Seedling aerial seeding density identification method based on pneumatic launching device

Abstract

The invention relates to the technical field of seed counting, in particular to a seedling aerial seeding density identification method based on a pneumatic launching device. The method comprises the following steps: determining a flow state feature vector based on a trajectory deviation degree and a form residual degree of a waveform mapping path between an upstream observation sequence and a downstream observation sequence passing through an event; dividing the working time period of the device into a starting time period and a running time period; based on the flow state feature vector of the previous passing event of each passing event in the operation time period, the classification reference of the flow state feature vectors of the passing events in the starting time period is adjusted, and single grain and cluster feature judgment vectors are determined; and based on the deviation degree of the flow state feature vector of the passing event relative to the single-grain feature judgment vector and the cluster feature judgment vector in the operation period, determining a grain counting increment, and calculating the total seeding rate. According to the method, the single seed and the cluster are accurately divided by integrating the microscopic motion characteristics and the waveform morphological characteristics of the seeds, and the seed metering accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of seed counting technology, and specifically to a method for identifying seedling aerial seeding density based on a pneumatic launcher. Background Technology

[0002] In modern forestry ecological restoration and large-scale afforestation projects, pneumatic conveyor aerial seeding devices are widely used due to their high efficiency and wide coverage. These devices typically utilize high-speed airflow to transport seedlings and use photoelectric sensors to detect the number of seeds passing through to control the sowing density. To meet project schedule requirements, the conveying process is often conducted at high throughput.

[0003] However, existing detection technologies have significant shortcomings in accurately counting non-spherical seeds (especially winged seeds) under high-throughput conditions: on the one hand, a single winged seed tumbles violently in the airflow, and the intermittent shading of its wings causes high-frequency jitter in the photoelectric signal. Traditional counting methods based on level thresholds or pulse widths cannot effectively distinguish such jitter from the real multi-particle signal, thus producing false counts; on the other hand, multiple seeds tend to aggregate into tight clusters, and their signals have increased amplitude and smoothed contours due to overlap. Traditional comparison algorithms based on waveform geometry contours are difficult to extract the true number of particles from these clusters, leading to missed counts. Summary of the Invention

[0004] To address the technical problem of measurement inaccuracies caused by the tumbling of single winged seeds and the overlapping of multiple seed clusters, the present invention aims to provide a method for identifying seedling aerial seeding density based on a pneumatic launching device. The specific technical solution adopted is as follows: One embodiment of the present invention provides a method for identifying seedling aerial seeding density based on a pneumatic launching device, the method comprising: The signal amplitude of the upstream and downstream monitoring points of the straight pipe section of the pneumatic launcher is acquired in real time at each moment during the device's operating period; based on the signal amplitude, a passage event is determined, and the upstream and downstream observation sequences for each passage event are acquired; Determine the waveform mapping path between the upstream and downstream observation sequences of each passing event; based on the trajectory deviation and morphological residual of the waveform mapping path, determine the flow characteristic vector of each passing event; The working period of the device is divided into a startup period and a running period. Based on the flow feature vector of the previous passing event of each passing event in the running period, the classification criterion of the flow feature vector of the passing event in the startup period is adjusted, and the single-particle feature decision vector and cluster feature decision vector of each passing event in the running period are determined. Based on the degree of deviation of the flow feature vector of each passing event within the runtime period from the single-particle feature decision vector and the cluster feature decision vector, the particle count increment of each passing event is determined, and the total seeding amount after each passing event is completed is calculated.

[0005] Further, determining the waveform mapping path between the upstream and downstream observation sequences of each passing event includes: Construct a distance matrix between the upstream and downstream observation sequences for each event, where the element in the nth row and mth column of the distance matrix is ​​the square of the absolute difference between the nth element in the upstream observation sequence and the mth element in the downstream observation sequence; where n and m are the element indices of the upstream and downstream observation sequences, respectively. Obtain the theoretical shortest flight time delay, the theoretical longest flight time delay, and the data sampling frequency between the upstream and downstream monitoring points; calculate the difference between the theoretical longest flight time delay and the theoretical shortest flight time delay, and use the product of the difference and the data sampling frequency as the maximum relative index deviation; In the distance matrix, elements whose row index is greater than or equal to the column index, and whose column index is less than or equal to the sum of the deviations between the row index and the maximum relative index, are denoted as valid search elements. Within the region comprised of the effective search elements, a dynamic programming algorithm is used to search for the path with the minimum cumulative cost from the top-left element to the bottom-right element of the distance matrix. The coordinate pairs consisting of the row index and column index of each element traversed by the path are arranged in the order of traversal to obtain a sequence of coordinate pairs, which is denoted as the waveform mapping path between the upstream observation sequence and the downstream observation sequence of each passing event.

[0006] Further, determining the flow state feature vector for each passing event includes: For each waveform mapping path that passes through an event, the row index and column index of the coordinate pair in the waveform mapping path are linearly fitted to obtain the ideal transmission path; Calculate the shortest distance from each coordinate pair in the waveform mapping path to the ideal transmission path, and denote it as the transmission trajectory deviation; The square of the row index of each coordinate pair in the waveform mapping path corresponding to the element in the upstream observation sequence is used as the biomass weighting factor for each coordinate pair. The sum of the products of the transmission trajectory deviation and the biomass weighting factor for each coordinate pair in the waveform mapping path is calculated and recorded as the weighted deviation value. The ratio of the weighted deviation value as the numerator and the sum of the biomass weighting factors for each coordinate pair in the waveform mapping path and the sum of a preset positive number as the denominator is used as the transmission instability value of the event. The waveform morphology residual value is obtained by averaging the absolute differences between the corresponding elements of the row index in the upstream observation sequence and the corresponding elements of the column index in the downstream observation sequence for each coordinate pair in the waveform mapping path. The transmission instability value and the waveform morphology residual value are standardized to obtain the standard instability value and the standard residual value in turn; the standard instability value and the standard residual value of each passing event constitute a flow characteristic vector.

