Highland barley low-loss header real-time cutting depth adjusting method based on intelligent sensor

Abstract

The invention relates to the technical field of intelligent agricultural machinery, in particular to a highland barley low-loss header real-time cutting depth adjusting method based on an intelligent sensor. Comprising the following steps: designing a special combine harvester for highland barley with an intelligent real-time cutting depth adjusting function; collecting and preprocessing data; fusing and analyzing data; generating a cut depth adjustment instruction, wherein the cut depth adjustment instruction comprises a depth adjustment decision based on fusion data analysis and redundancy and verification of the instruction; and realizing automatic adjustment of the header, wherein the adjustment comprises driving of an actuating mechanism and real-time monitoring and feedback of the position of the header. Through an intelligent sensor system and a real-time feedback adjusting mechanism, the problems that a traditional harvester is insufficient in header adjusting precision, tedious in operation, poor in adaptability and the like in the highland barley harvesting process are solved, the highland barley harvesting quality is improved, grain loss is reduced, efficient operation is kept in the complex field environment, and the working efficiency is improved. And the overall effect of mechanical harvesting of highland barley is greatly improved.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent agricultural machinery, and in particular to a real-time cutting depth adjustment method of a highland barley low-loss cutting platform based on an intelligent sensor. Background Art

[0002] As the main food crop in the Tibet Autonomous Region, highland barley is of great significance to the local agricultural economy and residents' lives. However, the current mechanized harvesting of highland barley faces many challenges. On the one hand, due to the differences in the growth habits, plant morphology and grain characteristics of highland barley from common crops, it is difficult for existing general-purpose combine harvesters to achieve ideal harvesting effects when harvesting highland barley. For example, highland barley is prone to lodging during maturity, and its spikes are relatively fragile, which makes it easy for the spikes to fall off during the harvesting process, resulting in grain loss. On the other hand, the existing harvesting equipment lacks optimization for the characteristics of highland barley in terms of header design and operating parameter control, which makes it impossible for the header to accurately control the cutting depth when harvesting highland barley, resulting in a high header loss rate, which seriously affects the harvesting quality and yield of highland barley.

[0003] Most of the headers of combine harvesters adopt the traditional mechanical design, and the adjustment of the cutting depth mainly depends on the operator's experience and manual adjustment. This adjustment method has the following problems: First, the accuracy of manually adjusting the header depth is limited, and it is difficult to make real-time and accurate adjustments according to the actual growth status of barley and the field environment. When faced with complex field conditions, such as barley plants of different heights, uneven fields, and barley with different degrees of lodging, it is difficult for operators to quickly make accurate judgments and adjustments, resulting in the header either cutting too deep, causing unnecessary soil disturbance and energy waste; or cutting too shallow, unable to completely cut the barley plants, resulting in grain loss. Secondly, the traditional header lacks the perception and feedback mechanism of real-time data during the harvesting process, and cannot obtain key information such as the growth status of barley, lodging conditions, and the relative position of the header and barley in a timely manner, so it is impossible to realize intelligent cutting depth adjustment. This not only affects the efficiency and quality of the harvesting operation, but also limits the application effect of the combine harvester in the field of barley harvesting. Summary of the invention

[0004] Based on the above purpose, the present invention provides a real-time cutting depth adjustment method of a highland barley low-loss harvesting platform based on an intelligent sensor, comprising the following steps: Design a special highland barley combine harvester with intelligent real-time cutting depth adjustment function, including sensor deployment, data communication network construction, data processing center hardware and software construction, and improvement of the cutting platform adjustment actuator; Data collection and preprocessing, including barley plant height data collection, ground flatness data collection, barley lodging degree data collection, barley ear position data collection, crop density data collection, and data preprocessing; Data fusion and analysis, including data fusion preprocessing, fusion model establishment, and real-time data fusion calculation; Cutting depth adjustment instruction generation, including depth adjustment decision based on fusion data analysis and instruction redundancy and verification; Automatic adjustment of the header is achieved, including adjustment of the actuator drive and real-time monitoring feedback of the header position.

[0005] Preferably, designing a highland barley special combine harvester with intelligent real-time cutting depth adjustment function comprises the following steps: Sensor deployment: laser rangefinders are installed on both sides of the cutting platform to measure the height of barley plants and the flatness of the ground; cameras are installed on both sides of the cutting platform to capture the lodging of barley and the position of the ear head; pressure sensors are installed under the cutting platform to sense the change in contact pressure between the cutting platform and the ground, so as to estimate the density of barley plants and the hardness of the stems; Build a data communication network, select appropriate communication protocols and hardware devices according to sensor types and data transmission requirements, complete the construction of the data communication network, and realize real-time transmission and reception of sensor data; The data processing center is equipped with software and hardware, and high-performance computers are configured for data processing and calculation; data storage devices with sufficient capacity are equipped to store the massive data collected by sensors; and special data processing software is developed with data receiving, processing, analysis and instruction generation functions; The header adjustment actuator is improved, and the appropriate actuator is selected according to the header structure and adjustment requirements.

