A method and system for detecting missed planting of peanut seeds

Abstract

The present application relates to the technical field of peanut seeding, and particularly relates to a peanut seeding missing detection method and system. The method comprises the following steps: calculating a theoretical seed distribution beat signal of a seed distribution mechanism, obtaining a theoretical seed distribution period and a theoretical seed distribution time; collecting actual passing signals of seeds through a detection unit arranged in a seed distribution channel or a seed guide channel; collecting vibration interference signals in the seed distribution process; establishing a time mapping relationship between the theoretical seed distribution event and the actual seed passing event, and constructing an effective detection window; setting a multi-channel structure according to the actual passing signals, performing frequency domain analysis and time domain analysis on the actual passing signals and the interference signals, extracting missing seed identification features, so as to realize real-time monitoring of the seed distribution state; fusing the theoretical seed distribution event, the actual seed passing event and the identification features, constructing a missing seed comprehensive criterion function, and outputting a missing seed determination result. The accuracy of peanut missing seed identification is improved.

Description

Technical Field

[0001] This invention relates to the field of peanut technology, and in particular to a method and system for detecting missed planting in peanut cultivation. Background Technology

[0002] Peanuts, as an important oilseed and cash crop, are directly affected by the quality of their sowing, which in turn impacts seedling uniformity, canopy formation, and final yield. In precision sowing operations, missed sowing can lead to problems such as missing holes, uneven plant spacing, and inconsistent seedling conditions, thereby affecting field management effectiveness and crop yield. Therefore, real-time online detection of missed sowing during peanut sowing has significant engineering application value.

[0003] Existing missed seed detection technologies mostly employ a single sensor to detect seeds through events, such as photoelectric detection, Hall effect detection, or simple counting methods. While these detection technologies are relatively simple in structure, they still have the following shortcomings in practical applications:

[0004] (1) During the sowing process, there are vibrations, speed fluctuations, material flow fluctuations and random noise. Single threshold detection is prone to false detection and false negative detection.

[0005] (2) Relying solely on "whether a pulse is detected" to determine missed seeding makes it difficult to distinguish between normal seed passing signals and pseudo pulses, overlapping pulses, or abnormal collision signals;

[0006] (3) There is no precise time alignment mechanism between the theoretical seeding rhythm and the actual seeding events. When the seeding speed changes or the transmission delay fluctuates, the correspondence is easily confused.

[0007] (4) Existing methods are mostly limited to the simple identification of single events and lack the ability to comprehensively analyze the missed broadcast range, the degree of missed broadcast and the potential risk of missed broadcast;

[0008] (5) Under complex working conditions, there is a lack of a comprehensive detection model that combines time domain, frequency domain and time-frequency domain features, resulting in insufficient detection robustness. Summary of the Invention

[0009] The purpose of this invention is to overcome the above-mentioned defects in the prior art and to propose a method and system for detecting missed planting in peanut sowing, so as to solve the problems of low accuracy of missed planting detection, insufficient anti-interference ability, and inaccurate correspondence between theoretical events and actual events in the prior art, thereby improving the accuracy of peanut missed planting identification.

[0010] The technical solution of this invention is: a method for detecting missed planting in peanut sowing, comprising the following steps:

[0011] S1. Calculate the theoretical seeding cycle signal of the seeding mechanism to obtain the theoretical seeding cycle and theoretical seeding time;

[0012] S2. Collect the actual passing signal of the seeds by a detection unit set in the seed dispensing channel or seed guiding channel;

[0013] S3. Collect vibration interference signals during the seeding process;

[0014] S4. Establish the time mapping relationship between theoretical seeding events and actual seeding events, and construct an effective detection window;

[0015] S5. Based on the actual passing signal, set up a multi-channel structure, perform frequency domain analysis and time domain analysis on the actual passing signal and interference signal, extract the missed seeding identification features, so as to realize real-time monitoring of the seeding status;

[0016] S6. Integrate theoretical seeding events, actual seeding events, and identification features to construct a comprehensive criterion function for missed seeding and output the missed seeding judgment result.

[0017] In step S1 of the present invention, the seed metering disc of the seeding mechanism has Each seeding station has a real-time angular velocity of the seeding disc. The theoretical seeding cycle Calculate using the following formula:

[0018] ,

[0019] The kth theoretical seeding event time :

[0020] ,

[0021] in, This is the time of the (k-1)th theoretical seeding event.

[0022] In step S2, when the seed has not passed through the detection area, the light-receiving device of the detection unit receives the reference light intensity. ;

[0023] When a seed passes through the detection area, the seed blocks, absorbs, or scatters the detection beam, causing the received light intensity to change as follows:

[0024] ,

[0025] in, The occlusion coefficient of the seed on the light beam;

[0026] The light-receiving device converts the received light intensity into photocurrent:

[0027] ,

[0028] in, The photocurrent output by the light-receiving device; The photoelectric response sensitivity of the light-receiving device;

[0029] And it is converted into output voltage by the signal conditioning circuit:

[0030] ,

[0031] in, To detect the output voltage of the unit; The equivalent gain of the conditioning circuit; This is the circuit bias voltage;

[0032] With preset threshold When comparing, When a seed actually passes through, it is determined that a seed has passed through, thereby obtaining a signal indicating that the seed has actually passed through the seeding channel or seed guiding channel.

[0033] In step S3, vibration interference signals are acquired using an accelerometer.

