Multi-sensor fusion-based crop harvesting and stacking collaborative control method and system

Abstract

The application provides a multi-sensor fusion crop harvesting and stacking cooperative control method and system. The method comprises the following steps: extracting the morphological features of crops from crop image data, generating a crop maturity recognition result according to the difference between the reflectance spectrum data of two groups of wave bands; determining a crop quality recognition result according to the morphological features and the crop maturity recognition result; sending an opening and closing control instruction to the pneumatic guide plate at the stacking port corresponding to the crop quality recognition result in the multi-stage stacking port at the end of the conveying belt, so as to transport the crops on the conveying belt to the corresponding stacking area; when the pressure value of the stacking area is greater than a preset pressure threshold, adjusting the harvesting rate of the crop harvesting and stacking all-in-one machine according to the pressure value. The technical scheme provided by the application can analyze the crop morphology and maturity through image and spectrum data, intelligently control the stacking and harvesting, and realize accurate and efficient crop quality grading and harvesting management.

Description

Technical Field

[0001] This application relates to the field of agricultural production technology, and in particular to a multi-sensor fusion method and system for collaborative control of crop harvesting and stacking. Background Technology

[0002] In modern agricultural production, crop harvesting and sorting are crucial steps. To improve production efficiency and ensure crop quality, a technology capable of automatically identifying crop maturity and sorting crops is needed. This technology requires the ability to acquire real-time image and reflectance spectral data of the crops. By analyzing this data, the maturity and quality of the crops can be determined, and the sorting equipment can be controlled accordingly to transport the crops to the appropriate sorting areas. Furthermore, the harvesting rate needs to be dynamically adjusted based on the pressure values ​​in the sorting areas to ensure the stability and efficiency of the sorting process.

[0003] Existing solutions address the automation needs of crop harvesting and sorting, including a technology based on machine vision and spectral analysis. This solution uses cameras and spectrometers mounted on the conveyor belt to acquire real-time image and reflectance spectral data of the crops. Image processing algorithms extract morphological features of the crops, and spectral data analysis reveals their maturity. Based on this data, the system can automatically control the sorting equipment, transporting the crops to different sorting areas. Simultaneously, the system is equipped with pressure sensors to monitor pressure values ​​in the sorting areas and adjust the harvesting rate accordingly.

[0004] While existing solutions have achieved some degree of automation in crop harvesting and sorting, several issues remain. First, the reliance on a single spectral band for maturity identification can lead to inaccurate results, particularly given the poor stability of spectral data under varying lighting conditions. Second, existing image processing algorithms are prone to misidentification or underidentification when dealing with crop images in complex backgrounds or with multiple occlusions, affecting sorting accuracy. Finally, the limited response speed and accuracy of pressure sensors can result in untimely and imprecise adjustments to the harvesting rate, impacting overall production efficiency. Summary of the Invention

[0005] This application provides a multi-sensor fusion-based collaborative control method and system for crop harvesting and piling, which solves the problems in the prior art of inaccurately identifying maturity and quality, low piling efficiency, and mismatch between harvesting rate and piling pressure during crop harvesting and piling.

[0006] In a first aspect, embodiments of this application provide a multi-sensor fusion-based collaborative control method for crop harvesting and stacking, including:

[0007] Acquire image data of crops on the conveyor belt in the integrated crop harvester and stacker, as well as reflectance spectral data of crop stems and leaves in two bands;

[0008] The morphological features of crops are extracted from the crop image data, and the crop maturity identification result is generated based on the difference between the reflectance spectral data of the two sets of bands.

[0009] Based on the morphological characteristics and the crop maturity identification results, the crop quality identification results are determined;

[0010] Send an opening and closing control command to the pneumatic guide plate at the multi-stage sorting port at the end of the conveyor belt that corresponds to the crop quality identification result, so as to transport the crops on the conveyor belt to the corresponding sorting area.

[0011] When the pressure value in the splitting area exceeds the preset pressure threshold, the harvesting rate of the integrated crop harvester and splitting machine is adjusted according to the pressure value.

[0012] Optionally, the step of extracting morphological features of crops from the crop image data and generating crop maturity identification results based on the difference between the two sets of reflectance spectral data includes:

[0013] The crop image data is segmented into the outline of the stem and leaves of individual crops to extract the central axis of the stem and the leaf area of ​​each crop, generating a stem-leaf separation image.

[0014] The morphological features of crops are extracted from the stem-leaf separation images. The morphological features include: maximum stem length, maximum leaf unfolding width, and stem bending arc angle, in order to construct a set of morphological features for all crops.

[0015] Extract the first band reflection intensity value and the second band reflection intensity value of each crop at multiple same spatial coordinate points from the reflection spectral data of the two sets of bands. Calculate the absolute value difference between the first band reflection intensity value and the second band reflection intensity value of each crop at each spatial coordinate point to generate a band reflection difference sequence for each crop.

[0016] Based on the spatial coordinates of the band reflection difference points that are greater than the preset band reflection difference threshold in the band reflection difference sequence of each crop, chlorophyll abnormal regions are formed.

[0017] Identify the boundary coordinates of the chlorophyll abnormal regions corresponding to all crops, and generate crop maturity identification results containing maturity level indicators based on the boundary coordinates and the morphological feature set.

[0018] Optionally, the step of segmenting the stem and leaf contours of individual crops in the crop image data to extract the central axis of the stem and leaf area of ​​each crop and generate a stem-leaf separated image includes:

[0019] The crop image data is spatially transformed from a red-green-blue color space to a preset color space used to enhance the contrast of the stem and leaf areas;

[0020] Extract the set of difference pixels between crop plants and background from crop image data in the preset color space;

[0021] An initial binary mask is generated based on the set of difference pixels. The minimum bounding rectangle of a single crop is determined by a connected component labeling algorithm. The connected regions in the crop image of the preset color space are separated according to the minimum bounding rectangle.

[0022] The stem region is located based on the pixels within the minimum bounding rectangle of each individual crop plant, and the location result is obtained. The location result includes the base coordinates of the stem and the coordinates of the apical growth point.

[0023] Based on the base coordinates and apical growth point coordinates of the stem, a preset extension fitting function is used to determine the coordinate sequence of the stem midline, and the dynamic variation area of ​​the stem width is determined according to the coordinate sequence of the stem midline.

[0024] Extract the edge contour point set of the leaf area from the remaining pixels, and connect all the edge contour points in the edge contour point set by a preset polygon closure rule to generate the boundary data of the leaf area. The remaining pixels are the other pixels in the set of differences except for the area where the stem width changes dynamically.

[0025] The coordinate sequence of the central axis of the stem and the boundary data of the leaf area are superimposed on the crop image in a preset color space to generate a stem-leaf separation image containing spatial separation markers of stem and leaf.

[0026] Optionally, the step of extracting the edge contour point set of the leaf surface region from the remaining pixels and connecting all the edge contour points in the edge contour point set according to a preset polygon closure rule to generate the boundary data of the leaf surface region includes:

[0027] In the remaining pixels, the pixel brightness values ​​are scanned in blocks along the horizontal direction, and the boundary lines where the brightness difference between adjacent blocks exceeds a preset threshold are marked as candidate edge points of the leaf surface.

[0028] Based on the candidate edge points on the leaf surface, continuous detection is performed along the vertical direction, and candidate edge points that satisfy the condition that the left side is the stem region and the right side is the leaf surface region are retained to generate a preliminary edge contour point set.

[0029] From the initial set of edge contour points, select the starting point closest to the growth point at the top of the stem, and trace the connectivity of adjacent pixels point by point in a clockwise direction. During the tracing process, skip isolated points and fill in the broken points to generate an ordered edge contour sequence.

[0030] Based on the ordered edge contour sequence, when the distance between two adjacent points exceeds a preset distance, an interpolation point is inserted between the two adjacent points to make the density of contour points uniform.

[0031] According to the preset polygon closure rules, the interpolated contour points are connected end to end in a clockwise order to form a closed polygon boundary. The closed polygon is then filled with the mask markers of the leaf area to generate the boundary data of the leaf area.

[0032] Optionally, determining the crop quality identification result based on the morphological characteristics and the crop maturity identification result includes:

[0033] And extract the maturity level classification identifier from the crop maturity identification results;

[0034] Determine whether the stem bending arc angle in the morphological features exceeds the preset stem bending threshold. If so, determine that the upright morphological condition is met and generate a stem breakage mark.

[0035] Based on the maturity level classification indicator, and combined with the difference between the maximum leaf unfolding width and the preset lower limit of leaf width in the morphological characteristics, the leaf shrinkage index is calculated.

[0036] When the stem breakage mark is present and the leaf shrinkage index exceeds a preset shrinkage threshold, a crop quality identification result with an inferior quality label is generated; or, when the stem bending arc angle does not exceed the threshold and the maturity grade is marked as mature, a crop quality identification result with a superior quality label is generated.

