested area will increase, affecting the subsequent trial cutting results, so that the total rapeseed loss of the target rapeseed field calculated based on the trial cutting results can be more accurate, and at the same time, the generated optimal parameter interval can significantly reduce the total rapeseed loss of the target rapeseed field.

[0148] In one embodiment, controlling a rapeseed harvesting device to perform a rapeseed trial harvesting task in a rapeseed trial harvesting area, and counting a trial harvesting loss rate of the rapeseed trial harvesting area by the rapeseed harvesting device during the execution of the rapeseed trial harvesting task, and analyzing the total rapeseed loss of a target rapeseed field based on the trial harvesting loss rate and using a correlation analysis method include the following steps:

[0149] When the rapeseed harvesting equipment arrives at any grid rapeseed field in the rapeseed trial harvesting area, the equipment parameters of the rapeseed harvesting equipment are updated based on the rapeseed maturity of the grid rapeseed field to obtain the harvesting equipment parameters;

[0150] Control the rapeseed harvesting equipment to perform the rapeseed trial harvesting task in the rapeseed trial harvesting area according to the harvesting equipment parameters, and obtain the rapeseed trial harvesting data of all grid rapeseed fields in the rapeseed trial harvesting area;

[0151] The trial cutting loss rate of all grid rapeseed fields in the rapeseed trial cutting area is calculated based on the rapeseed trial cutting data;

[0152] The correlation analysis between the trial cutting loss rate and rapeseed maturity was carried out to obtain the correlation analysis results;

[0153] According to the correlation analysis results, the rapeseed loss rate of all grid rapeseed fields except the rapeseed trial cutting area was calculated;

[0154] The average rapeseed loss rate of the target rapeseed field is calculated by combining the total trial cutting loss rate and the rapeseed loss rate;

[0155] The total rapeseed loss of the target rapeseed field was calculated based on the average rapeseed loss rate and the estimated rapeseed yield obtained in advance.

[0156] In this embodiment, when the rapeseed harvesting equipment is controlled to reach different grid rapeseed fields in the rapeseed trial harvesting area, the rapeseed image information of the grid rapeseed fields is collected by using an image acquisition device (which can be a high-definition camera) arranged outside the rapeseed harvesting equipment. The rapeseed height information of the different grid rapeseed fields is calculated based on the rapeseed image information and the parameters of the image acquisition device. The equipment parameters of the rapeseed harvesting equipment are set in combination with the rapeseed height information and the rapeseed maturity, and the harvesting equipment parameters of the different grid rapeseed fields are obtained.

[0157] In addition, the rapeseed harvesting device is provided with a positioning system, which stores the coordinate ranges of different grid rapeseed fields in the rapeseed trial harvesting area, and the coordinates of each coordinate range. When the positioning system inside the rapeseed harvesting device shows that the rapeseed harvesting device has entered the coordinate range of other grid rapeseed fields, the rapeseed harvesting device will immediately determine the rapeseed maturity of the corresponding grid rapeseed field according to the current coordinates. For example, the grid rapeseed field coordinate range is (60-80, 170-200). When the positioning coordinates of the rapeseed harvesting device are (61, 171), it is determined that the rapeseed harvesting device has entered the grid rapeseed field coordinates, and then the rapeseed maturity corresponding to the grid rapeseed field is queried according to the coordinate range of the grid rapeseed field and the image acquisition device is used to collect the rapeseed image information of the grid rapeseed field. In addition, when the image acquisition device calculates the appropriate harvesting device parameters and adjusts the parameters, the calculated harvesting device parameters will be matched and stored with the coordinate range of the grid rapeseed field, and then the corresponding harvesting device parameters can be queried according to the coordinate range.

[0158] While storing the harvesting equipment parameters, the rapeseed trial harvesting data at the current time node is also stored and matched with the corresponding grid rapeseed field coordinate range. The rapeseed trial harvesting data includes the rapeseed harvesting weight and the rapeseed falling amount. The rapeseed harvesting weight is obtained by a weight sensor set inside the rapeseed harvester, which refers to the weight of the rapeseed harvested in the rapeseed harvester at the current time node. The rapeseed falling amount is obtained by an image acquisition device set at the tail of the rapeseed harvester. The rapeseed image left on the ground after harvesting is collected by the image acquisition device (which can be a high-definition camera), and the rapeseed density in the rapeseed image is calculated using image recognition technology. Specifically, the rapeseed image is enhanced, and the enhanced rapeseed image is grayed to obtain a gray rapeseed image. Then, the measurement tool in the image processing software is used. Commonly used image processing software includes GIMP (general image processing program), PIE (remote sensing image processing software), etc., to calculate the area occupied by the rapeseed in the identified gray rapeseed image, and then the rapeseed density in the rapeseed image is calculated according to the area occupied by the rapeseed. Then, the rapeseed drop amount in each grid rapeseed field is calculated according to the rapeseed density in the rapeseed image, the image acquisition equipment parameters and the actual area of ​​the grid rapeseed.

