A harvester working quality self-adaptive control system

Abstract

The application relates to the field of mechanical control, and relates to a harvester operation quality self-adaptive control system, which comprises the following modules: a region division module that divides the operation region of a harvester into multiple operation units; an image acquisition module that acquires front and rear view images of the harvester; an image processing module that analyzes the front view images to obtain a scene feature vector of the operation unit, and analyzes the rear view images to obtain a grain impurity content, a grain breakage rate and a grain loss rate of the operation of the harvester in the operation unit; a parameter adjustment module that obtains actual execution parameters of the operation unit according to the grain loss rate, the grain impurity content and the grain breakage rate of the operation unit that has completed operation, a scene feature vector of the operation unit that has not been operated and a scene category to which the operation unit belongs; and a vehicle control module that controls the operation of the harvester in the operation unit corresponding to the actual execution parameters according to the actual execution parameters, so as to solve the problems of operation adjustment hysteresis and poor self-adaptive control capability of the harvester.

Description

Technical Field

[0001] This application relates to the field of mechanical control, and in particular to an adaptive control system for the operation quality of a harvester. Background Technology

[0002] Throughout the harvester's field operations, various operating parameters such as the harvester's travel speed and roller speed are adjusted in a timely manner based on the real-time harvesting conditions, including the crop's harvesting status and the presence of impurities or breakage in the grain.

[0003] In existing technologies, the common approach is to analyze real-time images of harvested crops to determine their impurity and breakage rates, and then adjust the harvester's operating parameters accordingly. However, this adjustment method relies solely on images of the harvested crops, neglecting crop growth and the harvester's own operational capabilities. This leads to lag and bias in adjusting the harvester's operating parameters, making it difficult to achieve precise adaptive control. Summary of the Invention

[0004] This application provides an adaptive control system for harvester operation quality, which at least solves the problems of lag and one-sidedness in harvester adjustment and poor adaptive control capability in related technologies.

[0005] In a first aspect, embodiments of this application provide a harvester operation quality adaptive control system, including: The area division module is configured to divide the harvester's operating area into multiple operating units; The image acquisition module is configured to acquire a front view image of the harvester in the direction of travel of the harvester, and acquire a rear view image of the harvester in the opposite direction of travel, when the harvester is working in any of the working units. The image processing module is configured to analyze the front view image obtained in any of the said working units to obtain the scene feature vector of the working unit; and to analyze the rear view image obtained in any of the said working units to obtain the grain impurity rate, grain breakage rate and grain loss rate of the harvester in the said working unit. The parameter adjustment module is configured to calculate an adjustment amount based on the grain loss rate, grain impurity rate, and grain breakage rate of the completed work unit; determine the initial execution parameters of the harvester in the work unit based on the scene feature vector of the non-operational work unit and the scene category to which the work unit belongs; and calculate the sum of the adjustment amount and the initial execution parameters corresponding to the work unit when the harvester enters the work unit to obtain the actual execution parameters of the work unit. The vehicle control module is configured to control the harvester to operate in the work unit corresponding to the actual execution parameters according to the actual execution parameters.

[0006] In some embodiments, the image processing module is further configured to: Extract the grain region from the rear view image; Set the color range of the seed pixel and the color range of the impurity pixel; In the grain region, the sum of pixels within the color range of the grain pixels is calculated as the total grain pixels; In the grain region, the sum of pixels within the color range of the impurity pixels is calculated as the total impurity pixels; The impurity content of the grain is obtained by calculating the ratio of the total number of impurity pixels to the total number of grain pixels.

[0007] In some embodiments, the image processing module is further configured to: Extract the connected region corresponding to any seed from multiple frames of the rear view image; Obtain the pixel area and aspect ratio of the connected region corresponding to any seed grain; Set area and shape constraints for complete grains; A complete seed is defined as a seed whose pixel area satisfies the area constraint of the complete seed and whose aspect ratio satisfies the shape constraint. Seeds whose pixel area does not meet the area constraints of the complete seed, the aspect ratio, or the shape constraints are considered broken seeds. Count the number of intact grains and the number of broken grains; The grain breakage rate is obtained by calculating the ratio of the number of broken grains to the sum of the number of intact grains and the number of broken grains.

[0008] In some embodiments, the scene feature vector includes ear density, and the image processing module is further configured to: The expected yield of the operation unit is obtained by multiplying the ear density, the area of ​​the operation unit, and the average weight of a single ear. Obtain the instantaneous grain mass flow rate and integrate the instantaneous grain mass flow rate to obtain the grain entering the warehouse. The first loss mass is obtained based on the expected output and the incoming quality; Obtain the number of lost ears in the work unit from the rear view image; The second loss mass is obtained by multiplying the number of lost ears by the average mass of fallen ears. The first loss quality and the second loss quality are weighted and summed to obtain the comprehensive loss quality; The grain loss rate is obtained by calculating the ratio of the total loss mass to the sum of the warehouse entry mass and the total loss mass.

[0009] In some embodiments, the parameter adjustment module is further configured to: Establish the correspondence between the adjustment amount and the mass error vector and the engine load margin; Based on the grain loss rate, the grain impurity rate, and the grain breakage rate, a current quality error vector for the work unit is constructed; wherein, the quality error vector includes grain loss rate error, grain impurity rate error, and grain breakage rate error; The current engine load margin of the harvester is obtained by subtracting the ratio of the current engine output power to the rated engine power. Based on the correspondence, the adjustment amount applied to the harvester corresponding to the current quality error vector and the current transmitter load margin is obtained.

[0010] In some embodiments, the parameter adjustment module is further configured to: According to the calculation formula for the scene category, the scene category to which the unworked work unit belongs is obtained; The calculation formula for the scene category includes: ; in, The scene category to which the unperformed work unit belongs. The weighted distance between the scene feature vector of the un-operated work unit and the feature center corresponding to any scene category.

[0011] In some embodiments, the parameter adjustment module is further configured to: Construct the scene feature weight matrix; Obtain the feature center corresponding to any of the aforementioned scene categories; The weighted distance is obtained according to the weighted distance calculation formula. The formula for calculating the weighted distance includes: ; in, The scene feature vector for the unperformed task unit. The scene feature weight matrix is... The feature center is the one corresponding to the scene category of the cth class.

[0012] In some embodiments, it also includes: The template update module is configured to update the feature center corresponding to the scene category of the cth class according to the feature center update formula; The feature center update formula includes: ; in, The feature center corresponding to the scene category of class c before the update. The feature center corresponding to the updated c-th scene category, is the learning rate for the scene category described in class c.

[0013] In some embodiments, the template update module is further configured to: Based on the grain loss rate, grain impurity rate, grain breakage rate, and operational efficiency of the operational unit, a comprehensive evaluation index for the operational unit is obtained. Determine whether the comprehensive evaluation index of the work unit meets the requirements. If yes, then the optimal comprehensive evaluation index corresponding to the scenario category to which the work unit belongs is updated to the comprehensive evaluation index of the work unit; otherwise, the optimal comprehensive evaluation index corresponding to the scenario category to which the work unit belongs remains unchanged. in, The optimal comprehensive evaluation index is the one corresponding to the scenario category to which the work unit belongs. To determine the threshold, This is the comprehensive evaluation index for the work unit.

[0014] In some embodiments, the template update module is further configured to: The comprehensive evaluation index of the work unit is obtained according to the calculation formula of the comprehensive evaluation index. The calculation formula for the comprehensive evaluation index includes: ; in, This is the weighting coefficient for the grain loss rate. This is the weighting coefficient for the impurity content of the grain. This is the weighting coefficient for the grain breakage rate. This is a weighting coefficient for work efficiency. This is the normalized value of the grain loss rate. This is the normalized value of the impurity content of the grain. This is the normalized value of the grain breakage rate. This is the normalized value of the work efficiency.

[0015] Compared to related technologies, the harvester operation quality adaptive control system provided in this application analyzes the front and rear view images of the harvester to obtain multiple indicators such as grain impurity rate, breakage rate, and loss rate, so as to adjust the actual execution parameters of the harvester. This solves the problems of lag and one-sidedness in harvester adjustment and poor adaptive control capability, and provides full-dimensional data support for precise parameter adjustment. It realizes adaptive dynamic adjustment of harvester operation parameters without manual intervention, avoiding the experience dependence and lag of harvester adjustment.

[0016] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a structural block diagram of a harvester operation quality adaptive control system according to an embodiment of this application; Figure 2 This is a schematic diagram showing the positional relationship between a harvester operation quality adaptive control system and a harvester according to an embodiment of this application.

[0018] In the diagram: 101, Region Division Module; 102, Image Acquisition Module; 103, Image Processing Module; 104, Parameter Adjustment Module; 105, Vehicle Control Module. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0020] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0021] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0022] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0023] Corn combine harvesters, during field operations, need to simultaneously consider multiple indicators such as harvest loss rate, impurity rate, breakage rate, and operational efficiency. This makes it a multi-variable, strongly coupled, and highly uncertain agricultural machinery operation system. Currently, corn harvesters widely used both domestically and internationally are primarily mechanical or electro-hydraulic adjustable structures. They achieve matching and optimization of operational quality by adjusting the overall machine speed, header or stalk-pulling roller speed, ear-picking plate gap, conveyor speed, peeling device operation, and cleaning fan speed. In actual production, after stopping at the field edge or midway to check for straw residue, ear loss, impurity in the grain bin, and breakage, the driver adjusts the overall machine parameters based on experience. This adjustment method is highly experience-dependent and requires a high level of skill from the driver. Furthermore, operational quality evaluation methods mainly rely on manual sampling and subjective visual inspection, lacking continuous, objective, online quantitative evaluation.

