Harvester header height pre-judgment type adjusting system and method and harvester

Abstract

The invention discloses a harvester header height pre-judgment type adjusting system and method and a harvester, and relates to the technical field of harvester header height adjusting.In the system, a look-ahead sensing module is used for detecting terrain data in real time and generating a terrain profile map according to the terrain data in front of the harvester; the multi-parameter acquisition module is used for acquiring the vehicle speed of the harvester and the crop density and crop humidity data in front of the harvester in real time; the central control module is used for receiving and performing multi-parameter fusion and pre-judgment control on the terrain profile map, the vehicle speed, the crop density and the crop humidity data to generate a hydraulic control instruction; the hydraulic execution module is used for receiving the hydraulic control instruction and adjusting the header height of the harvester according to the hydraulic control instruction. According to the method, the problems of sensing range limitation and hydraulic delay are solved through terrain change pre-judgment and multi-parameter fusion control, and the harvesting quality and the operation efficiency under the complex working condition are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of harvesting header height adjustment, and particularly relates to a harvesting header height pre-judgment type adjustment system and method and a harvester. BACKGROUND

[0002] The height adjustment of the header of an agricultural harvester is a key link to ensure the quality and efficiency of harvesting, and its core function is to ensure that the crop is not missed (to avoid the crop not being cut due to the header being too high) and the machine is not damaged or too many impurities are mixed (to avoid the header touching the ground due to being too low) when cutting the crop. The traditional height adjustment of the header relies on manual operation by the driver, and the driver manually controls the hydraulic handle to realize the lifting of the header by observing the growth state of the crop and the changes in the terrain. With the improvement of the level of agricultural mechanization, automatic adjustment technology is gradually applied, and its basic principle is to collect terrain or crop information in real time through sensors, transmit the information to the controller, and drive the actuator (mostly a hydraulic system) to complete the height adjustment. Among them, the physical mechanical sensor (such as a ground-touching roller or a contact-type angle sensor) is the core sensing component of the early automatic adjustment system, which provides raw data for the controller by directly contacting the ground height. In recent years, there have been technical solutions such as the invention patent with the publication number CN120202829A and the subject name "a corn off-row harvester header intelligent lifting device", which uses a laser radar sensor and an ultrasonic sensor as a sensing means, processes data through a controller, and drives a mechanical transmission structure to realize the lifting of the header, and also contains auxiliary structures such as an anti-winding component and a lubricating component. The hydraulic control system is the core of the header height adjustment, and its working principle is to provide pressure oil through a hydraulic pump, adjust the flow and direction through a control valve, drive the hydraulic cylinder to extend and retract, and then drive the header to lift. The response speed of the system is affected by factors such as the viscosity of the hydraulic oil, the precision of the control valve, the size of the load, and the feedback time delay of the control system, and there is usually a certain action delay, which is a common characteristic of hydraulic transmission technology in fast response scenarios.

[0003] The current mainstream header height self-adaptive adjustment technology takes "real-time detection-immediate response" as the core logic, and its typical structure includes: 1. Perception module: mainly adopts physical mechanical sensors (such as ground-touching roller type or contact type angle sensor) installed under the header, which converts the mechanical displacement generated by the contact of the roller with the ground into an electrical signal transmitted to the controller, directly reflecting the current terrain height under the header. 2. Control module: after the controller receives the real-time signal of the sensor, it compares it with the preset target height (such as 10-15 cm above the bottom of the crop), calculates the height deviation, and generates a hydraulic control instruction (such as the length and direction of the control of the hydraulic cylinder). 3. Execution module: after the hydraulic system receives the instruction, it switches the oil circuit through the electromagnetic reversing valve to drive the hydraulic cylinder to act, completing the adjustment of the header height, and the sensor continuously feeds back the current height during the adjustment process, forming a closed-loop control. Although this patent uses advanced sensing means such as laser radar and ultrasonic sensor, its core control logic still belongs to the "real-time detection-immediate response" mode. The same technology of imported agricultural machinery improves the response speed by optimizing hydraulic components (such as using high-precision electro-hydraulic proportional valve) and sensors, but the core logic is still "real-time detection-immediate response", with the following shortcomings: First, the perception range is limited to the area directly below or very close to the header, and only the terrain that has been contacted or will be contacted can be responded to; second, the hydraulic system has inherent action delay (usually 0.5-2 seconds), when the harvester is running at a high speed (such as ≥5 km / h) or the terrain is highly undulating, the time for the hydraulic system to complete the adjustment action after the sensor detects the terrain change is within the time for the header to travel to that terrain area with the machine, resulting in adjustment lag and problems such as header touching the ground and missing cutting. Although this patent and the technology of imported agricultural machinery improve the response speed by optimizing hardware performance, they do not fundamentally solve the problem of the time difference between hydraulic delay and terrain change, and their technical solutions require high hardware performance and are expensive, making it difficult to popularize in domestic low-end models. The existing technology also has the problem of single control logic, which only generates adjustment instructions based on the current terrain height deviation, without considering key factors such as the running speed of the harvester (the faster the speed, the farther the distance traveled in the same time, the more significant the lag impact), the density of the crop (high-density crops may hinder sensor detection, miscontact the sensor, or require a higher header height to avoid entanglement), and the humidity of the crop (high-humidity crop straw is strong in toughness, and a too-low header may cause straw accumulation), resulting in large scene influence on adjustment accuracy. At the same time, the compensation mechanism for the action delay of the hydraulic system is missing, and only by optimizing hardware (such as high-precision valve group) to improve response speed, the cost is high and the effect is limited, and the negative effects of delay cannot be fundamentally eliminated.

[0004] In summary, the existing header height self-adaptive adjustment technology is difficult to meet the operation requirements under high speed and complex working conditions due to limited sensing range, significant hydraulic delay, single control parameter and lack of effective compensation mechanism. Therefore, there is an urgent need for a new header height adjustment scheme that can realize forward-looking perception, multi-parameter fusion and predictive control, and effectively compensate for system delay, so as to improve the adaptability, operation quality and efficiency of the harvester. SUMMARY

[0005] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art. Specifically, a harvester header height predictive adjustment system, method and harvester are provided, as follows. 1) In a first aspect, the present application provides a harvester header height predictive adjustment system, and the specific technical solutions are as follows: It comprises a forward-looking perception module, a multi-parameter acquisition module, a central control module, a hydraulic execution module and a human-computer interaction module. The forward-looking perception module is used to detect and generate a terrain profile according to the terrain data in front of the harvester in real time. The multi-parameter acquisition module is used to acquire the speed of the harvester, the crop density and the crop humidity data in front of the harvester in real time. The central control module is used to receive and perform multi-parameter fusion and predictive control on the terrain profile, the speed, the crop density and the crop humidity data, and generate a hydraulic control instruction. The hydraulic execution module is used to receive and adjust the header height of the harvester according to the hydraulic control instruction.

[0006] The harvester header height predictive adjustment system provided by the present application has the following beneficial effects: The terrain profile generated by the forward-looking perception module can accurately predict the terrain changes in the distance, effectively overcoming the limitation of the traditional sensing range. The speed, crop density and humidity data are acquired in real time by the multi-parameter acquisition module, and the multi-parameter fusion and predictive control are performed by the central control module, which improves the adaptability and adjustment accuracy of the system under different working conditions. Through compensation and predictive control of the hydraulic system action delay characteristics, the adjustment lag phenomenon caused by the inherent delay of the hydraulic execution module is effectively reduced, and the header ground contact or missed cutting problem is avoided, thereby improving the harvesting quality and operation efficiency under high speed and complex terrain operation conditions.

[0007] Based on the above-mentioned scheme, the harvester header height predictive adjustment system of the present application can also be improved as follows.

[0008] Further, the forward-looking perception module comprises a radar, an ultrasonic sensor and a preprocessing unit, and the forward-looking perception module detects the terrain data in front of the harvester in real time through the radar and the ultrasonic sensor. The preprocessing unit filters and stitches together the terrain data in front of the harvester, which is detected in real time by radar and ultrasonic sensors, to generate a terrain contour map.

[0009] The beneficial effects of adopting the above-mentioned further scheme are as follows: by combining radar and ultrasonic sensors, multi-source acquisition of terrain data in front of the harvester is realized, expanding the detection range and reliability; the preprocessing unit filters the raw terrain data, effectively eliminating environmental noise and measurement errors, improving data quality; and then the multi-sensor data is fused through stitching to generate a continuous and accurate terrain contour map, providing the system with an accurate and complete foundation of terrain information in front, enhancing the system's ability to perceive and predict terrain changes.

[0010] Furthermore, the horizontal axis of the terrain contour map represents the horizontal distance forward, and the vertical axis represents the height difference between the ground and the cutter reference plane; the preprocessing unit is specifically used for: After filtering out clutter from the terrain data obtained by radar and ultrasonic sensors, a weighted fusion is performed in the overlapping area to generate a terrain contour map with the cut-off reference plane as the zero point.

[0011] The beneficial effects of adopting the above-mentioned further scheme are: the filtering process effectively eliminates environmental noise and abnormal data points, improving the accuracy and reliability of terrain data; by performing weighted fusion in the overlapping area detected by the sensors, the complementary advantages of multi-source data are fully utilized to generate a continuous, consistent and accurate terrain height curve; the height representation method with the cutter reference plane as the zero point enables the generated terrain contour map to directly and intuitively reflect the relative height relationship between the terrain ahead and the cutter, providing an accurate and reliable data foundation for subsequent height prediction and control.