[0159] The trial cutting loss rate of all grid rapeseed fields in the rapeseed trial cutting area is calculated based on the rapeseed harvesting weight and the rapeseed dropping amount. Specifically, the rapeseed harvesting amount in each rapeseed trial cutting area is calculated based on the difference between the rapeseed harvesting weights stored at adjacent time nodes, and the rapeseed harvesting amount and the corresponding rapeseed dropping amount are added to obtain the total amount of rapeseed. The rapeseed dropping amount is divided by the total amount of rapeseed to obtain the trial cutting loss rate. Then, a correlation analysis is performed on the trial cutting loss rate of each grid rapeseed field and the corresponding rapeseed maturity to obtain the correlation analysis result. The correlation analysis can be performed by calculating the Pearson correlation coefficient, the Spearman rank correlation coefficient, the Kendall rank correlation coefficient, etc. Taking the Pearson correlation coefficient as an example, the Pearson correlation coefficient formula is used to calculate the Pearson correlation coefficient between the trial cutting loss rate of all grid rapeseed fields and the corresponding rapeseed maturity, and the linear relationship between the trial cutting loss rate and the corresponding rapeseed maturity is obtained, that is, the correlation analysis result. The Pearson correlation coefficient formula is as follows:

[0160]

[0161] Among them, X i It is a grid rapeseed field i of rapeseed harvest, is the average rapeseed harvest of all grid rapeseed fields, Y i It is a grid rapeseed field i The amount of rapeseed dropped, is the average rapeseed drop amount of all grid rapeseed fields, and n is the number of grid rapeseed fields.

[0162] Then, based on the results of the correlation analysis, the rapeseed loss rate of all grid rapeseed fields except the rapeseed trial cutting area is calculated. The average rapeseed loss rate of the target rapeseed field is calculated by combining the total trial cutting loss rate and the rapeseed loss rate. The total rapeseed loss of the target rapeseed field is calculated based on the average rapeseed loss rate and the pre-acquired estimated rapeseed yield. The estimated rapeseed yield is calculated based on the historical rapeseed yield of the target rapeseed field. The historical rapeseed yield refers to the yield of the same variety of rapeseed planted in the same month. The average yield of all historical rapeseed yields in the world is calculated, and the average yield is used as the estimated rapeseed yield. The average rapeseed loss rate is the average loss rate of each grid rapeseed field in the target rapeseed field. The average loss rate is used as the overall loss rate of the target rapeseed field. The average rapeseed loss rate is multiplied by the estimated rapeseed yield to obtain the total rapeseed loss of the target rapeseed field.

[0163] The rapeseed loss rate refers to the rapeseed falling during the rapeseed harvesting process due to reasons such as unripe or overripe rapeseed, insufficient performance of rapeseed harvesting equipment, etc., resulting in rapeseed harvesting losses. Therefore, the rapeseed loss rate is also the rapeseed drop rate. By estimating the rapeseed loss rate and adjusting the equipment parameters of the rapeseed harvesting equipment according to the rapeseed loss rate, the rapeseed loss rate can be reduced, the rapeseed yield can be increased, and the economic benefits for rapeseed growers can be increased.

[0164] In one embodiment, updating the equipment parameters of rapeseed harvesting equipment based on the rapeseed maturity of the grid rapeseed field to obtain the harvesting equipment parameters includes the following steps:

[0165] The rapeseed image information of the grid rapeseed field is obtained by using an image acquisition device pre-installed outside the rapeseed harvesting device;

[0166] Preprocessing rapeseed image information;

[0167] The pre-processed rapeseed image information is binarized using the threshold segmentation method to obtain the target rapeseed image;

[0168] Obtaining device parameters of the image acquisition device, and calculating rapeseed height information of the rapeseed trial cutting area according to the device parameters and the target rapeseed image;

[0169] The equipment parameters of the rapeseed harvesting equipment are updated by combining the rapeseed height information and the rapeseed maturity corresponding to the grid rapeseed field to obtain the harvesting equipment parameters.

[0170] In this embodiment, when the rapeseed harvesting equipment is controlled to arrive at different grid rapeseed fields in the rapeseed trial harvesting area, the rapeseed image information of the grid rapeseed fields is collected by using an image acquisition device (which can be a high-definition camera) set outside the rapeseed harvesting equipment, and the rapeseed height information of different grid rapeseed fields is calculated according to the rapeseed image information and the parameters of the image acquisition device. The equipment parameters of the rapeseed harvesting equipment are set in combination with the rapeseed height information and the rapeseed maturity, and the harvesting equipment parameters of different grid rapeseed fields are obtained. Specifically, the height of the cutting table of the rapeseed harvesting equipment is adjusted according to the rapeseed height information. If the rapeseed is high, the height of the cutting table needs to be increased accordingly. The cutting knife speed and feeding speed of the rapeseed harvesting equipment are adjusted according to the rapeseed maturity. If the rapeseed maturity is too high, the cutting knife speed and feeding speed should be appropriately reduced to avoid the rapeseed from falling off due to impact, resulting in an increased rapeseed loss rate.