[0024] With the development of sensing and information technology, some grain combine harvesters have begun to be equipped with loss sensors, flow sensors, and grain bin status monitoring devices. These sensors, installed on the vibrating plate, screen box, or grain bin wall, detect the amount of material impact per unit time and the grain bin filling height, which are used to roughly determine the loss level and loading status.

[0025] In terms of structure and perception, most monitoring devices only monitor a single or a few quality indicators, and the detection locations are mostly concentrated in a local area of ​​the grain bin or the cleaning channel. They lack a system quality calibration mechanism that refines the operation area of ​​the entire machine operation process, making it difficult to accurately quantify the loss rate, impurity rate and breakage rate on a certain section of the operation path.

[0026] Existing technologies have also proposed solutions that guide parameter adjustments based on image analysis. These solutions use a single industrial camera or image sensor to image a localized area within the grain bin, identifying the proportion of impurities or broken particles on the grain pile surface. This information assists the driver in adjusting the speed of the cleaning fan or drum. While these solutions improve the online monitoring of single indicators such as impurity content or breakage rate, they primarily rely on alarms or alerts, lacking a systematic and coordinated control strategy for the multiple actuators of the harvester.

[0027] Based on this approach, even when online detection methods are introduced, most only limit or adjust a single variable for a single actuator, such as the speed of the cleaning fan, the travel speed, or the speed of a certain drum. There is a lack of coordinated adjustment strategies for multiple work units such as cutting table, conveying, peeling and cleaning, making it difficult to simultaneously take into account the comprehensive optimization of loss rate, impurity rate, breakage rate and work efficiency under complex working conditions.

[0028] In addition, some intelligent combine harvesters use yield monitoring devices and position sensors to spatially locate the operation process, establishing a correlation between yield and location for yield mapping and field management decisions. However, these systems primarily focus on the mapping between yield and space, processing information about the crop ahead, the results of subsequent operations, and the overall machine parameters in isolation. They lack a mechanism to establish a unified correlation between information on the growth of unharvested crops ahead, information on the quality of subsequent grain bin operations, and the parameters of the machine's multiple actuators. For various indicators such as loss rate, impurity rate, breakage rate, and operating efficiency, they only perform coarse-grained statistics on the time axis, making it difficult to accurately trace back and evaluate the overall machine's operating quality across different operating areas. Existing systems generally lack a knowledge base that links operational scenarios, equipment parameters, and quality results. Operational experience is difficult to structure and reuse, resulting in the need for repeated manual adjustments under different varieties, moisture contents, and plot conditions. Parameters are difficult to update adaptively with changes in the scenario, leading to insufficient stability and consistency in operational quality.

[0029] To address the aforementioned issues, this application proposes an adaptive control system for harvester operation quality. This system enables online quantitative evaluation of indicators such as loss rate, impurity rate, breakage rate, and operation efficiency of the operation unit. It establishes a mapping relationship between the operation scenario and the corresponding operation quality of the harvester's execution parameters. Based on this, it adaptively and collaboratively adjusts the multiple actuators of the whole machine, enabling the corn harvester to stably meet the preset operation quality targets under different operation scenarios.

[0030] like Figure 1 As shown in the figure, this application provides an adaptive control system for harvester operation quality, including: The area division module 101 is configured to divide the harvester's working area into multiple working units.

[0031] Specifically, according to unit length The harvester's operating path is divided into multiple segments, thereby dividing the operating area into multiple discrete operating units, and each operating unit is assigned a corresponding number.

[0032] Furthermore, after the harvester enters the field and begins operation, its speed and position are collected in real time, with each advance distance recorded as follows: The path defines a work unit. Assign a number to the work unit Record the timestamp of the work unit entering the header position of the harvester. And GNSS coordinates.

[0033] The Global Navigation Satellite System (GNSS) is a global navigation satellite system that can be used to locate, measure speed, and provide time synchronization for harvesters.

[0034] according to Calculate the ground area of ​​the work unit.

[0035] in, Number The floor area of ​​the work unit. The unit length is along the direction of travel. This refers to the effective working width of the cutting table.

[0036] The harvester is numbered During the operation of a work unit, the real-time travel speed is collected from the time the work unit begins to enter the header until the header has completely passed through the work unit. ,according to Calculate the average speed.

[0037] in, For the harvester, in the numbering The average travel speed of the work unit, Number The time required for the work unit to be completely passed by the cutting table. In order to be in Real-time travel speed at any given moment Number The start time of the work unit entering the cutting table area.

[0038] Calculate the number based on the transmission path and additional latency. The time delay of material transport by the harvester in the working unit. The calculation formula is: .

[0039] in, For numbering The time delay in the material transfer from the header to the grain drop observation window on the side of the harvester's grain bin within the working unit. This is the equivalent material transport path length from the header to the grain drop observation window on the side of the harvester's grain bin. Additional delays caused by conveyor chains, cleaning mechanisms, etc.

[0040] according to Calculate the numbering Reference time for evaluating the results in the work unit .

[0041] Define the backward detection time window as .

[0042] in, The time window width is half the width and can be set according to actual needs.

[0043] The image acquisition module 102 is configured to acquire a front view image of the harvester in the direction of travel when the harvester is operating in any work unit, and a rear view image of the harvester in the opposite direction of travel.

[0044] Specifically, the image acquisition module 102 detects that the geometric center of the working unit spans a distance of [distance missing] in front of the harvester's header. When the virtual trigger line is triggered, the front view image and the rear view image are acquired.

[0045] In the backward detection time window Multiple frames of images are captured internally, with the field of view covering the grain entering the warehouse or the area where grain falls.

[0046] The image processing module 103 is configured to analyze the front view image obtained in any working unit to obtain the scene feature vector of that working unit. It also analyzes the rear view image obtained in any working unit to obtain the grain impurity rate, grain breakage rate, and grain loss rate of the harvester in that working unit.

[0047] Specifically, a front view image is captured. Its field of view covers the current work unit. .

[0048] Front view image Based on target detection or segmentation networks, the system can perform row and column structure detection, including identifying the center line of the plant row and estimating the row spacing; perform ear detection, including identifying each ear target and its pixel coordinates and height position; and perform crop posture estimation, including calculating the crop tilt angle to obtain the degree of lodging.

[0049] Calculate the number based on the results of each test. The scene feature vector of the work unit, the scene feature vector is represented as .

[0050] in, This represents the scene feature vector of the work unit. The average ear height is obtained by averaging the height of the ear's center of gravity. The density of corn ears is given by the ratio of the number of ears to the ground area of ​​the work unit. The ratio, The line spacing deviation index is obtained by calculating the standard deviation or coefficient of variation of the difference between the measured line spacing and the designed line spacing. The lodging index, This refers to the density of the leaf canopy.

[0051] The parameter adjustment module 104 is configured to calculate the adjustment amount based on the grain loss rate, grain impurity rate, and grain breakage rate of the completed work unit. Based on the scene feature vector of the unoperated work unit and the scene category to which the work unit belongs, the initial execution parameters of the harvester in that work unit are determined. When the harvester enters the work unit, the sum of the adjustment amount and the initial execution parameters corresponding to that work unit is calculated to obtain the actual execution parameters of that work unit.

[0052] The vehicle control module 105 is configured to control the harvester to operate in the work unit corresponding to the actual execution parameters according to the actual execution parameters.

[0053] This application subdivides the work area into multiple work units, using these units as the control unit. This breaks through the limitations of traditional harvester's extensive, whole-area operation, enabling refined and traceable control of work quality and facilitating precise location of work areas corresponding to quality issues. Simultaneous acquisition of forward and rear-view images captures both the crop scene characteristics of the area to be worked in front of the harvester and the actual quality results of the already worked area behind the harvester. This dual-view image joint analysis method combines crop growth and operational status, providing comprehensive data support for precise parameter adjustment and solving the problem of single-view perception and incomplete information in existing technologies. Adjustments are calculated based on the quality feedback of completed work units, while initial parameters are preset based on the scene characteristics of unworked units. This achieves dual parameter tuning based on historical quality feedback and future scene prediction, resulting in stronger adaptability of work parameters. The actual execution parameters are obtained through parameter superposition and drive the operation, enabling adaptive dynamic adjustment of harvester work parameters without manual intervention. This significantly improves the level of operational intelligence and avoids the reliance on experience and lag of manual adjustments.

[0054] In some embodiments, the lodging index is calculated using the following methods: By using skeleton extraction or stem fitting algorithms, the direction vector of the main stem of each crop in the front view image is obtained, and the angle between it and the vertical direction is calculated. This angle is used as the tilt angle of the crop.

[0055] Statistical Number Number of crop plants within the work unit and the plant tilt angle The tilt angle ranges from [0°, 90°], where... This indicates that the crop is basically upright. This indicates that the crop has completely collapsed.