[0012] Furthermore, the multi-parameter acquisition module includes a camera installed above the header, an infrared humidity sensor installed near the cutting blade on the header, and a data processing unit; The data processing unit is used to: perform image recognition on images captured by the camera in front of the harvester to obtain crop density; The data processing unit is also used to: obtain crop humidity by analyzing the infrared reflectance of crop straw in front of the harvester detected by the infrared humidity sensor.

[0013] The beneficial effects of adopting the above-mentioned further solutions are as follows: real-time acquisition and recognition of crop images in front by camera enables accurate and non-contact acquisition of crop density data, providing important operating parameters for header height control; analysis of the infrared reflectivity of crop straw detected by infrared humidity sensor enables real-time and accurate acquisition of crop humidity information; the synchronous processing of multi-source information by data processing unit realizes parallel acquisition and fusion of crop density and humidity data, providing the system with comprehensive and accurate crop status information, and enhancing the system's adaptability and adjustment accuracy to complex crop conditions.

[0014] Furthermore, the central control module is specifically used for: The time it takes for terrain feature points in front of the harvester to reach the header is calculated based on the vehicle speed and terrain contour map. The pre-stored adjustment height-delay time mapping table is then consulted based on this time to determine the advance of the action and generate the initial hydraulic control command. The initial hydraulic control command is corrected based on crop density and crop humidity data to obtain the final hydraulic control command.

[0015] The beneficial effects of adopting the above-mentioned further solution are as follows: by calculating the arrival time of feature points and querying the mapping table, the required advance amount of the hydraulic system can be accurately predicted, effectively compensating for the inherent delay of the system; by combining crop density and humidity data to perform multi-parameter fusion correction on the initial command, the hydraulic control command can adapt to the operation requirements under different crop conditions, improve the accuracy and adaptability of the header height adjustment, and effectively avoid problems such as missed cutting, entanglement or impurity mixing caused by changes in crop condition, thereby improving harvest quality and operation efficiency.

[0016] Furthermore, it also includes a human-machine interaction module, which includes a touch screen for displaying real-time terrain contours, vehicle speed, crop density, crop humidity, and header height, and receiving user-defined target header height values ​​and crop type parameters.

[0017] The beneficial effects of adopting the above-mentioned further solutions are as follows: by displaying multiple key operating parameters in real time, the system provides drivers with comprehensive and intuitive operating information, enhancing the transparency and visibility of system operation; by receiving user-set target header height and crop type parameters, the system can be personalized according to specific operational needs, improving the system's adaptability to different crops and agronomic requirements; the touch screen provides an intuitive and convenient human-machine interaction method, effectively reducing operational complexity and improving the system's ease of use and operational efficiency.

[0018] 2) Secondly, the present invention also provides a method for predictive adjustment of the header height of a harvester, the specific technical solution of which is as follows: The method of using any of the above-mentioned harvester header height prediction adjustment systems includes: The forward-looking perception module detects and generates a terrain contour map in real time based on the terrain data in front of the harvester; The multi-parameter acquisition module collects real-time data on the harvester's speed, crop density in front of the harvester, and crop moisture content. The central control module receives and performs multi-parameter fusion and predictive control on terrain contour map, vehicle speed, crop density and crop humidity data to generate hydraulic control commands. The hydraulic actuator module receives and adjusts the header height of the harvester according to the hydraulic control commands.

[0019] 3) In a third aspect, the present invention also provides a harvester, including any of the above-mentioned harvester header height prediction adjustment systems.

[0020] 4) In a fourth aspect, the present invention also provides an electronic device, the electronic device including a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so as to enable the electronic device to implement any of the above-mentioned harvester header height prediction adjustment methods.

[0021] 5) In a fifth aspect, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-described harvester header height prediction adjustment methods.

[0022] It should be noted that the beneficial effects of the technical solutions of the second to fifth aspects of the present invention and their corresponding possible implementations can be found in the above description of the technical effects of the first aspect and its corresponding possible implementations, and will not be repeated here. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below: Figure 1 This is a schematic diagram of a harvester header height prediction adjustment system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a harvester header height prediction adjustment method according to an embodiment of the present invention. Detailed Implementation

[0024] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0025] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0026] like Figure 1 As shown, an embodiment of the present invention provides a harvester header height prediction adjustment system, comprising: a forward-looking sensing module, a multi-parameter acquisition module, a central control module, a hydraulic execution module, and a human-machine interaction module; The forward-looking perception module is used to: detect and generate a terrain contour map in real time based on terrain data in front of the harvester; The multi-parameter acquisition module is used to: collect real-time data on the harvester's speed, crop density in front of the harvester, and crop moisture content; The central control module is used to: receive and perform multi-parameter fusion and predictive control on terrain contour map, vehicle speed, crop density and crop moisture data, and generate hydraulic control commands; The hydraulic actuator module is used to receive and adjust the header height of the harvester according to hydraulic control commands.

[0027] Optionally, in the above technical solution, the forward-looking perception module includes radar, ultrasonic sensors and a preprocessing unit. The forward-looking perception module detects the terrain data in front of the harvester in real time through radar and ultrasonic sensors. The terrain data in front of the harvester includes terrain data obtained by radar and terrain data obtained by ultrasonic sensors.

[0028] The forward-looking perception module employs a composite detection scheme combining millimeter-wave radar and ultrasonic sensors. Installed at the front of the harvester, its detection direction aligns with the harvester's travel direction. The horizontal detection angle covers the width of the machine, while the vertical detection range is 0–1.5 meters above ground (including crops and terrain). The radar uses a millimeter-wave band at 77 GHz, with a detection range of 10–15 meters and a ranging accuracy of ±3 cm. It has the ability to penetrate crop stalks, can focus on ground-reflected signals, and is unaffected by rain, fog, or dust. It generates distance-velocity two-dimensional data using FMCW (Frequency Modulated Continuous Wave) technology, and can distinguish between ground and crop-reflected signals by utilizing the difference in reflectivity. The ultrasonic sensors use 40 kHz high-frequency probes, arranged in an array of 3–5 units, to assist in detecting near-ground terrain details (such as small bumps and depressions) within a 5–10 meter range, with a ranging accuracy of ±5 cm, complementing the millimeter-wave radar.

[0029] In another embodiment, lidar is used instead of millimeter-wave radar for forward detection. Specifically, lidar (e.g., 16-line lidar) has higher ranging accuracy (up to ±2cm) and terrain detail resolution, and can generate three-dimensional point cloud terrain data, thereby improving the accuracy of terrain prediction.

[0030] It should be noted that millimeter-wave radar is preferred. Although the millimeter-wave radar used in this invention is slightly less accurate than lidar, it has strong anti-interference capabilities, is less affected by weather and dust, and has a moderate cost, making it more suitable for application needs in complex farmland operating environments.

[0031] The terrain data acquired by millimeter-wave radar mainly includes: two-dimensional distance-velocity data within a 10–15 meter range forward, generated based on FMCW technology, capable of distinguishing between ground and crop reflection signals, with a ranging accuracy of ±3cm. The terrain data acquired by ultrasonic sensors mainly includes: high-precision near-ground distance information within a 5–10 meter range, capable of detecting small bumps, depressions, and other detailed terrain features, with a ranging accuracy of ±5cm.

[0032] The preprocessing unit filters and stitches together the terrain data in front of the harvester, which is detected in real time by radar and ultrasonic sensors, to generate a terrain contour map.

[0033] In this model, the horizontal axis of the terrain contour map represents the forward horizontal distance, and the vertical axis represents the height difference between the ground and the cutter reference plane; the preprocessing unit is specifically used for: After filtering out clutter from the terrain data obtained by radar and ultrasonic sensors, a weighted fusion is performed in the overlapping area to generate a terrain contour map with the cut-off reference plane as the zero point.

[0034] The preprocessing unit processes the terrain data in front of the harvester, which is detected in real time by radar and ultrasonic sensors. This includes filtering the radar and ultrasonic signals to remove clutter caused by crop straw, and stitching and fusing terrain data at different distances to generate a terrain contour map within a range of 10–15 meters in front of the harvester. This terrain contour map is a continuous curve with distance as the horizontal axis and terrain height (the height difference between the ground and the cutter reference plane) as the vertical axis, and is transmitted to the central control module via the CAN bus.

[0035] The process of generating the terrain contour map is as follows: ① Terrain data acquisition: The millimeter-wave radar acquires terrain data within a range of 10-15 meters in front of the harvester at a cycle of 0.1 seconds, and outputs 100-150 pairs of horizontal distance-terrain height (height difference between the ground and the header reference plane) data in front of the harvester per cycle; the ultrasonic sensor simultaneously acquires terrain data within a range of 5-10 meters in front of the harvester, including multiple pairs of horizontal distance-terrain height data, and outputs 3-5 parallel terrain data. The terrain data acquired by the millimeter-wave radar and the terrain data acquired by the ultrasonic sensor are both accompanied by acquisition timestamps.

[0036] ② Spatiotemporal alignment processing: Based on the timestamp of the terrain data collected by the millimeter-wave radar, the terrain data collected by the ultrasonic sensor is linearly interpolated to synchronize the terrain data of the 5-10 meter overlapping area in time and match the space, eliminating the spatiotemporal deviation caused by the difference in sampling frequency and installation position.

[0037] ③ Coordinate System 1 and Height Conversion: The slant range (straight-line distance from the radar installation point to the ground) detected by the millimeter-wave radar is converted into a vertical height relative to the cutter reference plane using trigonometric functions. The conversion formula is: Height = Slant Range × sinθ. This step unifies all terrain data into a vertical coordinate system with the cutter reference plane as the zero point, where θ represents the installation tilt angle of the millimeter-wave radar. Using the cutter reference plane as the zero point means that the height of the cutter reference plane is defined as 0 in the vertical coordinate system.