[0037] Optionally, the step of extracting the first band reflection intensity value and the second band reflection intensity value of each crop at multiple corresponding spatial coordinate points from the reflection spectral data of the two sets of bands, and calculating the absolute value difference between the first band reflection intensity value and the second band reflection intensity value of each crop at each spatial coordinate point to generate a band reflection difference sequence for each crop includes:

[0038] Based on the reflectance spectral data of two bands, spatial coordinate point mapping matching is performed to locate the first band reflectance intensity matrix and the second band reflectance intensity matrix corresponding to each crop in the crop image data.

[0039] According to the row and column index order of the spatial coordinate points, the pixel-level first band reflection intensity value is extracted from the first band reflection intensity matrix, and the pixel-level second band reflection intensity value with the same row and column index is obtained from the second band reflection intensity matrix simultaneously.

[0040] For each spatial coordinate point, the absolute value of the first band reflection intensity value and the second band reflection intensity value are subtracted point by point to generate a single-point band reflection difference value.

[0041] For each crop, the single-point band reflection difference values ​​of all spatial coordinate points within the minimum bounding rectangle of a single crop are arranged in row priority order to form a band reflection difference sequence for each crop.

[0042] Optionally, the step of performing continuous detection along the vertical direction based on the candidate leaf edge points, retaining candidate edge points that satisfy the condition that the left side is a stem region and the right side is a leaf region, to generate a preliminary edge contour point set, includes:

[0043] Traverse each candidate edge point in the set of candidate edge points on the leaf surface, obtain the pixel coordinates of the candidate edge points on the leaf surface, and extend the preset pixel width to the left and right sides along the vertical direction to form a horizontal detection band;

[0044] Based on the horizontal detection band, scan the mask markers of N consecutive pixels to the left. If all pixels belong to the mask markers of the stem region, then mark the left side as the stem region. Scan the mask markers of N consecutive pixels to the right. If all pixels belong to the mask markers of the leaf region, then mark the right side as the leaf region.

[0045] Candidate points that satisfy the condition that the left side is a stem region and the right side is a leaf region will be retained in the temporary point set;

[0046] Based on the temporary point set, the distance between adjacent points is checked. If the distance between adjacent points exceeds a preset fracture threshold, linear interpolation points are inserted along the fracture direction to achieve contour continuity and generate the preliminary edge contour point set.

[0047] Secondly, embodiments of this application provide a multi-sensor fusion-based collaborative control system for crop harvesting and piling, comprising:

[0048] The acquisition module is used to acquire image data of crops on the conveyor belt in the integrated crop harvester and stacker, as well as reflectance spectral data of crop stems and leaves in two bands.

[0049] The generation module is used to extract morphological features of crops from the crop image data and generate crop maturity identification results based on the difference between the two sets of reflectance spectral data.

[0050] The identification module is used to determine the crop quality identification result based on the morphological features and the crop maturity identification result;

[0051] The control module is used to send opening and closing control commands to the pneumatic guide plate at the multi-stage sorting port at the end of the conveyor belt that corresponds to the crop quality identification result, so as to transport the crops on the conveyor belt to the corresponding sorting area.

[0052] The adjustment module is used to adjust the harvesting rate of the integrated crop harvester and stacker based on the pressure value when the pressure value in the stacking area is greater than a preset pressure threshold.

[0053] Thirdly, embodiments of this application provide a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are to be invoked and executed by the processing component to implement a multi-sensor fusion-based crop harvesting and stacking collaborative control method as described in any of the first aspects.

[0054] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a computer, implements a multi-sensor fusion-based collaborative control method for crop harvesting and piling as described in any of the first aspects.

[0055] This application provides a multi-sensor fusion collaborative control method for crop harvesting and stacking. The method includes: acquiring crop image data and reflectance spectral data of crop stems and leaves in two sets of wavebands on the conveyor belt of a crop harvesting and stacking machine; extracting morphological features of the crop from the crop image data; generating a crop maturity identification result based on the difference between the reflectance spectral data of the two sets of wavebands; and determining a crop quality identification result based on the morphological features and the crop maturity identification result.

[0056] An opening and closing control command is sent to the pneumatic guide plate at the multi-stage sorting port at the end of the conveyor belt, which corresponds to the crop quality identification result, so as to transport the crops on the conveyor belt to the corresponding sorting area; when the pressure value in the sorting area is greater than the preset pressure threshold, the harvesting rate of the integrated crop harvesting and sorting machine is adjusted according to the pressure value.

[0057] This application embodiment acquires image data of crops on a conveyor belt and reflectance spectral data of two bands. Morphological features of the crops are extracted from the image data, and the difference between the two bands of reflectance spectral data is used to generate a crop maturity identification result, thereby determining the crop quality identification result. Based on the quality identification result, an opening and closing control command is sent to the pneumatic guide plate at the corresponding piling port to achieve precise piling of crops. Simultaneously, by monitoring the pressure value in the piling area, the harvesting rate is dynamically adjusted when the pressure exceeds a preset threshold to ensure the stability and efficiency of the piling process. This method achieves automated identification and piling of crop maturity and quality, improves the intelligence level and operational efficiency of the harvesting and piling integrated machine, and solves the problems of inaccurate piling and mismatch between harvesting rate and piling pressure in traditional methods.

[0058] Furthermore, by segmenting the stem and leaf contours of individual plants in crop image data, the color space is converted from red-green-blue to a preset color space to enhance contrast. The set of difference pixels between the crop plant and the background is extracted, an initial binary mask is generated, and a connected component labeling algorithm is used to separate adhered regions. The coordinates of the stem base and apical growth points are located, the stem central axis is fitted, and the dynamic variation region of stem width is determined. Simultaneously, the edge contour point set of the leaf area is extracted from the remaining pixels to generate leaf boundary data. The stem central axis and leaf boundary data are superimposed to generate a stem-leaf separation image. Morphological features such as stem length, leaf width, and curvature are extracted from the stem-leaf separation image. A band reflectance difference sequence is generated by combining the difference between two sets of band reflectance spectral data to identify chlorophyll abnormal regions and their boundary coordinates. Finally, a crop maturity identification result containing maturity grading indicators is generated. Through image processing and spectral analysis techniques, accurate segmentation and morphological feature extraction of crop stem and leaf regions are achieved. Combined with the band reflectance difference sequence, chlorophyll abnormal regions are identified, generating high-precision crop maturity grading results. This technology solves the problems of inaccurate extraction of crop morphological features and low efficiency of maturity identification in traditional methods, providing reliable data support for crop quality grading and automated sorting, and improving the intelligence level and operational accuracy of integrated crop harvesting and sorting machines.

[0059] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1A flowchart illustrating a multi-sensor fusion-based collaborative control method for crop harvesting and piling is provided in this application embodiment;

[0062] Figure 2 A flowchart for generating crop maturity identification results is provided in an embodiment of this application;

[0063] Figure 3 This is a schematic diagram of the structure of a multi-sensor fusion collaborative control system for crop harvesting and piling, provided in an embodiment of this application.

[0064] Figure 4 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0065] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0066] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 11, 12, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0067] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] To address the problems of inaccurate identification of maturity and quality, low sorting efficiency, and mismatch between harvesting rate and sorting pressure in existing technologies for crop harvesting and sorting, this application provides a multi-sensor fusion collaborative control method for crop harvesting and sorting. This method employs the following concept: By integrating multi-source data acquisition and intelligent decision-making technologies, visual sensors and spectrometers are first deployed to acquire crop image data and near-infrared and visible dual-band reflectance spectral data on the conveyor belt. Image processing algorithms are used to extract morphological features such as stem and leaf length and curvature. Simultaneously, a maturity discrimination model is constructed based on the difference in reflectance between the two bands. Then, morphological feature quantification indicators and maturity levels are weighted and fused to generate a comprehensive quality score, which is then mapped to a preset sorting level. The controller sends sorting commands to the corresponding pneumatic guide vanes to achieve graded flow guidance based on quality. Simultaneously, pressure sensors are integrated to monitor the load on the sorting bins in real time. When the pressure exceeds the limit, an algorithm dynamically adjusts the harvesting motor speed, forming a dynamic closed-loop control link of "perception-analysis-execution-feedback," ultimately achieving collaborative optimization of intelligent graded harvesting and sorting operations for crops.

[0069] Figure 1 A flowchart illustrating a multi-sensor fusion-based collaborative control method for crop harvesting and piling, as provided in this application embodiment, is shown below. Figure 1 As shown, the method includes:

[0070] S11. Acquire crop image data and crop stem and leaf reflectance spectral data in two bands on the conveyor belt of the integrated crop harvester and stacker.

[0071] The crops can include leeks, celery, corn, wheat, rice, sugarcane, etc. The crop image data includes red, green, and blue images of the crops on the conveyor belt, as well as depth information. The two sets of reflectance spectral data refer to the difference in reflectance between the crop stems and leaves in the visible and near-infrared bands. The image data is a 3D point cloud generated by a multi-view stereo vision system, used to calculate the stacking height and density distribution of the crops. The spectral data uses a spectrophotometer to measure the reflectance of the two bands in real time.