[0056] Set multiple tilt angle thresholds .For example, The specific values ​​can be determined through field trials and can be adjusted appropriately according to different crop varieties or operating areas.

[0057] Establish crop grading rules, including: if Then determine the first The crop is an upright crop, lodging level .like Then determine the first The crop was slightly lodged, lodging level .like Then determine the first The crop was moderately lodged, lodging level .like Then determine the first The crop was severely lodged, lodging grade .

[0058] Based on the tilt angle of each crop, the set tilt angle threshold, and the crop grading rules, the lodging level corresponding to each crop is determined.

[0059] For number Calculate the average lodging level for all crops within the work unit. .

[0060] based on The formula for calculating the average lodging level was obtained.

[0061] The formula for calculating the average lodging level is: .

[0062] in, Number The total number of crops included in the statistics within the work unit. For the first lodging severity of the crop To determine the number of crop plants that are considered upright, To determine the number of crop plants that are considered to have suffered slight lodging, To determine the number of crop plants classified as moderately lodged, The number of crop plants determined to be severely lodged.

[0063] according to Normalize the average lodging level to the [0, 1] interval to obtain the lodging index of the work unit: in, , Number The lodging index of the work unit, The closer to 0, the more upright the crop is as a whole in that work unit. The closer the value is to 1, the higher the proportion of severely lodged crops in that work unit, and the more severe the lodging. This indicates the total number of crops included in the statistics within the work unit. At that time, it can be Set to 0 or record as an invalid value, and it will not participate in subsequent processing.

[0064] In some embodiments, the method for calculating the canopy density includes: When number When the geometric center of the working unit crosses the preset virtual trigger line in front of the cutter table, one or more frames of front view images are acquired. Based on the calibrated camera intrinsic and extrinsic parameters and the geometric dimensions of the cutter table, the effective area of ​​the working unit in the image is cropped out as the working unit image area. ,statistics Total number of pixels .

[0065] Image region of the work unit Convert from RGB space to HSV or Lab space, and extract vegetation areas based on a pre-defined threshold range.

[0066] For example, in the HSV color space, the threshold range for hue H can be set to... The threshold range of saturation S is The threshold range of brightness V is Pixels that fall within the threshold ranges of hue (H), saturation (S), and brightness (V) are identified as leaf pixels, while other pixels that do not meet these threshold ranges are identified as non-leaf pixels.

[0067] Count the number of pixels of the leaf pixels within the statistical work unit area The green pixel coverage ratio is defined as: .

[0068] in, The percentage of green pixels covered. The larger the value, the denser the leaf canopy coverage within the field of view of the working unit. In the simplest implementation, it can be... This serves as an indicator of the canopy density for this work unit.

[0069] When there are complex occlusion situations such as overlapping leaves and messy branches and leaves within the working unit, a texture complexity index is further introduced to improve the ability to respond to complex occlusion situations.

[0070] Specifically, in the area of ​​green pixels, the pixels are converted into grayscale values ​​to form a grayscale image of the leaf area.

[0071] The texture complexity of the green pixel region can be quantized using gray-level co-occurrence matrix entropy, local variance, or other texture operators. For example, using gray-level co-occurrence matrix entropy can yield the raw texture complexity value for that task unit. .

[0072] Determine the empirical minimum value of texture complexity based on field trials or historical samples. and maximum value ,according to ,right Perform linear normalization to obtain the texture complexity index. .

[0073] Among them, when At that time, it can be set ,when At that time, it can be set , , The larger the value, the more complex the leaf texture and the more severe the overlapping and occlusion.

[0074] To simultaneously consider both the degree of leaf canopy coverage and the complexity of occlusion, this application uses a weighted fusion of the green pixel ratio and the texture complexity index to obtain the number. Leaf density of the working unit .

[0075] The formula for calculating the canopy density is: .

[0076] in, The weighting coefficient for the proportion of green pixels covered. The weighting coefficients of the texture complexity index satisfy the following conditions: The weighting coefficients for the green pixel coverage ratio and the texture complexity index are determined experimentally based on actual working conditions.

[0077] , The larger the value, the denser the canopy and the more severe the occlusion by branches and leaves within the field of view of the working unit. In a simplified implementation, when texture complexity is not introduced, it can be set to... Then at this time .

[0078] In some embodiments, the image processing module 103 is further configured to: Extract the grain region from the rear view image.

[0079] Specifically, within the time window Inside, multiple frames of back view images are captured at a fixed frame rate, for example... First, based on the calibration results and installation location of the rear-view industrial camera, determine the effective area for monitoring the incoming grain or fallen grain in the rear view image. Then, crop this area into a monitoring sub-image, denoted as [image name missing]. .right Noise filtering and brightness equalization are performed to reduce the impact of noise and uneven lighting.

[0080] Converting the monitoring sub-image from the RGB color space to the HSV or Lab color space facilitates more stable extraction of the color features of corn kernels.

[0081] For example, in the HSV color space, corn kernels typically appear in the yellow-orange region, with their hue (H), saturation (S), and brightness (V) distributed within a certain range. Therefore, screening criteria for corn kernels, such as kernel hue, can be set based on field-collected sample data. Must meet Grain saturation It needs to be greater than the saturation threshold. Grain brightness It needs to be greater than the brightness threshold. For any pixel in the monitored sub-image When the hue, saturation, and brightness of a pixel meet the screening criteria, the pixel is identified as a candidate grain pixel and assigned a value of 1 in the binary image. Pixels that do not meet the screening criteria are assigned a value of 0, thus obtaining the initial binary image of grain.

[0082] Morphological opening and closing operations are performed on the initial binary image of grain to remove isolated noise points and fill small holes. Furthermore, connected component analysis can be used to remove regions with excessively small areas, resulting in a smooth and continuous binary image of the grain region. The grain region is then masked. To characterize, when any pixel point When identified as a grain-producing area, ,otherwise, .

[0083] The grain region mask is used to define a reference area for calculating impurity content. Its functions include the following: First, it distinguishes the effective grain area in the rear view image from the background, organism structure, and non-detection areas, allowing only the grain to be detected within the target area. First, it counts impurity pixels within a specific area to avoid mistakenly including the machine surface or background in the impurity content. Second, it provides a search range for subsequent impurity identification. Judging impurities only within the grain area helps reduce the probability of false detection and the amount of computation. It also provides a denominator reference for impurity content calculation, that is, the total number of pixels in the grain area serves as the basis for measuring the total material.

[0084] Set the color range of the seed pixels and the color range of the impurity pixels.

[0085] Corn kernels typically appear as yellow-orange in images, with the hue concentrated within a certain range and high saturation. In contrast, impurities are mostly green, brown, or dark brown, such as stalks, leaves, and broken husks, and their hue and saturation distribution differ significantly from that of the kernels.

[0086] For example, in the HSV color space, the feature ranges of the two types of pixels can be defined separately.

[0087] Seed pixel color range set to: Hue saturation ,brightness .

[0088] The color range of the impurity pixels is set separately for green impurities and brown / sepia impurities. Specifically, the color range of the impurity pixels for green impurities is set as follows: Hue and saturation The color range of the impurity pixels for brown / brown impurities is set to: hue. And brightness .

[0089] In the grain region, the sum of pixels within the color range of the grain pixels is calculated as the total grain pixel sum.

[0090] In the grain region, the sum of pixels within the color range of the impurity pixels is calculated as the total impurity pixels.

[0091] For satisfying pixels The pixels are classified based on the relationship between their HSV components and the color ranges of grain and impurity pixels. If the hue, saturation, and brightness of a pixel are all within the color range of grain pixels, then the pixel is classified as a normal grain pixel. If the hue, saturation, and brightness of a pixel fall within the color range of green or brown / tan impurity pixels, then it is classified as an impurity candidate pixel. For pixels with indistinct colors, secondary classification can be performed by combining information from the neighborhood mean, local contrast, etc., to avoid misclassification due to isolated noise.

[0092] Furthermore, pre-trained pixel-level classification models can be introduced, such as decision trees or shallow classifiers, using the color features, brightness features, and neighborhood texture of pixels as input features to classify seeds and impurities.

[0093] Constructing impurity pixel masks When pixel When a pixel is identified as an impurity and is located within the grain area, it is treated as an impurity pixel. .otherwise, Similarly, it can be used for... Morphological processing is performed to remove isolated false-detection pixels and retain connected regions whose area and shape conform to the characteristics of impurities.

[0094] according to The total number of pixels with a statistical value of 1 on the grain area mask is used as the number. The sum of grain pixels in the rear view image corresponding to the work unit.

[0095] according to The total number of pixels with a statistical value of 1 on the impurity pixel mask is used as the number. The sum of impurity pixels in the rear view image corresponding to the working unit.

[0096] The impurity content of the grain is obtained by calculating the ratio of the total number of impurity pixels to the total number of grain pixels.

[0097] grain impurity content The calculation formula is: .

[0098] First, this application avoids including pixels from non-grain areas such as the background and machine body in the statistics by extracting the grain area first and then counting pixels, thus eliminating the interference of non-grain areas on the calculation of grain impurity rate from the source and significantly improving the accuracy of grain impurity rate calculation. Second, it sets the pixel range of grains and impurities based on color space, and uses the inherent color difference between grains and impurities to distinguish them. The method is simple, efficient, and computationally inefficient, adapting to the real-time calculation needs of the harvester's on-board unit. The grain impurity rate is calculated as the ratio of the total number of pixels, realizing the quantitative calculation of grain impurity rate, replacing the traditional subjective evaluation method of manual visual inspection, and providing an objective and quantifiable quality indicator for parameter adjustment. Furthermore, the entire calculation process is based on pixel analysis of the rear view image, requiring no additional hardware sensors, reducing the system's hardware cost and integration complexity.