[0038] ④ Filtering, noise reduction, and anomaly handling: For terrain data acquired by millimeter-wave radar, based on a preset ground reflection intensity threshold (set in 3 levels according to soil type), terrain data with reflection intensity below the threshold (such as crop straw clutter) is removed from the terrain data acquired by millimeter-wave radar. Specifically: During system initialization, the central control module's human-machine interface offers three typical soil type options for the driver to choose from: A, B, and C. A represents sandy soil with low reflectivity; B represents ordinary loam with medium reflectivity; and C represents moist clay with high reflectivity. Each soil type corresponds to a pre-calibrated ground reflectance intensity threshold, determined through extensive experimentation. This calibration experiment is conducted under typical operating conditions: on pristine ground of a specific soil type (without crop cover), the system's millimeter-wave radar is used to collect a large amount of ground reflectance intensity data. After statistical analysis (e.g., subtracting twice the standard deviation from the mean), a threshold value that effectively distinguishes ground reflectance signals from crop reflectance signals is determined. For example, the threshold for A might be set to 40%, for B to 55%, and for C to 70%. These thresholds are pre-stored as parameters in the central control module's non-volatile memory.

[0039] When millimeter-wave radar is operating, it outputs a series of data points every 0.1 seconds. Each data point includes its horizontal distance to the front, slant range, and a key reflection intensity value. Upon receiving this raw data, the preprocessing unit (usually an embedded signal processing chip) immediately initiates a filtering algorithm. The filtering algorithm iterates through each data point in the current cycle, comparing its reflection intensity value with a preset ground reflection intensity threshold retrieved based on the currently selected soil type. If the reflection intensity value of a data point is lower than the current threshold, it is determined that the point is not generated by ground reflection and is highly likely to be interference signals caused by crop straw clutter, weeds, or other low-reflectivity objects. The preprocessing unit then removes this data point from the current valid dataset, excluding it from subsequent coordinate transformation and data fusion. If the reflection intensity value of a data point is equal to or higher than the current threshold, it is determined that the point has a very high confidence of originating from ground reflection and is retained as a valid surface point. For the removed data points, the system does not simply discard them but marks them as invalid or clutter and may selectively record their location information for subsequent data quality assessment or system diagnosis.

[0040] To further improve reliability, a continuity verification step can be added after the simple thresholding method. Specifically, the distribution of valid data points retained after thresholding is analyzed along the distance dimension. For data gaps caused by the removal of multiple consecutive points, linear interpolation or spline interpolation algorithms can be used for data compensation based on the trend of valid data points before and after them (such as the rate of change of height) to ensure the continuity of the generated terrain contour map and avoid interruptions in terrain information due to dense local clutter.

[0041] For terrain data acquired by ultrasonic sensors, a moving average filter is used to suppress signal jumps caused by crop movement or momentary obstacles, retaining stable terrain features with a duration of not less than 0.3 seconds. Specifically: The preprocessing unit allocates a first-in, first-out (FIFO) data buffer in memory for each ultrasonic sensor channel. The depth of this buffer (i.e., the window size) N is directly determined by the requirement of "retention duration of no less than 0.3 seconds." Specifically, the ultrasonic sensor's data sampling frequency is 5Hz. Therefore, to cover a duration of 0.3 seconds, the required buffer depth is calculated as: N = 0.3 seconds / 0.2 seconds = 1.5, where 0.2 seconds is the sampling interval. To ensure coverage, N is rounded down to N = 2. That is, the buffer retains all effective shape and height data acquired by the ultrasonic sensor within the most recent two sampling periods, forming a sliding window containing two data points.

[0042] Once a new ultrasonic terrain data point (containing a timestamp and the height value H_new converted from the ranging value) is acquired and its coordinates are transformed, the preprocessing unit performs the following operations: The new data point H_new is placed at the end of the data buffer corresponding to this channel. If the number of data points in the data buffer exceeds the window size N, the oldest data point is removed from the head of the data buffer. The arithmetic mean of all N data points in the current data buffer is immediately calculated and used as the filtered output value H_filtered of the ultrasonic sensor at the current moment. The calculated filtered output value H_filtered is the terrain height value at the current moment after moving average filtering, and it will replace the original instantaneous value H_new for subsequent data stitching and fusion.

[0043] Signal jumps caused by crop swaying or momentary obstacles typically manifest as short-duration spikes or troughs (far shorter than 0.3 seconds). Since the moving average calculates the average over a time window, a single anomalous pulse value entering the window is diluted by another normal value within that window, and its effect is immediately eliminated once it leaves the window. This prevents drastic jumps in the filtered output H_filtered, effectively smoothing the data curve. Real-world terrain changes (such as a field ridge or a gentle slope) are characterized by stable terrain features lasting at least 0.3 seconds. When such changes occur, new height values ​​continuously enter the moving window, while older low values ​​are removed, allowing the output value H_filtered to smoothly track and represent this real, continuous terrain change trend without being filtered out as noise.

[0044] The above process is executed cyclically with each new ultrasonic data point, and the sliding window moves forward continuously, achieving real-time, continuous data stream filtering. The preprocessing unit ultimately outputs this series of smoothed H_filtered values, appends a timestamp, and sends it to the central control module.

[0045] Furthermore, if the height deviation of three consecutive sampling points in the terrain data acquired by the millimeter-wave radar is greater than 10cm, it is considered noise and replaced by the trend prediction value of the previous five points; if the terrain data acquired by the ultrasonic sensor is missing for more than 0.5 seconds, it automatically switches to interpolated data from other sensors, specifically: When terrain data acquired by the ultrasonic sensor remains missing for more than 0.5 seconds, the system automatically switches to interpolated data from the millimeter-wave radar. Specifically, the central control module retrieves the millimeter-wave radar terrain data received via the CAN bus within the same time period. This data has already undergone filtering and coordinate unification processing. The system selects data points within a 5-10 meter overlap area of ​​the radar that are spatiotemporally aligned with the missing ultrasonic data points. After performing coordinate system transformation, the system directly replaces or weights and fuses these data points into the missing ultrasonic data sequence to ensure the continuity and integrity of the terrain contour map.

[0046] ⑤ Multi-source terrain data stitching and fusion: Within the 10-15 meter range, millimeter-wave radar topographic data is dominant. In the 5-10 meter overlapping area, weighted fusion is used (radar weight 0.7, ultrasonic weight 0.3), with ultrasonic topographic data dominating within 5 meters. At the splicing points, first-order derivative constraints are used to achieve a smooth transition in the rate of change of height. Cubic spline curves are fitted to the fused discrete sampling points to generate a continuous topographic profile function H(x). The independent variable of the topographic profile function H(x) is the horizontal distance x in front. H(x) represents the relative height value of the ground at the horizontal distance x in front of the header (which can also be considered as the harvester) (the height difference between the ground and the header reference plane). The relative height value of the ground is based on a vertical coordinate system with the header reference plane as the zero point. Connecting each horizontal distance x in front and the corresponding height value H(x) forms a continuous curve, which is the topographic profile map, visually showing whether the terrain in front is uphill, downhill, flat, or uneven. Specifically: Based on the effective detection range of the sensors and using the horizontal distance in front as a reference, the preprocessing unit divides the detection area in front of the harvester into three sections and identifies the main data source that each section relies on: Section 1 (10-15 meter range): This area is covered only by millimeter-wave radar (ultrasonic effective range ≤ 10 meters). Therefore, millimeter-wave radar topographic data is the primary data source in this area. The preprocessing unit directly uses the filtered and coordinate-transformed radar data points as the final effective data for this area without fusion. Section 2 (5-10 meter overlap area): This area is the shared effective detection range of both millimeter-wave radar and ultrasonic sensors. A weighted fusion strategy is used in this area. For each identical or interpolated horizontal distance point in the fusion area, its fused height value H_fused is calculated using the formula: H_fused = 0.7 × H_radar + 0.3 × H_ultrasonic. H_radar represents the elevation in terrain data acquired by millimeter-wave radar, and H_ultrasonic represents the elevation in terrain data acquired by ultrasonic sensors. This weighting (radar weight 0.7, ultrasonic weight 0.3) is based on sensor characteristics: millimeter-wave radar maintains high accuracy and strong anti-interference capabilities within the 5-10 meter range, hence its higher weight; ultrasonic sensors offer extremely high accuracy at this distance and can serve as an effective supplement, but are susceptible to transient interference, hence their slightly lower weight. Segment 3 (0-5 meter range): While millimeter-wave radar can still detect this area, ultrasonic sensors, due to their physical characteristics, offer higher ranging accuracy and detail resolution in the very near field. Therefore, ultrasonic terrain data is dominant in this area. The preprocessing unit prioritizes ultrasonic sensor data after moving average filtering. Radar data is only used for interpolation when ultrasonic data is missing.

[0047] Simply splicing the data from the three sections directly may cause abrupt changes in height or slope at the section boundaries (especially at the 5-meter and 10-meter points) due to data source switching. This is non-physical in terms of topography and will introduce high-frequency disturbances to the control system. To solve this problem, the system uses first-order derivative constraints at the splicing points to achieve a smooth transition of the rate of change of height.