[0072] In this embodiment, crop image data is acquired by an industrial camera installed directly above the conveyor belt of the integrated crop harvester and stacker. Simultaneously, two sets of spectral sensors, operating in the near-infrared and visible light bands respectively, are symmetrically arranged on both sides of the conveyor belt. A synchronous triggering mechanism is used to collect the reflectance spectral data of the crop stems and leaves. The acquisition process includes preprocessing steps such as sensor calibration, illumination compensation, and outlier removal to ensure that the image data resolution reaches 0.5 mm / pixel and the spectral sampling interval is 10 ms / sample.

[0073] S12. Extract morphological features of crops from crop image data, and generate crop maturity identification results based on the difference between the reflectance spectral data of the two sets of bands.

[0074] Morphological features include crop length, stem and leaf uprightness, root cut smoothness, stem diameter, leaf curvature, and leaf tip integrity. Maturity identification results are classification labels based on spectral differences and preset thresholds. Maturity level classifications in the results can be extremely mature, moderately mature, or underripe. Stem diameter is measured using a skeleton extraction algorithm to measure the stem midline width. Leaf curvature can refer to the degree of curvature quantified based on curvature integrals. Maturity determination uses partial least squares regression to establish a mapping relationship between the difference in near-infrared and visible light reflectance spectral data and sugar content and fiber content, outputting the maturity level. In crop image data on the conveyor belt after harvesting, since the lower surface of the crop cannot be directly observed, relying solely on upper surface image data and reflectance spectral data may lead to incomplete or inaccurate maturity identification results. Therefore, to further improve the overall accuracy of crop maturity identification, the following methods can be adopted: 1. Utilize multi-angle image acquisition by installing multiple cameras on both sides or above the conveyor belt to collect crop image data from different angles. First, multi-angle image fusion technology can be used to reconstruct the three-dimensional morphology of crops, thereby indirectly obtaining information about the lower surface. Second, based on reflectance spectrum inference, the maturity of the lower surface can be inferred by utilizing the characteristics of reflectance spectrum data. Assuming that the upper and lower surfaces of crops have consistent or correlated spectral reflectance characteristics, the maturity of the lower surface can be inferred from the spectral data of the upper surface. Third, combining physical models and prior knowledge, the morphological characteristics and maturity of the lower surface can be inferred by utilizing physical models and prior knowledge of crop growth. For example, the stems and leaves of crops exhibit certain symmetry and regularity during growth, and the characteristics of the lower surface can be inferred from the characteristics of the upper surface.

[0075] In this embodiment, a convolutional neural network is used to extract features from crop image data, and an edge detection algorithm is used to calculate morphological features such as leaf length and leaf spread. For two sets of reflectance spectral data, the reflectance difference at specific wavelength points is calculated after baseline correction. When the difference exceeds 0.35, it is determined to be in the mature stage; 0.25-0.35 is in the suitable harvest stage; and below 0.25 is in the immature stage. Finally, a probability distribution vector containing maturity levels is generated.

[0076] S13. Based on the morphological characteristics and crop maturity identification results, determine the crop quality identification results.

[0077] The quality identification results include a weighted evaluation value of the overall morphological score and maturity level.

[0078] In this embodiment, multimodal data fusion technology is used to weight and fuse morphological feature vectors with maturity identification results: total quality score = 0.6 × maturity coefficient + 0.4 × morphological coefficient. When the total score is ≥0.85, it is classified as premium grade; 0.7-0.84 is classified as first grade; and others are classified as ordinary grade. At the same time, a mapping table between quality identification results and stacking points is established.

[0079] Alternatively, the coefficients for crop maturity identification results can be adaptively adjusted. For example, extremely mature, moderately mature, and undermature crops correspond to maturity coefficients in the ranges of 0.8-1.0, 0.5-0.79, and 0-0.49, respectively.

[0080] Alternatively, the crop quality identification results can be classified according to the following rules: (1) Special grade, score ≥ 0.9: suitable for ripening or above, leaf damage ≤ 5%, and no breakage; (2) Grade 1, 0.7 ≤ score < 0.9: suitable for ripening or above, leaf damage ≤ 15%, and normal bending or upright; (3) Qualified, 0.5 ≤ score < 0.7: underripe or above, leaf damage ≤ 30%, and slight breakage is allowed, such as an arc angle of 60°-90°; (4) Inferior quality, score < 0.5: underripe, leaf damage > 30%, and severe breakage, such as an arc angle > 90°.

[0081] S14. Send an opening and closing control command to the pneumatic guide plate at the multi-stage sorting port at the end of the conveyor belt that corresponds to the crop quality identification result, so as to transport the crops on the conveyor belt to the corresponding sorting area.

[0082] The pneumatic guide vane is a retractable baffle driven by a solenoid valve. The sorting openings are divided into three levels according to quality: A (high quality), B (qualified), and C (substandard). In this embodiment, the sorting logic is triggered based on the quality result Q: if Q ≥ 85, the A-level sorting opening is opened; if 70 ≤ Q < 85, the B-level sorting opening is opened; if Q < 70, the C-level sorting opening is opened. The guide vane adjusts its opening angle using a proportional-integral-derivative (PID) controller to ensure that crops accurately fall into the corresponding areas. The crops on the conveyor belt are transported to their corresponding piles via a multi-level pile configuration: three pile openings are set at the end of the conveyor belt: a premium zone, a standard zone, and a secondary zone. Each pile opening is equipped with a pneumatic guide plate. The corresponding guide plate is triggered based on the quality identification result. The priority pile logic is as follows: when the maturity meets the standard and the morphological characteristics are excellent, the crops are guided to the premium pile area first; crops that meet the maturity standard but do not meet the morphological standard enter the standard zone; crops that do not meet the standard or contain impurities enter the secondary zone. A high-pressure air spray device is added in the middle section of the conveyor belt to perform targeted blowing based on image recognition of impurity location, thereby improving the impurity removal rate.

[0083] In this embodiment, a pulse width modulation control signal is sent to the pneumatic guide vane of the target stacking port via a programmable logic controller to control the opening duration of the solenoid valve. Specifically, based on the quality identification results, a preset stacking rule base is queried, and a control instruction packet containing position codes and action parameters is sent via bus to the actuator of the corresponding stacking port.

[0084] S15. When the pressure value in the splitting area is greater than the preset pressure threshold, adjust the harvesting rate of the integrated crop harvester and splitting machine according to the pressure value.

[0085] The pressure threshold is the full-load pressure of the pile area, which can be set to 20 kPa, 15 kPa, 10 kPa, etc. The harvesting rate is adjusted linearly based on the pressure sensor data to regulate the speed of the harvesting motor.

[0086] In this embodiment, a piezoresistive pressure sensor is used to monitor the load pressure in the stacking area in real time. When the detected pressure value exceeds a preset threshold, the harvesting rate is dynamically adjusted through a PID control algorithm: adjustment Δv = Kp × (P_actual - P_threshold) + Ki × ∫(P_actual - P_threshold) dt, where P_actual refers to the actual load pressure value in the stacking area detected in real time by the piezoresistive pressure sensor, and P_threshold refers to the maximum allowable pressure threshold preset by the system. The proportional coefficient Kp = 0.5 rpm / kg, and the integral coefficient Ki = 0.2 rpm / (kg·s). The calculated rotational speed command is sent to the harvesting motor driver via a protocol.

[0087] Alternatively, in S15, this embodiment can also set a calibration model for the pile size and pressure: Step 1: Deploy a lidar in the pile area, with a sampling frequency of 10Hz, to collect the crop pile volume in real time, and establish a three-dimensional mapping relationship based on the pressure sensor data: mass m = k1 × pressure P + k2 × volume V, where k1 and k2 are crop type coefficients, for example, for leafy vegetables k1 = 0.8, k2 = 0.2; for stem vegetables k1 = 0.6, k2 = 0.4; Step 2: Dynamic threshold setting: Set the mass threshold of the pile area (e.g., full silo mass M) based on the calibration model. When the calculated mass m ≥ 0.9M, the harvesting rate adjustment is triggered; Step 3, the rate adjustment algorithm can adopt PID control combined with mass prediction: adjustment amount Δv = Kp × (m - 0.9M) + Ki × ∫ (m - 0.9M) dt + Kd × d(m) / dt, where Kp, Ki, and Kd are proportional, integral, and derivative coefficients, which can be dynamically adapted according to the crop loft. When the loft is high, Kd is increased to reduce overshoot; Step 4, the model coefficients and PID parameters are updated in real time according to the crop type of the splitting area to ensure that the rate adjustment matches the actual splitting amount.

[0088] Here is a specific implementation example: In a crop planting base, a crop harvesting and sorting machine begins operation. First, a camera and a spectrometer acquire images and spectral data of the crops on the conveyor belt, respectively. The central processing unit identifies the morphological characteristics and maturity of the crops through image processing and spectral analysis, and assesses the quality grade of the crops accordingly. Based on the quality grade, pneumatic guide vanes are controlled to guide the crops to different sorting areas. When a pressure sensor in a sorting area detects that the pressure value exceeds a threshold, the central processing unit automatically reduces the harvesting rate to ensure the smooth progress of the sorting process.