[0007] Further, determining the single-particle feature decision vector and cluster feature decision vector for each passing event within the runtime segment includes: The flow feature vectors of all events passing through during the startup period are subjected to binary clustering to obtain two cluster center vectors; Calculate the magnitude of the two cluster center vectors, denote the cluster center vector with the smallest magnitude as the single-particle initiation feature vector, and denote the cluster center vector with the largest magnitude as the cluster initiation feature vector; For the first pass event within the runtime segment, the single-particle initiation feature vector and the cluster initiation feature vector are used as the single-particle feature decision vector and the cluster feature decision vector, respectively, for the pass event. For each passing event after the first passing event within the runtime segment, the single-particle initiation feature vector and the cluster initiation feature vector are adjusted using the flow state feature vector of the previous passing event, thereby obtaining the single-particle feature decision vector and the cluster feature decision vector for each passing event in sequence.

[0008] Further, determining the particle count increment for each passing event includes: For all passing events during the runtime period, calculate the distances between the flow feature vector of the passing event and its single-particle feature decision vector and cluster feature decision vector, respectively, and denoted as single-particle reference deviation distance and cluster reference deviation distance; When the single-particle reference deviation distance of the event is less than the cluster reference deviation distance, the particle count increment of the event is set to 1. When the single-particle reference deviation distance of the event is greater than or equal to the cluster reference deviation distance, the ratio obtained by taking the distance between the flow characteristic vector of the event and the single-particle characteristic decision vector as the numerator and the distance between the single-particle characteristic decision vector and the cluster characteristic decision vector of the event as the denominator is denoted as the characteristic deviation ratio of the event. Substituting the feature deviation ratio into the logarithmic compensation function yields the particle count increment for each event.

[0009] Furthermore, the formula for calculating the logarithmic compensation function is as follows: ; In the formula, For the increment of the particle count through the event; This is the preset cluster overlap compensation coefficient; is the characteristic deviation ratio of the event; ln is the logarithmic function with the natural constant e as the base; Round is the rounding function.

[0010] Further, the calculation of the total seeding amount after each event is completed includes: Set the initial total seeding quantity to zero; use the sum of the particle count increment of the first pass event within the runtime segment and the initial seeding quantity as the total seeding quantity after the first pass event is completed; For each pass event following the first pass event within the runtime segment, the sum of the particle count increment of each pass event and the seeding amount after the completion of the previous pass event is taken as the total seeding amount after the completion of each pass event.

[0011] Further, the step of adjusting the single-particle initiation feature vector and the cluster initiation feature vector using the flow state feature vector of the preceding passing event for each passing event, to sequentially obtain the single-particle feature decision vector and the cluster feature decision vector for each passing event, includes: The first passing event after the first passing event in any runtime segment is denoted as the example event. The difference between the constant 1 and the preset update rate coefficient is used as the weight of the single-particle start feature vector, and the preset update rate coefficient is used as the weight of the flow feature vector of the previous passing event of the example event. The single-particle start feature vector and the flow feature vector of the previous passing event of the example event are weighted and summed. The resulting new vector is denoted as the single-particle feature decision vector of the example event. The single-particle initiation feature vector in the calculation process of the single-particle feature decision vector of the example event is updated to the cluster initiation feature vector to obtain the cluster feature decision vector of the example event.

[0012] Furthermore, obtaining the upstream and downstream observation sequences for each passing event includes: The first moment is recorded as the initial analysis moment. When the signal amplitude of the upstream monitoring point at the analysis moment is less than or equal to the preset background noise threshold, the next adjacent moment is recorded as the new analysis moment. When the signal amplitude of the upstream monitoring point at the analysis time is greater than the background noise threshold, the analysis time is taken as the trigger time of the passing event. Starting from the trigger time of the event, the signal amplitudes of each time within the preset event time are arranged in chronological order to obtain the upstream signal sequence of the corresponding passing event. The difference between the theoretical longest flight time delay and the theoretical shortest flight time delay is taken as the theoretical flight time tolerance; the downstream monitoring point takes the theoretical shortest flight time delay as the starting point at the trigger time, and arranges the signal amplitudes at each time within the total time of the preset event time and the theoretical flight time tolerance in chronological order to obtain the downstream signal sequence of the event. The first moment after adding the preset event time to the trigger time of each event is recorded as the new analysis moment; Determine whether the signal amplitude at the new analysis moment is greater than the background noise threshold, determine the trigger moment of the next passing event, and obtain the upstream and downstream signal sequences of the passing event; The upstream and downstream signal sequences are subjected to baseline removal and amplitude normalization processes to obtain the upstream and downstream observation sequences of the passing event.

[0013] Furthermore, the method for performing binary clustering on the flow feature vectors of all passing events during the startup period is the K-means clustering algorithm, where the value of K is equal to 2.

[0014] The present invention has the following beneficial effects: In this embodiment of the invention, the degree of trajectory deviation of the waveform mapping path can capture the microscopic fluctuation characteristics of the fluid unit's velocity during transmission, making the feature sensitive to the motion of the seed body and filtering out false fluctuations from multiple particles; the degree of morphological residual of the waveform mapping path can capture the changes in the physical shape of the fluid unit before and after transmission, constituting a microscopic dynamic description feature, i.e., a flow state feature vector, capable of single-unit tumbling and cluster rearrangement, effectively distinguishing between rigid single-unit tumbling and loose cluster deformation; the division of the start-up period and the running period aims to establish an initial classification benchmark through unsupervised learning during the start-up period, and to realize the system's adaptability to environmental drift through dynamic adjustment during the running period. The long-term adaptive method incorporates the flow characteristic vector of the previous passing event for each passing event during the runtime into the decision criterion, obtaining single-grain feature decision vectors and cluster feature decision vectors that can dynamically track the slow-changing process of the operating conditions, thereby maintaining high-precision recognition capability in long-term operation. By quantifying the deviation of the flow characteristic vector of the passing event from the single-grain feature decision vector and the cluster feature decision vector, the micro kinematic features are mapped to the macro equivalent number of seeds, thereby overcoming the counting distortion caused by seed wing tumbling noise and cluster overlap saturation effect, and finally achieving adaptive and accurate measurement of seeding density in high-throughput aerodynamic aerial seeding environment. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the steps of a seedling aerial seeding density identification method based on a pneumatic launching device, as provided in one embodiment of the present invention. Figure 2 This is a system structure diagram of a seedling aerial seeding density identification system based on a pneumatic launching device, provided in one embodiment of the present invention. Figure 3 This is a schematic diagram of a computer device for identifying seedling aerial seeding density based on a pneumatic launcher, provided as an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a seedling aerial seeding density identification method based on a pneumatic launching device proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a seedling aerial seeding density identification method based on a pneumatic launching device provided by the present invention.