[0099] In the calculation of grain impurity rate in this application, the main focus is on the proportion of impurity area to grain area. Therefore, either the pixel ratio or the physical area ratio can be used, and the two are equivalent.

[0100] When it is necessary to convert to actual physical area, the actual area coefficient corresponding to a single pixel is obtained based on camera calibration. .

[0101] The formula for calculating the area of ​​impurity pixels is: .

[0102] The total area of ​​the grain-producing region is: .

[0103] Within the backward detection time window corresponding to the work unit, a total of [number] items were selected. Valid image frame, denoted as frame number 1 Frame is ( For each frame of the image, after extracting the grain region and identifying the impurity pixels, the pixel area of ​​the grain region in that frame is obtained. and the pixel area of ​​the noise region in this frame .

[0104] In the time dimension, The pixel areas of the grain region and the pixel areas of the impurity region in the effective frame image are summed to obtain the total pixel area of ​​the grain region and the pixel area of ​​the impurity region accumulated by the working unit.

[0105] The total pixel area of ​​the grain region in multiple frames of images corresponding to the work unit The calculation formula is: .

[0106] The pixel area of ​​the impurity region in the multi-frame image corresponding to the working unit The calculation formula is: .

[0107] The formula for calculating the impurity content of grain is now updated as follows: .

[0108] It should be noted that the statistics here are the statistical average impurity level of the grain flow within the time window, rather than geometric deduplication of the same physical area: the materials (grains and impurities) in the rear camera's field of view are in continuous motion, and different frames correspond to material cross sections at different times. Even if some grains or impurities are repeatedly observed in adjacent frames, they are accumulated synchronously in the cumulative grain area and cumulative impurity area. Therefore, it will not change the statistical expectation value of the grain impurity rate, but only reflects the degree to which the impurity continues to exist in the grain flow within the time window.

[0109] In some embodiments, the image processing module 103 is further configured to: Extract the connected region corresponding to any seed from multiple frames of back view images.

[0110] Obtain the pixel area and aspect ratio of the connected region corresponding to any seed.

[0111] Set area and shape constraints for complete grains.

[0112] A complete seed is defined as a seed whose pixel area satisfies the area constraint of a complete seed and whose aspect ratio satisfies the shape constraint.

[0113] Grains whose pixel area does not meet the area constraints and aspect ratio / shape constraints of a complete grain are considered broken grains.

[0114] Count the number of whole kernels and the number of broken kernels.

[0115] The grain breakage rate is obtained by calculating the ratio of the number of broken grains to the sum of the number of intact grains and the number of broken grains.

[0116] This application extracts grains based on multi-frame rear-view images, avoiding the problem of missed grain flow in single-frame images, achieving comprehensive grain statistics, and improving the completeness of breakage rate calculation. Single grain segmentation is achieved through connected component analysis, elevating breakage rate calculation from regional area statistics to single grain target-level identification, significantly improving the accuracy of breakage rate calculation. Furthermore, grain classification is performed by combining pixel area and aspect ratio dual morphological features, avoiding misclassification of small broken grains as intact grains, and also avoiding misclassification of long and thin broken grains as intact grains, resulting in high classification accuracy. The use of a rule-based classification method with set constraints makes the computational logic simple and highly real-time, adapting to the rapid computational needs of harvester on-board units, without requiring complex deep learning model training and deployment.

[0117] Specifically, in the rear view images captured by the rearward industrial camera, image recognition and classification are performed on individual corn kernel targets, rather than just a rough statistical analysis of the area. Within the rearward detection time window corresponding to a work unit, each kernel is segmented at the target level, features are extracted, and whole / broken kernels are classified through the processing of multiple frames of rear view images, and statistics are performed in the time dimension.

[0118] First, within the backward detection time window of the job unit, select according to the set frame rate. Frame valid image ( Using the camera calibration results, sub-images corresponding to the grain monitoring area are cropped from each frame, such as the elevator outlet or the grain bin inlet section, and the sub-images are preprocessed with noise reduction, brightness equalization, etc.

[0119] Secondly, analyze and process each frame of the image. Obtain the grain area mask Non-grain areas, including the background and organism structure, are removed. Then, impurity pixels are identified within the grain areas to obtain the corresponding impurity mask. In the grain region of this frame, by subtracting the impurity region from the grain region, a binary image of the grain candidate region is obtained. This image is then used to determine the grain candidate region. and The pixel is set to 1 and the rest are set to 0 to obtain a binary image. The binary image is then labeled with connected components to obtain a series of independent connected regions. Each connected region corresponds to a seed or seed fragment.

[0120] Again, according to the first Calculate the pixel area of ​​the connected regions corresponding to each particle target. The aspect ratio is calculated based on the lengths of the major and minor axes of the circumscribed rectangle. Aspect Ratio It reflects whether the particle shape is complete or whether it is a thin, elongated fragment. Furthermore, depending on the implementation requirements, shape features such as roundness and rectangularity can be selected for calculation.

[0121] Based on field trials or calibration results, pre-set area and shape constraints for intact grains. Area constraints include a lower area limit. Shape constraints include upper limit of aspect ratio. The rules are categorized as follows: When and When the area is significantly smaller or the shape is significantly longer or irregular, i.e., it does not meet the area and shape constraints of a complete grain, it is judged as a broken grain or fragment.

[0122] Statistics in this frame Number of whole grains in and the number of broken grains .

[0123] Furthermore, based on the above rules, a simple target detection / segmentation network is introduced, which can further classify each connected region into whole particles / fragmented particles.

[0124] Finally, in the number Within the backward detection time window of the work unit, according to and Calculations show that Number of complete seeds in the frame and the number of broken kernels .

[0125] according to Calculation number Grain breakage rate of the work unit .

[0126] It should be noted that the cumulative statistics here reflect the statistical breakage ratio of the grain flow within that time window. Since the grain flow moves continuously within the monitoring area, different frames correspond to material cross-sections at different times. Even if individual grains are repeatedly observed in adjacent frames, their contribution to the number of intact and broken grains is accumulated synchronously and does not change the statistically expected value of the breakage rate; it only reflects the degree to which the broken particle persists in the grain flow. Therefore, this application uses backward industrial camera images, through connected component analysis and shape feature determination, or target detection / segmentation models, to identify and count individual grain targets as either intact or broken grains, rather than using other hardware counters or merely rough estimations based on region area.

[0127] In some embodiments, the scene feature vector includes ear density, and the image processing module 103 is further configured to: The expected yield of a work unit is obtained by multiplying the ear density, the area of ​​the work unit, and the average weight of a single ear.

[0128] Obtain the instantaneous grain mass flow rate and integrate the instantaneous grain mass flow rate to obtain the grain entering the warehouse.

[0129] The first loss quality is determined based on the expected output and the quality of goods entering the warehouse.

[0130] Obtain the number of lost ears in the work cell from the rear view image.

[0131] The second loss mass is obtained by multiplying the number of lost ears by the average mass of fallen ears.

[0132] The weighted sum of the first and second loss masses yields the comprehensive loss mass.

[0133] The grain loss rate is obtained by calculating the ratio of the total loss mass to the sum of the warehouse quality and the total loss mass.

[0134] This application estimates the expected yield of the operational unit by extracting ear density from the forward-view image, integrating forward scene prediction into the loss rate calculation process. This overcomes the limitation of traditional techniques that rely solely on backward detection to calculate the loss rate, enriching the data sources for loss rate calculation. Simultaneously, it employs two methods—combining theoretical expected yield and grain-entry quality, as well as a grain-fall detection algorithm—to calculate the loss quality separately. A weighted summation is then used to obtain the comprehensive loss quality, balancing the comprehensiveness of theoretical estimation with the accuracy of actual visual detection. This solves the bias problem inherent in single-method loss rate calculations, significantly improving the accuracy and reliability of loss rate calculation. The weighting of the comprehensive loss quality can be dynamically adjusted according to the operational scenario, allowing the loss rate calculation to flexibly adapt to different field operation conditions, improving the system's scenario adaptability.

[0135] Furthermore, the number of lost ears can be directly identified and extracted from the rear view image without the need for additional hardware such as ground loss sensors. This fully utilizes existing modules to achieve simultaneous calculation of multiple quality indicators, further reducing the hardware cost and integration difficulty of the system.

[0136] Specifically, according to Calculate the expected output of the work unit.

[0137] in, This represents the projected output of the work unit. For ear density, This refers to the ground area of ​​the work unit. This represents the average weight of a single ear.

[0138] The average weight of a single ear is a parameter that is pre-calibrated or periodically updated for the current operating conditions, and it is used as a known parameter in the calculation method of this application to calculate the expected yield.

[0139] The average weight of a single ear can be obtained by sampling and weighing corn ears under the current working plot, current variety, and moisture content conditions. For example, several ear samples can be randomly collected, weighed, and the arithmetic mean can be taken as the average weight of a single ear for this working stage. Alternatively, the weight can be assigned based on existing variety test data or typical single ear weights recorded in the database, and the parameter can be corrected according to the necessary sampling results during the actual operation.