[0089] By executing S11 to S15, the embodiments of this application realize the automatic identification and accurate piling of crop quality through the integrated crop harvesting and piling machine. At the same time, the harvesting rate is dynamically adjusted according to the pressure of the piling area, which improves the harvesting efficiency and piling accuracy, reduces manual intervention, and enhances the overall automation level of the operation.

[0090] In one possible embodiment, such as Figure 2 As shown, S12, extracting morphological features of crops from crop image data, and generating crop maturity identification results based on the difference between the reflectance spectral data of two sets of bands, including:

[0091] S121. Perform stem and leaf contour segmentation on the crop image data to extract the central axis of the stem and leaf area of ​​each crop, and generate a stem and leaf separation image.

[0092] The crop image data consists of images of crop plants acquired through cameras or other image acquisition devices. Stem-leaf contour segmentation is used to separate the stem and leaf regions in the crop image, thereby extracting their contours. The stem midline is the centerline of the stem, used to describe its morphological features. The leaf region is the area of ​​the crop leaf, used for subsequent extraction of leaf morphological features. The stem-leaf separated image is the image after segmentation, with the stem and leaf labeled separately.

[0093] In this embodiment, deep learning methods are used to preprocess the original image to enhance the contrast between stems, leaves, and background. Semantic segmentation algorithms are used to segment the outlines of individual stems and leaves in the crop image data. Convolutional neural networks are used to identify and segment the stems and leaf areas of individual crops. The central axis of the stem is extracted using a skeletonization algorithm, while the leaf area is separated using edge detection and region filling techniques, generating a binary image containing clear stem and leaf structures. This ensures that the stem and leaf boundaries do not overlap, providing basic data for subsequent morphological analysis.

[0094] S122. Extract the morphological features of crops from the stem-leaf separation image. The morphological features include: maximum stem length, maximum leaf spread width, and stem bending angle, in order to construct a set of morphological features for all crops.

[0095] Among these, the maximum stem length is the maximum length of the stem from the root to the tip. The maximum leaf spread width is the maximum width of the leaf in the horizontal direction. The stem curvature angle is the arc angle of the curved portion of the stem, used to describe the degree of curvature of the stem. The morphological feature set is the collection of morphological feature data for all crops.

[0096] In this embodiment, morphological features can be extracted based on image processing technology. Pixel coordinate tracking of the stem's central axis in the stem-leaf separation image is performed, and the geometric distance between consecutive points is calculated to determine the maximum stem length. The maximum unfolded width of the leaf can be obtained by calculating the maximum lateral span of the bounding rectangle of the leaf area. The stem bending arc angle can be analyzed using curve fitting methods on the central axis, extracting the segment with the largest curvature change and calculating its tangent angle difference. These features collectively constitute the morphological feature set of each crop, used to characterize its physical growth state.

[0097] S123. Extract the first band reflection intensity value and the second band reflection intensity value of each crop at multiple spatial coordinate points from the two sets of reflection spectrum data. Calculate the absolute value difference between the first band reflection intensity value and the second band reflection intensity value of each crop at each spatial coordinate point to generate a band reflection difference sequence for each crop.

[0098] The reflectance spectral data consists of reflectance intensity data of crop plants in different spectral bands acquired by a spectrometer. The first band reflectance intensity value is the reflectance intensity value in the first spectral band. The second band reflectance intensity value is the reflectance intensity value in the second spectral band. The band reflectance difference sequence is the sequence of absolute differences between the two band reflectance intensity values ​​for each crop at each spatial coordinate point.

[0099] In this embodiment, a reflectance difference sequence is generated through spectral data registration and difference calculation. Reflectance spectral data collected by visible and near-infrared band sensors are aligned according to spatial coordinates. The absolute value difference of the two band reflectance intensity values ​​at each coordinate point is performed to generate a sequence reflecting local spectral differences. This process ensures a one-to-one correspondence between data points through time synchronization and spatial interpolation techniques. The difference sequence is used to quantify the differences in spectral response characteristics in different regions.

[0100] Optionally, this embodiment of the application can supplement the sensor calibration and dynamic synchronization mechanism in S123. The image sensor and spectrometer are jointly calibrated using a checkerboard calibration plate (5mm corner spacing) to obtain the intrinsic parameter matrix (focal length, principal point coordinates) and extrinsic parameter matrix (rotation angle, translation). Perspective transformation is used to map the spatial coordinates of the spectral data to the image pixel coordinate system, compensating for differences in field of view and resolution. High-precision timestamps are configured for the image sensor and spectrometer, with a synchronization accuracy ≤1ms. Crop displacement is calculated based on conveyor belt encoder data (e.g., real-time speed v): if the sampling interval is Δt, the crop displacement on the conveyor belt is v×Δt. Positional offset caused by motion is compensated through coordinate translation. For coordinate points in the edge region, bilinear interpolation is used to correct the spatial alignment deviation between the spectral data and image data. Based on the calibrated coordinate mapping relationship, the first band reflection intensity matrix and the second band reflection intensity matrix corresponding to each crop are located. Pixel-level reflection intensity values ​​are extracted synchronously according to the row and column index order of the spatial coordinate points, and the difference is calculated.

[0101] S124. Based on the spatial coordinates of the band reflection difference points that are greater than the preset band reflection difference threshold in the band reflection difference sequence of each crop, chlorophyll abnormal regions are formed.

[0102] The preset band reflectance difference threshold is used to determine whether chlorophyll is abnormal, and it is usually determined experimentally. An abnormal chlorophyll region is defined as an area where the band reflectance difference is greater than the threshold, indicating an abnormal chlorophyll content.

[0103] In this embodiment, chlorophyll aberration regions are screened based on dynamic thresholds. Each data point in the band reflectance difference sequence is thresholded, and coordinate points exceeding a preset threshold are marked as aberration points. Adjacent aberration points are clustered into continuous regions using a region growing algorithm, and discrete noise points are removed using morphological filtering. Finally, closed regions characterizing chlorophyll distribution abnormalities or lesions are formed, and their spatial distribution range is recorded.

[0104] S125. Identify the boundary coordinates of the chlorophyll abnormal regions corresponding to all crops, and generate crop maturity identification results containing maturity level indicators based on the boundary coordinates and morphological feature set.

[0105] The boundary coordinates are the coordinates of the boundary points of the chlorophyll abnormal region. The maturity level classification is based on the chlorophyll abnormal region and morphological characteristics to classify the maturity of crops, such as immature, mature, and overripe.

[0106] In this embodiment, maturity grading is achieved through feature fusion. The boundary coordinates and morphological feature parameters of chlorophyll-abnormal regions are combined, and a machine learning model is used to analyze the correlation between the spatial distribution pattern of the abnormal regions and the stem and leaf morphology. A trained classifier maps the morphological-spectral features of each crop to a preset maturity level, generating identification results containing grading identifiers, thus providing a basis for sorting decisions.

[0107] Here is a specific example: In a crop planting base, images and reflectance spectral data of crop plants are first acquired using high-definition cameras and spectrometers. Next, image segmentation algorithms are used to separate the stems and leaves, extracting the stem midline and leaf areas. Then, morphological features such as stem length, leaf width, and stem bending angle are extracted from the separated images. Simultaneously, the reflectance difference sequence between two spectral bands is calculated from the spectral data, and chlorophyll aberration regions are marked according to preset thresholds. Finally, combining morphological features and chlorophyll aberration regions, a classification model is used to grade the maturity of each crop plant, outputting the maturity identification result.

[0108] By executing steps S121–S125, this application achieves refined analysis of the growth status of individual crops by integrating high-precision image segmentation and multispectral analysis technologies. Precise separation of stem and leaf structures provides a reliable foundation for morphological feature extraction, dynamic spectral difference analysis effectively captures chlorophyll distribution anomalies, and a machine learning-based grading strategy improves the accuracy of maturity assessment. The system can adapt to morphological variations in crops under different growth conditions, reducing the risk of quality misjudgment while ensuring sorting efficiency, and providing comprehensive quality assessment criteria for automated sorting.

[0109] In one possible embodiment, step 131, segmenting the stem and leaf contours of individual crops in the crop image data to extract the central axis of the stem and leaf area of ​​each crop, generating a stem-leaf separation image, includes:

[0110] Step a1: Spatial transformation of crop image data is performed to convert it from red-green-blue color space to a preset color space used to enhance the contrast of stem and leaf areas.

[0111] The red-green-blue color space is a way of representing colors in an image, consisting of three channels: red, green, and blue. The preset color space is a selected color space used to enhance the contrast between the stem / leaf area and the background.

[0112] In this embodiment, crop image data is converted from a red-green-blue color space to a preset enhanced contrast color space. The specific process is as follows: First, the characteristics of the stem and leaf regions in the red, green, and blue channels are analyzed, and a linear or non-linear color conversion formula is designed to generate new color components. This preset color space may include dimensions such as brightness, hue, and saturation, or combinations thereof, maximizing the pixel differences between the stems and leaves and the soil, shadows, and other background elements, providing an optimized data foundation for subsequent pixel separation.