[0020] Example 1: This invention proposes a method for identifying seedling aerial seeding density based on a pneumatic launcher. Please refer to [link / reference]. Figure 1 The diagram illustrates a flowchart of a seedling aerial seeding density identification method based on a pneumatic launching device, according to an embodiment of the present invention. The method includes: Step S1: Real-time acquisition of the signal amplitude of the upstream and downstream monitoring points of the straight pipe section of the pneumatic launcher at each moment during the device's operating period; determination of passing events based on signal amplitude, and acquisition of the upstream and downstream observation sequences for each passing event.

[0021] In the straight pipe section of the pneumatic conveying device, upstream and downstream monitoring points are sequentially set at positions before and after the airflow along the straight pipe section, and photoelectric sensors are installed at both monitoring points. The photoelectric sensors at the upstream and downstream monitoring points synchronously collect the output voltage at each moment during the device's operating period; the output voltage is the signal amplitude. The device's operating period refers to the entire continuous working time from power-on startup to task completion or power-off. In this embodiment of the invention, the data sampling frequency is set to 500 Hz; however, the implementer can set it according to specific circumstances.

[0022] In this embodiment of the invention, the method for obtaining an event and its upstream and downstream observation sequences includes: recording the first moment as the initial analysis moment; when the signal amplitude of the upstream monitoring point at the analysis moment is less than or equal to a preset background noise threshold, recording the next adjacent moment as the new analysis moment; when the signal amplitude of the upstream monitoring point at the analysis moment is greater than the background noise threshold, using the analysis moment as a trigger moment for a passing event; starting from the trigger moment of the event, arranging the signal amplitudes of each moment within a preset event time in chronological order to obtain the corresponding upstream signal sequence for the passing event; and calculating the difference between the theoretical longest flight time delay and the theoretical shortest flight time delay. The theoretical flight time tolerance is used as the starting point. Starting from the downstream monitoring point at the trigger time plus the theoretical shortest flight time delay, the signal amplitudes at each moment within the total time between the preset event time and the theoretical flight time tolerance are arranged in time sequence to obtain the downstream signal sequence of the passing event. The first moment after adding the preset event time to the trigger time of each passing event is recorded as the new analysis time. It is determined whether the signal amplitude at the new analysis time is greater than the background noise threshold to determine the trigger time of the next passing event, and the upstream and downstream signal sequences of the passing event are obtained. The upstream and downstream signal sequences are subjected to baseline removal and amplitude normalization processing respectively to obtain the upstream and downstream observation sequences of the passing event.

[0023] It should be noted that when the signal amplitude at each moment exceeds the background noise threshold, a true seed occlusion signal is generated at that moment; the trigger moment marks the moment when the seed begins to occlude the upstream sensor. By adding the theoretical minimum flight time delay to the downstream monitoring point at the trigger moment as the starting point, invalid physical dead zones can be actively eliminated, ensuring that the starting point of the downstream signal sequence is physically aligned with the earliest possible arrival time of the seed. During the acquisition of the downstream signal sequence, the preset event time can cover the time it takes for the slowest seed to pass through the sensor, ensuring the integrity of the upstream signal. The theoretical flight time tolerance covers the maximum time jitter caused by speed fluctuations. The total time of both ensures that regardless of the actual flight speed of the seed, its complete downstream passage signal will inevitably be included within the acquisition window, fundamentally avoiding signal truncation.

[0024] For both the upstream and downstream signal sequences, the difference between each element in the sequence and the background noise threshold is calculated. These differences are then arranged to obtain a baseline-removed sequence, which eliminates DC offset caused by sensor zero drift and ambient light, accurately reflecting shading changes. The root mean square (RMS) value of each element in the baseline-removed sequence is calculated, and the ratio of each element to the RMS value is calculated. These ratios are then arranged to obtain the observation sequence. The observation sequences corresponding to the upstream and downstream signal sequences are denoted as the upstream observation sequence and the downstream observation sequence, respectively. The elements in the observation sequence reflect the relative shading intensity caused by the shading object (seed or seed wing) at each moment.

[0025] By calculating the average signal amplitude of the photoelectric sensor at the upstream monitoring point during all moments of the silent period when the device is unloaded (i.e., no seeds passing through), a preset background noise threshold is established. This threshold can reliably distinguish between environmental electrical noise and actual seed occlusion signals. Experiments were conducted to release typical or extreme seed samples under controlled conditions. A high-precision time sensor was used to measure the duration from the upstream monitoring point to the downstream monitoring point's complete departure. The maximum value was taken as the preset event time through repeated experiments, ensuring complete coverage of the maximum time span required for a single seed or cluster to pass between the upstream and downstream monitoring points. The methods for obtaining the theoretical shortest and longest flight time delays include: using the airflow velocity generated by the pneumatic launcher's fan at maximum output power and under unload conditions as the maximum wind speed boundary, and the minimum airflow velocity required to prevent pipe blockage at rated output power as the minimum wind speed boundary. The straight-line distance between the optical axis centers of the upstream and downstream sensors was measured using calipers and recorded as the sensor spacing. The ratio of the sensor spacing to the maximum wind speed boundary was used as the theoretical shortest flight time delay, and the ratio of the sensor spacing to the minimum wind speed boundary was used as the theoretical longest flight time delay.

[0026] Step S2: Determine the waveform mapping path between the upstream and downstream observation sequences of each passing event; based on the trajectory deviation and morphological residual of the waveform mapping path, determine the flow characteristic vector of each passing event.