[0048] In this context, the first derivative represents the terrain slope. The preprocessing unit calculates the terrain height change rate S (slope) for several data points at the end of segment two (fusion zone) (near the 5-meter point) and the beginning of segment three (ultrasonic main data area). The system mandates that the difference between the starting slope of segment three and the slope at the end of segment two cannot exceed a preset reasonable threshold (e.g., a maximum slope change rate difference of 2% / meter). If a sudden change is detected, a smoothing algorithm is activated: data points with slope differences exceeding the limit are searched near the splicing point (e.g., within the range of 4.8 meters to 5.2 meters). Minor linear or polynomial adjustments are made to the initial portion of the data (ultrasonic data) in segment three to slightly correct its height value, thus smoothly connecting its first derivative (slope) with the slope trend at the end of segment two. This ensures that the height change of the terrain contour is continuous and gentle from the fusion zone to the near-field ultrasonic zone, avoiding drastic fluctuations in the cutter height control command due to sudden changes in the terrain data ahead. Through the aforementioned partitioning and gradient smoothing processes, the system ultimately outputs a high-quality terrain contour map with consistent height and continuous slope across the entire range (0-15 meters), providing reliable and stable terrain information input for the central control module to perform precise predictive control. Furthermore, after the aforementioned filtering optimization, it ensures that the height difference between two adjacent points on the horizontal axis is ≤3cm, ultimately outputting a terrain contour map with distance on the horizontal axis and terrain height on the vertical axis, which is then transmitted to the central control module via the CAN bus.

[0049] The "fused discrete sampling points" refer to the final effective data set within the 0-15 meter detection range, which exists in the form of a series of "horizontal distance-relative height" data pairs after multi-source data partitioning fusion and smoothing transition processing. These data points come from three parts: pure millimeter-wave radar data within the 10-15 meter range, weighted fusion data of radar and ultrasound within the 5-10 meter overlapping area, and data mainly composed of ultrasound within the 5 meter range. Together, they constitute a spatially continuous but not yet curve-connected discrete spatial sampling sequence before the generation of the terrain contour map.

[0050] In another embodiment, the process of obtaining the terrain contour map is as follows: ① The preprocessing unit receives terrain data (including timestamps, horizontal distance, and slant distance data) within a 10-15 meter range from the millimeter-wave radar output via the CAN bus, and multiple terrain data (including timestamps, horizontal distance, and ranging values) within a 5-10 meter range from the ultrasonic sensor array output. First, it performs initial screening based on signal strength and physical probability. Specifically, for millimeter-wave radar data, data points with reflection intensity below a preset threshold (e.g., 30%) are removed, as these data points are highly likely to originate from crop clutter. For ultrasonic data, anomalies with instantaneous ranging values ​​exceeding 20cm (compared to the previous point) are removed, as these anomalies are usually caused by small, airborne debris.

[0051] ② Using the system's high-precision clock as a reference, time synchronization is performed on the data from the millimeter-wave radar and ultrasonic sensor. The timestamps of the ultrasonic data are matched with the timestamps of the radar data. For ultrasonic data at asynchronous times, a linear interpolation algorithm is used to interpolate it to the sampling time point of the radar data. Subsequently, coordinate system one is established. Specifically, the slant range data of the millimeter-wave radar is converted into a vertical height with the cutting platform reference plane as the zero point according to its installation tilt angle θ (θ is the angle between the radar beam centerline and the horizontal plane) using the formula height = slant range × sinθ. The ranging value of the ultrasonic sensor, which is itself a vertical distance, is directly converted into a vertical height with the cutting platform reference plane as the zero point.

[0052] ③ An adaptive median filter based on elevation change rate is employed. The elevation change rate of adjacent data points is calculated. If the change rate exceeds a preset reasonable threshold (e.g., the change rate corresponding to a 10% slope), the point is determined to be residual clutter or noise, and the median of the three points before and after it is used as the replacement. A sliding window weighted average filter is employed. Within a sliding window containing five consecutive sampling points, the center point is given the highest weight, and the edge points are given lower weights. A weighted average is calculated to effectively smooth out minor fluctuations in distance measurement caused by slight crop swaying.

[0053] ④ For the 5-10 meter overlap area where both radar and ultrasonic data exist, data fusion is performed. This implementation uses a dynamic weighted fusion algorithm based on sensor confidence, rather than fixed weights. Specifically, the weight W_radar(d) of the millimeter-wave radar at a certain distance d is dynamically calculated based on its historical ranging stability (variance) at that distance; the smaller the variance, the higher the weight. The weight W_ultrasonic(d) of the ultrasonic sensor is calculated based on its signal strength; the higher the signal strength, the higher the weight. The weights are normalized so that W_radar(d) + W_ultrasonic(d) = 1. For the same horizontal distance d within the overlap area, the fused height value H_fused(d) is calculated as follows: H_fused(d) = W_radar(d) × H_radar(d) + W_ultrasonic(d) × H_ultrasonic(d). This method relies more on radar when the radar data is reliable and more on ultrasonic when the quality of the nearby ultrasonic data is extremely high, achieving adaptive optimization.

[0054] ⑤ In the 10-15 meter range (non-overlapping area), filtered millimeter-wave radar data is used directly; in the 5-10 meter range (overlapping area), the dynamically weighted fused data described above is used. In the 0-5 meter range, filtered ultrasonic sensor data is used directly (due to its higher near-field accuracy). The data points in the three areas are sorted according to their horizontal distance from the front, forming a set of discrete (distance, height) sampling points. Finally, a cubic spline interpolation algorithm is used to fit these discrete sampling points to generate a smooth, continuous terrain contour curve H(x), which is the final terrain contour map. This curve clearly shows the height change of the terrain in front relative to the cutter reference plane, with the horizontal axis representing the horizontal distance x and the vertical axis representing the relative height H(x).

[0055] In this implementation method, the process further improves the anti-interference ability of complex field environments and the generation accuracy and reliability of terrain contour maps through dynamic weight allocation and more refined adaptive filtering.

[0056] Optionally, in the above technical solution, the multi-parameter acquisition module includes a camera installed above the cutting table, an infrared humidity sensor installed near the cutting blade of the cutting table, and a data processing unit. The data processing unit is used to: perform image recognition on images captured by the camera in front of the harvester to obtain crop density; A high-definition camera (1920×1080 resolution) is installed above the header to capture images of the top of the crop in front of the harvester. A deep learning-based image segmentation algorithm is used to identify the quantity and distribution of crop stalks in the image, and then the crop density per unit area (unit: plants / m²) is calculated. The crop density data is sampled at a frequency of 5Hz, specifically: A total of 50,000 top images of three typical crops—wheat, corn, and rice—at different growth stages were collected, and the center point coordinates of each crop stalk in each image were labeled to form a dataset for model training. A deep learning model with an encoder-decoder structure was employed. An attention mechanism was introduced in the encoding stage to enhance the extraction of crop stalk features, and a high-precision stalk probability heatmap was output in the decoding stage, effectively improving recognition accuracy in densely planted crop scenarios. A camera captured one frame of crop top image at a 0.2-second interval, which was input into the trained model to infer candidate stalk regions. The number of valid stalks in the image was counted. Combining the camera's installation height (1.2 meters above the ground) and its horizontal field of view (e.g., 60°), the counted stalk count was converted into stalk density per unit area (1m × 1m), i.e., crop density, in units of stalks / m². The recognition error of this method was controlled within ±5 stalks / m².

[0057] In another embodiment, the process of obtaining crop density is as follows: ① The raw RGB image captured by the camera is first converted into a grayscale image to reduce computational complexity. Then, based on the camera's fixed installation height (1.2 meters above the ground) and horizontal field of view (60°), a fixed region of interest (ROI) corresponding to a 1m × 1m unit area on the actual ground is determined using a perspective transformation model. This ROI is typically located in the lower center of the image to minimize the impact of perspective distortion and ensure that the analyzed area falls within the working range of the cutting table.

[0058] ② Calculate the Gray-Level Co-occurrence Matrix (GLCM) of the grayscale ROI image. The GLCM is an effective method for describing image texture by studying the second-order statistical properties of the grayscale space. In this implementation, four directions—0°, 45°, 90°, and 135°—are selected, and the GLCM for each direction is calculated. Four key texture features are extracted from each matrix: contrast, energy, homogeneity, and correlation. Finally, the average of each feature value in the four directions is taken to obtain a set (4) of numerical feature vectors that can comprehensively characterize the texture properties of the crop canopy within the ROI.

[0059] ③ A mapping model between texture features and crop density was established in advance through extensive field experiments. Top images of wheat, corn, and rice fields at different density levels (e.g., 10-100 plants / m²) were collected, their true density values ​​were manually labeled, and the GLCM feature vectors of the ROIs in each image were calculated. A support vector machine (SVM) regression model was trained using this dataset. The input of this model is a 4-dimensional texture feature vector, and the output is the predicted crop density value (unit: plants / m²). The trained model was then embedded into the algorithm library of the data processing unit.

[0060] ④ In real-time operation, for each newly acquired image frame, the data processing unit executes steps 1 and 2 above to extract the texture feature vector of the current ROI. Subsequently, this feature vector is input into the pre-trained SVM regression model for real-time inference, directly outputting a continuous crop density estimate. This process is executed every 0.2 seconds, with an output frequency of 5Hz.

[0061] In this embodiment, by analyzing the macroscopic texture features of the crop population rather than accurately identifying individual crops, the demand for computing resources is reduced, and the problem of inaccurate segmentation of individual plants in complex scenarios is avoided. It is particularly suitable for scenarios where plants are severely intertwined in the later stages of crop growth, and provides stable and reliable crop density parameters for the central control module.

[0062] The data processing unit is also used to: obtain crop humidity by analyzing the infrared reflectance of crop straw in front of the harvester detected by the infrared humidity sensor.