[0006] Preferably, data collection and preprocessing includes the following steps: To collect data on the height of highland barley plants, the laser ranging sensor is activated to work in high-speed measurement mode, generating a series of real-time data streams reflecting the changes in the height of highland barley plants; Ground flatness data collection: the laser rangefinder sensor emits a laser downward from the bottom of the header chassis to measure the distance from the ground to the sensor, and obtains a data stream describing the undulation of the ground, recording the ground height fluctuation value at a frequency of 50Hz; Data collection on the degree of lodging of highland barley: cameras on both sides of the harvesting platform regularly take images of highland barley plants, obtain image data containing information on lodging of highland barley, and perform preliminary recognition to analyze the lodging angle θ and lodging ratio of highland barley; The data of the position of the barley head is collected by processing and analyzing the barley image taken by the camera to locate the position of the barley head in space and express it in three-dimensional coordinates; Calculate the number and spatial density distribution of highland barley plants in the highland barley image captured by the camera to obtain the crop density; Data preprocessing uses a Kalman filter to remove noise, and performs temporal or spatial calibration on barley plant height data, ground flatness data, barley lodging degree data, barley ear position data, and crop density data to ensure that data collected by different sensors can be synchronously matched.

[0007] Preferably, data fusion and analysis includes the following steps: Data fusion preprocessing normalizes the barley plant height data, ground flatness data, barley lodging degree data, barley ear position data, and crop density data after noise removal to eliminate data differences caused by different sensor measurement ranges or units; According to the influence of various factors on the cutting depth of the header during the barley harvesting process, the weights of different data are determined; The fusion model based on Gaussian process regression was established, with the normalized barley plant height data, ground flatness data, barley lodging degree data, barley ear position data, and crop density data as input variables, and the ideal cutting depth of the cutting platform as the output variable to construct a Gaussian process regression model. Real-time data fusion calculation, substitute the real-time collected data into the Gaussian process regression model, perform prediction calculation, and obtain the cutting depth adjustment amount of the cutting table.

[0008] Preferably, the weights of different data are determined according to the influence of various factors on the cutting depth of the header during the highland barley harvesting process, and the weights are as follows: Highland barley plant height: 0.4, ground flatness: 0.3, highland barley lodging degree: 0.2, ear position: 0.06, grain maturity: 0.02, crop density: 0.02; Collect historical data on highland barley plant height, ground flatness, highland barley lodging, highland barley ear position, crop density and corresponding cutting depth; The radial basis function is selected as the kernel function of Gaussian process regression, and its specific form is as follows: ; in, represents the signal variance, is the length scale parameter, and denote the i-th and j-th input samples respectively, is the number of input features, m is the index parameter, is the kernel function; The normalized and weighted sensor data is used as input features to construct a Gaussian process regression model. The specific input form is as follows: ; in, is the weight vector of highland barley plant height data, ground flatness data, highland barley lodging degree data, highland barley ear position data, and crop density data. It is the normalized data of highland barley plant height, ground flatness, highland barley lodging degree, highland barley ear position and crop density vector; For the training dataset , calculate the covariance matrix based on the kernel function : ; in, and is the weighted input vector ; Using historical data, we can maximize the log-likelihood function to estimate the Gaussian process regression kernel function. and , and the noise variance .

[0009] Preferably, according to the training and , and the noise variance Calculate the mean of the output from the training data and variance , according to the mean Determine the cutting depth adjustment of the header, and refer to the variance To assess the credibility of the forecast; The real-time collected data on highland barley plant height, ground flatness, highland barley lodging, highland barley ear position, and crop density are normalized and weighted to obtain , and input it into the trained Gaussian process regression model to calculate the predicted value of the cutting depth adjustment amount and its confidence interval.

[0010] Preferably, in the depth adjustment decision based on fusion data analysis, the predicted value of the cutting depth adjustment amount of the header and its confidence interval are input and calculated to generate the cutting depth adjustment instruction of the header. The specific calculation process is as follows: Adjust the value according to the predicted cutting depth of the header , combined with the actual cutting depth of the current header , calculate the target cutting depth , the formula is: ; Consider the adjustment speed and accuracy of the header adjustment actuator and set an appropriate adjustment speed.

[0011] Preferably, in the redundancy and verification of the instructions, the target cutting depth The confidence interval of the adjustment speed V and the cutting depth adjustment amount of the header is used as input to perform redundancy verification on the adjustment command to ensure the accuracy and reliability of the command. The specific verification process is as follows: A dual-channel instruction generation mechanism is adopted, that is, the adjustment instructions are generated through two independent channels respectively, and the instructions generated by different channels are compared to see if they are consistent. If the instructions generated by the two channels are consistent, the instructions are considered to be reliable. If the instructions are inconsistent, further verification is required; If the confidence interval is large, it means that the predicted cutting depth adjustment value of the header is The uncertainty of the value is high, which triggers a further verification process and calls the backup sensor data for secondary calculation to verify the predicted The value is accurate, and the header adjustment speed is temporarily slowed down to wait for subsequent data to be updated and then recalculated.

[0012] Preferably, in the implementation of automatic adjustment of the header, the following steps are included: The target cutting depth adjustment command and adjustment speed V that have been verified to be correct are used as input to drive the cutting table adjustment actuator to move. For the electric actuator, the motor speed and time required to rotate are calculated according to the target cutting depth; for the hydraulic actuator, the opening and pressure of the hydraulic valve are controlled according to the difference between the target depth and the current depth, so that the hydraulic cylinder pushes the cutting table to move to the target position; The motor current after the header adjustment actuator is actuated, the displacement information fed back by the position encoder, or the displacement sensor data of the hydraulic cylinder is used as input to monitor the actual position of the header in real time. If there is a deviation between the actual position and the target position, the actuator parameters are adjusted in time.

[0013] Preferably, data storage and backtracking are also included, which specifically includes the following steps: All collected raw data, pre-processed data, fused analysis results, header adjustment instructions, and header actual action feedback information are stored in data storage devices according to time sequence and data type, providing basic data resources for subsequent data mining, machine learning model training, and harvesting strategy optimization. Through retrospective analysis of historical data, we discovered problems and potential areas for improvement in the harvesting process. By comparing the adjustment depth and loss rate of the harvesting table under different degrees of lodging, we established a more sophisticated cutting depth adjustment strategy, identified areas where specific sensor data showed frequent abnormalities, and reminded farmers to pay attention to problems in barley planting and field management.