[0027] Due to the influence of airflow turbulence, pipe wall friction, inter-particle collisions, and their own attitude, the flight velocity of seeds in a pneumatic pipe fluctuates, causing nonlinear time distortion in the signals generated by upstream and downstream sensors. Establishing a waveform mapping path between the upstream and downstream observation sequences can eliminate the influence of velocity fluctuations on waveform analysis, laying a reliable analytical foundation for subsequent analysis of the dynamic characteristics of fluid units (seeds or clusters).

[0028] During pneumatic transport, the tumbling of rigid seeds is mainly manifested as changes in the projection of thin parts such as seed wings into the sensor's field of view, while the deformation of clusters is manifested as changes in the relative position of the seed body. Therefore, the degree of trajectory deviation of the waveform mapping path can capture the microscopic fluctuation characteristics of the fluid unit's velocity during transport. Even when a rigid single unit rotates, its physical shape remains unchanged, and the upstream and downstream waveforms should have a high degree of similarity after alignment. However, rearrangement within loose clusters leads to irreversible morphological changes, resulting in poor similarity between upstream and downstream waveforms after alignment. Therefore, the degree of morphological residual of the waveform mapping path can capture the changes in the physical shape of the fluid unit before and after transport. Combining these two factors constitutes the microscopic dynamic description features, i.e., the flow characteristic vector, that can effectively distinguish between the tumbling of rigid single units and the deformation of loose clusters.

[0029] Step S3: Divide the device's working period into a startup period and a running period; based on the flow characteristic vector of the previous passing event for each passing event in the running period, adjust the classification criterion of the flow characteristic vector of the passing events in the startup period, and determine the single-particle feature decision vector and cluster feature decision vector of each passing event in the running period.

[0030] In the initial stage of system startup, there is a lack of any prior information on current seed characteristics, wind speed conditions, and pipeline status, making reliable judgments impossible. The division into startup and runtime phases aims to establish an initial classification benchmark through unsupervised learning during the startup phase, and to achieve long-term adaptation of the system to environmental drift through dynamic updates during the runtime phase. This ensures and continuously improves the accuracy of seed density identification even in the absence of prior knowledge.

[0031] The classification criterion based on the flow pattern feature vectors of passing events during the startup period can divide seed samples into two categories: rigid seeds and clusters. However, due to wear and tear on pneumatic conveying pipes and fluctuations in fan air pressure, the feature criterion slowly drifts. Relying solely on the static classification criterion established during the startup period leads to a significant decrease in accuracy over time. By incorporating the flow pattern feature vector of the preceding passing event into the decision criterion for each passing event during the runtime, the decision criteria can be dynamically adjusted. This results in single-seed and cluster feature decision vectors that can dynamically track the slowly changing operating conditions, thus maintaining high-precision identification capabilities over long-term operation.

[0032] Step S4: Based on the deviation of the flow feature vector of each passing event from the single-particle feature decision vector and the cluster feature decision vector during the runtime, determine the particle count increment of each passing event, and calculate the total seeding amount after each passing event is completed.

[0033] In high-throughput pneumatic seeding, an increase in the number of seeds within a cluster can lead to a saturation effect in signal features (such as overlap and occlusion). By quantifying the deviation of the flow characteristic vector of the passing event from the single-seed feature decision vector and the cluster feature decision vector, the microscopic kinematic features are mapped to the macroscopic equivalent number of seeds. This overcomes the counting distortion caused by seed wing tumbling noise and cluster overlap saturation effect, ultimately achieving adaptive and accurate measurement of seeding density in high-throughput pneumatic aerial seeding environments.

[0034] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the waveform mapping path includes: constructing a distance matrix between the upstream observation sequence and the downstream observation sequence for each passing event, wherein the element in the nth row and mth column of the distance matrix is ​​the square of the absolute difference between the nth element in the upstream observation sequence and the mth element in the downstream observation sequence; wherein n and m are the element indices of the upstream observation sequence and the downstream observation sequence, respectively; obtaining the theoretical shortest flight time delay, the theoretical longest flight time delay, and the data sampling frequency between the upstream monitoring point and the downstream monitoring point; and calculating the theoretical longest flight time delay and the theoretical shortest flight time delay. The difference is multiplied by the data sampling frequency as the maximum relative index deviation. In the distance matrix, elements with row indices greater than or equal to column indices, and column indices less than or equal to the sum of row indices and maximum relative index deviations, are denoted as valid search elements. Within the region formed by valid search elements, a dynamic programming algorithm is used to search for the path with the minimum cumulative cost from the top-left element to the bottom-right element of the distance matrix. The coordinate pairs consisting of the row and column indices of each element traversed by the path are arranged in the order of traversal to obtain a coordinate pair sequence, which is denoted as the waveform mapping path between the upstream and downstream observation sequences of each passing event.

[0035] It should be noted that, in order to quantify the local morphological similarity between the upstream and downstream observation sequences, a point-to-point distance matrix is ​​constructed. The element in the nth row and mth column of the distance matrix reflects the degree of matching between the signal amplitude of the nth element in the upstream observation sequence and the mth element in the downstream observation sequence; the smaller the value, the better the alignment of the point pair. The seed flight event delay is constrained within the theoretical shortest flight time delay and the theoretical longest flight time delay. By multiplying it by the data sampling frequency, the time constraint is converted into a constraint on the number of sampling point indices. Elements with row indices greater than or equal to column indices: ensure that the downstream signal cannot precede the upstream signal, satisfying the physical causality of time sequence; elements with column indices less than or equal to the sum of the row index and the maximum relative index deviation: ensure that the time delay does not exceed the maximum possible delay determined by the minimum wind speed, satisfying the maximum time delay limit; the above two conditions are to prevent the algorithm from producing non-physical matching results such as time reversal or ultra-slow delay, ensuring that the mapping path conforms to the actual transportation conditions.

[0036] It is important to note that the element indices of both the upstream and downstream observation sequences are positive indices from left to right, starting from 1. For all paths from the top-left element (1,1) to the bottom-right element (N,M) in the distance matrix, the cumulative cost of the path refers to the sum of all elements traversed by that path in the distance matrix; where N is the number of elements in the upstream observation sequence and M is the number of elements in the downstream observation sequence.