[0113] Step a2: Extract the set of difference pixels between crop plants and background from crop image data in a preset color space.

[0114] The set of differing pixels is the set of pixels that differ from the background in a preset color space. The majority of these differing pixels belong to the crop area; a small number of misidentified background pixels may be mixed in.

[0115] In this embodiment, based on the generated image in a preset color space, a set of differential pixels is extracted using threshold segmentation or clustering algorithms. Specifically, the statistical characteristics of each pixel in the preset color space are calculated, a dynamic threshold or classification boundary is set, and the pixels are divided into two categories: "crop plants" and "background," forming a set of differential pixels containing all possible plant components. This set, by excluding background interference pixels and retaining candidate stem and leaf regions, provides data input for generating the initial binary mask. It should be noted that, to achieve the practical feasibility of single-plant crop segmentation, this embodiment can introduce a complex adhesion processing mechanism in step a2. For example, through adaptive threshold segmentation and morphological opening and closing operations, erosion followed by dilation can remove noise points and fine adhesions.

[0116] Step a3: Generate an initial binary mask based on the set of difference pixels, and determine the minimum bounding rectangle of a single crop using a connected component labeling algorithm, so as to separate the connected regions in the crop image in the preset color space according to the minimum bounding rectangle.

[0117] The initial binary mask is a matrix of the same size as the original image, where each pixel can only be 0 (background) or 1 (crop). The connected component labeling algorithm is used to identify connected foreground regions in the image. The minimum bounding box is the smallest rectangular region that can completely enclose a single crop. Adhesive regions are areas in the image where multiple crops overlap or touch.

[0118] In this embodiment, the set of differing pixels is first converted into a binary image to generate an initial binary mask. Next, a connected component labeling algorithm is used to identify all connected components in the mask, and the minimum bounding box of each connected component is calculated. For overlapping or connected bounding boxes, individual plants are separated by merging overlapping regions or setting an area threshold, ultimately outputting a precise set of bounding boxes for each individual crop, providing spatial constraints for subsequent stem localization.

[0119] Alternatively, step a3 can combine distance transformation and watershed algorithm, and use an improved connected component labeling algorithm to handle densely stacked scenes: calculate the shortest distance from the pixel to the background for the overlapping area, generate a distance transformation map, determine the seed point of a single plant through local minima, and then separate the severely overlapping stem and leaf areas through watershed segmentation; determine the minimum bounding rectangle of a single crop through the connected component labeling algorithm, and further separate the single plants for areas with a bounding box overlap of more than 30% based on the stem basal coordinate clustering algorithm.

[0120] Step a4: Locate the stem region based on the pixels within the minimum bounding rectangle of each individual crop plant to obtain the positioning results, which include the base coordinates of the stem and the coordinates of the apical growth point.

[0121] The stem region localization step determines the position of the stem in the image. The stem base coordinates are the coordinates of the starting point where the stem contacts the soil, and the apical growth point coordinates are the coordinates of the growth point at the top of the stem. This is done by scanning the pixel density distribution curve along a path perpendicular to the stem's central axis or its extension direction. Specifically, this step aims to determine the stem base coordinates and apical growth point coordinates by analyzing the pixel density distribution curve.

[0122] In this embodiment, the stem region is located within the bounding box of a single plant. The specific process is as follows: based on color or morphological features, projection histogram analysis or Hough transform is used to detect vertical straight lines, determining the coordinates of the stem's base and the coordinates of its apical growth point. For example, the base is determined by statistically analyzing pixel density peaks along the vertical direction, while the apex is located using gradient descent to find color abrupt changes. The positioning results form a pair of start and end point coordinates for the stem, providing endpoint constraints for fitting the central axis.

[0123] Step a5: Based on the base coordinates and apical growth point coordinates of the stem, a preset extension fitting function is used to determine the coordinate sequence of the stem midline, and the dynamic change area of ​​the stem width is determined according to the coordinate sequence of the stem midline.

[0124] The preset extended fitting function is a mathematical function used to fit the central axis of the stem. The coordinate sequence of the central axis of the stem is the set of coordinates of all points on the central axis of the stem. The dynamic variation region of stem width is the region where the stem width changes with height.

[0125] In this embodiment, a central axis coordinate sequence is generated using polynomial curve fitting or spline interpolation based on the stem endpoint coordinates. For example, starting from the base coordinates and ending at the top coordinates, the offset of the intermediate points is calculated according to a preset step size to form a continuous central axis. Subsequently, pixels are scanned along the normal direction of the central axis, the boundaries on both sides of the stem are determined through edge detection, the width at each position is dynamically calculated, and areas where the width change exceeds a threshold are marked, ultimately dividing the dynamically changing area of ​​the stem.

[0126] Step a6: Extract the edge contour point set of the leaf area from the remaining pixels, and connect all the edge contour points in the edge contour point set using a preset polygon closure rule to generate the boundary data of the leaf area. The remaining pixels are the pixels in the difference pixel set other than the area where the stem width changes dynamically.

[0127] Among them, the remaining pixels are pixels in the difference pixel set that do not belong to the area where the stem width dynamically changes. The edge contour point set is the set of edge points of the leaf surface region. The preset polygon closure rule is the rule used to connect the edge points to generate closed polygons. The boundary data of the leaf surface region is the set of coordinates of the boundary points of the leaf surface region.

[0128] In this embodiment, the marked stem region is first removed from the set of differing pixels, and the remaining pixels are the candidate leaf region. A set of leaf contour points is extracted using an edge detection algorithm and sorted clockwise. A polygon closure rule is used to connect the discrete points into a closed boundary, generating a sequence of vertex coordinates for the leaf polygon, forming boundary data. This data, together with the stem's central axis, constitutes the stem-leaf separation marker.

[0129] Step a7: Overlay the coordinate sequence of the stem midline with the boundary data of the leaf area onto the crop image in the preset color space to generate a stem-leaf separation image containing stem-leaf spatial separation markers.

[0130] The stem-leaf spatial separation marker is the result of marking the separation of stem and leaf regions in the image. The stem-leaf separation image contains images with stem and leaf separation markers.

[0131] In this embodiment, the stem midline coordinate sequence and leaf boundary data are superimposed onto a crop image in a preset color space. An image fusion algorithm preserves the original color information while highlighting markers, generating a stem-leaf separated image. In this image, the stem and leaf areas are distinguished by different colors and geometric shapes, and can be directly used for morphological measurement or automated harvesting control.

[0132] Here's a concrete example: In a crop planting base, red, green, and blue image data of crop plants are first acquired using high-definition cameras. Next, the images are converted from the red-green-blue space to a hue, saturation, and brightness color space to enhance the contrast between stems / leaves and the background. Then, thresholding is used to extract the set of differing pixels, generating an initial binary mask. A connected component labeling algorithm is used to determine the minimum bounding rectangle of each crop plant, separating adhered regions. Within each bounding box, morphological operations and skeletonization algorithms are used to locate the stem region, and the stem's central axis is fitted. Next, the edge contour point set of the leaf region is extracted, generating leaf boundary data. Finally, the stem's central axis and leaf boundary data are superimposed onto the hue, saturation, and brightness image to generate a stem-leaf separated image.

[0133] By executing steps a1 to a7, this application achieves accurate separation of stem and leaf regions in crop images through techniques such as color space conversion, threshold segmentation, connected component labeling, skeletonization algorithms, and edge detection. First, color space conversion enhances contrast. Then, threshold segmentation and connected component labeling separate adhered regions. Next, a skeletonization algorithm locates the stem region and fits the central axis. Finally, edge detection generates leaf boundary data. This method effectively improves the accuracy and efficiency of stem and leaf separation, providing a reliable foundation for subsequent morphological feature extraction and maturity analysis.

[0134] In one possible embodiment, step a6, extracting the edge contour point set of the leaf area from the remaining pixels, and connecting all the edge contour points in the edge contour point set using a preset polygon closure rule to generate the boundary data of the leaf area, includes:

[0135] Step a61: In the remaining pixels, scan the pixel brightness values ​​in blocks along the horizontal direction, and mark the boundary lines where the brightness difference between adjacent blocks exceeds a preset threshold as candidate edge points of the leaf surface.

[0136] Pixel brightness value refers to the grayscale or brightness value of each pixel in the image. The preset threshold is used to determine whether there is a difference in brightness between adjacent blocks, and is usually determined experimentally. The "horizontal direction" of scanning pixel brightness values ​​along the horizontal direction is relative to the image coordinate system. Specifically, the "horizontal direction" refers to the row direction of the image, while the "vertical direction" refers to the column direction. Blocking refers to dividing the image horizontally into several small regions, each containing a certain number of pixels. The purpose of blocking is to decompose the image into smaller units, facilitating the analysis of pixel brightness values ​​block by block, thereby detecting edge points of brightness differences.