[0014] Beneficial effects of the present invention: 1. Through real-time data collection and analysis based on intelligent sensors, the growth status, lodging degree and terrain undulation of highland barley can be accurately sensed. Intelligent sensors can monitor the height and lodging of highland barley and the relative position between the cutting platform and highland barley in real time, and automatically adjust the depth of the cutting platform, thereby avoiding errors caused by manual experience and improving the adjustment accuracy of the cutting depth.

[0015] 2. This technical solution can adjust the working state of the harvesting platform in real time according to different field conditions (such as uneven ground, plants of different heights, barley of different lodging degrees, etc.), so that it can operate stably in a changing environment and reduce low harvesting efficiency and losses caused by improper operation or environmental changes.

[0016] 3. The precise adjustment of cutting depth by intelligent sensors avoids cutting too deep or too shallow, reduces unnecessary soil disturbance and energy waste, helps save fuel and improve work efficiency.

[0017] 4. Real-time adjustment and precise control of cutting depth can avoid over-cutting or incomplete cutting, significantly reduce grain loss, and improve the harvesting quality of highland barley. The application of intelligent adjustment system can help improve the overall yield and quality of the harvesting process and improve agricultural production efficiency.

[0018] 5. Through real-time data feedback and automatic adjustment, the burden of manual operation is reduced, the degree of automation of operations is improved, the impact of human factors is reduced, and the efficiency and stability of the harvesting process are ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 is a flow chart of the steps of the method of the present invention; Figure 2 A flowchart of the steps of designing a special highland barley combine harvester with an intelligent real-time cutting depth adjustment function in the method of the present invention; Figure 3 Flow chart of the steps of data collection and preprocessing in method S1 of the present invention. DETAILED DESCRIPTION

[0021] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0022] See also Figure 1-Figure 3 The embodiment of the present invention provides a real-time cutting depth adjustment method for a low-loss barley harvester based on an intelligent sensor. The method realizes the automatic adjustment of the depth of the harvester of a barley combine harvester through intelligent sensors and real-time data analysis technology. By real-time acquisition and processing of multiple sensor data, the system can intelligently adjust the depth of the harvester according to the actual growth status of the barley (such as plant height, lodging degree, etc.), avoiding too deep or too shallow cutting depth, and improving the harvesting efficiency and crop yield rate. Especially in complex terrain and uneven fields, automatic adjustment can effectively reduce the loss of barley, ensure the quality of harvesting, and improve the comprehensive output value of crops.

[0023] Specifically, first of all, the designed combine harvester needs to be equipped with multiple sensors, including sensors for measuring the height of barley plants, ground flatness, lodging degree, ear position and crop density. These sensors transmit the collected data to the data processing center in real time through the data communication network. The data processing center consists of hardware and software systems, which can perform real-time data processing and depth adjustment decisions. At the same time, the header adjustment actuator needs to be improved to ensure that it can respond quickly according to instructions and adjust the height and angle of the header.

[0024] Furthermore, the sensors are responsible for collecting real-time data on the growth of highland barley, including information such as the height of highland barley plants, ground flatness, degree of lodging, ear position, and crop density. These data are pre-processed to remove noise and outliers to ensure data accuracy and reliability.

[0025] Through data fusion preprocessing, the data collected by various sensors are comprehensively analyzed. For example, the height and lodging degree of highland barley plants directly affect the adjustment requirements of the cutting table depth, while the crop density may affect the tool pressure during cutting. By establishing a data fusion model, the system can comprehensively consider various factors and perform real-time data fusion calculations to obtain accurate depth adjustment decisions.

[0026] Based on the results of data fusion and analysis, the system will generate adjustment instructions for the header. These instructions will take multiple factors into consideration and ensure the accuracy and effectiveness of the instructions through redundancy and verification mechanisms. The instruction generation process requires the integration of multiple data (such as the height of the barley, ground conditions, etc.) to determine the depth and position of the header in real time.

[0027] According to the generated depth adjustment command, the adjustment actuator will drive the height adjustment device of the header. By real-time monitoring of the header position and feeding it back to the system, the header depth is always kept at the optimal state to adapt to the growth of barley and changes in the ground environment.

[0028] In one possible implementation, laser rangefinder sensors are installed on both sides in front of the harvesting platform, and their main function is to measure the height of the barley plants and the flatness of the ground in real time. The laser rangefinder sensor has high measurement accuracy and can accurately reflect the growth of barley plants under different terrain conditions. In addition, the laser sensor can overcome the errors that may be caused by traditional touch sensors and reduce the problem of inaccurate data caused by uneven terrain or plant lodging. By monitoring the ground conditions in real time, the laser sensor can help the system intelligently adjust the depth of the harvesting platform to avoid harvesting losses caused by uneven ground or plant lodging.

[0029] Cameras are installed on both sides of the cutting platform to capture the lodging of barley and the position of the ear head. Through computer vision technology, the camera can obtain and analyze image data, and judge in real time whether the barley plant has lodged and the changes in the position of the ear head. This information directly affects the fine adjustment of the cutting platform depth. When the barley lodges, the system will issue adjustment instructions based on the image recognition results to ensure that the cutting platform can smoothly pass through the lodging area and minimize the loss of barley.