[0063] An infrared humidity sensor is installed near the cutting blade of the header. By detecting the infrared reflectance of the crop straw (the infrared reflectance of straw with different moisture contents varies significantly), the crop humidity data (unit: %) is calculated. The sampling frequency of the crop humidity data is 2Hz. This parameter is used to identify crops with high humidity and avoid the problem of the header getting tangled due to the toughness of the crops.

[0064] The specific process for obtaining crop humidity is as follows: A dual-path infrared sensor with 850nm (moisture-sensitive band) and 1300nm (reference band) wavelengths is used to simultaneously collect the reflected light intensity of crop straw in front of the harvester. The reflectance ratio is calculated. , This indicates the reflectivity of crop straw under infrared light at a wavelength of 850 nanometers. This indicates the reflectivity of crop straw under infrared light at a wavelength of 1300 nanometers.

[0065] The ratio is converted into a humidity value based on a pre-established calibration curve: this calibration curve is obtained by measuring the R value of straw samples with different humidity levels (30%-90%) in the laboratory and fitting the relationship H=62.3-45.8×R (where H is the humidity percentage and R is the reflectance ratio). During on-site collection, by setting a reflectance threshold (only retaining signals with reflectance >20% to effectively eliminate interference from soil background), crop humidity values ​​are output every 2 seconds, with a detection error ≤5%.

[0066] The system calculates the real-time vehicle speed (unit: km / h) by acquiring the rotational speed signal from the harvester's gearbox or receiving the speed signal from the GPS module. The sampling frequency of the vehicle speed data can be 10Hz, enabling dynamic capture of speed changes such as acceleration and deceleration during driving.

[0067] In another embodiment, the specific process for obtaining crop humidity is as follows: ① An active infrared reflectivity measurement unit is constructed using a near-infrared light source with a center wavelength of 1450nm (a band representing a strong absorption peak for water molecules) and a corresponding infrared photodetector. The sensor emits a modulated 1450nm infrared beam at a frequency of 2Hz (every 0.5 seconds) towards the crop straw in front of the harvester and simultaneously detects the intensity of the reflected light signal. The photoelectric conversion circuit inside the sensor converts the light signal intensity into an electrical signal and further calculates the reflectivity measured. (Defined as the ratio of reflected light intensity to emitted light intensity). Simultaneously, a high-precision temperature sensor integrated within the sensor probe collects the ambient temperature T (unit: °C) in real time.

[0068] ② The absorption intensity of moisture in the 1450nm infrared band is affected by ambient temperature. To eliminate this interference, the data processing unit receives the original reflectivity... After adjusting for ambient temperature T, standardization is first performed. The system pre-stores a baseline reflectance-temperature curve R_dry(T) measured at different temperatures for absolutely dry straw samples (0% humidity). The data processing unit queries this curve based on the current temperature T to obtain the theoretical baseline reflectance value R_dry for dry straw at that temperature. Subsequently, the temperature-compensated standardized reflectance R_comp is calculated using the following formula: This step aims to uniformly convert the reflectance measured at different temperatures to a relative value based on dry straw at the current temperature, thereby effectively eliminating measurement bias introduced by temperature changes.

[0069] ③ The higher the moisture content of crop straw, the stronger its absorption of infrared light in the 1450nm band, resulting in a lower measured reflectance R_comp value. The system converts the R_comp value into the final crop moisture value (unit: %) using a pre-established standardized reflectance-moisture calibration curve. This calibration curve is obtained by measuring straw samples with different known moisture contents (from 30% to 90%) at different temperatures in a constant-temperature laboratory. The data is then fitted after the above standardization process. Its functional relationship can be expressed as H = a × exp(-b × R_comp) + c (where a, b, and c are fitting coefficients, and H is the humidity percentage). The data processing unit substitutes the calculated R_comp value into this function to calculate the current crop humidity H in real time and outputs it at a frequency of 2Hz.

[0070] In this embodiment, by utilizing the strong absorption characteristics of water molecules in a specific wavelength band and a precise temperature compensation algorithm, a high-sensitivity perception of crop humidity is achieved, avoiding the challenges of synchronization and calibration complexity that may be brought about by multi-band sensors, and providing stable and reliable crop humidity parameters for the central control module.

[0071] The multi-parameter acquisition module incorporates a high-precision timestamp unit. This unit, based on the system clock, adds a unified timestamp to the real-time acquired vehicle speed, crop density, and crop humidity data. Simultaneously, the forward-looking perception module adds the same timestamp to its generated terrain contour map data. Through this mechanism, all source data carries a precise timestamp when transmitted to the central control module via the CAN bus. The central control module uses this timestamp to perform time alignment processing on the received multi-source heterogeneous data, ensuring that vehicle speed, crop density, crop humidity, and the preceding terrain contour data are strictly matched in the time dimension, providing a consistent time reference for subsequent fusion calculations and predictive control.

[0072] Optionally, in the above technical solution, the central control module is specifically used for: 1) Calculate the time it takes for terrain feature points in front of the harvester to reach the header based on vehicle speed and terrain contour map, and then query the pre-stored adjustment height-delay time mapping table based on this time to determine the action lead, generating initial hydraulic control commands, specifically: ① The central control module receives a 10-15 meter terrain contour map generated by the forward-looking perception module and, combined with real-time vehicle speed data, calculates the estimated time for each terrain feature point to reach the cut-off platform. The specific calculation process includes: Let the current vehicle speed be v (km / h), and convert it to v' = v × 1000 / 3600 (unit: m / s); let the horizontal distance from a certain terrain feature point ahead to the leading edge of the cutter be d (meters), then the time required for that point to reach the bottom of the cutter is t = d / v' (seconds).

[0073] The slope S = ΔH / Δx (where x is the horizontal distance ahead) is calculated using the first-order difference of the terrain profile function H(x). When S > 5% (steep slope), a quadratic polynomial H(x) = ax² + bx + c is used to fit the terrain change trend and predict the terrain height change within the next 5 seconds; when S ≤ 5% (gentle slope), a linear fitting H(x) = kx + b is used to simplify the calculation.

[0074] The prediction model parameters are revised every 0.5 seconds based on newly acquired terrain data to ensure that the terrain trend prediction error does not exceed 2cm.

[0075] ② The central control module has a pre-stored height adjustment-delay time mapping table, which was calibrated through hydraulic system experiments (e.g., a 10cm height adjustment corresponds to a delay time of 0.8 seconds, and a 20cm height adjustment corresponds to a delay time of 1.2 seconds). The calculation process is as follows: Determine the required adjustment height ΔH based on the terrain prediction results (e.g., increase by 20cm); look up the mapping table based on ΔH to obtain the corresponding hydraulic system delay time t_delay (e.g., 1.2 seconds); calculate the advance trigger time: t_trigger = t_arrive - t_delay, where t_arrive is the arrival time of the terrain feature point; advance the timing of the cutter's action by t_delay seconds to ensure that the hydraulic adjustment action is completed exactly when the cutter reaches the terrain change point.

[0076] The initial hydraulic control command is then obtained, which includes the hydraulic cylinder extension / retraction direction, action duration, flow control parameters, target height value, and trigger timestamp. The hydraulic cylinder extension / retraction direction can be specifically represented by Boolean values ​​or specific control codes to indicate the direction of movement of the hydraulic cylinder, thereby controlling the lifting and lowering of the header. Its generation directly depends on the terrain trend prediction results: when a protrusion is predicted in front of the harvester (i.e., the predicted height value is greater than the current height of the header), the command direction is "lift" (or code 1), controlling the header to rise to avoid collision; when a depression is predicted in front (the predicted height value is less than the current height), the command direction is "lower" (or code 0), controlling the header to lower to maintain the cutting height. The action duration refers to the length of time the hydraulic cylinder needs to continuously move (unit: seconds, accurate to 0.1 seconds), directly determining the adjustment range of the header. The values ​​are derived from the calibration data of the hydraulic system's delay characteristics: The central controller first calculates the required height difference ΔH based on the terrain contour data, then queries the pre-stored adjustment height-delay time mapping table to obtain the hydraulic delay time t_delay corresponding to ΔH (for example, a 20cm height adjustment corresponds to a delay time of 1.2 seconds). This t_delay value is then set as the action duration. Flow control parameters are used to set the opening degree of the electro-hydraulic proportional control valve (usually corresponding to a 0-5V control current or percentage opening), controlling the speed of the header adjustment by adjusting the hydraulic oil flow. Their values ​​are primarily determined by crop humidity parameters: The controller queries the "humidity-speed correction coefficient table" based on real-time data detected by the infrared humidity sensor to obtain a correction coefficient (for example, a coefficient of 0.7 when humidity > 70%), and multiplies the default flow parameter by this coefficient to finally generate the flow parameter used for speed control, thereby reducing the action speed under high humidity conditions and preventing straw entanglement. The target height value is the absolute height the header needs to reach (unit: cm, resolution 0.1cm), with the header reference plane as the zero point. The calculation process integrates multi-parameter corrections: First, the basic target height is determined based on the crop type (e.g., 12cm for wheat). Then, the density correction value is obtained by consulting the "Density-Height Correction Table" based on the crop density obtained from image recognition (e.g., +5cm when density > 50 plants / m²). Simultaneously, the humidity correction value is obtained by consulting the "Humidity-Height Correction Table" based on the crop humidity (e.g., +3cm when humidity > 70%). Finally, the final target height is calculated using the formula "Final Target Height = Basic Height + Density Correction Value + Humidity Correction Value". The precise absolute time (accurate to 0.01 seconds) at which the trigger timestamp hydraulic system begins its action is crucial for achieving predictive adjustment.Its calculation integrates travel speed and system delay: the controller calculates the arrival time t_arrive=d / v based on the horizontal distance d of the terrain feature point and the real-time vehicle speed v, and obtains the hydraulic delay time t_delay by querying the delay mapping table. Finally, it combines the environmental correction amount t_correct (such as compensation under low temperature or heavy load conditions) and calculates the accurate trigger time through the formula "trigger time=current time+t_arrive-t_delay+t_correct" to ensure that the action command is issued in advance so that the adjustment action is completed exactly at the point of terrain change.