[0037] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the flow characteristic vector includes: for each waveform mapping path of a passing event, performing linear fitting on the row index and column index of the coordinate pairs in the waveform mapping path to obtain the ideal transmission path; calculating the shortest distance from each coordinate pair in the waveform mapping path to the ideal transmission path, denoted as the transmission trajectory deviation; using the square of the corresponding element of the row index of each coordinate pair in the upstream observation sequence as the biomass weighting factor for each coordinate pair; and calculating the sum of the products of the transmission trajectory deviation and the biomass weighting factor for each coordinate pair in the waveform mapping path, denoted as... The weighted deviation value is used as the numerator, and the sum of the biomass weighting factors of the coordinate pairs in the waveform mapping path is used as the denominator. The ratio obtained is the transmission instability value of the passing event. The absolute difference between the corresponding element of the row index in the upstream observation sequence and the corresponding element of the column index in the downstream observation sequence of each coordinate pair in the waveform mapping path is averaged to obtain the waveform morphology residual value. The transmission instability value and the waveform morphology residual value are standardized to obtain the standard instability value and the standard residual value, respectively. The flow characteristic vector is composed of the standard instability value and the standard residual value of each passing event. The preset positive number is a very small positive number, set to 0.01, to prevent the denominator from being zero and making the fraction meaningless.

[0038] It should be noted that the ideal transmission path represents the theoretical spatiotemporal trajectory of a rigid object passing through a straight pipe section at a constant speed, serving as a baseline reference for measuring whether speed fluctuations exist in the actual transmission process. In the linear fitting process using the least squares method, the row index is the independent variable, and the column index is the dependent variable. The transmission trajectory deviation quantifies the degree of deviation of each coordinate pair in the waveform mapping path from the ideal uniform velocity trajectory. This deviation reflects the instantaneous velocity distortion of the seed during flight caused by airflow disturbances, spin, or rearrangement within the cluster. In photoelectric detection, the signal amplitude directly reflects the area of ​​the cross-section obstructed by the object, i.e., the biomass. Winged seeds have thin, translucent wings, resulting in low signal amplitudes, while the main body of the seed is thick and dense, producing high signal amplitudes. By taking the square of the signal amplitude, the influence of the high-density seed body is significantly amplified, thereby achieving focused attention on the effective biological entity.

[0039] The weighted deviation value reflects the total cumulative motion deviation considering the distribution of material entities within the event, characterizing the absolute intensity of the overall motion instability of the event. The sum of the biomass weighting factors for coordinate pairs in the waveform mapping path reflects the cumulative effective biomass of the fluid unit along the spatiotemporal mapping path. The ratio of these two values ​​eliminates the absolute magnitude differences caused by variations in seed or cluster physical size and biomass, making events of different scales comparable in terms of motion stability. Transport instability indicates the internal structural properties of the fluid unit; the lower the transport instability value, the more stable the internal structure of the fluid unit, and the greater the likelihood of it being a rigid single entity. Conversely, the higher the transport instability value, the more loose and variable the internal structure of the fluid unit, and the greater the likelihood of it being a cluster.

[0040] The waveform morphology residual value characterizes the degree of distortion of the geometry or internal structure of the fluid unit (seed or cluster) during transmission after eliminating the influence of velocity fluctuations. Rigid individuals exhibit lower distortion, while clusters show higher distortion. Therefore, a smaller waveform morphology residual value indicates lower distortion of the fluid unit during transmission, and a greater likelihood that the event represents the passage of a rigid individual; conversely, a larger residual value indicates a greater likelihood that the event represents the passage of a cluster.

[0041] The transmission instability value and waveform residual value are only sensitive to instability caused by cluster rearrangement, but have a strong inhibitory effect on spurious instability caused by seed wing tumbling. Thus, the flow characteristic vector formed by the two can effectively distinguish between rigid single-cell tumbling and loose cluster deformation.

[0042] In this embodiment of the invention, based on the transmission instability value and waveform residual value of all passing events before each passing event during the device's working period, the Z-score standardization method is used to standardize the transmission instability value and waveform residual value of each passing event.

[0043] In this embodiment of the invention, the method for dividing the start-up period and the running period includes: recording the i-th passing event within the device's working period as the initial segmented event; performing binary clustering on the flow feature vectors of the segmented event and all previous passing events to obtain two cluster center vectors; when the distance between the two cluster center vectors is less than or equal to the discrimination threshold, determining that the current sample library is singular, i.e., no clusters appear, and taking the next adjacent passing event of the segmented event as the new segmented event; when the distance between the two cluster center vectors is greater than the discrimination threshold, taking the time period corresponding to the segmented event and all previous passing events as the start-up period; and recording the time period after the start-up period within the device's working period as the running period.

[0044] The value of 'i' must meet the minimum sample size requirement to ensure that the initial cluster analysis has basic statistical significance. To ensure that the binary clustering converges to a stable pattern, the initial sample size should be significantly larger than the number of feature dimensions, and a margin should be reserved to cope with data perturbations. In this embodiment, 'i' is set to 30. The discrimination threshold is used to quantify the separability of the clustering results. During the system design phase, representative sample data is collected, and the statistical distribution (such as the minimum value or 10th percentile) of the distance between the flow characteristic vectors of typical single events and typical cluster events is calculated, which is denoted as the discrimination threshold.

[0045] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the single-particle feature decision vector and the cluster feature decision vector includes: performing binary clustering on the flow state feature vectors of all passing events during the startup period to obtain two cluster center vectors; calculating the magnitude of the two cluster center vectors, and recording the cluster center vector with the smallest magnitude as the single-particle startup feature vector and the cluster center vector with the largest magnitude as the cluster startup feature vector; for the first passing event during the runtime period, using the single-particle startup feature vector and the cluster startup feature vector as the single-particle feature decision vector and the cluster feature decision vector of the passing event, respectively; for passing events after the first passing event during the runtime period, using the flow state feature vector of the previous passing event for each passing event, adjusting the single-particle startup feature vector and the cluster startup feature vector, and sequentially obtaining the single-particle feature decision vector and the cluster feature decision vector of each passing event.