[0030] A pressure sensor is installed under the header to monitor the pressure changes between the header and the ground in real time. By sensing the pressure changes, the pressure sensor can estimate the density of the barley plants and the hardness of the stems, helping the system to determine the harvesting status of the crop. If the sensor detects a sharp fluctuation in ground pressure, the system can respond quickly and adjust the header depth to avoid crop damage due to the header being set too low.

[0031] Furthermore, to ensure that the data collected by the sensors can be transmitted accurately and in real time, the system has designed a dedicated data communication network. According to the types of different sensors and data transmission requirements, select appropriate communication protocols, such as wireless communication, Bluetooth, Wi-Fi or dedicated industrial communication protocols, to ensure efficient transmission of data under various environmental conditions. In terms of network hardware, select high-bandwidth, low-latency communication equipment to ensure real-time transmission and reception of sensor data. The establishment of a data communication network can not only effectively connect sensors and data processing centers, but also automatically switch to alternative communication paths when encountering network obstacles, ensuring high stability and reliability of the system.

[0032] In order to achieve accurate header depth adjustment, the data processing center needs to have powerful computing power. High-performance computer configuration ensures that data can be processed in a short time and generate accurate adjustment instructions. In terms of storage devices, large-capacity, high-speed data storage devices are used to store the massive data collected by sensors and ensure that no important information is lost under high-speed data flow. In terms of data processing software, the developed special software can receive and analyze data from various sensors, use algorithms for real-time analysis, identify key factors such as ground conditions, plant height, lodging conditions, and automatically generate header depth adjustment instructions.

[0033] The header adjustment actuator is responsible for adjusting the depth and angle of the header according to the adjustment instructions generated by the data processing center. According to the structural requirements of the header, electric actuators or hydraulic actuators are selected to ensure that the adjustment response is fast and accurate. Through the feedback loop system, the actual position of the header is constantly monitored and the action of the actuator is adjusted to ensure that the header depth is always maintained at the optimal state. The choice of electric or hydraulic actuators can provide appropriate adjustment force according to different terrain and crop conditions to avoid damage to the header or crop loss caused by too fast or too slow adjustment.

[0034] In one possible implementation, after the laser ranging sensor is started, it will work in high-speed measurement mode to collect real-time data on the height of highland barley plants. These data generate a real-time data stream reflecting the height changes of highland barley plants through the principle of laser reflection. Since laser ranging technology can provide high-precision data, it can accurately capture the slight changes in the height of highland barley plants during their growth process. Therefore, in different terrains or crop-dense areas, the laser ranging sensor can help the system dynamically adjust the depth of the cutting platform to ensure that the cutting platform is always in the optimal harvesting position for highland barley.

[0035] The laser distance sensor is installed under the chassis of the header, emitting laser downwards to measure the distance from the ground to the sensor and record the changes in ground undulations. By recording the ground height fluctuations at a frequency of 50Hz, it can capture the slight unevenness of the ground and avoid crop damage caused by uneven ground on the header. This step can accurately sense the terrain changes in the area where the header is located, and then adjust the depth of the header in real time to ensure that it maintains optimal contact with the ground and reduce barley losses.

[0036] The cameras installed on both sides of the header periodically take images of highland barley plants and obtain image data containing information about highland barley lodging. Computer vision technology is used to perform a preliminary analysis of the images and extract the lodging angle θ and lodging ratio of highland barley. These data reflect the lodging of highland barley plants and provide a basis for subsequent header depth adjustment. When the lodging situation is serious, the system can predict and adjust the header depth in advance to reduce the loss of lodging highland barley and ensure the smooth progress of the harvesting operation.

[0037] By processing the barley images taken by the camera, the position of the barley head in space can be accurately located to obtain three-dimensional coordinate data. Accurately grasping the spatial position of the barley head makes the adjustment of the cutting table more precise, effectively avoiding areas that are not suitable for harvesting, and avoiding cutting too low to immature or damaged heads, thereby improving harvesting efficiency and barley harvest quality.

[0038] The barley images captured by the camera are further processed to calculate the number of barley plants and the spatial density distribution of the area, and then the crop density is obtained. The crop density directly affects the sensitivity of the depth adjustment of the header. When the crop density is too high, the header may need to increase the depth to avoid damaging the plants in the dense area; when the density is low, the header can be adjusted to a lower depth for more accurate harvesting.

[0039] In order to eliminate the noise in the data collection process, the Kalman filter is used to calibrate the collected data such as highland barley plant height, ground flatness, lodging degree, ear position and crop density in time or space. The Kalman filter can effectively suppress sensor errors and environmental noise, making the data more accurate. The filtered data can be more synchronously matched in time and space, ensuring that the signals collected by various sensors can be coordinated to provide a more reliable basis for the system, thereby achieving accurate header depth adjustment.

[0040] In a possible implementation, during the data collection process, the barley plant height, ground flatness, barley lodging degree, barley ear position and crop density data obtained by the sensor may have errors caused by inconsistent measurement ranges and units or external environmental factors. Therefore, before data fusion, all data are first denoised to ensure data accuracy. Then, through normalization, all data are converted to a unified standard range (such as between 0 and 1) to eliminate differences caused by different data sources. This step effectively ensures the compatibility of different types of data and provides a stable and balanced input for subsequent analysis.

[0041] According to the degree of influence of different factors on the cutting depth during the barley harvesting process, the system needs to assign corresponding weights to each data. By analyzing the sensitivity of different variables to the cutting depth, the contribution of each data in the model is determined. For example, the height change of barley plants may have a greater impact on the cutting depth, while the crop density may have a smaller impact on the depth adjustment. By reasonably determining the weight, ensure that the impact of each data is accurately reflected in the fusion model.