[0077] The process of constructing the height-delay time mapping table is as follows: Ten different adjustment heights (from 5cm to 50cm, in 5cm intervals) were set on a hydraulic test bench. Each height was tested 30 times repeatedly. The delay time data collected from all experiments were accurately fitted using the least squares method to obtain the functional relationship between delay time T and adjustment height h: T = 0.2 + 0.05 × h + 0.001 × h² (where h is in cm and T is in seconds). This fitted curve characterizes the overall law of the hydraulic system delay time variation with adjustment height. The time elapsed from the issuance of the control command to the actual arrival of the hydraulic cylinder piston rod at the target position was recorded. The adjustment height h was divided into segments of 5cm intervals, and the average delay time and its standard deviation for each height segment were stored. During actual control, when the required adjustment height is between two stored segments (e.g., 12cm), a linear interpolation algorithm was used to calculate its delay time. Specifically, the delay times corresponding to two adjacent height segments (10cm and 15cm) were weighted and averaged to ensure accurate delay time predictions for any adjustment height.

[0078] When the real-time vehicle speed increases (e.g., from 5 km / h to 8 km / h), the time it takes for terrain feature points at the same horizontal distance ahead to reach the cutter is shortened. The central control module dynamically calculates the arrival time t=d / v′ based on the real-time vehicle speed v (km / h) and the converted speed v′=v×1000 / 3600 (unit m / s), combined with the horizontal distance d between the terrain feature point and the leading edge of the cutter. Based on this time, the lead time of the predicted action is shortened, thereby ensuring that the triggering time of the cutter height adjustment is synchronized with the actual terrain changes.

[0079] In another embodiment, the specific process of generating the initial hydraulic control command is as follows: ① After receiving the terrain contour function H(x), the central control module does not perform calculations on all data points. Instead, it first performs real-time analysis to identify the key terrain feature points that truly require triggering the cutter height adjustment. This identification is achieved by calculating the curvature of the terrain contour. When the curvature value exceeds a preset threshold (e.g., 0.05 m⁻¹), the point (x, H(x)) is determined to be a key terrain feature point (such as a slope top, slope bottom, or inflection point of concavity and convexity changes).

[0080] When acceleration a≠0, the current vehicle speed is set to v (m / s), and its acceleration a (m / s²) is calculated in real time. For each identified key feature point, its instantaneous distance to the leading edge of the cutter is d (meters). The central control module uses a uniformly accelerated kinematic model to dynamically predict the time t for the feature point to reach the cutter. arrive If the acceleration a is 0 (uniform motion), then the formula t is used. arrive The calculation is performed using d / v. This model can more accurately reflect the arrival time of the harvester under acceleration or deceleration conditions.

[0081] ② The central control module no longer stores a simple "height adjustment-delay time mapping table," but rather a hydraulic system state-space model. This model stores the hydraulic delay time t. delay The adjustment height ΔH (unit: cm) and the current hydraulic oil temperature T are expressed as follows: oil (Unit: °C) and system load L (unit: kg). ΔH is determined by the difference between the current cutter height and the predicted height of key terrain feature points. Oil temperature T oil The temperature is provided in real time by a temperature sensor installed on the hydraulic circuit; the load L is estimated by a pressure sensor installed in the hydraulic cylinder inlet circuit.

[0082] Set the current ΔH, T oil Substituting L into the state-space model (which is fitted from a large amount of system identification experimental data and can be a polynomial or neural network model), a precise predicted delay time t that perfectly matches the current system operating state is calculated in real time. delay .

[0083] ③ Calculate the action advance trigger time (i.e., action advance amount) t_trigger: t_trigger = t_arrive - t delay The central control module plans a hydraulic control event at the corresponding future time point based on the calculated t_trigger. This event includes the hydraulic cylinder extension / retraction direction (lift / lower) corresponding to ΔH, and the event based on ΔH and t. delayThe calculated duration of the action, along with a default flow control parameter initially set based on the magnitude of ΔH, are used to encapsulate a complete initial hydraulic control command. This command, accompanied by a precise trigger timestamp, is then temporarily stored in a command queue awaiting transmission.

[0084] In this embodiment, key terrain change points are accurately captured by dynamic curvature recognition, and a state-space-based delay prediction model is adopted, making the system's compensation for hydraulic delay more accurate and adaptive. This is especially suitable for complex operating environments with large vehicle speed fluctuations and frequent changes in hydraulic system conditions, further improving the accuracy and robustness of predictive control.

[0085] 2) Based on crop density and crop moisture data, the initial hydraulic control command is corrected to obtain the hydraulic control command, specifically: ① Crop density correction is achieved by establishing a mapping relationship between crop density and target height correction value. This mapping relationship is quantified by obtaining the optimal header height compensation values ​​for different density ranges through field trials, forming a correction parameter table. The specific correction logic is as follows: When the crop density is ≤30 plants / m², the target header height correction value is 0cm. In low-density areas, the crop straw is sparse, and the baseline target height (e.g., 10-15cm) is maintained. When the crop density is in the range of 30-50 plants / m², the target header height correction value is +2cm. In medium-density areas, the header height needs to be slightly increased to reduce the contact area with the straw. When the crop density is >50 plants / m², the target header height correction value is +5cm. In high-density areas, the straw is dense and easily tangled, and the header height needs to be significantly increased to reduce the risk of blockage.

[0086] The correction process involves three steps: a. Baseline value setting: Preset the initial target height according to the crop type (e.g., wheat baseline value 12cm, corn baseline value 15cm), and store it in the parameter library of the central control module; b. Real-time calculation: After the camera captures images and identifies the number of crop plants per unit area, the controller calls the above mapping table data and automatically adds correction values ​​(e.g., target height of high-density cornfield = 15cm baseline value + 5cm correction value = 20cm). c. Dynamic calibration: The system indirectly judges the actual blockage situation every 30 seconds through the header load sensor. If blockage still occurs, it automatically increases the correction value of the current density range by 1cm (upper limit is +8cm) to ensure that it adapts to the growth characteristics of different crop varieties.

[0087] At the same time, in areas with high-density crops, the duration of the adjustment action is extended accordingly to reduce the load on the hydraulic system.

[0088] ② Crop humidity correction is mainly for crops with high humidity (e.g., humidity > 70%). It prevents straw from tangling around the header components due to its toughness by reducing the header adjustment speed. Specifically, based on real-time crop humidity data, the hydraulic flow is reduced proportionally (e.g., by 10%), thereby lowering the header adjustment speed. Humidity correction works in conjunction with density correction, achieving final speed control by adjusting flow control parameters.

[0089] By comprehensively correcting for the crop density and crop moisture, the final hydraulic control commands suitable for the current operating environment are generated. These include: the final hydraulic cylinder extension / retraction direction (lift / lower), the final action duration (accurate to 0.1 seconds), the final flow control parameters (used to adjust the action speed), the final target height value, and the final trigger timestamp.

[0090] In another embodiment, the initial hydraulic control command is modified based on crop density and crop moisture data to obtain the final hydraulic control command. The specific implementation process is as follows: ① The central control module receives real-time data on crop density (unit: plants / m²) and crop humidity (unit: %). First, these precise inputs are converted into fuzzy variables that the fuzzy logic system can process. Specifically, crop density (D) is divided into three fuzzy sets: {"sparse", "medium", "dense"}. Its universe of discourse (range) is set, for example, to [0, 100] plants / m². The membership function uses a trapezoidal or triangular function; for example, the "dense" set has high membership above 50 plants / m². Crop humidity (H) is divided into three fuzzy sets: {"dry", "medium", "moist"}. Its universe of discourse is set to [30%, 90%]. The membership function also uses a trapezoidal or triangular function; for example, the "moist" set has high membership above 70%.

[0091] ② The system incorporates a fuzzy rule base based on expert knowledge and extensive field trials. This rule base defines the fuzzy relationship between crop conditions (density and humidity) and the required control strategies (height correction ΔZ and velocity coefficient correction coefficient K_v). Rules are described in "IF-THEN" format. For example, Rule 1: IF density is "dense" AND humidity is "wet", THEN height correction is "significantly increased" AND velocity coefficient is "significantly decreased". Rule 2: IF density is "sparse" AND humidity is "dry", THEN height correction is "slightly increased" AND velocity coefficient is "normal". Rule 3: IF density is "medium" AND humidity is "medium", THEN height correction is "moderately increased" AND velocity coefficient is "slightly decreased".

[0092] The fuzzy inference engine activates all relevant fuzzy rules based on the fuzzy sets of the current input crop density and crop humidity, and uses the Mamdani inference method to calculate the fuzzy set of output variables (height correction ΔZ and velocity correction coefficient K_v).

[0093] ③ The result obtained through fuzzy inference is a fuzzy quantity, which needs to be converted into a precise control value. This system uses the centroid method for defuzzification calculation. That is, it calculates the abscissa value corresponding to the center of the area enclosed by the membership function of the output fuzzy set and the abscissa, and obtains the total height correction value ΔZ_final (unit: cm) and the total speed correction coefficient K_v_final (dimensionless, between 0 and 1).