[0171] Analyzing the height information of rapeseed according to the image information of rapeseed includes: removing the noise in the image information of rapeseed by using the mean filter, median filter, Gaussian filter and other technologies to improve the image quality, converting the image information of rapeseed into a grayscale image by using the threshold segmentation method, and segmenting the rapeseed plant in the image information of rapeseed from the background to obtain the target rapeseed image without the background. Then the target rapeseed image is subjected to the connected region analysis, and the connected region refers to the set of white pixels connected to each other in the image. Through the analysis, each connected region in the image can be determined, and the largest connected region can be found, which usually corresponds to the main part of the rapeseed plant. For the largest connected region found, its minimum circumscribed moment is calculated. The minimum circumscribed moment refers to a rectangle that can completely contain the connected region and has the smallest area. The height of this rectangle (i.e., the vertical side length of the rectangle) can approximately represent the image height of the rapeseed plant, and then the device parameters of the image acquisition device are obtained, and the device parameters include focal length, principal point coordinates, distortion coefficient (such as radial distortion and tangential distortion), etc. The image height of the rapeseed plant is converted into the actual height according to the device parameters to obtain the height information of the rapeseed.

[0172] In one embodiment, determining the rapeseed maturity node of the target rapeseed field in combination with the rapeseed remote sensing characteristics and the rapeseed total loss, and judging whether the predicted harvesting time node is the optimal harvesting node according to the rapeseed maturity node includes the following steps:

[0173] If the total loss of rapeseed is less than or equal to the preset loss threshold, the predicted harvesting time node is determined to be the optimal harvesting node;

[0174] If the total loss of rapeseed is greater than the loss threshold, the rapeseed maturity node of the target rapeseed field is predicted by combining the rapeseed image information and the rapeseed remote sensing characteristics;

[0175] If the rapeseed maturity node is later than the predicted harvesting time node, it is determined that the predicted harvesting time node is not the optimal harvesting node;

[0176] If the rapeseed maturity node is not later than the predicted harvesting time node, the predicted harvesting time node is determined to be the optimal harvesting node.

[0177] In this embodiment, if the predicted total rapeseed loss of the target rapeseed field is less than or equal to the preset loss threshold, the predicted harvesting time node can be directly determined as the optimal harvesting node because the total rapeseed loss is small enough, and the subsequent formal harvesting work can be started. If the total rapeseed loss is greater than the loss threshold, the rapeseed maturity node of the target rapeseed field is re-predicted in combination with the rapeseed image information and the rapeseed remote sensing features. Specifically, the rapeseed rhizome area in the rapeseed image information is first extracted to obtain the rapeseed rhizome image. As rapeseed matures, its rhizomes will gradually become thicker and harder, and the color may also change from light green to dark green or yellow, etc. Therefore, the maturity of rapeseed can be predicted based on the color characteristics of the rhizomes. In addition, since the height information of rapeseed can also reflect the maturity of rapeseed to a certain extent, for example, generally speaking, rapeseed can be harvested when the height information is between 30-50 cm. Based on the neural network model and the introduction of the self-attention mechanism, a rapeseed maturity prediction model is constructed. The rapeseed maturity prediction model is iteratively trained using a training set with different rhizomes, rapeseed heights and remote sensing features and pre-annotated. After reaching the maximum number of training times, the rhizomes color features, rapeseed height information and rapeseed remote sensing features are input into the trained rapeseed maturity prediction model, and the predicted rapeseed maturity of the target rapeseed field is output through the rapeseed maturity prediction model. The predicted rapeseed maturity refers to how long it will take for the rapeseed to mature. The rapeseed maturity is added to the current time node (predicted harvesting time node) to obtain the rapeseed maturity node of the target rapeseed field.