[0140] Within the backward detection time window of the work unit, acquire the instantaneous grain mass flow rate. ,according to The quality of the incoming goods in this work unit is calculated.

[0141] in, For the quality of goods entering the warehouse for the work unit, This refers to the instantaneous grain mass flow rate output by the grain flow sensor.

[0142] according to Thus, the first loss mass is obtained.

[0143] in, The first loss mass [kg] of the work unit. Construct the work unit based on the frame geometry in the rear view. The spatial mask area corresponding to the projection is used to count the number of fallen ears / grains and the quality of grains entering the warehouse only within this area. and in accordance with Calculate the second loss mass.

[0144] in, The second loss mass of the work unit, Calculate the number of ears of grain corresponding to the shattered grains for each work unit. This represents the average mass of fallen ears.

[0145] The average mass of fallen ears is a parameter obtained through sampling and calibration for materials such as ears or ear segments that have fallen to the ground. It is used as a known quantity in the algorithm to estimate the amount of loss.

[0146] The average mass of fallen ears is used to characterize the average mass of fallen ears or ear segments that have landed on the ground under current operating conditions. It can be obtained by sampling fallen ears or ear segments on the ground after harvesting within a representative plot, counting and weighing them, and calculating the average mass of each fallen ear or equivalent ear segment as the average mass of fallen ears. Alternatively, an initial value can be given based on existing experimental data or a database, and this parameter can be corrected in subsequent experiments based on field sampling results.

[0147] according to Calculate the overall loss mass .

[0148] in, For the first loss of quality, For the second loss of mass, The weighting coefficient for the first loss quality. This is the weighting coefficient for the second loss quality.

[0149] in, ,and .

[0150] , Adjustable parameters are determined through experiments or historical data, based on factors such as crop variety, moisture content, sensor calibration accuracy, and the reliability of ear drop detection. For example, when the grain mass flow sensor is calibrated accurately and the theoretical yield estimation deviation is small, the parameters can be appropriately increased. When the accuracy of the fallen grain detection device is high and the surface residue better reflects the actual loss, the accuracy can be appropriately increased. .

[0151] In another preferred embodiment, the two weighting coefficients can be adaptively adjusted by combining the historical data of the harvester during operation, so that the loss mass calculation method with smaller deviations can obtain higher weight in subsequent operations.

[0152] according to The grain loss rate was calculated. .

[0153] In some embodiments, the grain impurity rate, grain breakage rate, and grain loss rate are combined to generate a quality index vector for the work unit, which is then stored together.

[0154] Specifically, the quality index vector is .in, For grain loss rate, For the impurity content of grain, For grain breakage rate, This is a vector of quality indicators.

[0155] In some embodiments, the formula for calculating job efficiency is as follows: .

[0156] in, The efficiency of the work unit. The time it takes for the work unit to be completely passed by the cutting table.

[0157] In some embodiments, the actual execution parameters corresponding to the work unit that has completed the task are obtained. .

[0158] in, These are the actual execution parameters corresponding to the work unit. The average travel speed of the work unit. This refers to the fan speed. The rotational speed of the stem-pulling roller. The rotational speed of the screw conveyor. For the gap of the picking plate, The conveying speed before peeling.

[0159] In some embodiments, the template update module is further configured to: The comprehensive evaluation index of the work unit is obtained according to the calculation formula of the comprehensive evaluation index.

[0160] The calculation formula for the comprehensive evaluation index includes: .

[0161] in, This is a weighting coefficient for the grain loss rate. This is the weighting coefficient for the impurity content of grain. This is the weighting coefficient for grain breakage rate. This is a weighting coefficient for work efficiency. This is the normalized value of the grain loss rate. This is the normalized value of the impurity content of the grain. This is the normalized value of the grain breakage rate. This is the normalized value of the work efficiency.

[0162] The four weighting coefficients mentioned above are manually tuned parameters that satisfy... and .

[0163] Grain loss rate, grain impurity rate, grain breakage rate, and operational efficiency are normalized, converting indicators with different dimensions and numerical ranges to the [0, 1] interval. This eliminates quantitative differences between indicators and enables unified quantitative comparison of multi-dimensional indicators, making the calculation of comprehensive evaluation indicators more reasonable and comparable. Furthermore, weight coefficients for each indicator are preset according to actual operational needs, and these coefficients can be flexibly adjusted. For example, in quality-first mode, the weight of quality indicators is increased, and in efficiency-first mode, the weight of operational efficiency is increased, allowing comprehensive evaluation indicators to accurately meet different operational needs and improving the practicality of the evaluation results. The comprehensive evaluation indicators are calculated using a weighted summation method, which has simple calculation logic, high computational efficiency, and strong real-time performance. It does not require complex algorithm models, adapts to the rapid calculation needs of the harvester's onboard terminal, and facilitates subsequent parameter debugging and optimization.

[0164] This application achieves a quantitative comprehensive evaluation of the quality and efficiency of harvester operations through comprehensive evaluation indicators, providing an objective and unified evaluation standard for optimizing and adjusting operation parameters and updating the optimal indicators for different scenario categories. It fundamentally solves the problem that existing technologies have difficulty comprehensively considering multi-dimensional operation indicators and lack a unified judgment basis, providing core support for the adaptive optimization of the entire control system.

[0165] Specifically, calculation , , , The grain loss rate, grain impurity rate, and grain breakage rate are converted to the interval [0, 1].

[0166] in, This is the upper limit of the loss rate. This represents the upper limit of the impurity content. This is the upper limit of the breakage rate. All parameters are pre-tuned to achieve the highest desired efficiency.

[0167] Furthermore, the scenario feature vector, actual execution parameters, and comprehensive evaluation index of the work unit are constructed as a sample record and written into the system's work knowledge base.

[0168] The sample record is represented as follows: .

[0169] Knowledge base updates are represented as: .

[0170] in, This is a sample record for the work unit. This is a sample set for the homework knowledge base.

[0171] To achieve scene adaptation, the system performs scene classification and cluster center updates after writing samples.

[0172] In some embodiments, the parameter adjustment module 104 is further configured to: Based on the calculation formula for scene categories, the scene category to which the unworked work unit belongs is obtained.

[0173] The calculation formula for scene category includes: .

[0174] in, This refers to the scenario category to which the unperformed work units belong. It is the weighted distance between the scene feature vector of the unworked work unit and the feature center corresponding to any scene category.

[0175] The minimum distance criterion based on weighted distance is used to determine the scene category of inactive work units. Compared with traditional Euclidean distance, this method fully considers the differences in the importance of different scene features, making the scene classification results more consistent with actual work scenarios and improving the accuracy of scene classification. Simultaneously, it enables early scene identification of inactive work units, providing a precise basis for subsequently retrieving the optimal work parameter template for the corresponding scene. This allows for proactive adjustment of the harvester's execution parameters, avoiding the passive adaptation problem of traditional technologies and improving the timeliness of parameter adjustments.

[0176] Furthermore, the scene classification process is automatically completed based on the scene feature vectors of the work units and historical clustering results, requiring no manual intervention. This achieves automation and intelligence in scene recognition, reducing the professional requirements for operators. Managing harvester operation parameters by scene category allows for the structured and standardized conversion of scattered and difficult-to-reuse historical operational experience into scene parameter templates. This solves the problems of ineffective transfer of operational experience and the need for repeated trial and error adjustments of parameters in different scenarios, improving the reusability and adjustment efficiency of operational parameters.

[0177] In some embodiments, the parameter adjustment module 104 is further configured to: Constructing the scene feature weight matrix .

[0178] in, Ear height Feature weights, ear density The feature weights, row spacing deviation λ, lodging index θ, and canopy density η, are used to represent the importance of each feature. The weights can be flexibly adjusted according to actual needs.

[0179] For each scene category Set the corresponding feature center as Obtain the feature center corresponding to any scene category.

[0180] The weighted distance is obtained according to the formula for calculating the weighted distance.

[0181] The formula for calculating the weighted distance includes: .

[0182] in, This represents the scene feature vector of the unfinished work units. This is the scene feature weight matrix. is the feature center corresponding to the c-th scene category.

[0183] A diagonal scene feature weight matrix is ​​constructed based on the actual importance of scene features, assigning differentiated weight coefficients to different scene features. This highlights the influence of key scene features and weakens the interference of secondary features, solving the problem of unreasonable distance calculation caused by the equal emphasis on all features in traditional Euclidean distance, thus improving the scientific rigor and rationality of weighted distance calculation. Furthermore, the matrix-based weighted distance calculation formula achieves unified quantitative calculation of multi-dimensional scene features, with rigorous calculation logic and accurate results, providing a reliable quantitative basis for scene category determination. The entire weighted distance calculation process is automatically completed based on the preset weight matrix and historical scene feature centers, requiring no manual intervention. It boasts high calculation efficiency and strong real-time performance, fully adapting to the real-time calculation needs of harvester field operations. Accurate weighted distance calculation provides solid data support for subsequent scene classification, significantly improving the accuracy of scene classification results. Accurate scene classification, in turn, ensures the precise retrieval of subsequent initial execution parameters, forming a virtuous cycle and further improving the adaptability and reliability of the entire control system parameter adjustment.