[0042] Gaussian process regression is a powerful non-parametric regression method that can capture the complex nonlinear relationship between variables by modeling the predicted distribution of input variables (such as plant height, ground flatness, etc.). All normalized data are used as input variables, and the ideal cutting depth of the cutting table is used as the output variable, and the Gaussian process regression model is used for modeling. This step enables the system to update the depth prediction of the cutting table in real time according to the changes in different factors, and has strong adaptive capabilities.

[0043] During the actual harvesting process, the sensor will continuously collect data and input the real-time data into the trained Gaussian process regression model. The model performs prediction calculations based on the latest input data and outputs the depth that the header needs to be adjusted. The real-time data fusion calculation process enables the header to automatically adjust to the most appropriate cutting depth under different travel states and terrain conditions, thereby ensuring the best harvesting effect of highland barley.

[0044] In a possible implementation, the various sensor data of highland barley are first normalized to ensure that the dimensions of different data are consistent and to avoid deviations in the model caused by data of different scales. The normalized data are then given different weights according to their importance in the depth adjustment of the cutting platform, such as the weight of highland barley plant height is 0.4, the ground flatness is 0.3, the degree of highland barley lodging is 0.2, the ear position is 0.06, and the grain maturity and crop density have a smaller weight of 0.02. The setting of weights reflects the relative importance of each factor in actual harvesting. For example, plant height and ground flatness have a greater impact on the depth of the cutting platform, while factors such as crop density have a smaller impact. Therefore, the weighted step can accurately reflect the contribution of each data in the data fusion process, thereby improving the accuracy of depth adjustment.

[0045] Gaussian process regression (GPR) was chosen as the main modeling tool for this method because it can capture complex nonlinear relationships between data and make predictions based on input features. The input features consist of weighted normalized sensor data. The radial basis function (RBF) was selected as the kernel function in this method. The form of the RBF kernel function is: ; in, represents the signal variance, is the length scale parameter, and denote the i-th and j-th input samples respectively, is the number of input features, m is the index parameter, is the kernel function; through this kernel function, the similarity between the input features can be effectively captured, so as to more accurately predict the cutting platform depth.

[0046] For the training dataset , the method of maximizing the log-likelihood function is used to estimate the hyperparameters of the Gaussian process regression model, including the signal variance and length scale , and the noise variance Through this step, the model parameters can be optimized so that the model can better fit the historical data and make accurate predictions.

[0047] During the actual harvesting process, sensors will continuously collect real-time data, and after normalization and weighting, input the data into the Gaussian process regression model. Based on the model's prediction results, the depth of the harvesting platform is adjusted in real time, so that it can be flexibly changed according to the growth status of the barley and ground conditions, thereby maximizing the harvesting effect.

[0048] By combining data weighting with the Gaussian process regression model, the depth of the harvesting table can be accurately adjusted in real time according to the different growth stages of highland barley, ground conditions, and other key factors of the crop. The advantage of this method is that it is highly adaptable and can make adjustments based on real-time environmental changes, reducing errors caused by human intervention while improving harvesting efficiency and quality. In addition, the kernel function design of the model enables this method to handle complex nonlinear relationships, further enhancing the accuracy of the system. In general, this method can not only improve the level of automation in highland barley harvesting, but also effectively reduce losses and promote the development of agricultural mechanization.

[0049] In one possible implementation, real-time data is processed by a Gaussian process regression (GPR) model to achieve accurate adjustment of the header depth, and the credibility of the prediction is evaluated based on the mean and variance of the model output.

[0050] Specifically, first of all, the real-time data collected, such as the height of highland barley plants, ground flatness, degree of highland barley lodging, ear position and crop density, will be collected through sensors. In order to ensure that data of different dimensions can work together in the model, these data need to be normalized. The normalized data makes all input features in the same dimension, avoiding errors caused by inconsistent dimensions.

[0051] The normalized data is weighted to assign different weights to each feature according to its importance in adjusting the depth of the header. For example, the height of the barley plant and the flatness of the ground play a more significant role in adjusting the depth of the header, while the density of the crop may have a smaller impact on the depth. After weighting, the data can better reflect the relative importance of each factor, thereby improving the accuracy of the model prediction.

[0052] The normalized and weighted real-time data (expressed as ) is input into a pre-trained Gaussian process regression model. The model has optimized hyperparameters (such as signal variance) by maximizing the log-likelihood function based on historical data. and length scale ), and can handle the complex nonlinear relationship between input data and header depth adjustment.

[0053] The Gaussian process regression model calculates the predicted output (i.e., the amount of header depth adjustment) based on the input real-time data. Because Gaussian process regression has the ability to model uncertainty, the model not only provides a predicted value for the header adjustment, but also gives a confidence interval that indicates the reliability of the prediction. Specifically, the model calculates the mean (i.e., the predicted header depth) and the variance (i.e., the uncertainty of the prediction). The mean is used to determine the amount of header adjustment, while the variance is used to assess the credibility of the prediction. If the variance is large, it means that the uncertainty of the prediction is high and further adjustments or other supplementary measures are needed.

[0054] In a possible implementation, a depth adjustment decision process based on fusion data analysis is further introduced. By inputting the predicted value of the cutting depth adjustment amount of the header calculated based on the Gaussian process regression (GPR) model and its confidence interval, a header depth adjustment instruction is generated, and finally the header depth is accurately adjusted.

[0055] Specifically, first, after real-time data collection and Gaussian process regression model prediction, the adjustment value of the cutting depth of the cutting table is obtained (i.e. ) and the confidence interval of the prediction. These data reflect the depth adjustment amount that the header needs to make and the confidence of the adjustment based on the actual cutting depth of the current header. and the predicted adjustment , calculate the target cutting depth , the formula is: ; By adding the current header depth to the adjustment amount, a new target depth is obtained. This target depth will be used as a reference value for the next adjustment.