[0094] ④ Add the total height correction value ΔZ_final output by the fuzzy logic controller to the basic target height (determined by terrain prediction) in the initial hydraulic control command to obtain the final target height value applicable to the current crop conditions. Final target height = Basic target height + ΔZ_final; Multiplying the total speed correction coefficient K_v_final output by the fuzzy logic controller by the default flow parameter in the initial command (determined by the adjustment range) yields the final flow control parameter used to adjust the action speed. Final flow parameter = default flow parameter × K_v_final; The central control module combines the calculated final target height value and final flow control parameters with the hydraulic cylinder extension / retraction direction and corrected action duration (fine-tuned according to the final adjustment range) determined in the initial command, and the trigger timestamp, into a final hydraulic control command that has undergone multi-parameter intelligent correction, and sends it to the hydraulic execution module.

[0095] In this embodiment, the complex coupling relationship between crop density and humidity is handled by fuzzy logic. Its output is no longer a simple linear superposition, but a more nonlinear and intelligent comprehensive decision that is closer to actual operation experience. This enables it to cope with complex and ever-changing crop conditions more delicately and adaptively, and further improves the robustness and operation quality of the system.

[0096] The hardware components of the hydraulic actuator module include a variable displacement hydraulic pump, an electro-hydraulic proportional control valve, a double-acting hydraulic cylinder, and a displacement sensor. The module's workflow is as follows: Upon receiving instructions from the central control module, the electro-hydraulic proportional control valve adjusts its valve opening according to the flow parameters in the instructions, precisely controlling the flow of hydraulic oil into the hydraulic cylinder, thereby achieving precise control of the cutting table adjustment speed; the displacement sensor monitors the extension and retraction of the hydraulic cylinder in real time and transmits this feedback signal to the central control module, forming a closed-loop control circuit to ensure that the actual adjustment height matches the height required by the instruction.

[0097] This application achieves a three-closed-loop control of "position-velocity-acceleration", which is different from traditional hydraulic control that only provides feedback on the final position: a) The displacement sensor provides real-time position information at a sampling frequency of 1kHz, and the deviation between the displacement sensor and the commanded position is used for proportional adjustment (P parameter); b) The real-time velocity (v=Δs / Δt) is calculated through position difference, and the deviation between this velocity and the commanded velocity is used for integral adjustment (I parameter); c) The acceleration (a=Δv / Δt) is calculated through velocity difference, and the maximum acceleration is limited to ≤0.5m / s² to avoid hydraulic shock (D parameter). Through the coordinated control of these three parameters, the system improves the adjustment accuracy from ±5mm in the traditional scheme to ±2mm, with an overshoot of ≤3%.

[0098] Regarding delay characteristic calibration, the system's hydraulic delay time under different load and oil temperature conditions is experimentally measured before leaving the factory and stored in the mapping table of the central control module to ensure the accuracy of delay compensation calculations. The innovation of this calibration work lies in the following: traditional calibration methods only consider static load, while this invention establishes a three-dimensional mapping table of "load-oil temperature-delay time." Specifically, a) experimental variables include load (500kg-2000kg, interval 250kg) and oil temperature (20℃-60℃, interval 5℃); b) 20 step response tests are performed under each set of conditions, recording the 90% response time (i.e., the time from command issuance to the hydraulic cylinder reaching 90% of the target height); c) a three-dimensional interpolation algorithm is used to store 16×9=144 sets of data. The controller automatically calls the corresponding delay time based on real-time load (detected by a pressure sensor) and oil temperature, improving the compensation accuracy to ±0.1 seconds. This enables the hydraulic actuator module to accurately and efficiently execute the cutting table height adjustment command.

[0099] In another embodiment, an electric actuator is used instead of a hydraulic system. Specifically, an electric motor (such as a servo motor) is used in conjunction with a ball screw mechanism to drive the cutting table to rise and fall. By taking advantage of the fast response speed and short action delay (usually ≤0.3 seconds) of the motor, the hydraulic delay problem is directly eliminated from the execution stage.

[0100] It should be noted that this invention prioritizes the use of hydraulic actuators and actively compensates for hydraulic delays through a predictive control algorithm. This not only preserves the high load capacity and high reliability of the hydraulic system, but also effectively solves the control lag caused by the delay, making it more in line with the actual working conditions of the harvester.

[0101] Optionally, the above technical solution also includes a human-machine interaction module, which includes a touch screen for displaying real-time terrain contours, vehicle speed, crop density, crop humidity and header height, and receiving the target header height value and crop type parameters set by the user.

[0102] The human-machine interface module is equipped with a touchscreen display that shows the driver multiple real-time system parameters, including the current header height, a simulated terrain outline in front of the harvester (display range 10 meters), real-time vehicle speed, crop density, and crop moisture content. Simultaneously, the module allows the driver to manually set target operating parameters via the touchscreen, including setting the allowable operating range for the target header height (e.g., 8-20cm) and selecting the crop type (e.g., wheat, corn; different crop types correspond to different density and moisture correction coefficients, which are pre-stored in the system parameter library). Furthermore, the module features a fault alarm function. When the system detects a sensor malfunction or an abnormality in the hydraulic system, it issues an audible and visual alarm, prompting the driver to switch the system to manual operation mode.

[0103] After the harvester header height prediction adjustment system of the present invention is started, each module begins to work according to a preset process. The specific steps are as follows: Step 1, Module initialization, specifically: In the forward-looking perception module, the millimeter-wave radar preheats (locks onto the 77GHz band), the ultrasonic sensor array performs a self-test (transmits test signals), and the lens dust removal device is activated to remove surface straw debris. Key parameters include radar detection angle calibration (horizontal ±15° covering the width of the machine body) and confirmation of ultrasonic probe spacing (adjacent interval 50cm). In the multi-parameter acquisition module, the GPS module searches for satellites for positioning (ensuring signals from ≥6 satellites), the camera focuses (focusing on crops at a height of 1–1.2m), and the infrared humidity sensor preheats to eliminate ambient temperature interference. Key parameters include resetting the initial value of vehicle speed acquisition and locking the image sampling frame rate to 5Hz. In the central control module, preset parameters (including correction coefficients corresponding to crop types and hydraulic delay mapping tables) are loaded, and the CAN bus communication protocol is initialized. The controller's main frequency is increased to 500MHz, entering high-speed operation mode.

[0104] Step 2, Data Acquisition and Preprocessing, specifically: The millimeter-wave radar transmits frequency-modulated continuous waves every 0.1 seconds and receives reflected signals from the ground and crops. It separates the ground contour data by utilizing the difference in reflectivity (ground reflectivity > 80%, crop reflectivity < 30%). The ultrasonic sensor detects synchronously every 0.2 seconds, supplementing the terrain details (such as protrusions ≤ 10 cm) within a range of 5–10 meters. After being stitched with the radar data, it generates a two-dimensional "distance-height" curve (the horizontal axis is the distance in front, and the vertical axis is the relative height). The preprocessing unit removes interference signals such as weeds and birds through Kalman filtering, while retaining the continuous terrain contour. Every 0.1 seconds, the system reads the transmission speed pulse or GPS speed value and converts it to km / h (e.g., 5000 rpm corresponds to a vehicle speed of 6 km / h). The camera captures an image every 0.2 seconds, and an image recognition algorithm counts the number of straw stalks to calculate the crop density per unit area (e.g., 100 stalks / 2m² corresponds to 50 stalks / m²). The infrared sensor emits 8–14μm infrared waves every 0.5 seconds, and the reflected energy is used to calculate crop humidity (e.g., 30% energy attenuation corresponds to 60% humidity). All collected data is timestamped (accurate to 0.01 seconds) and transmitted to the central control module via the CAN bus, with a transmission delay controlled to ≤10ms. The central control module uses the radar data timestamp as a reference to time-align the received multi-source data, ensuring strict matching of terrain, vehicle speed, and crop parameters in the time dimension, providing a consistent time reference for subsequent predictive calculations.

[0105] Step 3: The central control module receives and synchronizes multi-source data, and calculates the time it takes for the terrain to reach the cutter based on the vehicle speed. Specifically: a. Coordinate synchronization: Convert the terrain distance detected by forward sensing (with the radar installation point as the origin) into a distance with the leading edge of the cutter as the origin (the radar installation point is 40cm ahead of the cutter, so the actual distance d′=d–0.4m). b. Dynamic time calculation: Terrain arrival time t_arrive = d′ / v′ (v′ is the real-time vehicle speed, in m / s); when the vehicle speed change rate is >10% / s (rapid acceleration / deceleration), the first-order inertial filter t_filtered = 0.7 × t_arrive + 0.3 × t_prev (t_prev is the calculated value of the previous moment) is enabled to avoid time jumps; c. Safety margin: For close-range terrain with t_arrive < 0.5 seconds, automatically add 0.1 seconds of safety time to reserve system response redundancy.

[0106] Step 4: Based on the delay characteristics of the hydraulic system, calculate the advance trigger time of the cutting table movement. Specifically: Determine the advance trigger time t_trigger, t_trigger = t_arrive – t_delay + t_correct, where t_arrive represents the terrain arrival time (s); t_delay represents the hydraulic delay time (s) retrieved from the mapping table; and t_correct represents the environmental correction (+0.2s when oil temperature < 25℃, +0.1s when load > 1500kg). For example, if the terrain arrival time is 4.8s, the delay time is 1.2s, and the low temperature correction is +0.2s, then t_trigger = 4.8 – 1.2 + 0.2 = 3.8s (i.e., the action command is issued 3.8 seconds in advance).