[0137] In this embodiment, candidate edge points of the leaf surface are extracted by horizontally scanning the brightness differences of the remaining pixels. Specifically, the remaining pixels are divided into blocks of fixed width, and the average brightness value of the pixels within each block is calculated. The average brightness difference between adjacent blocks is compared; if the difference exceeds a preset threshold, all pixels on the boundary line of the adjacent blocks are marked as candidate edge points of the leaf surface. This step uses local brightness abrupt change detection to initially locate the boundary region between the leaf surface and the background or stem.

[0138] Step a62: Based on the candidate edge points on the leaf surface, perform continuity detection along the vertical direction, retain candidate edge points that satisfy the condition that the left side is the stem region and the right side is the leaf surface region, so as to generate a preliminary edge contour point set.

[0139] The continuity detection checks the vertical continuity of candidate edge points to ensure their validity. The left side represents the stem region, indicating that the left pixels of candidate edge points belong to the stem region. The right side represents the leaf surface region, indicating that the right pixels of candidate edge points belong to the leaf surface region. This initial set of edge contour points, after filtering, forms the preliminary set of leaf surface edge points.

[0140] In this embodiment, vertical continuity filtering is performed on candidate edge points. Specifically, for each generated candidate edge point, the system scans column by column along the vertical direction, checking the left and right region attributes of each point. If the left pixel of the current point belongs to the stem region and the right pixel belongs to the leaf candidate region, the point is retained; otherwise, it is discarded. This process uses spatial constraints to filter edge points that conform to the stem-leaf boundary characteristics, forming a preliminary edge contour point set and eliminating noise and false positives.

[0141] Step a63: From the initial edge contour point set, select the starting point closest to the growth point at the top of the stem, and trace the connectivity of adjacent pixels point by point in a clockwise direction. During the tracing process, skip isolated points and fill in the broken points to generate an ordered edge contour sequence.

[0142] The starting point is the starting point for edge contour tracing, typically chosen as the point closest to the stem tip growth point. Clockwise tracing connects edge points sequentially in a clockwise direction to form a closed contour. Isolated points are points not connected to other edge points. Breakpoints are missing points in the edge contour, requiring interpolation to complete them. An ordered edge contour sequence is a set of edge points arranged in order, used to generate a closed contour. The closest point is determined by calculating the Euclidean distance. Choosing the starting point closest to the stem tip growth point ensures that edge contour tracing begins at the junction of the leaf surface and stem, thus more accurately describing the leaf's edge contour.

[0143] In this embodiment, an ordered edge contour sequence is generated through connectivity tracing. The specific process is as follows: From the initial set of edge contour points, the point with the closest Euclidean distance to the stem tip growth point is selected as the starting point. Connectivity between adjacent pixels is traced point by point in a clockwise direction. An 8-neighborhood connectivity rule is used; if the distance between the current point and the next candidate point exceeds 1 pixel or a break exists, missing points are filled using linear interpolation. Finally, a continuous and ordered edge contour point sequence is generated, ensuring contour closure and topological integrity.

[0144] Step a64: Based on the ordered edge contour sequence, when the distance between two adjacent points exceeds the preset distance, interpolation points are inserted between the two adjacent points to make the contour point density uniform.

[0145] The preset distance is a threshold used to determine whether the distance between adjacent points is too large, and it is usually determined experimentally. The interpolation point is a new point inserted between two adjacent points to homogenize the density of contour points, using the maximum distance between two adjacent points as a reference.

[0146] In this embodiment, the contour point sequence is subjected to density homogenization processing. Specifically, the ordered edge contour sequence is traversed, and the Euclidean distance between adjacent points is calculated. If the distance exceeds a preset value, an interpolation point is inserted between the two points. Piecewise linear interpolation ensures a uniform distribution of contour points, eliminating uneven density caused by tracking and providing a smooth foundation for subsequent polygon closure.

[0147] Step a65: According to the preset polygon closure rules, connect the interpolated contour points end to end in a clockwise order to form a closed polygon boundary, and fill the closed polygon with the mask markers of the leaf area to generate the boundary data of the leaf area.

[0148] In this context, the closed polygon boundary is a closed boundary formed by connecting contour points. The mask identifier is a binary image used to mark the leaf surface region.

[0149] In this embodiment, a closed polygon boundary and a leaf surface mask are generated. The specific process is as follows: the interpolated contour points are connected in a clockwise order, and the connection order is adjusted using a preset polygon closure rule to ensure that the polygons do not self-intersect. Subsequently, a scanline filling algorithm is used to fill the closed polygons with specific identifier values, generating a binary mask for the leaf surface region. This mask, together with the contour coordinates, constitutes the leaf surface boundary data, used to distinguish the stem and leaf regions.

[0150] Here is a concrete example: In a crop image processing system, the remaining pixels are first scanned horizontally in blocks, marking candidate edge points on the leaf surface with brightness differences. Next, reasonable edge points are selected through vertical continuity detection, generating a preliminary edge contour point set. Then, the starting point closest to the stem tip growth point is selected from the point set, and edge points are traced clockwise, skipping isolated points and filling in breakpoints to generate an ordered edge contour sequence. Subsequently, an interpolation algorithm is used to homogenize the contour point density. Finally, the contour points are connected to form closed polygon boundaries, and mask markers are filled in to generate boundary data for the leaf surface region.

[0151] By executing steps a61 to a65, this application implements methods such as horizontal block scanning, vertical continuity detection, edge point tracking, interpolation completion, and polygon closure to efficiently and accurately extract the edge contours of the leaf surface region and generate complete boundary data. This method effectively solves problems such as edge point breakage and isolated point interference, improves the accuracy and completeness of leaf surface region extraction, and provides a reliable foundation for subsequent morphological feature analysis and maturity identification.

[0152] In one possible embodiment, S125, identify the boundary coordinates of chlorophyll-abnormal regions corresponding to all crop plants, and generate crop maturity identification results containing maturity level classification indicators based on the boundary coordinates and morphological feature set, including:

[0153] Step 1251: Extract the maturity level classification identifier from the crop maturity identification results.

[0154] Among them, the crop maturity identification result is the crop maturity classification result generated by the maturity identification model.

[0155] Step 1252: Determine whether the stem bending arc angle in the morphological features exceeds the preset stem bending threshold. If so, determine that the upright morphological condition is met and generate a stem breakage mark.

[0156] The stem bending arc angle, a parameter describing the degree of stem bending, is typically extracted using image processing techniques. A preset stem bending threshold is a standard value used to determine whether the stem is upright; exceeding this value indicates stem breakage. An upright posture is indicated when the stem bending arc angle does not exceed the threshold. A stem breakage marker is used to indicate whether the stem is broken.

[0157] Step 1253: Based on the maturity level classification identifier, and combined with the difference between the maximum leaf unfolding width and the preset lower limit of leaf width in the morphological characteristics, calculate the leaf shrinkage index.

[0158] The leaf shrinkage index is a parameter used to quantify the degree of leaf shrinkage. The calculation formula is: Leaf shrinkage index = (Preset lower limit of leaf width - maximum unfolded width of leaf) / Preset lower limit of leaf width.

[0159] Step 1254: When the stem breakage mark exists and the leaf shrinkage index exceeds the preset shrinkage threshold, generate a crop quality identification result with a poor quality label; or, when the stem bending arc angle does not exceed the threshold and the maturity grade is marked as mature, generate a crop quality identification result with a high quality label.

[0160] The preset shrinkage threshold is a standard value used to determine the severity of leaf shrinkage. The inferior quality label is used to mark crops of poor quality. The superior quality label is used to mark crops of good quality.

[0161] Here's a specific example: In a crop quality inspection system, the system first extracts a maturity level classification indicator, such as "ripe," from the maturity identification results. Next, it extracts the stem bending angle from morphological features and determines if it exceeds a preset stem bending threshold. If it exceeds the threshold, a stem breakage marker is generated. Then, the difference between the maximum leaf unfolding width and a preset lower limit for leaf width is calculated to obtain the leaf shrinkage index. If a stem breakage marker exists and the leaf shrinkage index exceeds a preset shrinkage threshold, a poor quality marker is generated; otherwise, if the stem bending angle does not exceed the threshold and the maturity level classification indicator is "ripe," a high-quality marker is generated. Finally, the system outputs the crop quality identification results.

[0162] By executing steps 1251-1254, this application implements a method that efficiently and accurately identifies crop quality by extracting maturity grading indicators, determining stem bending angles, calculating leaf shrinkage indices, and comprehensively judging quality indicators. This method combines morphological characteristics and maturity information, effectively distinguishing between high-quality and low-quality crops, providing a scientific basis for crop grading and sales.

[0163] In one possible embodiment, step 133 involves extracting the first band reflection intensity value and the second band reflection intensity value corresponding to each crop at multiple spatial coordinate points from the two sets of reflectance spectral data, calculating the absolute value difference between the first band reflection intensity value and the second band reflection intensity value of each crop at each spatial coordinate point, and generating a band reflection difference sequence for each crop, including:

[0164] Step b1: Based on the reflectance spectral data of the two bands, perform spatial coordinate point mapping matching to locate the first band reflectance intensity matrix and the second band reflectance intensity matrix corresponding to each crop in the crop image data.