[0056] In order to ensure the smooth adjustment of the header, the adjustment speed and accuracy of the header adjustment actuator must also be considered. Since the adjustment mechanism has certain speed limitations and accuracy errors in actual operation, a reasonable adjustment speed must be set so that it can complete the adjustment within the specified time and ensure that the adjustment accuracy is within the allowable range. Therefore, based on the actual adjustment capability of the header, select the appropriate adjustment speed to adjust the target depth.

[0057] Based on the calculated target cutting depth and adjustment speed, the system generates a header depth adjustment command and transmits the command to the actuator for specific adjustment operations. This command contains the specific value required for adjustment and the speed limit during execution, ensuring that the adjustment process is efficient and accurate.

[0058] In a possible implementation, in order to ensure the accuracy of the instructions, the system adopts a dual-channel independent generation mechanism. The two channels calculate the instructions for the depth adjustment of the cutting table respectively. Each channel generates an independent adjustment instruction based on the input target cutting depth, adjustment speed and adjustment amount confidence interval. The two channels are independent of each other and are calculated by different sensors or algorithms. Finally, the instructions generated by the two channels are compared. If the instructions are consistent, the instructions are considered reliable and no further verification is required; if the instructions are inconsistent, the next step of the redundant verification procedure is triggered.

[0059] If the instructions are inconsistent, the system needs to perform further redundancy checks. At this time, first make a judgment based on the confidence interval: if the confidence interval is large, it indicates that the predicted header depth adjustment amount is The uncertainty is high. In this case, the system will start the secondary calculation of the backup sensor data to verify whether the predicted adjustment is accurate. The backup sensor may obtain data through a sensor with a different working principle (such as a different type of sensor or a backup position sensor). Using these backup data for secondary calculations helps to eliminate the inconsistency caused by the measurement error of the main sensor.

[0060] When there is a high degree of uncertainty in the predicted adjustment, the system will temporarily slow down the adjustment speed of the header to reduce the potential risk caused by inaccurate adjustment. At this time, the system pauses or slows down the depth adjustment of the header and waits for subsequent data updates to recalculate. This slow adjustment process ensures the stable operation of the header under uncertain conditions and avoids the adverse consequences caused by over-adjustment.

[0061] During the redundancy check, the system will recalculate the header adjustment instructions based on the backup sensor data until a consistent and reliable instruction is obtained. At this time, the accurate instructions that have been verified will be used to drive the header adjustment to ensure that the header harvests at the optimal depth.

[0062] The redundancy and verification mechanism ensures the high reliability of the instructions, and through flexible verification processes and data updates, it guarantees the accuracy and stability of the depth adjustment of the barley cutter, significantly improving the overall performance and automation level of the system.

[0063] In a possible implementation, first, the target cutting depth adjustment instruction and adjustment speed that have been verified and confirmed to be correct are used as input to drive the actuator of the header to adjust. The target cutting depth is the depth value calculated by the sensor, and the adjustment speed is set by the system according to the current state of the header, terrain conditions and crop characteristics, ensuring that no excessive vibration or excessive impact is generated during the adjustment process, thereby reducing crop losses.

[0064] Furthermore, for electric actuators, once the target cutting depth is determined, the system calculates the motor speed and time required to rotate. This process is optimized based on factors such as the motor's rotation characteristics, the difference between the target depth and the current depth, and the operating load of the header. The electric actuator ensures that the header can be adjusted smoothly to the target depth by precisely controlling the motor speed, avoiding too fast or too slow adjustments that may cause crop damage or equipment failure.

[0065] If the header uses a hydraulic actuator, the system will control the opening and pressure of the hydraulic valve according to the difference between the target depth and the current depth. The core of this process is to accurately control the hydraulic cylinder to push the header to the target position. The opening and pressure of the hydraulic valve will be adjusted according to the required depth to ensure that the hydraulic system outputs stable and accurate force to push the header to complete fine adjustment.

[0066] Furthermore, after the header adjustment actuator moves, the system will monitor the actuator's feedback signal in real time, including motor current, position encoder displacement data or hydraulic cylinder displacement sensor data. These feedback data help the system monitor the actual position in real time and compare it with the target position. If a deviation is detected between the actual position and the target position, the system will make adjustments in a timely manner.

[0067] When the feedback signal shows position deviation, the system will automatically adjust the control parameters of the actuator, adjust the motor current or the opening and pressure of the hydraulic valve, etc., to ensure that the header continues to move accurately toward the target position. This real-time dynamic adjustment mechanism ensures the accuracy of the header depth adjustment and effectively avoids adjustment deviations caused by mechanical errors, load changes or interference from the external environment.

[0068] In one possible implementation, during the operation of the header, the system will continuously collect multiple types of data, including raw sensor data (such as soil moisture, crop height, lodging angle, etc.), pre-processed data (such as sensor signals after noise filtering), fused analysis results (such as header adjustment depth, header working status, etc.), and header adjustment instructions and actual feedback information (such as electric / hydraulic system action, header actual position, etc.). All of these data will be stored in chronological order to ensure the integrity and traceability of the information. The data storage device uses an efficient database management system that can classify and store different types of data and maintain efficient retrieval, which is convenient for subsequent analysis and processing.

[0069] Furthermore, through retrospective analysis of stored data, the system can identify the changing trend of header depth adjustment in different harvesting stages, as well as the impact of different crop lodging degrees on cutting depth and loss rate. For example, by comparing the header adjustment depth under different lodging degrees, the system can find out in which cases the header adjustment strategy is not suitable or abnormal, resulting in the loss of highland barley crops. These data provide valuable basis for analyzing the actual effect of header adjustment and optimizing harvesting strategies.