[0184] In some embodiments, it also includes: The template update module is configured to update the feature center corresponding to the c-th scene category according to the feature center update formula.

[0185] The feature center update formula includes: .

[0186] in, The feature center corresponding to the c-th scene category before the update. The feature center corresponding to the updated c-th scene category, Let be the learning rate for the c-th scene category.

[0187] This application employs a recursive feature center update method, making minor adjustments to the original feature centers only based on the scene feature vectors of newly completed work units. It eliminates the need to re-cluster all historical work data, resulting in low computational cost, simple operation, and suitability for the real-time requirements of online harvester operations. A learning rate is introduced and set to decrease as the number of scene category samples increases, gradually reducing the impact of new samples on the feature centers. This ensures that the scene feature centers are updated in real-time to reflect the latest changes in work scene features while effectively avoiding excessive interference from a few abnormal samples, thus improving the stability and reliability of the feature centers. Online, self-learning dynamic updates of scene feature centers are achieved, allowing the system to continuously adapt to constantly changing field work scenarios, such as differences in different plots, crop growth, and work periods. This significantly improves the system's scenario adaptability and robustness. The dynamic updates of scene feature centers also ensure that subsequent scene classification is always based on the latest clustering results, guaranteeing the accuracy of scene classification and ensuring the adaptability of initial execution parameters for unfinished work units, providing continuous assurance for the stable quality of harvester operations.

[0188] It can decrease as the number of samples in class c increases, allowing the center vector of each scene category to be gradually adjusted with new samples, making it closer to the latest true distribution of that category.

[0189] In some embodiments, the template update module is further configured to: Based on the grain loss rate, grain impurity rate, grain breakage rate, and operational efficiency of the work unit, a comprehensive evaluation index for the work unit is obtained.

[0190] Determine whether the comprehensive evaluation indicators of the work unit are met. If yes, then the optimal comprehensive evaluation index corresponding to the scenario category to which the work unit belongs will be updated to the comprehensive evaluation index of that work unit; otherwise, the optimal comprehensive evaluation index corresponding to the scenario category to which the work unit belongs will remain unchanged.

[0191] in, The optimal comprehensive evaluation index is the one that corresponds to the scenario category to which the work unit belongs. To determine the threshold and prevent frequent template updates caused by minor fluctuations, It serves as a comprehensive evaluation indicator for work units.

[0192] Specifically, the update rules are as follows: , .

[0193] in, Scene category The optimal parameter template vector. This is the optimal comprehensive evaluation index for the current record in this category. If... If the sample is not significantly better than the template in the current assignment knowledge base, the template will not be updated; the sample will only be retained in the assignment knowledge base.

[0194] This application introduces a comprehensive evaluation index as the criterion for judging operational effectiveness. It integrates multiple dimensions of indicators, including grain loss rate, impurity rate, breakage rate, and operational efficiency, achieving a comprehensive assessment of harvester operational quality and efficiency. This effectively avoids the problem of existing technologies where a single optimal indicator results in poor overall operational effectiveness, ensuring optimal overall operational results. By setting a judgment threshold as the trigger condition for updating the optimal comprehensive evaluation index, frequent updates of the optimal index due to minor index fluctuations are effectively avoided, reducing invalid update operations, lowering the system's computational load, and improving system stability. The optimal index is only updated when the comprehensive evaluation index of a new operational unit is significantly better than the optimal index for the current scenario category. This ensures that the optimal comprehensive evaluation index for each scenario category always represents the best operational effect in that scenario, providing an objective and reliable evaluation basis for subsequent updates to operational parameter templates. Furthermore, it enables online self-learning updates of the optimal comprehensive evaluation index for scenario categories, allowing the system to continuously accumulate optimal operational experience during operation, constantly optimizing the criteria for judging operational effectiveness in different scenarios, and improving the overall quality and efficiency of harvester operations in various scenarios.

[0195] When determining the initial execution parameters of any task unit, based on the scene feature vector of the unworked task unit and the scene category to which the task unit belongs, the optimal parameter template vector for that scene category is found and used as the initial execution parameters of the task unit.

[0196] Specifically, when the harvester is about to enter any work unit, the scene features of the current work unit are extracted based on the front view image. The scene category corresponding to the current work unit is calculated. The system reads the optimal parameter template vector for the corresponding category from the system's job knowledge base and uses this parameter template vector as the initial execution parameter for the current job unit. .

[0197] in, These are the initial execution parameters for the current job unit. Scene category The optimal parameter template vector.

[0198] In some embodiments, the parameter adjustment module 104 is further configured to: Set the correspondence between the adjustment amount, the quality error vector, and the engine load margin.

[0199] Based on the grain loss rate, grain impurity rate, and grain breakage rate, a current quality error vector is constructed for the work unit. This quality error vector includes errors in grain loss rate, grain impurity rate, and grain breakage rate.

[0200] The current engine load margin of the harvester is obtained by subtracting the ratio of the current engine output power to the rated engine power.

[0201] Based on the correspondence, the adjustment amount applied to the harvester corresponding to the current quality error vector and the current transmitter load margin is obtained.

[0202] This application constructs a quality error vector to integrate and consider the deviations of three core quality indicators: grain loss rate, grain impurity rate, and grain breakage rate. This enables multi-indicator synergistic guidance for operation parameter adjustment, effectively avoiding the problem of deterioration of other quality indicators caused by single-indicator adjustment in existing technologies, and ensuring the comprehensive optimization of harvester operation quality.

[0203] By incorporating engine load margin into the calculation process of adjustment amount as a hard constraint condition for parameter adjustment, it is ensured that all adjustments to operating parameters are carried out within the safe operating range of the engine, avoiding problems such as engine overload and failure caused by blindly increasing parameters, thereby improving the safety of harvester operation and the stability of equipment operation.

[0204] The adjustment range is determined by various deviations in actual operation quality and the actual load capacity of the harvester engine. This ensures that the adjustment of operation parameters is both precisely aligned with the actual needs for improving operation quality and matches the harvester's own operation capabilities. This fundamentally solves the problem of unreasonable parameter adjustments and poor operation results caused by existing technologies neglecting the equipment's own capabilities.

[0205] Specifically, when number The work unit is completed, and the quality indicators are obtained. and warehousing quality Then, for the number The execution parameters of the work unit are fine-tuned.

[0206] according to Construct the quality error vector .

[0207] according to Construct engine load margin .

[0208] in, For the error of grain loss rate, To account for the error in grain impurity content, To account for the error in grain breakage rate, This serves as a reference upper limit for grain loss rates. This serves as a reference upper limit for the impurity content of grains. This serves as a reference upper limit for grain breakage rate. The engine load margin for the current work unit, if This indicates full load. This represents the current engine output power. This refers to the engine's rated power.

[0209] All of these are target control levels selected based on standards and field trials.

[0210] Taking vehicle speed as an example, set the allowable adjustment step size. It is pre-tuned based on the overall power and driving experience of the machine. These parameters are selected offline or configured according to the machine's power performance, control cycle, and operating mode. They are used to limit the range of single speed adjustment and ensure sufficient responsiveness of speed adjustment without causing significant impact.

[0211] Specifically, the method for setting the adjustable step size includes: First, determining a reasonable speed change per unit time by referring to the engine power margin and the acceleration / deceleration capacity allowed by the transmission system. Second, combining the control cycle or the duration of the work unit to ensure that speed changes on that time scale do not cause a shock to the driver, while allowing the speed to be brought to a suitable range within a few adjustments. Finally, different values ​​can be selected when the harvester is in different operating modes, such as in the quality-priority mode. The set value is relatively small to allow for smoother adjustments, in efficiency-first mode. The set value is relatively large in order to speed up the convergence.

[0212] The correspondence can be set as follows: .

[0213] in, This is the speed increment for the next work unit. The tolerance band difference for grain loss rate error. This is the tolerance band difference for the grain impurity content error. The tolerance band difference for grain breakage rate error, the tolerance band difference for grain loss rate error, the tolerance band difference for grain impurity rate error, and the tolerance band difference for grain breakage rate error are all manually tuned control parameters. This is the minimum allowable value for engine load margin.

[0214] The tolerance bands for grain loss rate error, grain impurity rate error, and grain breakage rate error can be set using the following methods.

[0215] First, multiple sets of stable operation tests were conducted under typical working conditions. Without adjusting parameters, the data from multiple work units were continuously recorded. Observe their natural fluctuation range under normal and acceptable conditions, for example, the grain loss rate fluctuating between 0.8% and 1.2%. Secondly, based on this natural fluctuation and the measurement error of sensor / visual inspection, define a fixed range for each indicator, for example... This can be taken as the upper limit of reference. A certain percentage, or a multiple of the standard deviation of multiple trials. Similarly, if the indicator deviates... , , Falling into the corresponding tolerance zone Within the tolerance range, fluctuations are considered normal and do not trigger parameter adjustments. Only when the deviation exceeds the tolerance band will incremental adjustments with a limited step size be triggered. The tolerance band can also be tightened or loosened accordingly under different operating modes.