[0178] If the rapeseed maturity node is later than the predicted harvesting time node, it means that the optimal harvesting node has not been reached. Therefore, the predicted harvesting time node is determined not to be the optimal harvesting node. After reaching the optimal harvesting node, the rapeseed maturity prediction model is used to predict the overall maturity of the target rapeseed field to obtain the target maturity. Then, all the grid rapeseed fields in the rapeseed trial harvesting area with a rapeseed loss rate less than the loss rate threshold are screened out as the target grid rapeseed fields. Then, the harvesting equipment parameters corresponding to all the target grid rapeseed fields are obtained, and the average value of all the harvesting equipment parameters is calculated to obtain the average equipment parameters. Based on the average equipment parameters, the average equipment parameters are adjusted according to the target maturity to obtain the optimal average equipment parameters. The rapeseed harvesting equipment is controlled according to the optimal average equipment parameters to carry out the formal rapeseed harvesting work. If the rapeseed maturity node is not later than the predicted harvesting time node, it means that the current time node may be the optimal harvesting node, or it may have exceeded the optimal harvesting node. Therefore, the predicted harvesting time node is directly determined to be the optimal harvesting node.

[0179] In one embodiment, combining rapeseed image information and rapeseed remote sensing features to predict the rapeseed maturity node of a target rapeseed field includes the following steps:

[0180] A rapeseed maturity prediction model was constructed based on a neural network model and a self-attention mechanism, and the rapeseed maturity prediction model was trained using a pre-constructed training set.

[0181] The target detection algorithm is used to identify the root and stem region of the rapeseed image information, and the rapeseed image information is segmented according to the root and stem region identification result to obtain the rapeseed root and stem image in the rapeseed image information;

[0182] Clustering algorithm is used to extract the rhizome color features of rapeseed rhizome images;

[0183] The root and stem color features, rapeseed height information and rapeseed remote sensing features are input into the trained rapeseed maturity prediction model, and the rapeseed maturity prediction model is used to output the predicted rapeseed maturity of the target rapeseed field;

[0184] The predicted harvesting time node is corrected according to the predicted rapeseed maturity period to obtain the rapeseed maturity node of the target rapeseed field.

[0185] In this embodiment, a rapeseed maturity prediction model is constructed based on a neural network model and a self-attention mechanism. Taking a multi-layer perceptron (MLP) model as an example, an input layer is designed to receive root and stem color features, rapeseed height information, and rapeseed remote sensing features. A self-attention layer is introduced in the model so that the model can pay attention to the relationship and importance between different features. According to the relationship and importance between different features, weights are assigned to the root and stem color features, rapeseed height information, and rapeseed remote sensing features. The self-attention layer is constructed based on the self-attention mechanism. A query, a key, and a value are generated by linearly transforming the input features. The dot product between the query vector and the key vector is calculated to obtain an attention score matrix. The dot product result is divided by a scaling factor (which can be the square root of the key vector) to stabilize the gradient. The scaled attention score matrix is ​​converted into a probability value between [0,1] using a normalization function to obtain an attention weight. The normalization function can be a softmax function. The obtained attention weight is multiplied by the value vector to complete the weight allocation of the root and stem color features, the rapeseed height information, and the rapeseed remote sensing features. The hidden layer of the rapeseed maturity prediction model uses an activation function to map the input data to the output layer. Softmax (normalized exponential function) can be selected as the activation function of the hidden layer. The hidden layer can be set to two layers. The sample sets with different root and stem colors, rapeseed heights and remote sensing features and pre-annotated are divided into training sets and validation sets. The rapeseed maturity prediction model is iteratively trained until the maximum number of training times is reached. Cross-validation can be used to evaluate the performance of the rapeseed maturity prediction model. Cross-validation includes K-fold cross-validation and leave-one-out cross-validation. Taking K-fold cross-validation as an example, the sample set is divided into K subsets, and K-1 subsets are selected as training sets each time, and the remaining subset is used as the validation set. Repeat K times, and each time a different subset is selected as the validation set. Leave-one-out cross-validation is to leave only one sample as the validation set each time, and the remaining samples are used as training sets. Repeat this process until each sample has been used as a validation set. In each cross-validation iteration, the model is trained using the training set, and the model performance is evaluated using the validation set. Then, the model parameters are adjusted or the optimal model is selected according to the performance indicators on the validation set (such as accuracy, recall rate, F1 score, etc.). After all cross-validation iterations are completed, the average performance indicator of the model on the validation set is calculated. If the average performance indicator is greater than or equal to the preset indicator threshold, it means that the rapeseed maturity prediction model training is completed.