[0070] Furthermore, based on the results of retrospective analysis, the system can propose targeted improvement plans. For example, when sensor data anomalies frequently occur in certain areas, the system can automatically identify these areas and remind farmers to pay attention to the growth status of barley or potential problems in field management, such as too much or too little water, uneven soil, etc. By optimizing the header depth adjustment strategy, the system can finely adjust the depth to reduce crop losses and improve harvesting efficiency.

[0071] Furthermore, by further mining historical data and combining it with machine learning algorithms, the system can continuously optimize the strategy of adjusting the harvesting platform. For example, by analyzing the impact of different climate conditions, soil types, crop growth status and other factors on the harvesting process, a more adaptive machine learning model can be trained. In this way, the system can not only cope with current environmental changes, but also continue to learn and improve to adapt to future harvesting needs under different conditions.

[0072] Combining data analysis and backtracking results, the system can also provide farmers with intelligent advice and warnings. For example, when sensors detect excessive lodging of crops in a certain area, the system will give timely management suggestions and prompt farmers on what measures to take (such as adjusting the depth of the cutting table, changing the cutting time, or increasing field management, etc.). In addition, the system will also warn of potential production problems based on the backtracking results of historical data, helping farmers take measures in advance to avoid losses.

[0073] The storage and retrospective analysis of data form a closed-loop feedback mechanism, providing continuous support for the optimization of harvesting strategies. Through retrospective analysis of historical data, the system can continuously identify potential problems in the harvesting process and make timely adjustments. Through data storage and retrospective analysis, combined with deep mining of historical data and machine learning optimization, it is possible to achieve continuous improvement of the header depth adjustment strategy, improve the accuracy and efficiency of harvesting operations, and provide farmers with intelligent field management support.

[0074] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0075] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A real-time cutting depth adjustment method for a low-loss barley harvesting platform based on an intelligent sensor, characterized in that: The following steps are involved: Design a special highland barley combine harvester with intelligent real-time cutting depth adjustment function, including sensor deployment, data communication network construction, data processing center hardware and software construction, and improvement of the cutting platform adjustment actuator; Data collection and preprocessing, including barley plant height data collection, ground flatness data collection, barley lodging degree data collection, barley ear position data collection, crop density data collection, and data preprocessing; Data fusion and analysis, including data fusion preprocessing, fusion model establishment, and real-time data fusion calculation; Cutting depth adjustment instruction generation, including depth adjustment decision based on fusion data analysis and instruction redundancy and verification; Automatic adjustment of the header is achieved, including adjustment of the actuator drive and real-time monitoring feedback of the header position.

2. The method for adjusting the cutting depth of low-loss highland barley harvesting platform based on intelligent sensor according to claim 1 is characterized in that: Designing a highland barley combine harvester with intelligent real-time cutting depth adjustment function includes the following steps: Sensor deployment: laser rangefinders are installed on both sides of the cutting platform to measure the height of barley plants and the flatness of the ground; cameras are installed on both sides of the cutting platform to capture the lodging of barley and the position of the ear head; pressure sensors are installed under the cutting platform to sense the change in contact pressure between the cutting platform and the ground, so as to estimate the density of barley plants and the hardness of the stems; Build a data communication network, select appropriate communication protocols and hardware devices according to sensor types and data transmission requirements, complete the construction of the data communication network, and realize real-time transmission and reception of sensor data; The data processing center is equipped with software and hardware, and high-performance computers are configured for data processing and calculation; data storage devices with sufficient capacity are equipped to store the massive data collected by sensors; and special data processing software is developed with data receiving, processing, analysis and instruction generation functions; The header adjustment actuator is improved, and the appropriate actuator is selected according to the header structure and adjustment requirements.

3. The real-time cutting depth adjustment method of highland barley low-loss header based on intelligent sensor according to claim 2 is characterized in that: Data collection and preprocessing include the following steps: To collect data on the height of highland barley plants, the laser ranging sensor is activated to work in high-speed measurement mode, generating a series of real-time data streams reflecting the changes in the height of highland barley plants; Ground flatness data collection: the laser rangefinder sensor emits a laser downward from the bottom of the header chassis to measure the distance from the ground to the sensor, and obtains a data stream describing the undulation of the ground, recording the ground height fluctuation value at a frequency of 50Hz; Data collection on the degree of lodging of highland barley: cameras on both sides of the harvesting platform regularly take images of highland barley plants, obtain image data containing information on lodging of highland barley, and perform preliminary recognition to analyze the lodging angle θ and lodging ratio of highland barley; The data of the position of the barley head is collected by processing and analyzing the barley image taken by the camera to locate the position of the barley head in space and express it in three-dimensional coordinates; Calculate the number and spatial density distribution of highland barley plants in the highland barley image captured by the camera to obtain the crop density; Data preprocessing uses a Kalman filter to remove noise, and performs temporal or spatial calibration on barley plant height data, ground flatness data, barley lodging degree data, barley ear position data, and crop density data to ensure that data collected by different sensors can be synchronously matched.