[0216] This is a manually set control threshold, representing the minimum safe value of the engine load margin. Its purpose is to ensure that the engine is not subjected to prolonged extreme operating conditions during automatic parameter adjustment. Based on the engine's rated power, torque characteristics, and the manufacturer's recommended safe operating range, first determine a load range for long-term operation. For example, the actual load generally does not exceed 80% to 90% of the rated load. Then, convert this range into a load margin form and take a slightly conservative lower limit as the load margin. If the real-time calculated load margin is lower than If the speed is too high or the workload is too heavy, adjustments to reduce the load are permitted. This applies to different machine models and operating modes. It can be set to different values.

[0217] During operation, the harvester also requires adjustments to the actuators such as the stalk-pulling roller, the screw conveyor, the gap between the ear-picking plates, and the lifting speed. Similar rules can be formulated for each actuator to ensure that the increment of each actuator is a finite set of three values. . This refers to the step size parameter for each execution variable. When setting the step size parameter, physical and safety constraints, convergence speed, and perceptibility need to be considered.

[0218] Physical and safety constraints dictate that single adjustments should not be too large to avoid causing significant impact or exceeding the mechanism's allowable range. For example, the vehicle speed should only be adjusted by 0.1–0.3 m / s each time, the fan speed by 20–50 r / min each time, and the gap between the picking plates by 1–2 mm each time.

[0219] In terms of convergence speed and perceptibility, a step size that is too small will require many work units to adjust to the appropriate range, while a step size that is too large will cause the work status to fluctuate. Therefore, the step size should be selected according to a certain proportion of the effective adjustment range of each execution quantity. For example: ,in, These are empirical coefficients, determined through field trials or empirical tuning.

[0220] A fixed set of parameters can also be pre-defined for different execution volumes. While remaining unchanged under the same machine model and operating mode, different step length tables are configured according to operating modes such as quality priority and efficiency priority when flexible adjustments are required. This ensures that the entire adjustment process can achieve sufficient adjustment effect while maintaining smoothness.

[0221] according to and Adjusting the number The actual execution parameters of the work unit.

[0222] in, For scene category The initial execution parameters, For scene category The optimal parameter template vector, The adjustment amount for the j-th execution quantity, Number The actual execution parameters of the working unit. Vehicle control module 105 After applying upper / lower limits, the instruction becomes the target instruction for the next work unit.

[0223] The harvester's adaptive control system for operational quality also includes a dual-vision operational quality monitoring module, an operational condition and location information acquisition module, an operational unit calibration and quality assessment module, a scene knowledge base and parameter template management module, an operational quality adaptive control module, an actuator drive and interface module, a human-machine interaction and data storage module, and an onboard communication and power supply guarantee module. Among these, the image acquisition module 102 and the image processing module 103 rely on the dual-vision operational quality monitoring module. The dual-vision operational quality monitoring module includes a forward-looking operational scene perception submodule and a backward-looking operational quality detection submodule, used to acquire operational scene characteristics and quality information such as impurity rate, breakage rate, and loss rate. The operational condition and location information acquisition module is used to collect basic operational condition data such as travel speed, engine condition, grain flow rate, and GNSS location information. The area division module 101 relies on the operational unit calibration and quality assessment module, which is used to divide the operational path into operational units and complete the calculation of yield, loss rate, and comprehensive evaluation for each unit. The scenario knowledge base and parameter template management module stores historical samples of "operation scenario - operation parameters - operation quality" and maintains parameter templates for different scenarios. The parameter adjustment module 104 relies on the operation quality adaptive control module, which generates adaptive adjustment commands for actuators such as vehicle speed, fan, and pull roller based on template parameters output from the knowledge base and the current unit quality evaluation results. The vehicle control module 105 relies on the actuator drive and interface module, which is responsible for sending control commands to the walking system and various operating components. The human-machine interaction and data storage module displays the operation status, sets control parameters, and records operation data. The vehicle communication and power supply module provides data communication channels and stable power support for the above modules.

[0224] The dual-vision operation quality monitoring module includes a forward-looking operation scene perception submodule and a backward-looking operation quality detection submodule, which together constitute the core device for online monitoring and testing of operation quality. The forward-looking operation scene perception submodule includes a forward-looking industrial camera, lens, and dustproof, waterproof, and shockproof mounting structure installed outside the cab, approximately 1.8m above the ground. The forward-looking camera's field of view covers the operation area a certain distance in front of the header. The harvester's central controller is positioned at a preset distance in front of the header. A virtual trigger line is set at the work unit. When the geometric center of the work unit passes through this trigger line, image acquisition is triggered. The front view image is processed by the built-in or external embedded computing unit to complete row and column structure recognition, ear target detection, plant height and ear height estimation, lodging degree estimation, and leaf canopy density analysis, and output scene feature vectors to provide input for subsequent work unit calibration and scene classification. The backward operation quality detection submodule includes a backward industrial camera installed at the grain fall observation window on the side of the grain silo. This camera acquires multiple frames of images within the time window issued by the central controller. The field of view covers the grain entering the silo or the grain fall area. The backward image is used to segment the grain area, identify the impurity area, classify whole grains / broken grains, and identify and count fallen ears / grains to obtain the grain impurity rate, grain breakage rate, and ear fall count of the work unit. This data is combined with the output of the grain flow sensor to calculate the loss mass and grain loss rate. The dual-vision operation quality monitoring module works in conjunction with the work unit calibration and quality assessment module to realize non-contact online monitoring and calibration of the operation quality of each work unit.

[0225] The working condition and location information acquisition module includes a travel speed sensor, an engine speed and torque acquisition unit, a grain flow sensor, a GNSS antenna, and status sensors for each actuator, which are used to collect basic working condition data of the harvester's operating status and working environment.

[0226] The system includes a travel speed sensor for real-time data acquisition. An engine speed and torque acquisition unit reads ECU data via the CAN bus to calculate the current engine output power and load margin. A grain flow sensor, installed at the elevator or grain bin inlet, outputs the instantaneous grain mass flow rate, which is integrated over time to obtain the grain mass entering the bin for each work unit. A GNSS antenna acquires the harvester's real-time position and trajectory, providing a basis for work unit division and spatial distribution of work mass. Status sensors for each actuator acquire data such as the stalk-pulling roller speed, screw conveyor speed, waste-removal fan speed, pre-peeling elevator speed, and the gap position of the ear-picking plate. The work condition and position information acquisition module connects to the central controller via the CAN bus, providing real-time work condition input for the control algorithm.

[0227] like Figure 2 As shown, the harvesters used in this system include corn combine harvesters. The direction of travel of the corn combine harvester is the front, and the opposite direction is the rear. A forward-looking industrial camera is set in front of the corn combine harvester, and a rearward-looking industrial camera is set in the rear of the corn combine harvester. A speed sensor, a GNSS antenna, and a grain flow sensor are respectively set on both sides of the corn combine harvester.

[0228] The work unit calibration and quality assessment module, deployed within the central controller's internal software, is a key functional unit of this system. Its main task is to divide the field area along the work path into discrete work units and perform work quality calibration and comprehensive evaluation at the work unit scale. It primarily divides the work path into work units according to a preset length based on the travel distance and the effective width of the header, and records the time and GNSS coordinates of each unit entering the header. It calculates the corresponding material transport time delay using travel speed, transport path length, and additional lag time to obtain the reference time and backward detection time window. It calculates the expected yield by combining scene feature vectors, the ground area of ​​the work unit, and the average weight of a single ear, and calculates the grain storage quality based on the grain flow sensor output. It obtains the loss quality and grain loss rate through weighted estimation of multi-source losses. It normalizes the grain loss rate, grain impurity rate, grain breakage rate, and work efficiency, calculates the comprehensive evaluation index according to the corresponding weighting coefficients, and outputs the quality-efficiency evaluation results for each work unit.

[0229] The scene feature vector and grain loss rate output by the work unit calibration and quality assessment module grain impurity content Data such as grain breakage rate, operational efficiency, and comprehensive evaluation indicators will be used for updating the operational knowledge base and making control decisions.

[0230] The scenario knowledge base and parameter template management module, deployed within the central controller, is the core component for achieving self-learning functionality. Its main function is to store sample records composed of the scenario feature vector, actual execution parameters, and comprehensive evaluation indicators for each work unit into the work knowledge base, forming a historical dataset of work scenarios, work parameters, and work quality results. Based on weighted Euclidean distance, scenario cluster centers, and feature weight matrices, new sample records are classified into work scenarios, and the scenario centers for each category are updated recursively. An optimal parameter template vector and corresponding optimal comprehensive evaluation are maintained for each scenario category c. When the evaluation indicators of a new sample record in that category are significantly better than the current template performance, the parameter template for that category is automatically updated, enabling online self-learning and optimization of recommended work parameters for different scenarios. The knowledge base and parameter templates are permanently stored in non-volatile memory, enabling experience accumulation and reuse across plots and years.

[0231] The scene knowledge base and parameter template management module provide a basis for subsequent adaptive control, enabling the system to have the ability of scene recognition, template calling, and continuous optimization.

[0232] The adaptive control module for job quality is the core of the entire system. It implements a two-layer collaborative control strategy, namely parameter template feedforward control and rule-based incremental feedback control.

[0233] The parameter template feedforward control involves reading the scene feature vector when the harvester is about to enter a new work unit, calling the scene knowledge base and template management module to perform scene identification, and obtaining the scene category to which the current work unit belongs. The optimal operation parameters are read from the parameter template of the scene category as the initial execution parameters of the work unit, and then distributed through the actuator driver and interface module to quickly retrieve the parameter combination suitable for the current scene from historical experience.