[0186] The target detection algorithm is used to identify the root and stem area of ​​rapeseed image information. Common target detection algorithms include R-CNN series and YOLO series. Taking YOLO (target detection algorithm based on deep learning) as an example, the rapeseed image information is input into the neural network, and the key frame is divided into multiple grids through operations such as convolution pooling. Each grid predicts whether there is a root and stem area in the grid. According to the root and stem area recognition result, the root and stem area is segmented from the rapeseed image information to obtain the rapeseed root and stem image in the rapeseed image information. First, the color space conversion of the rapeseed root and stem image is performed, and the RGB (red, green, and blue) color space is converted to the HSV (hue, saturation, and brightness) color space, and the RGB value is normalized to the range. Then, the hue, saturation, and brightness are calculated based on R, G, and B to complete the color space conversion. After completing the color space conversion, the main color in the rapeseed rhizome image is identified by a clustering algorithm. A commonly used clustering algorithm can be the k-means (mean clustering) algorithm. The pixels in the rapeseed rhizome image are clustered, and pixels of similar colors are classified into the same class. The center of each class, i.e., the main color, is calculated, the central color of each class is extracted, and the frequency of occurrence of each color is calculated. The color with the highest frequency of occurrence is taken as the main color, i.e., the rhizome color feature. The rhizome color feature, rapeseed height information, and rapeseed remote sensing features are input into the trained rapeseed maturity prediction model, and the predicted rapeseed maturity of the target rapeseed field is output through the rapeseed maturity prediction model. The predicted rapeseed maturity refers to how long it will take for the rapeseed to mature. For example, if the rapeseed maturity is 10 days, the rapeseed maturity is added to the current time node (predicted harvest time node) to obtain the rapeseed maturity node of the target rapeseed field.

[0187] In one embodiment, combining rapeseed maturity and trial harvesting loss rate and using parameter optimization algorithm to generate an optimal parameter range for rapeseed harvesting equipment includes the following steps:

[0188] All grid rapeseed fields in the rapeseed trial harvesting area where the rapeseed loss rate is less than the loss rate threshold are taken as standard rapeseed fields;

[0189] Obtain harvesting equipment parameters corresponding to rapeseed harvesting equipment in all standard rapeseed fields;

[0190] Calculating parameter intervals of rapeseed harvesting equipment according to all harvesting equipment parameters to obtain multiple initial parameter intervals of different types;

[0191] Generate multiple random parameters according to all initial parameter intervals, and encode all random parameters to obtain multiple initial parameter populations;

[0192] Based on the genetic algorithm, all initial parameter populations are optimized respectively until the maximum number of parameter optimization times is reached, and multiple target device parameters are output;

[0193] Communicate information about all target device parameters, and calculate target parameter ranges based on all target device parameters that have completed information communication;

[0194] The update iteration steps are repeated for the target parameter interval until a preset maximum number of iterations is reached, thereby obtaining the optimal parameter interval for the rapeseed harvesting equipment.

[0195] In this embodiment, all grid rapeseed fields in the rapeseed trial harvesting area whose rapeseed loss rate is less than the loss rate threshold are first screened out to obtain a plurality of standard rapeseed fields, and harvesting equipment parameters of the rapeseed harvesting equipment within the corresponding standard rapeseed field range are obtained according to the coordinate range of each standard rapeseed field. The harvesting equipment parameters include header height, cutter speed, and feeding speed, etc., and the maximum and minimum values ​​of each type of parameters are screened out from all the harvesting equipment parameters, namely, the maximum header height, the minimum header height, the fastest cutter speed, the slowest cutter speed, the fastest feeding speed, and the slowest feeding speed, etc., and a plurality of different types of initial parameter intervals are constructed according to the maximum and minimum values ​​of the above parameters, for example, the initial parameter interval of the header height is 8-14 cm, the initial parameter interval of the cutter speed is 1.8-2.7 m / s, and the initial parameter interval of the feeding speed is 0.6-1.3 m / s. First, multiple random parameters are generated according to all initial parameter intervals. The generated multiple random parameters are all located in the corresponding initial parameter intervals. Multiple random parameters are generated multiple times to ensure that the initial population has sufficient diversity to avoid the algorithm from converging to the local optimal solution too early. Real number coding can be used. Each gene is a real number, representing a specific parameter value. Therefore, each random parameter is used as a gene of an individual. Each individual is composed of multiple genes. Several individuals form a population, that is, multiple initial parameter populations are obtained. According to the pre-constructed fitness function, the individual fitness of all population individuals in the initial parameter population is calculated respectively. According to the individual fitness and the selection operator, the next generation population is selected. The tournament selection operator can be selected. The core idea of ​​the tournament selection is to randomly select several chromosomes (individuals) at a time, and select the individual with the largest fitness in this group of chromosomes into the next generation. Repeat this process until the preset population size is reached, and then use the crossover operator to perform a crossover operation on the next generation population to obtain the offspring population. The crossover operator can select single-point crossover, multi-point crossover and uniform crossover, etc. Taking uniform crossover as an example, each gene position is exchanged with a certain probability to generate a new offspring individual. After completing the crossover operation, the mutation operator is used to mutate the offspring population to obtain the updated parameter population. Differential mutation or equal mutation can be used. Taking differential mutation as an example, an individual is randomly selected as the base individual, and then two different individuals are randomly selected. The difference vector between them is calculated, the difference vector is multiplied by a scaling factor, and then added to the base individual to generate a new individual.Continue to perform the above optimization iterations on the updated parameter population until the maximum number of optimization iterations is reached, and output multiple different types of optimal equipment parameters, that is, multiple target equipment parameters, and exchange information for all target equipment parameters to achieve collaborative evolution. Mutation operations can be performed, that is, new genetic information is introduced to exchange information for target equipment parameters. Similarly, the maximum and minimum values ​​of different types of parameters in the target equipment parameters are selected to construct a target parameter interval. Repeat the above iteration steps for the target parameter interval to further reduce the target parameter interval until the preset maximum number of iterations is reached, and the optimal parameter interval for the rapeseed harvesting equipment is obtained. In the subsequent formal harvesting process, the rapeseed harvesting equipment is controlled to make small parameter adjustments within the optimal parameter interval according to the rapeseed maturity and rapeseed height in the target rapeseed field to ensure that the harvesting equipment parameters of the rapeseed harvesting equipment can continue to remain within the optimal parameter interval, thereby reducing the rapeseed loss rate.