[0046] It should be noted that, in this embodiment of the invention, based on the distance between different flow characteristic vectors, the K-means clustering algorithm is used to cluster the flow characteristic vectors of all passing events during the startup period. Because there are two types of physically different events in the pneumatic transport process: high-frequency single-particle passing and occasional cluster passing, the K value is set to 2. Rigid single-particle seeds have high transport stability and small morphological residuals, resulting in smaller modulus lengths of their characteristic vectors; while loose clusters, due to internal rearrangement and complex morphology, have poor transport stability and large morphological distortions, resulting in larger modulus lengths of their characteristic vectors. Therefore, the cluster center vector with the smallest modulus length represents a single-particle seed, and the cluster center vector with the largest modulus length represents a cluster. The average vector of all flow characteristic vectors within each cluster obtained by binary clustering is denoted as a cluster center vector.

[0047] During pneumatic conveying, minor changes in pipe inner wall wear, fan thermal drift, or seed batch humidity can cause a slow overall shift in the flow characteristic distribution. By adjusting the decision vector using the flow characteristic vector of the previous passing event, we obtain the single-particle characteristic decision vector and cluster characteristic decision vector for each passing event. This ensures that the decision benchmark continuously tracks the slow changes in the system state, guaranteeing decision accuracy during continuous operation.

[0048] In this embodiment of the invention, the method for obtaining the single-particle feature decision vector and the cluster feature decision vector includes: selecting any event following the first passing event within a runtime segment as an example event; using the difference between a constant 1 and a preset update rate coefficient as the weight of the single-particle initiation feature vector; using the preset update rate coefficient as the weight of the flow state feature vector of the preceding passing event of the example event; and performing a weighted summation of the single-particle initiation feature vector and the flow state feature vector of the preceding passing event of the example event, recording the resulting new vector as the single-particle feature decision vector of the example event; updating the single-particle initiation feature vector during the calculation of the single-particle feature decision vector of the example event to the cluster initiation feature vector, thus obtaining the cluster feature decision vector of the example event. It should be noted that the method for obtaining the single-particle feature decision vector and the cluster feature decision vector of all passing events and example events is the same.

[0049] It should be noted that the update rate coefficient determines the sensitivity of the reference point to new samples and the rate at which historical information is forgotten; its value ranges from (0, 1). Fast Response Mode: If the operating environment fluctuates significantly, a larger update rate coefficient should be selected, for example, 0.05 to 0.1, to enable the system to quickly adapt to environmental changes. Stable Operation Mode: If the operating environment is relatively stable, a smaller update rate coefficient should be selected, for example, 0.001 to 0.01, to enhance the stability of the benchmark and smooth out occasional measurement noise. In this embodiment, the update rate coefficient is set to 0.01, ensuring that the benchmark can track hourly environmental temperature drift while effectively filtering out second-level instantaneous airflow disturbances.

[0050] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the particle count increment includes: for all passing events during the runtime, calculating the distances between the flow characteristic vector of the passing event and its single-particle characteristic decision vector and its cluster characteristic decision vector, respectively, and recording them as the single-particle reference deviation distance and the cluster reference deviation distance; when the single-particle reference deviation distance of the passing event is less than the cluster reference deviation distance, setting the particle count increment of the passing event to 1; when the single-particle reference deviation distance of the passing event is greater than or equal to the cluster reference deviation distance, the ratio obtained by using the distance between the flow characteristic vector of the passing event and the single-particle characteristic decision vector as the numerator and the distance between the single-particle characteristic decision vector and the cluster characteristic decision vector of the passing event as the denominator is recorded as the feature deviation ratio of the passing event; substituting the feature deviation ratio into the logarithmic compensation function to obtain the particle count increment of the passing event.

[0051] It should be noted that the single-particle baseline deviation distance and the cluster baseline deviation distance represent the similarity between the flow characteristics of the passing event and typical single-particle and cluster events under the current environment, respectively, with smaller distances indicating greater similarity. When the single-particle baseline deviation distance of the passing event is less than the cluster baseline deviation distance, it indicates that its motion characteristics are more consistent with the rigid single-cell pattern, representing a single-seed passing event, with a particle count increment of 1. When the single-particle baseline deviation distance of the passing event is greater than or equal to the cluster baseline deviation distance, it indicates that its motion characteristics are more consistent with the cluster pattern, representing a cluster passing event. In this case, it is known that it is not a single-particle event, but the specific number of seeds is unknown. Using the distance between the single-particle feature decision vector and the cluster feature decision vector as a benchmark, the degree of deviation of the fluid unit of the passing event relative to the rigid single-cell pattern is measured. The feature deviation ratio quantifies the multiple relationship between the complexity and overlap severity of the cluster of the passing event relative to the typical cluster, eliminating the influence of environmental variables on the feature values ​​and making clusters under different operating conditions comparable.

[0052] Under high-throughput conditions in pneumatic conveying, the increase in the number of seeds within a cluster can lead to a saturation effect in signal characteristics (such as overlap and occlusion). The growth rate of characteristic values ​​such as transmission instability, waveform morphology residuals, and characteristic deviation ratios is slower than the growth rate of the actual number of seeds, exhibiting a non-linear relationship. Using a logarithmic compensation function to correct for the physical saturation effect can restore the compressed characteristic values ​​to the true equivalent particle number, thereby significantly improving the accuracy of seed density measurement under congested conditions. The formula for calculating the logarithmic compensation function is: In the formula, For the increment of the particle count through the event; This is the preset cluster overlap compensation coefficient; is the characteristic deviation ratio of the event; ln is the logarithmic function with the natural constant e as the base; Round is the rounding function.

[0053] Preset cluster overlap compensation coefficient The method for obtaining the data is as follows: On an experimental platform simulating actual working conditions, using seeds of the same variety, a series of typical cluster samples containing different known numbers of seeds (such as 2, 3, 4, 5, etc.) are artificially configured; these samples are sequentially transported using a pneumatic launcher, data is collected synchronously, and the feature deviation ratio of each cluster event is calculated; based on the known actual number of seeds, the least squares method is used to fit and optimize the parameter λ in the above logarithmic compensation function to minimize the root mean square error between the model prediction value and the actual value; finally, the optimal λ value obtained by fitting is used as the cluster overlap compensation coefficient.