[0107] Step 5, Correction, specifically: When crop moisture H ≤ 50%, the header speed correction factor is 1.0 and the height correction is 0 cm; when 50% < crop moisture H ≤ 70%, the header speed correction factor is 0.9 and the height correction is +2 cm; when crop moisture H > 70%, the header speed correction factor is 0.7 and the height correction is +3 cm. Final target height = base height + density correction value + humidity correction value; final velocity coefficient = density flow coefficient × humidity velocity coefficient (multi-parameter coupling adjustment is achieved by using the product rule).

[0108] Step 6: Generate hydraulic control commands and send them to the hydraulic actuation module at the predicted trigger time. Specifically: The hydraulic control command is a 32-bit binary code, where bits 1–8: target height (resolution 0.1cm); bits 9–16: flow parameters (corresponding to the electro-hydraulic proportional valve control current 0–5V); bits 17–24: trigger timestamp (accurate to 0.01s); bits 25–32: checksum (CRC16 check). The system employs time-triggered CAN bus communication, sending a pre-command to the hydraulic module buffer 0.1 seconds before the trigger time. The trigger is activated by the local clock of the hydraulic module, ensuring a time accuracy of ≤5ms.

[0109] Step 7: The hydraulic system executes the action, and the displacement sensor provides feedback on the actual adjustment result, specifically: After receiving the command, the hydraulic module converts the flow parameters into proportional valve control current and drives the valve group to move through the PWM signal. The displacement sensor outputs a position signal (4–20mA current signal) every 1ms, which the hydraulic module converts into a digital quantity (resolution 0.01cm) and sends feedback data to the central controller every 10ms. When the deviation between the feedback position and the target position is >2cm, the central controller immediately generates a compensation command (adjustment amount = deviation × 0.8) and sends it in the next control cycle (0.1 seconds) to ensure that the dynamic error is ≤3cm.

[0110] Step 8: The central control module compares the actual results with the target value. If there is a deviation (>2cm), compensation and adjustment will be carried out in the next cycle.

[0111] Step 9: Repeat steps 1–8 to achieve continuous predictive height adjustment.

[0112] This invention employs a composite detection scheme combining millimeter-wave radar and ultrasonic sensors to achieve long-range, precise terrain detection within a 10-15 meter range in front of the harvester. This overcomes the limitation of traditional technologies where sensors can only detect the area near the header, providing a crucial data foundation for predictive header height adjustment. Furthermore, based on the action delay characteristics of the hydraulic system, an "adjustment height-delay time mapping table" is established through experimental calibration, thereby constructing a correlation model of "terrain arrival time-action lead time." By triggering header adjustment actions in advance based on prediction results, the inherent delay of the hydraulic system is actively compensated, solving the inherent adjustment lag problem of the "real-time detection-instant response" mode from the control logic level. A comprehensive correction mechanism integrating multiple parameters such as vehicle speed, crop density, and crop humidity can dynamically adjust the target height and action speed of the header according to different operating scenarios, significantly improving the system's adaptability and adjustment accuracy under different crop conditions and environments. A complete control logic of "forward detection - trend prediction - delay compensation - multi-parameter correction - closed-loop verification" has been constructed, forming a full-chain technical solution from the prediction of the terrain ahead to the precise execution of the cutter platform. It is different from the "real-time detection - instant response" logic adopted by existing technologies at the control architecture level.

[0113] This invention fundamentally solves the problem of adjustment lag caused by hydraulic system action delay by using long-range forward terrain detection and multi-parameter fusion to predict adjustment timing. In high-speed operation scenarios (e.g., vehicle speed 8 km / h), a 10-meter forward detection distance provides approximately 4.5 seconds of prediction time, completely covering the maximum delay time of the hydraulic system (2 seconds), thus ensuring that the header height adjustment action is completed when it reaches the terrain change point, effectively avoiding header touching the ground or missing the cut. In hilly areas with complex terrain, the terrain trend prediction algorithm and vehicle speed dynamic correction can identify and respond to continuous undulating terrain in advance, reducing the number and magnitude of frequent hydraulic system adjustments and reducing system wear. At the same time, the multi-parameter fusion correction mechanism improves the system's adaptability to different operation scenarios: in high-density crop areas, by increasing the target header height and reducing the adjustment speed, the risk of straw blockage is effectively reduced; in high-humidity environments, speed control effectively avoids crop entanglement problems. Compared with existing technologies, this invention has significant improvements in harvesting quality (such as a reduction of ≥8% in missed cutting rate and a reduction of ≥10% in impurity contamination rate) and operating efficiency (an increase of ≥15% in continuous operation time without failure). Moreover, it does not rely on high-cost high-end hydraulic components, and has the characteristics of cost control and high reliability, making it more suitable for widespread application in harvesters.

[0114] like Figure 2 As shown, an embodiment of the present invention provides a harvester header height prediction adjustment method, employing any of the above-mentioned harvester header height prediction adjustment systems. The method includes: S1. Through the forward-looking perception module, the terrain contour map is generated in real time based on the terrain data in front of the harvester. S2. Through the multi-parameter acquisition module, the harvester's speed, crop density in front of the harvester, and crop moisture data are collected in real time. S3. Through the central control module, receive and perform multi-parameter fusion and predictive control on terrain contour map, vehicle speed, crop density and crop humidity data to generate hydraulic control commands; S4. Enable the hydraulic actuator module to receive and adjust the header height of the harvester according to the hydraulic control command.

[0115] It should be noted that the beneficial effects of the harvester header height prediction adjustment method provided in the above embodiments are the same as the beneficial effects of the harvester header height prediction adjustment system provided in the above embodiments. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0116] A harvester according to an embodiment of the present invention includes any of the above-mentioned harvester header height prediction adjustment systems.

[0117] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-described harvester header height prediction adjustment methods. The electronic device can also be a terminal device, which can be any device capable of installing applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0118] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-described harvester header height prediction adjustment methods.

[0119] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0120] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A harvester header height prediction adjustment system, characterized in that, include: The system includes a forward-looking perception module, a multi-parameter acquisition module, a central control module, a hydraulic actuation module, and a human-machine interaction module. The forward-looking perception module is used to: detect and generate a terrain contour map in real time based on terrain data in front of the harvester; The multi-parameter acquisition module is used to: collect real-time data on the harvester's speed, crop density in front of the harvester, and crop humidity. The central control module is used to: receive and perform multi-parameter fusion and predictive control on the terrain contour map, the vehicle speed, the crop density and the crop humidity data, and generate hydraulic control commands; The hydraulic actuator module is used to: receive and adjust the header height of the harvester according to the hydraulic control command.

2. The harvester header height prediction adjustment system according to claim 1, characterized in that, The forward-looking perception module includes radar, ultrasonic sensors and a preprocessing unit. The forward-looking perception module uses radar and ultrasonic sensors to detect terrain data in front of the harvester in real time. The preprocessing unit filters and stitches together the terrain data in front of the harvester, which is detected in real time by radar and ultrasonic sensors, to generate the terrain contour map.

3. The harvester header height prediction adjustment system according to claim 2, characterized in that, The horizontal axis of the terrain contour map represents the forward horizontal distance, and the vertical axis represents the height difference between the ground and the cutter reference plane; the preprocessing unit is specifically used for: After filtering out clutter from the terrain data obtained by radar and ultrasonic sensors, a weighted fusion is performed in the overlapping area to generate a terrain contour map with the cut-off reference plane as the zero point.

4. The harvester header height prediction adjustment system according to claim 1, characterized in that, The multi-parameter acquisition module includes a camera installed above the cutting table, an infrared humidity sensor installed near the cutting blade on the cutting table, and a data processing unit. The data processing unit is used to: perform image recognition on the image captured by the camera in front of the harvester to obtain the crop density; The data processing unit is also used to: analyze the infrared reflectance of the crop straw in front of the harvester detected by the infrared humidity sensor to obtain the crop humidity.

5. The harvester header height prediction adjustment system according to claim 1, characterized in that, The central control module is specifically used for: Based on the vehicle speed and the terrain contour map, the time it takes for the terrain feature points in front of the harvester to reach the header is calculated, and the pre-stored adjustment height-delay time mapping table is queried according to the time to determine the advance of the action and generate the initial hydraulic control command. The initial hydraulic control command is modified based on the crop density and crop humidity data to obtain the final hydraulic control command.

6. The harvester header height prediction adjustment system according to claim 5, characterized in that, It also includes a human-machine interaction module, which includes a touch screen for displaying real-time terrain contours, vehicle speed, crop density, crop humidity and header height, and receiving user-set target header height value and crop type parameters.

7. A method for predictive adjustment of header height in a harvester, characterized in that, The harvester header height prediction adjustment system according to any one of claims 1 to 6 includes the following method: The forward-looking perception module detects and generates a terrain contour map in real time based on the terrain data in front of the harvester; The multi-parameter acquisition module collects real-time data on the harvester's speed, crop density in front of the harvester, and crop moisture content. The central control module receives and performs multi-parameter fusion and predictive control on the terrain contour map, vehicle speed, crop density and crop humidity data to generate hydraulic control commands. The hydraulic actuator module receives and adjusts the header height of the harvester according to the hydraulic control command.

8. A harvester, characterized in that, Includes a harvester header height prediction adjustment system as described in any one of claims 1 to 6.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the harvester header height prediction adjustment method of claim 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the harvester header height prediction adjustment method of claim 7.

Images