[0196] The present application also discloses an intelligent control device for mechanized rapeseed harvesting based on online monitoring, which is characterized by comprising:

[0197] a memory configured to store instructions; and

[0198] A processor is configured to call instructions from a memory and implement the method for intelligent control of mechanized rapeseed harvesting based on online monitoring according to any of the above items when executing the instructions.

[0199] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0200] Among them, the memory can be an internal storage unit of a computer device, such as a hard disk or memory of a computer device, or an external storage device of a computer device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the computer device, etc., and the memory can also be a combination of an internal storage unit and an external storage device of a computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0201] An embodiment of the present application also provides a machine-readable storage medium having instructions stored thereon, the instructions being used to enable a machine to execute the above-mentioned method for intelligent control of mechanized rapeseed harvesting based on online monitoring.

[0202] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0203] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0204] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0205] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0206] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0207] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0208] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0209] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0210] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. An intelligent control method for mechanized rapeseed harvesting based on online monitoring, characterized in that: The method comprises the following steps: Acquire rapeseed growth-related information of a target rapeseed field, wherein the rapeseed growth-related information includes soil information of the target rapeseed field, meteorological information of a location of the target rapeseed field, and sowing time information and rapeseed variety information of the rapeseed planted in the target rapeseed field; Inputting the rapeseed variety information, the meteorological information and the soil information into a pre-built rapeseed growth model to predict the rapeseed growth cycle, and determining the predicted maturity time period of the target rapeseed field according to the rapeseed growth cycle prediction result; Determining a predicted harvesting time node of the target rapeseed field according to the sowing time information and the predicted maturity time period; At the predicted harvest time node, collecting rapeseed remote sensing information of the target rapeseed field by a remote sensing drone; Acquire rapeseed field size information of the target rapeseed field based on the rapeseed remote sensing information, and grid-divide the target rapeseed field according to the rapeseed field size information to obtain a plurality of grid rapeseed fields; Performing feature analysis on all the grid rapeseed fields according to the rapeseed remote sensing information to obtain rapeseed remote sensing features, and calculating rapeseed maturity of all the grid rapeseed fields based on the rapeseed remote sensing features; A reference grid rapeseed field is selected from all the grid rapeseed fields according to the rapeseed maturity, and adjacent grid rapeseed fields of the reference grid rapeseed field are subjected to iterative diffusion screening with the reference grid rapeseed field as the center and the maximization of the rapeseed maturity difference as the screening target, until the number of the selected rapeseed fields is greater than or equal to a preset number threshold, thereby obtaining a rapeseed trial harvesting area; Controlling the rapeseed harvesting equipment to perform a rapeseed trial harvesting task in the rapeseed trial harvesting area, and during the execution of the rapeseed trial harvesting task, using the rapeseed harvesting equipment to count the trial harvesting loss rate of the rapeseed trial harvesting area, and analyzing the total rapeseed loss of the target rapeseed field based on the trial harvesting loss rate and using a correlation analysis method; Determine the rapeseed maturity node of the target rapeseed field in combination with the rapeseed remote sensing characteristics and the rapeseed total loss, and judge whether the predicted harvesting time node is the optimal harvesting node according to the rapeseed maturity node; If the predicted harvesting time node is the optimal harvesting node, the rapeseed maturity and the trial harvesting loss rate are combined and a parameter optimization algorithm is used to generate an optimal parameter range for the rapeseed harvesting equipment, and the rapeseed harvesting equipment is controlled to harvest the rapeseed in the target rapeseed field according to the optimal parameters.