[0165] The two sets of reflectance spectral data are obtained from a spectrometer, representing the reflectance intensity of crop plants in two different bands. Spatial coordinate mapping and matching aligns the spatial coordinates of the reflectance spectral data with the spatial coordinates of the crop image data to ensure data consistency. The first band reflectance intensity matrix is ​​a matrix composed of the reflectance intensity values ​​at each spatial coordinate point in the first band. The second band reflectance intensity matrix is ​​a matrix composed of the reflectance intensity values ​​at each spatial coordinate point in the second band.

[0166] Step b2: Extract the pixel-level first band reflection intensity value from the first band reflection intensity matrix according to the row and column index order of the spatial coordinate points, and simultaneously obtain the pixel-level second band reflection intensity value with the same row and column index from the second band reflection intensity matrix.

[0167] The row and column indexing order is determined by traversing the spatial coordinate points row by row and column by column according to the matrix's row and column order. The pixel-level first-band reflection intensity value is the reflection intensity value of each pixel in the first band. The pixel-level second-band reflection intensity value is the reflection intensity value of each pixel in the second band.

[0168] Step b3: Perform point-by-point absolute value subtraction on the first band reflection intensity value and the second band reflection intensity value at each spatial coordinate point to generate a single-point band reflection difference value.

[0169] The point-by-point absolute value subtraction operation performs an absolute value subtraction operation on the reflection intensity values ​​of two bands at each spatial coordinate point. The single-point band reflection difference is the absolute difference between the reflection intensity values ​​of two bands at each spatial coordinate point.

[0170] Step b4: For each crop, based on the minimum bounding rectangle of a single crop, arrange the single-point band reflection difference values ​​of all spatial coordinate points within the bounding rectangle in row priority order to form a band reflection difference value sequence for each crop.

[0171] The minimum bounding rectangle is the smallest rectangular region that can completely enclose a single crop plant. Row priority is the data arranged row by row in the matrix. The band reflection difference sequence is the sequence of band reflection differences of all spatial coordinate points of a single crop plant within the bounding box, arranged in order.

[0172] Here's a specific example: In a crop maturity detection system, the system first acquires reflectance spectral data in the red and near-infrared bands using a spectrometer, and maps these data to the spatial coordinates of the crop image data. Next, it locates the region corresponding to each crop in the image data and extracts the reflectance intensity matrix of that region in both bands. Then, it iterates through each spatial coordinate point in row and column index order, extracting the reflectance intensity values ​​of the two bands and calculating their absolute difference to generate a single-point band reflectance difference. Finally, for the minimum bounding rectangle of each crop, the band reflectance differences of all spatial coordinate points within the bounding box are arranged in row-major order to form a band reflectance difference sequence. The system uses this sequence to further analyze the crop's maturity.

[0173] By executing steps b1 to b4, this application implements a method that efficiently and accurately quantifies the difference in reflection intensity of crops in two bands through spatial coordinate point mapping and matching, band reflection intensity value extraction, point-by-point difference calculation, and band reflection difference sequence generation. This method provides a reliable data foundation for crop maturity identification and can effectively improve the accuracy of maturity analysis.

[0174] In one possible embodiment, step a62, based on candidate edge points on the leaf surface, performs continuity detection along the vertical direction, retaining candidate edge points that satisfy the condition that the left side is a stem region and the right side is a leaf surface region, to generate a preliminary edge contour point set, including:

[0175] Step c1: Traverse each candidate edge point in the candidate edge point set, obtain the pixel coordinates of the candidate edge points, and extend the preset pixel width to the left and right sides along the vertical direction to form a horizontal detection band.

[0176] The pixel coordinates represent the position coordinates of the candidate leaf edge points in the image. The preset pixel width, the number of pixels extending to the left and right, defines the width of the horizontal detection band. The horizontal detection band is a strip-shaped region extending to the left and right from the candidate leaf edge points.

[0177] Step c2: Based on the horizontal detection band, scan the mask markers of N consecutive pixels to the left. If all pixels belong to the mask markers of the stem region, then mark the left side as the stem region. Scan the mask markers of N consecutive pixels to the right. If all pixels belong to the mask markers of the leaf region, then mark the right side as the leaf region.

[0178] Here, N consecutive pixels refers to the number of pixels continuously scanned within the horizontal detection band. The stem region is the area in the image that belongs to the stem. The leaf region is the area in the image that belongs to the leaf.

[0179] Step c3: Reserve candidate points that satisfy the condition that the left side is a stem region and the right side is a leaf region to a temporary point set. The temporary point set is a collection used to store candidate edge points that meet the condition.

[0180] Step c4: Based on the temporary point set, perform adjacent point spacing verification. If the adjacent point spacing exceeds a preset breakage threshold, insert linear interpolation points along the breakage direction to achieve contour continuity and generate a preliminary edge contour point set. The adjacent point spacing verification checks whether the distance between adjacent points in the temporary point set exceeds the preset breakage threshold. The preset breakage threshold is a standard value used to determine whether the edge contour is broken.

[0181] Here's a concrete example: In a crop image processing system, firstly, each point in the candidate leaf edge point set is traversed to obtain its pixel coordinates, and then extended 5 pixels vertically to the left and right to form a horizontal detection band. Next, within the horizontal detection band, the mask markers of three consecutive pixels are scanned to the left. If all pixels belong to the stem region, the left side is marked as the stem region; similarly, the mask markers of three consecutive pixels are scanned to the right. If all pixels belong to the leaf region, the right side is marked as the leaf region. Candidate points that meet the criteria are retained in a temporary point set. Then, the temporary point set is traversed, and the distance between adjacent points is checked. If the distance exceeds 10 pixels, linear interpolation points are inserted along the break direction to generate a preliminary edge contour point set. Finally, the system outputs a continuous leaf edge contour.

[0182] By executing steps c1 to c4, this application achieves efficient and accurate extraction of leaf edge contours through horizontal detection band generation, mask marker scanning, conditional filtering, and interpolation completion. This method effectively solves the problems of edge point breakage and isolated point interference, generating continuous edge contours and providing a reliable foundation for subsequent leaf region segmentation and morphological feature extraction.

[0183] Figure 3 This is a schematic diagram of a multi-sensor fusion collaborative control system for crop harvesting and piling, provided in an embodiment of this application. Figure 3 As shown, the system includes:

[0184] The acquisition module 31 is used to acquire crop image data on the conveyor belt of the integrated crop harvester and stacker and the reflectance spectral data of crop stems and leaves in two bands.

[0185] The generation module 32 is used to extract the morphological features of crops from the crop image data and generate crop maturity identification results based on the difference between the two sets of reflectance spectral data.

[0186] The identification module 33 is used to determine the crop quality identification result based on the morphological characteristics and the crop maturity identification result.

[0187] The control module 34 is used to send opening and closing control commands to the pneumatic guide plate at the multi-stage sorting port at the end of the conveyor belt that corresponds to the crop quality identification result, so as to transport the crops on the conveyor belt to the corresponding sorting area.

[0188] The adjustment module 35 is used to adjust the harvesting rate of the integrated crop harvester and stacker according to the pressure value when the pressure value in the stacking area is greater than the preset pressure threshold.

[0189] Figure 3 The aforementioned multi-sensor fusion-based collaborative control system for crop harvesting and stacking can execute... Figure 1 The implementation principle and technical effects of the multi-sensor fusion-based crop harvesting and stacking collaborative control method described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit performs operations in the multi-sensor fusion-based crop harvesting and stacking collaborative control system described in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0190] In one possible design, Figure 3 The multi-sensor fusion collaborative control system for crop harvesting and piling shown in the embodiment can be implemented as a computing device, such as... Figure 4 As shown, the computing device may include a storage component 41 and a processing component 42.

[0191] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 42.

[0192] The processing component 42 is used to: acquire crop image data and crop stem and leaf reflectance spectral data in two bands on the conveyor belt of the integrated crop harvester and stacker; extract morphological features of the crop from the crop image data, and generate crop maturity identification results based on the difference between the reflectance spectral data in the two bands; determine crop quality identification results based on morphological features and crop maturity identification results; send opening and closing control commands to the pneumatic guide vanes at the multi-stage stacking ports at the end of the conveyor belt that correspond to the crop quality identification results, so as to transport the crops on the conveyor belt to the corresponding stacking area; and adjust the harvesting rate of the integrated crop harvester and stacker according to the pressure value when the pressure value in the stacking area is greater than a preset pressure threshold.

[0193] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0194] Storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as Random Access Memory (RAM), Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0195] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0196] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0197] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0198] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0199] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown is a multi-sensor fusion collaborative control method for crop harvesting and stacking.