[0054] In this embodiment of the invention, the method for obtaining the total seeding amount after the completion of a passing event includes: setting the initial total seeding amount to zero; taking the sum of the particle count increment of the first passing event within the runtime segment and the initial seeding amount as the total seeding amount after the completion of the first passing event; and for passing events after the first passing event within the runtime segment, taking the sum of the particle count increment of each passing event and the seeding amount after the completion of the previous passing event as the total seeding amount after the completion of each passing event.

[0055] This invention is now complete.

[0056] Example 2: This invention proposes a seedling aerial seeding density identification system based on a pneumatic launcher. Please refer to [link / reference]. Figure 2 The diagram illustrates a system structure of a seedling aerial seeding density identification system based on a pneumatic launching device, according to an embodiment of the present invention. The system includes: The data acquisition module 510 is used to acquire the signal amplitude of the upstream and downstream monitoring points of the straight pipe section of the pneumatic launcher at each moment during the device's operating period in real time; determine the passing event based on the signal amplitude, and acquire the upstream observation sequence and downstream observation sequence for each passing event; The flow feature recognition module 520 is used to determine the waveform mapping path between the upstream observation sequence and the downstream observation sequence of each passing event; and to determine the flow feature vector of each passing event based on the trajectory deviation and morphological residual of the waveform mapping path. The decision vector analysis module 530 is used to divide the working period of the device into a start-up period and a running period; based on the flow feature vector of the previous passing event of each passing event in the running period, the classification benchmark of the flow feature vector of the passing event in the start-up period is adjusted, and the single-particle feature decision vector and cluster feature decision vector of each passing event in the running period are determined. The seeding density identification module 540 is used to determine the particle count increment of each passing event based on the degree of deviation of the flow state feature vector of each passing event within the runtime period from the single-particle feature decision vector and the cluster feature decision vector, and to calculate the total seeding amount after each passing event is completed.

[0057] It should be noted that the devices provided in the above embodiments are only illustrative examples of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the seedling aerial seeding density identification system based on a pneumatic launcher and the seedling aerial seeding density identification method based on a pneumatic launcher provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0058] Example 3: Figure 3 This is a schematic diagram of a computer device for identifying seedling aerial seeding density based on a pneumatic launching device, as provided in one embodiment of the present invention. For example,... Figure 3 As shown, the computer device includes: a memory 601, a processor 602, and a computer program 603 stored in the memory 601 and running on the processor 602. When the processor 602 executes the computer program 603, the computer device can execute any of the seedling aerial seeding density identification methods based on a pneumatic launcher described above.

[0059] Furthermore, this application also protects an apparatus that may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to perform a seedling aerial seeding density identification method based on a pneumatic launching device provided in this application.

[0060] This embodiment can divide the device into functional modules based on the above method example. For example, each module can correspond to a separate function, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0061] It should be understood that the device provided in this embodiment is used to perform the above-described seedling aerial seeding density identification method based on a pneumatic launching device, and therefore can achieve the same effect as the above-described implementation method.

[0062] When using integrated units, the device may include a processing module and a storage module. When applied to a workpiece, the processing module can be used to control and manage the workpiece's operations. The storage module can be used to support the execution of program code by the workpiece.

[0063] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits contained in conjunction with the disclosure of this application. The processor may also be a combination of functions that implement computing capabilities, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and microprocessors, etc., and the storage module may be a memory.

[0064] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0065] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0066] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying seedling aerial seeding density based on a pneumatic launcher, characterized in that, The method includes: The signal amplitude of the upstream and downstream monitoring points of the straight pipe section of the pneumatic launcher is acquired in real time at each moment during the device's operating period; based on the signal amplitude, a passage event is determined, and the upstream and downstream observation sequences for each passage event are acquired; Determine the waveform mapping path between the upstream and downstream observation sequences of each passing event; based on the trajectory deviation and morphological residual of the waveform mapping path, determine the flow characteristic vector of each passing event; The working period of the device is divided into a startup period and a running period. Based on the flow feature vector of the previous passing event of each passing event in the running period, the classification criterion of the flow feature vector of the passing event in the startup period is adjusted, and the single-particle feature decision vector and cluster feature decision vector of each passing event in the running period are determined. Based on the degree of deviation of the flow feature vector of each passing event within the runtime period from the single-particle feature decision vector and the cluster feature decision vector, the particle count increment of each passing event is determined, and the total seeding amount after each passing event is completed is calculated.

2. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 1, characterized in that, Determining the waveform mapping path between the upstream and downstream observation sequences of each passing event includes: Construct a distance matrix between the upstream and downstream observation sequences for each event, where the element in the nth row and mth column of the distance matrix is ​​the square of the absolute difference between the nth element in the upstream observation sequence and the mth element in the downstream observation sequence; where n and m are the element indices of the upstream and downstream observation sequences, respectively. Obtain the theoretical shortest flight time delay, the theoretical longest flight time delay, and the data sampling frequency between the upstream and downstream monitoring points; calculate the difference between the theoretical longest flight time delay and the theoretical shortest flight time delay, and use the product of the difference and the data sampling frequency as the maximum relative index deviation; In the distance matrix, elements whose row index is greater than or equal to the column index, and whose column index is less than or equal to the sum of the deviations between the row index and the maximum relative index, are denoted as valid search elements. Within the region comprised of the effective search elements, a dynamic programming algorithm is used to search for the path with the minimum cumulative cost from the top-left element to the bottom-right element of the distance matrix. The coordinate pairs consisting of the row index and column index of each element traversed by the path are arranged in the order of traversal to obtain a sequence of coordinate pairs, which is denoted as the waveform mapping path between the upstream observation sequence and the downstream observation sequence of each passing event.

3. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 1, characterized in that, Determining the flow state feature vector for each passing event includes: For each waveform mapping path that passes through an event, the row index and column index of the coordinate pair in the waveform mapping path are linearly fitted to obtain the ideal transmission path; Calculate the shortest distance from each coordinate pair in the waveform mapping path to the ideal transmission path, and denote it as the transmission trajectory deviation; The square of the row index of each coordinate pair in the waveform mapping path corresponding to the element in the upstream observation sequence is used as the biomass weighting factor for each coordinate pair. The sum of the products of the transmission trajectory deviation and the biomass weighting factor for each coordinate pair in the waveform mapping path is calculated and recorded as the weighted deviation value. The ratio of the weighted deviation value as the numerator and the sum of the biomass weighting factors for each coordinate pair in the waveform mapping path and the sum of a preset positive number as the denominator is used as the transmission instability value of the event. The waveform morphology residual value is obtained by averaging the absolute differences between the corresponding elements of the row index in the upstream observation sequence and the corresponding elements of the column index in the downstream observation sequence for each coordinate pair in the waveform mapping path. The transmission instability value and the waveform morphology residual value are standardized to obtain the standard instability value and the standard residual value in turn; the standard instability value and the standard residual value of each passing event constitute a flow characteristic vector.

4. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 1, characterized in that, The determination of the single-particle feature decision vector and cluster feature decision vector for each passing event within the runtime segment includes: The flow feature vectors of all events passing through during the startup period are subjected to binary clustering to obtain two cluster center vectors; Calculate the magnitude of the two cluster center vectors, denote the cluster center vector with the smallest magnitude as the single-particle initiation feature vector, and denote the cluster center vector with the largest magnitude as the cluster initiation feature vector; For the first pass event within the runtime segment, the single-particle initiation feature vector and the cluster initiation feature vector are used as the single-particle feature decision vector and the cluster feature decision vector, respectively, for the pass event. For each passing event after the first passing event within the runtime segment, the single-particle initiation feature vector and the cluster initiation feature vector are adjusted using the flow state feature vector of the previous passing event, thereby obtaining the single-particle feature decision vector and the cluster feature decision vector for each passing event in sequence.

5. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 1, characterized in that, The determination of the particle count increment for each passing event includes: For all passing events during the runtime period, calculate the distances between the flow feature vector of the passing event and its single-particle feature decision vector and cluster feature decision vector, respectively, and denoted as single-particle reference deviation distance and cluster reference deviation distance; When the single-particle reference deviation distance of the event is less than the cluster reference deviation distance, the particle count increment of the event is set to 1. When the single-particle reference deviation distance of the event is greater than or equal to the cluster reference deviation distance, the ratio obtained by taking the distance between the flow characteristic vector of the event and the single-particle characteristic decision vector as the numerator and the distance between the single-particle characteristic decision vector and the cluster characteristic decision vector of the event as the denominator is denoted as the characteristic deviation ratio of the event. Substituting the feature deviation ratio into the logarithmic compensation function yields the particle count increment for each event.

6. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 5, characterized in that, The formula for calculating the logarithmic compensation function is as follows: ； In the formula, This is the increment of the particle count for the event; This is the preset cluster overlap compensation coefficient; is the characteristic deviation ratio of the event; ln is the logarithmic function with the natural constant e as the base; Round is the rounding function.

7. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 1, characterized in that, The calculation of the total seeding amount after each event is completed includes: Set the initial total seeding quantity to zero; use the sum of the particle count increment of the first pass event within the runtime segment and the initial seeding quantity as the total seeding quantity after the first pass event is completed; For each pass event following the first pass event within the runtime segment, the sum of the particle count increment of each pass event and the seeding amount after the completion of the previous pass event is taken as the total seeding amount after the completion of each pass event.

8. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 4, characterized in that, The process of adjusting the single-particle initiation feature vector and the cluster initiation feature vector using the flow state feature vector of the preceding event for each passing event, and sequentially obtaining the single-particle feature decision vector and the cluster feature decision vector for each passing event, includes: The first passing event after the first passing event in any runtime segment is denoted as the example event. The difference between the constant 1 and the preset update rate coefficient is used as the weight of the single-particle start feature vector, and the preset update rate coefficient is used as the weight of the flow feature vector of the previous passing event of the example event. The single-particle start feature vector and the flow feature vector of the previous passing event of the example event are weighted and summed. The resulting new vector is denoted as the single-particle feature decision vector of the example event. The single-particle initiation feature vector in the calculation process of the single-particle feature decision vector of the example event is updated to the cluster initiation feature vector to obtain the cluster feature decision vector of the example event.

9. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 2, characterized in that, The process of obtaining the upstream and downstream observation sequences for each passing event includes: The first moment is recorded as the initial analysis moment. When the signal amplitude of the upstream monitoring point at the analysis moment is less than or equal to the preset background noise threshold, the next adjacent moment is recorded as the new analysis moment. When the signal amplitude of the upstream monitoring point at the analysis time is greater than the background noise threshold, the analysis time is taken as the trigger time of the passing event. Starting from the trigger time of the event, the signal amplitudes of each time within the preset event time are arranged in chronological order to obtain the upstream signal sequence of the corresponding passing event. The difference between the theoretical longest flight time delay and the theoretical shortest flight time delay is taken as the theoretical flight time tolerance; the downstream monitoring point takes the theoretical shortest flight time delay as the starting point at the trigger time, and arranges the signal amplitudes at each time within the total time of the preset event time and the theoretical flight time tolerance in chronological order to obtain the downstream signal sequence of the event. The first moment after adding the preset event time to the trigger time of each event is recorded as the new analysis moment; Determine whether the signal amplitude at the new analysis time is greater than the background noise threshold, determine the trigger time of the next passing event, and obtain the upstream and downstream signal sequences of the passing event; The upstream and downstream signal sequences are subjected to baseline removal and amplitude normalization processes to obtain the upstream and downstream observation sequences of the passing event.

10. The method for identifying seedling aerial seeding density based on a pneumatic launching device according to claim 4, characterized in that, The method for performing binary clustering on the flow feature vectors of all passing events during the startup period is the K-means clustering algorithm, where K equals 2.

Images