2. The method according to claim 1, characterized in that The step of performing feature analysis on all the grid rapeseed fields according to the rapeseed remote sensing information to obtain rapeseed remote sensing features, and calculating the rapeseed maturity of all the grid rapeseed fields based on the rapeseed remote sensing features comprises the following steps: Preprocessing the rapeseed remote sensing information of all the grid rapeseed fields; For any of the grid rapeseed fields, extracting the reflectance spectrum in the rapeseed remote sensing information that has completed preprocessing; Calculate a spectral reflectance value according to the reflectance spectrum, perform feature extraction on the spectral reflectance value, and obtain a spectral feature of rapeseed; Performing principal component transformation on the rapeseed remote sensing information to obtain spatial color features of the rapeseed remote sensing information; Using a clustering algorithm to perform color classification on the spatial color features to obtain rapeseed color features of the rapeseed remote sensing information; Integrating the rapeseed spectral characteristics and the rapeseed color characteristics into the rapeseed remote sensing characteristics of the grid rapeseed field; A rapeseed maturity prediction model is constructed based on a neural network model, and all the rapeseed remote sensing features are input into the trained rapeseed maturity prediction model, and the rapeseed maturity prediction model is used to output the rapeseed maturity of all the grid rapeseed fields.

3. The method according to claim 2, characterized in that The method of constructing a rapeseed maturity prediction model based on a neural network model, inputting all the rapeseed remote sensing features into the trained rapeseed maturity prediction model, and outputting the rapeseed maturity of all the grid rapeseed fields through the rapeseed maturity prediction model comprises the following steps: Construct rapeseed maturity prediction model based on BP neural network model; Using a pre-constructed rapeseed training set to train the rapeseed maturity prediction model; For any of the grid rapeseed fields, the rapeseed spectral characteristics and the rapeseed color characteristics of the grid rapeseed field are sequentially input into the rapeseed maturity prediction model to perform maturity prediction, and obtain a first maturity prediction value and a second maturity prediction value; Using Kalman filtering to filter and correct the first maturity prediction value and the second maturity prediction value; The first maturity prediction value and the second maturity prediction value after filtering and correction are fused to obtain the rapeseed maturity of all the grid rapeseed fields.

4. The method according to claim 1, characterized in that The step of selecting a plurality of target grid rapeseed fields from all the grid rapeseed fields as rapeseed trial harvesting areas according to the rapeseed maturity comprises the following steps: According to the rapeseed maturity, all the grid rapeseed fields are divided into maturity grades using a threshold method to obtain the rapeseed maturity grades of all the grid rapeseed fields; If the rapeseed maturity level is the maximum rapeseed maturity level or the rapeseed maturity level is the minimum rapeseed maturity level, all the grid rapeseed fields corresponding to the maximum rapeseed maturity level and the minimum rapeseed maturity level are marked as trial rapeseed fields respectively; For any of the trial-cut rapeseed fields, the trial-cut rapeseed field marking step is repeatedly performed based on the rapeseed maturity level and with the trial-cut rapeseed field as the center until the number of the trial-cut rapeseed fields is greater than or equal to a preset number threshold, and the trial-cut rapeseed field marking step is: Calculating the maturity level difference between the adjacent rapeseed field of the trial-cut rapeseed field and the trial-cut rapeseed field, and recording the adjacent rapeseed field whose maturity level difference is greater than or equal to a preset difference threshold as the trial-cut rapeseed field; Integrate all adjacent trial-cut rapeseed fields to obtain a plurality of initial trial-cut areas; Constructing a target maturity level matrix of the target rapeseed field according to the rapeseed maturity levels of all the grid rapeseed fields; For any of the initial trial cutting areas, constructing a regional maturity grade matrix of the initial trial cutting area based on the rapeseed maturity grade; Based on a similarity calculation algorithm, the similarities between the target maturity level matrix and all the regional maturity level matrices are respectively calculated to obtain a plurality of matrix similarities; The initial trial cutting area corresponding to the largest data similarity is used as the rapeseed trial cutting area.

5. The method according to claim 1, characterized in that The controlling of the rapeseed harvesting device to perform the rapeseed trial harvesting task in the rapeseed trial harvesting area, counting the trial harvesting loss rate of the rapeseed trial harvesting area by the rapeseed harvesting device during the execution of the rapeseed trial harvesting task, and analyzing the total rapeseed loss of the target rapeseed field based on the trial harvesting loss rate and using a correlation analysis method comprises the following steps: When the rapeseed harvesting equipment arrives at any of the grid rapeseed fields in the rapeseed trial harvesting area, the equipment parameters of the rapeseed harvesting equipment are updated based on the maturity of the rapeseed in the grid rapeseed field to obtain harvesting equipment parameters; Controlling the rapeseed harvesting device to perform the rapeseed trial harvesting task in the rapeseed trial harvesting area according to the harvesting device parameters, and obtaining the rapeseed trial harvesting data of all the grid rapeseed fields in the rapeseed trial harvesting area; Calculating the trial cutting loss rate of all the grid rapeseed fields in the rapeseed trial cutting area according to the rapeseed trial cutting data; Performing a correlation analysis on the trial cutting loss rate and the rapeseed maturity to obtain a correlation analysis result; Calculating the rapeseed loss rate of all the grid rapeseed fields except the rapeseed trial harvesting area according to the correlation analysis result; Calculate the average rapeseed loss rate of the target rapeseed field by combining all the trial cutting loss rates and the rapeseed loss rate; The total rapeseed loss of the target rapeseed field is calculated according to the average rapeseed loss rate and the pre-acquired estimated rapeseed yield.