4. The method for adjusting the cutting depth of low-loss highland barley harvesting platform based on intelligent sensor according to claim 3 is characterized in that: Data fusion and analysis includes the following steps: Data fusion preprocessing normalizes the barley plant height data, ground flatness data, barley lodging degree data, barley ear position data, and crop density data after noise removal to eliminate data differences caused by different sensor measurement ranges or units; According to the influence of various factors on the cutting depth of the header during the barley harvesting process, the weights of different data are determined; The fusion model based on Gaussian process regression was established, with the normalized barley plant height data, ground flatness data, barley lodging degree data, barley ear position data, and crop density data as input variables, and the ideal cutting depth of the cutting platform as the output variable to construct a Gaussian process regression model. Real-time data fusion calculation, substitute the real-time collected data into the Gaussian process regression model, perform prediction calculation, and obtain the cutting depth adjustment amount of the cutting table.

5. The method for adjusting the real-time cutting depth of a low-loss barley harvesting platform based on an intelligent sensor according to claim 4 is characterized in that: According to the influence of various factors on the cutting depth of the header during the barley harvesting process, the weights of different data are determined as follows: Highland barley plant height: 0.4, ground flatness: 0.3, highland barley lodging degree: 0.2, ear position: 0.06, grain maturity: 0.02, crop density: 0.02; Collect historical data on highland barley plant height, ground flatness, highland barley lodging, highland barley ear position, crop density and corresponding cutting depth; The radial basis function is selected as the kernel function of Gaussian process regression, and the specific form is as follows: ； in, represents the signal variance, is the length scale parameter, and denote the i-th and j-th input samples respectively, is the number of input features, m is the index parameter, is the kernel function; The normalized and weighted sensor data is used as input features to construct a Gaussian process regression model. The specific input form is as follows: ； in, is the weight vector of highland barley plant height data, ground flatness data, highland barley lodging degree data, highland barley ear position data, and crop density data. It is the normalized data of highland barley plant height, ground flatness, highland barley lodging degree, highland barley ear position and crop density vector; For the training dataset , calculate the covariance matrix based on the kernel function : ； in, and is the weighted input vector ; Using historical data, we can maximize the log-likelihood function to estimate the Gaussian process regression kernel function. and , and the noise variance .

6. The method for adjusting the real-time cutting depth of a low-loss barley harvesting platform based on an intelligent sensor according to claim 5, characterized in that: Based on training and , and the noise variance Calculate the mean of the output from the training data and variance , according to the mean Determine the cutting depth adjustment of the header, and refer to the variance To assess the credibility of the forecast; The real-time collected data on highland barley plant height, ground flatness, highland barley lodging, highland barley ear position, and crop density are normalized and weighted to obtain , and input it into the trained Gaussian process regression model to calculate the predicted value of the cutting depth adjustment amount and its confidence interval.

7. The method for adjusting the real-time cutting depth of a low-loss highland barley harvesting platform based on an intelligent sensor according to claim 6, characterized in that: In the depth adjustment decision based on fusion data analysis, the predicted value of the cutting depth adjustment amount of the header and its confidence interval are calculated and the cutting depth adjustment instruction of the header is generated. The specific calculation process is as follows: Adjust the value according to the predicted cutting depth of the header , combined with the actual cutting depth of the current header , calculate the target cutting depth , the formula is: ； Consider the adjustment speed and accuracy of the header adjustment actuator and set an appropriate adjustment speed.

8. The method for adjusting the real-time cutting depth of a low-loss highland barley harvesting platform based on an intelligent sensor according to claim 7, characterized in that: In the redundancy and verification of instructions, the target is cut to depth The confidence interval of the adjustment speed V and the cutting depth adjustment amount of the header is used as input to perform redundancy verification on the adjustment command to ensure the accuracy and reliability of the command. The specific verification process is as follows: A dual-channel instruction generation mechanism is adopted, that is, the adjustment instructions are generated through two independent channels respectively, and the instructions generated by different channels are compared to see if they are consistent. If the instructions generated by the two channels are consistent, the instructions are considered to be reliable. If the instructions are inconsistent, further verification is required; If the confidence interval is large, it means that the predicted cutting depth adjustment value of the header is The uncertainty of the value is high, which triggers a further verification process and calls the backup sensor data for secondary calculation to verify the predicted The value is accurate, and the header adjustment speed is temporarily slowed down to wait for subsequent data to be updated and then recalculated.

9. The method for adjusting the real-time cutting depth of a low-loss highland barley harvesting platform based on an intelligent sensor according to claim 8, characterized in that: The following steps are included in the implementation of automatic adjustment of the header: The target cutting depth adjustment command and adjustment speed V that have been verified to be correct are used as input to drive the cutting table adjustment actuator to move. For the electric actuator, the motor speed and time required to rotate are calculated according to the target cutting depth; for the hydraulic actuator, the opening and pressure of the hydraulic valve are controlled according to the difference between the target depth and the current depth, so that the hydraulic cylinder pushes the cutting table to move to the target position; The motor current after the header adjustment actuator is actuated, the displacement information fed back by the position encoder, or the displacement sensor data of the hydraulic cylinder is used as input to monitor the actual position of the header in real time. If there is a deviation between the actual position and the target position, the actuator parameters are adjusted in time.

10. The method for adjusting the real-time cutting depth of a low-loss highland barley harvesting platform based on an intelligent sensor according to claim 9, characterized in that: It also includes data storage and backtracking, which specifically includes the following steps: All collected raw data, pre-processed data, fused analysis results, header adjustment instructions, and header actual action feedback information are stored in data storage devices according to time sequence and data type, providing basic data resources for subsequent data mining, machine learning model training, and harvesting strategy optimization. Through retrospective analysis of historical data, we discovered problems and potential areas for improvement in the harvesting process. By comparing the adjustment depth and loss rate of the harvesting table under different degrees of lodging, we established a more sophisticated cutting depth adjustment strategy, identified areas where specific sensor data showed frequent abnormalities, and reminded farmers to pay attention to problems in barley planting and field management.

Images