[0234] Rule-based incremental feedback control involves generating finite-step adjustment quantities for parameters such as vehicle speed, fan speed, and stalk-pulling roller speed according to a preset rule table after the completion of a work unit and quality evaluation. This is based on the errors of grain loss rate, grain impurity rate, and grain breakage rate relative to reference values, as well as the engine load margin. The initial execution parameters and adjustment quantities are then combined to obtain the actual execution parameters for the next work unit. This is implemented through the actuator drive and interface module, employing a discrete, finite-step rule-based control approach. This eliminates the need for fuzzy inference and PID control, facilitating engineering debugging and parameter tuning. Through this two-layer collaborative control, the system achieves online adaptive optimization and adjustment of work quality based on fully utilizing historical experience.

[0235] The actuator drive and interface module is responsible for converting the parameter settings output by the controller into specific drive commands for each actuator, including: travel drive, waste removal fan drive, stem pulling roller drive, screw conveyor drive, pre-peeling lifting device drive, and ear-picking plate adjustment mechanism. Specifically, the travel drive adjusts the hydrostatic transmission or gearbox control commands to regulate the travel speed. The waste removal fan drive controls the fan motor or hydraulic motor to regulate the fan speed. The stem pulling roller drive controls the stem pulling roller speed. The screw conveyor drive controls the conveyor speed. The pre-peeling lifting device drive controls the lifting belt speed. The ear-picking plate adjustment mechanism drives the motor to adjust the ear-picking plate gap. The actuator drive and interface module communicates with the vehicle's existing electronic control unit via a CAN bus to ensure that each actuator operates within safe ranges during system adjustments, such as speed limits, torque limits, and protection against frequent start-stop cycles.

[0236] The human-machine interaction and data storage module mainly includes: an industrial touchscreen or vehicle-mounted display, a simple parameter setting interface, an alarm and prompt unit, and a data storage unit. The industrial touchscreen or vehicle-mounted display shows in real time the current work scenario category, parameter template number, grain impurity rate, grain breakage rate, grain loss rate, curves of comprehensive evaluation indicators for each work unit, and current execution parameters and adjustments. The simple parameter setting interface allows the driver to select different work modes and adjust weights. The alarm and prompt unit provides visual or audible alarms when the grain loss rate, grain impurity rate, or engine load exceeds the set range. The data storage unit stores key work data and knowledge base incremental update records locally, providing a data foundation for subsequent analysis and parameter optimization.

[0237] The vehicle communication and power supply module provides communication and power support for all parts of the system, including: the vehicle CAN bus, DC / DC power module, wiring harness, and connectors. The vehicle CAN bus is used for real-time data exchange between the central controller and sensor nodes and actuator drive units. The DC / DC power module converts the vehicle's power supply system voltage to the voltage levels required by industrial cameras, computing units, displays, etc., while providing overvoltage, undervoltage, and short-circuit protection. The wiring harness and connector design ensures long-term reliable operation of the system in vibration, high humidity, and high dust environments.

[0238] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0239] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. An adaptive control system for the operation quality of a harvester, characterized in that, include: The area division module is configured to divide the harvester's operating area into multiple operating units; The image acquisition module is configured to acquire a front view image of the harvester in the direction of travel of the harvester when the harvester is working in any of the work units; Obtain a rear view image of the harvester in the opposite direction to the stated direction of travel; The image processing module is configured to analyze the front view image obtained in any of the said work units to obtain the scene feature vector of the work unit; By analyzing the rear view image obtained in any of the aforementioned work units, the grain impurity rate, grain breakage rate, and grain loss rate of the harvester in that work unit are obtained. The parameter adjustment module is configured to calculate an adjustment amount based on the grain loss rate, grain impurity rate, and grain breakage rate of the completed work unit; determine the initial execution parameters of the harvester in the work unit based on the scene feature vector of the non-operational work unit and the scene category to which the work unit belongs; and calculate the sum of the adjustment amount and the initial execution parameters corresponding to the work unit when the harvester enters the work unit to obtain the actual execution parameters of the work unit. The vehicle control module is configured to control the harvester to operate in the work unit corresponding to the actual execution parameters according to the actual execution parameters; The parameter adjustment module is further configured as follows: Establish the correspondence between the adjustment amount and the mass error vector and the engine load margin; Based on the grain loss rate, the grain impurity rate, and the grain breakage rate, a current quality error vector for the work unit is constructed; wherein, the quality error vector includes grain loss rate error, grain impurity rate error, and grain breakage rate error; The current engine load margin of the harvester is obtained by subtracting the ratio of the current engine output power to the rated engine power. Based on the correspondence, the adjustment amount applied to the harvester corresponding to the current quality error vector and the current transmitter load margin is obtained; The parameter adjustment module is further configured to: According to the calculation formula for the scene category, the scene category to which the unworked work unit belongs is obtained; The calculation formula for the scene category includes: ; in, The scene category to which the unperformed work unit belongs. The weighted distance between the scene feature vector of the unperformed work unit and the feature center corresponding to any scene category; The parameter adjustment module is further configured to: Construct the scene feature weight matrix; Obtain the feature center corresponding to any of the aforementioned scene categories; The weighted distance is obtained according to the weighted distance calculation formula. The formula for calculating the weighted distance includes: ; in, The scene feature vector for the unperformed task unit. The scene feature weight matrix is... The feature center is the one corresponding to the scene category of the cth class.

2. The adaptive control system for harvester operation quality according to claim 1, characterized in that, The image processing module is further configured to: Extract the grain region from the rear view image; Set the color range of the seed pixel and the color range of the impurity pixel; In the grain region, the sum of pixels within the color range of the grain pixels is calculated as the total grain pixels; In the grain region, the sum of pixels within the color range of the impurity pixels is calculated as the total impurity pixels; The impurity content of the grain is obtained by calculating the ratio of the total number of impurity pixels to the total number of grain pixels.

3. The adaptive control system for harvester operation quality according to claim 1, characterized in that, The image processing module is further configured to: Extract the connected region corresponding to any seed from multiple frames of the rear view image; Obtain the pixel area and aspect ratio of the connected region corresponding to any seed grain; Set area and shape constraints for complete grains; A complete seed is defined as a seed whose pixel area satisfies the area constraint of the complete seed and whose aspect ratio satisfies the shape constraint. Seeds whose pixel area does not meet the area constraints of the complete seed, the aspect ratio, or the shape constraints are considered broken seeds. Count the number of intact grains and the number of broken grains; The grain breakage rate is obtained by calculating the ratio of the number of broken grains to the sum of the number of intact grains and the number of broken grains.

4. The adaptive control system for harvester operation quality according to claim 1, characterized in that, The scene feature vector includes ear density, and the image processing module is further configured to: The expected yield of the operation unit is obtained by multiplying the ear density, the area of ​​the operation unit, and the average weight of a single ear. Obtain the instantaneous grain mass flow rate and integrate the instantaneous grain mass flow rate to obtain the grain entering the warehouse. The first loss mass is obtained based on the expected output and the incoming quality; Obtain the number of lost ears in the work unit from the rear view image; The second loss mass is obtained by multiplying the number of lost ears by the average mass of fallen ears. The first loss quality and the second loss quality are weighted and summed to obtain the comprehensive loss quality; The grain loss rate is obtained by calculating the ratio of the total loss mass to the sum of the warehouse entry mass and the total loss mass.

5. The adaptive control system for harvester operation quality according to claim 1, characterized in that, Also includes: The template update module is configured to update the feature center corresponding to the scene category of the cth class according to the feature center update formula; The feature center update formula includes: ; in, The feature center corresponding to the scene category of class c before the update. The feature center corresponding to the updated c-th scene category, is the learning rate for the scene category described in class c.

6. The adaptive control system for harvester operation quality according to claim 5, characterized in that, The template update module is further configured as follows: Based on the grain loss rate, grain impurity rate, grain breakage rate, and operational efficiency of the operational unit, a comprehensive evaluation index for the operational unit is obtained. Determine whether the comprehensive evaluation index of the work unit meets the requirements. If yes, then the optimal comprehensive evaluation index corresponding to the scenario category to which the work unit belongs is updated to the comprehensive evaluation index of the work unit; otherwise, the optimal comprehensive evaluation index corresponding to the scenario category to which the work unit belongs remains unchanged. in, The optimal comprehensive evaluation index is the one corresponding to the scenario category to which the work unit belongs. To determine the threshold, This is the comprehensive evaluation index for the work unit.

7. The adaptive control system for harvester operation quality according to claim 6, characterized in that, The template update module is further configured as follows: The comprehensive evaluation index of the work unit is obtained according to the calculation formula of the comprehensive evaluation index. The calculation formula for the comprehensive evaluation index includes: ; in, This is the weighting coefficient for the grain loss rate. This is the weighting coefficient for the impurity content of the grain. This is the weighting coefficient for the grain breakage rate. This is a weighting coefficient for work efficiency. This is the normalized value of the grain loss rate. This is the normalized value of the impurity content of the grain. This is the normalized value of the grain breakage rate. This is the normalized value of the work efficiency.

Images