6. The method according to claim 5, characterized in that The updating of the equipment parameters of the rapeseed harvesting equipment based on the maturity of the rapeseed in the grid rapeseed field to obtain the harvesting equipment parameters comprises the following steps: Acquiring rapeseed image information of the grid rapeseed field using an image acquisition device pre-arranged outside the rapeseed harvesting device; Preprocessing the rapeseed image information; The rapeseed image information that has completed the preprocessing is binarized using a threshold segmentation method to obtain a target rapeseed image; Acquire device parameters of the image acquisition device, and calculate rapeseed height information of the rapeseed trial cutting area according to the device parameters and the target rapeseed image; The equipment parameters of the rapeseed harvesting equipment are updated in combination with the rapeseed height information and the rapeseed maturity corresponding to the grid rapeseed field to obtain harvesting equipment parameters.

7. The method according to claim 6, characterized in that The step of determining the rapeseed maturity node of the target rapeseed field by combining the rapeseed remote sensing characteristics and the rapeseed total loss, and judging whether the predicted harvesting time node is the optimal harvesting node according to the rapeseed maturity node comprises the following steps: If the total rapeseed loss is less than or equal to a preset loss threshold, determining the predicted harvesting time node as the optimal harvesting node; If the total loss of rapeseed is greater than the loss threshold, predicting the rapeseed maturity node of the target rapeseed field by combining the rapeseed image information and the rapeseed remote sensing characteristics; If the rapeseed maturity node is later than the predicted harvesting time node, determining that the predicted harvesting time node is not the optimal harvesting node; If the rapeseed maturity node is not later than the predicted harvesting time node, the predicted harvesting time node is determined to be the optimal harvesting node.

8. The method according to claim 7, characterized in that The method of predicting the rapeseed maturity node of the target rapeseed field by combining the rapeseed image information and the rapeseed remote sensing characteristics comprises the following steps: Constructing a rapeseed maturity prediction model based on a neural network model and a self-attention mechanism, and using a pre-constructed training set to perform model training on the rapeseed maturity prediction model; Using a target detection algorithm to identify the root and stem region of the rapeseed image information, and performing image segmentation on the rapeseed image information according to the root and stem region identification result to obtain a rapeseed root and stem image in the rapeseed image information; Extracting rhizome color features of the rape rhizome image using a clustering algorithm; Inputting the root and stem color features, the rapeseed height information and the rapeseed remote sensing features into the trained rapeseed maturity prediction model, and outputting the predicted rapeseed maturity of the target rapeseed field through the rapeseed maturity prediction model; The predicted harvesting time node is corrected according to the predicted rapeseed maturity period to obtain the rapeseed maturity node of the target rapeseed field.

9. The method according to claim 6, characterized in that The step of combining the rapeseed maturity and the trial harvesting loss rate and using a parameter optimization algorithm to generate the optimal parameter range of the rapeseed harvesting equipment comprises the following steps: All the grid rapeseed fields in the rapeseed trial harvesting area where the rapeseed loss rate is less than the loss rate threshold are used as standard rapeseed fields; Acquire the harvesting equipment parameters corresponding to the rapeseed harvesting equipment in all the standard rapeseed fields; Calculating the parameter interval of the rapeseed harvesting equipment according to all the harvesting equipment parameters to obtain a plurality of different types of initial parameter intervals; Generate multiple random parameters according to all the initial parameter intervals, and encode all the random parameters to obtain multiple initial parameter populations; Based on the genetic algorithm, all the initial parameter populations are optimized respectively until a maximum number of parameter optimization times is reached, and a plurality of target device parameters are output; Performing information exchange on all the target device parameters, and calculating a target parameter range based on all the target device parameters that have completed the information exchange; The update iteration step is repeated for the target parameter interval until a preset maximum number of iterations is reached, thereby obtaining the optimal parameter interval of the rapeseed harvesting equipment.

10. An intelligent control device for mechanized rapeseed harvesting based on online monitoring, characterized in that: include: a memory configured to store instructions; as well as A processor is configured to call the instructions from the memory and to implement the method for intelligent control of mechanized rapeseed harvesting based on online monitoring according to any one of claims 1 to 9 when executing the instructions.

Images