[0200] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0201] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0202] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0203] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A multi-sensor fusion method for collaborative control of crop harvesting and stacking, characterized in that, include: Acquire image data of crops on the conveyor belt in the integrated crop harvester and stacker, as well as reflectance spectral data of crop stems and leaves in two bands; The morphological features of crops are extracted from the crop image data, and the crop maturity identification result is generated based on the difference between the reflectance spectral data of the two sets of bands. Based on the morphological characteristics and the crop maturity identification results, the crop quality identification results are determined; Send an opening and closing control command to the pneumatic guide plate at the multi-stage sorting port at the end of the conveyor belt that corresponds to the crop quality identification result, so as to transport the crops on the conveyor belt to the corresponding sorting area. When the pressure value in the pile-dividing area is greater than the preset pressure threshold, the harvesting rate of the integrated crop harvester and pile-dividing machine is adjusted according to the pressure value. The step of extracting morphological features of crops from the crop image data and generating crop maturity identification results based on the difference between the two sets of reflectance spectral data includes: The crop image data is segmented into the outline of the stem and leaves of individual crops to extract the central axis of the stem and the leaf area of ​​each crop, generating a stem-leaf separation image. The morphological features of crops are extracted from the stem-leaf separation images. The morphological features include: maximum stem length, maximum leaf unfolding width, and stem bending arc angle, in order to construct a set of morphological features for all crops. Extract the first band reflection intensity value and the second band reflection intensity value of each crop at multiple same spatial coordinate points from the reflection spectral data of the two sets of bands. Calculate the absolute value difference between the first band reflection intensity value and the second band reflection intensity value of each crop at each spatial coordinate point to generate a band reflection difference sequence for each crop. Based on the spatial coordinates of the band reflection difference points that are greater than the preset band reflection difference threshold in the band reflection difference sequence of each crop, chlorophyll abnormal regions are formed. Identify the boundary coordinates of the chlorophyll abnormal regions corresponding to all crops, and generate crop maturity identification results containing maturity level indicators based on the boundary coordinates and the morphological feature set.

2. The method according to claim 1, characterized in that, The step of segmenting the stem and leaf contours of individual crops from the crop image data to extract the central axis of the stem and leaf area of ​​each crop and generate a stem-leaf separation image includes: The crop image data is spatially transformed from a red-green-blue color space to a preset color space used to enhance the contrast of the stem and leaf areas; Extract the set of difference pixels between crop plants and background from crop image data in the preset color space; An initial binary mask is generated based on the set of difference pixels. The minimum bounding rectangle of a single crop is determined by a connected component labeling algorithm. The connected regions in the crop image of the preset color space are separated according to the minimum bounding rectangle. The stem region is located based on the pixels within the minimum bounding rectangle of each individual crop plant, and the location result is obtained. The location result includes the base coordinates of the stem and the coordinates of the apical growth point. Based on the base coordinates and apical growth point coordinates of the stem, a preset extension fitting function is used to determine the coordinate sequence of the stem midline, and the dynamic variation area of ​​the stem width is determined according to the coordinate sequence of the stem midline. Extract the edge contour point set of the leaf area from the remaining pixels, and connect all the edge contour points in the edge contour point set by a preset polygon closure rule to generate the boundary data of the leaf area. The remaining pixels are the other pixels in the set of differences except for the area where the stem width changes dynamically. The coordinate sequence of the central axis of the stem and the boundary data of the leaf area are superimposed on the crop image in a preset color space to generate a stem-leaf separation image containing spatial separation markers of stem and leaf.

3. The method according to claim 2, characterized in that, The step of extracting the edge contour point set of the leaf surface region from the remaining pixels and connecting all the edge contour points in the edge contour point set according to a preset polygon closure rule to generate the boundary data of the leaf surface region includes: In the remaining pixels, the pixel brightness values ​​are scanned in blocks along the horizontal direction, and the boundary lines where the brightness difference between adjacent blocks exceeds a preset threshold are marked as candidate edge points of the leaf surface. Based on the candidate edge points on the leaf surface, continuous detection is performed along the vertical direction, and candidate edge points that satisfy the condition that the left side is the stem region and the right side is the leaf surface region are retained to generate a preliminary edge contour point set. From the initial set of edge contour points, select the starting point closest to the growth point at the top of the stem, and trace the connectivity of adjacent pixels point by point in a clockwise direction. During the tracing process, skip isolated points and fill in the broken points to generate an ordered edge contour sequence. Based on the ordered edge contour sequence, when the distance between two adjacent points exceeds a preset distance, an interpolation point is inserted between the two adjacent points to make the density of contour points uniform. According to the preset polygon closure rules, the interpolated contour points are connected end to end in a clockwise order to form a closed polygon boundary. The closed polygon is then filled with the mask markers of the leaf area to generate the boundary data of the leaf area.

4. The method according to claim 1, characterized in that, The process involves identifying the boundary coordinates of the chlorophyll-abnormal regions corresponding to all crop plants, and generating crop maturity identification results containing maturity level classification indicators based on the boundary coordinates and the morphological feature set. And extract the maturity level classification identifier from the crop maturity identification results; Determine whether the stem bending arc angle in the morphological features exceeds the preset stem bending threshold. If so, determine that the upright morphological condition is met and generate a stem breakage mark. Based on the maturity level classification indicator, and combined with the difference between the maximum leaf unfolding width and the preset lower limit of leaf width in the morphological characteristics, the leaf shrinkage index is calculated. When the stem breakage mark is present and the leaf shrinkage index exceeds a preset shrinkage threshold, a crop quality identification result with an inferior quality label is generated; or, when the stem bending arc angle does not exceed the threshold and the maturity grade is marked as mature, a crop quality identification result with a superior quality label is generated.

5. The method according to claim 1, characterized in that, The process involves extracting the first and second band reflection intensity values ​​for each crop at multiple spatial coordinate points from the reflection spectral data of the two sets of bands, calculating the absolute difference between the first and second band reflection intensity values ​​for each crop at each spatial coordinate point, and generating a band reflection difference sequence for each crop, including: Based on the reflectance spectral data of two bands, spatial coordinate point mapping matching is performed to locate the first band reflectance intensity matrix and the second band reflectance intensity matrix corresponding to each crop in the crop image data. According to the row and column index order of the spatial coordinate points, the pixel-level first band reflection intensity value is extracted from the first band reflection intensity matrix, and the pixel-level second band reflection intensity value with the same row and column index is obtained from the second band reflection intensity matrix simultaneously. For each spatial coordinate point, the absolute value of the first band reflection intensity value and the second band reflection intensity value are subtracted point by point to generate a single-point band reflection difference value. For each crop, the single-point band reflection difference values ​​of all spatial coordinate points within the minimum bounding rectangle of a single crop are arranged in row priority order to form a band reflection difference sequence for each crop.

6. The method according to claim 3, characterized in that, The step of performing continuous detection along the vertical direction based on the candidate leaf edge points, retaining candidate edge points that satisfy the condition that the left side is a stem region and the right side is a leaf region, to generate a preliminary edge contour point set, includes: Traverse each candidate edge point in the set of candidate edge points on the leaf surface, obtain the pixel coordinates of the candidate edge points on the leaf surface, and extend the preset pixel width to the left and right sides along the vertical direction to form a horizontal detection band; Based on the horizontal detection band, scan the mask markers of N consecutive pixels to the left. If all pixels belong to the mask markers of the stem region, then mark the left side as the stem region. Scan the mask markers of N consecutive pixels to the right. If all pixels belong to the mask markers of the leaf region, then mark the right side as the leaf region. Candidate points that satisfy the condition that the left side is a stem region and the right side is a leaf region will be retained in the temporary point set; Based on the temporary point set, the distance between adjacent points is checked. If the distance between adjacent points exceeds a preset fracture threshold, linear interpolation points are inserted along the fracture direction to achieve contour continuity and generate the preliminary edge contour point set.

7. A multi-sensor fusion collaborative control system for crop harvesting and stacking, characterized in that, A method for implementing a multi-sensor fusion-based collaborative control method for crop harvesting and stacking as described in any one of claims 1 to 6, comprising: The acquisition module is used to acquire image data of crops on the conveyor belt in the integrated crop harvester and stacker, as well as reflectance spectral data of crop stems and leaves in two bands. The generation module is used to extract morphological features of crops from the crop image data and generate crop maturity identification results based on the difference between the two sets of reflectance spectral data. The identification module is used to determine the crop quality identification result based on the morphological features and the crop maturity identification result; The control module is used to send opening and closing control commands to the pneumatic guide plate at the multi-stage sorting port at the end of the conveyor belt that corresponds to the crop quality identification result, so as to transport the crops on the conveyor belt to the corresponding sorting area. The adjustment module is used to adjust the harvesting rate of the integrated crop harvester and stacker based on the pressure value when the pressure value in the stacking area is greater than a preset pressure threshold.

8. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a multi-sensor fusion-based crop harvesting and stacking collaborative control method as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a multi-sensor fusion-based collaborative control method for crop harvesting and stacking as described in any one of claims 1 to 6.

Images