[0034] Let the output voltages of the accelerometer in the x, y, and z directions be respectively... , , The zero bias voltages of the accelerometer in the x, y, and z directions are respectively , , The sensitivity coefficients of the accelerometer in the x, y, and z directions are respectively , , Then the corresponding triaxial acceleration can be expressed as:

[0035] ,

[0036] ,

[0037] ,

[0038] in, The vibration acceleration is in the x-axis direction; The y-axis is the vibration acceleration. The vibration acceleration is in the z-axis direction; This is the real-time output voltage signal of the accelerometer in the x-axis measurement channel; This is the real-time output voltage signal of the accelerometer in the y-axis measurement channel; This is the real-time output voltage signal of the accelerometer in the z-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the x-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the y-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the z-axis measurement channel; This represents the sensitivity coefficient of the accelerometer in the x-axis measurement channel. This represents the sensitivity coefficient of the accelerometer in the y-axis measurement channel. This represents the sensitivity coefficient of the accelerometer in the z-axis measurement channel.

[0039] The above three-axis accelerations are combined into a comprehensive vibration signal, which is the vibration interference signal:

[0040] .

[0041] In step S4, there is a propagation delay τ from the seeding position to the detection position. An effective detection window is constructed corresponding to the k-th theoretical seeding event.

[0042] ,

[0043] in, For effective detection window; To allow for time deviation;

[0044] Search for candidate events within the window: if there are no candidate events in the window, it is directly determined that the k-th theoretical seeding event has been missed; if there are candidate events in the window, further feature extraction and discrimination are performed on the candidate events.

[0045] The specific implementation process of step S5 is as follows:

[0046] S5.1 Set up the actual multi-channel structure of the signal and extract the total energy characteristics, main characteristic value characteristics, energy concentration, and channel correlation characteristics;

[0047] S5.2 Frequency domain analysis to obtain frequency domain energy characteristics;

[0048] S5.3 Time domain analysis: extract time location features, peak features, and pulse width features.

[0049] In step S5.1, a signal segment is extracted within the valid detection window of the k-th theoretical priming event. ,in , This represents the number of sampling points within the window.

[0050] An array of photoelectric detection units is arranged in the seed dispensing channel, seed guiding channel, or seed landing area. These array of photoelectric detection units form a multi-channel structure, and each channel synchronously collects the seed passing signal.

[0051] ,

[0052] Where m is the total number of detection channels in the multi-channel structure; n is the sampling point number of the time series signal of each channel;

[0053] Arrange the signals from each channel in rows to construct a seed-through signal matrix:

[0054] ,

[0055] To eliminate DC bias and amplitude differences between channels, the signal matrix is ​​subjected to mean-reduction processing. If the zero-mean signal matrix is ​​a column vector of all 1s, then... :

[0056] ,

[0057] Among them, the mean vector for:

[0058] ,

[0059] Based on zero-mean signal matrix Construct the covariance matrix :

[0060] ,

[0061] Based on covariance matrix Extract total energy features, principal eigenvalue features, energy concentration, and channel correlation features:

[0062] S5.1.A. Extracting Total Energy Characteristics:

[0063] The expression for the total energy characteristic is:

[0064] ,

[0065] Among them, E k This represents the total energy of all channels;

[0066] Total energy characteristics are used to characterize the overall fluctuation intensity of multi-channel signals within the current detection window;

[0067] S5.1.B. Extracting principal feature values:

[0068] Based on the covariance matrix, eigenvalue decomposition is performed, and the largest eigenvalue is extracted as the principal eigenvalue. Its expression is as follows:

[0069] ,

[0070] in, For feature vectors; For eigenvalues;

[0071] The principal eigenvalues ​​are obtained using the above formula:

[0072] ,

[0073] The largest eigenvalue Characterizes the main energy direction;

[0074] The principal eigenvalue is used to characterize the intensity of the dominant change pattern of the multi-channel signal within the current detection window;

[0075] S5.1.C, Energy Concentration Extraction:

[0076] Energy Concentration The calculation formula is:

[0077] ,

[0078] Energy concentration is used to characterize the degree of concentration of signal energy in the dominant feature direction within the current detection window;

[0079] S5.1.D. Extracting channel correlation features:

[0080] Choose the average of the off-diagonal elements of the covariance matrix As a channel correlation feature:

[0081] ,

[0082] Channel correlation features are used to characterize the consistency and degree of coordinated change in response among multiple detection channels.

[0083] In step S5.2, a signal segment is extracted from the window. Perform a Fast Fourier Transform:

[0084] ,

[0085] Establish the relationship between the discrete frequency point number d and the actual physical frequency:

[0086] ,

[0087] in, Where is the sampling frequency; L is the window length; This is the actual frequency corresponding to the d-th discrete frequency point;

[0088] Obtain the power spectrum of the corresponding d-th discrete frequency point :

[0089] ,

[0090] Let the target frequency band be Then the frequency domain energy characteristics Defined as:

[0091] ,

[0092] Frequency domain energy characteristics can reflect the degree of energy concentration of candidate events within the target frequency range, and can be used to distinguish normal signals from vibration interference signals.

[0093] In step S5.3, the specific feature extraction process is as follows:

[0094] S5.3.A. Extraction of temporal and location features:

[0095] By finding the sampling point location corresponding to the largest pulse within the window. :

[0096] ,

[0097] Obtain the actual time of passing :

[0098] ,

[0099] in: ΔT is the start time of the window; ΔT is the sampling period.

[0100] S5.3.B. Peak Feature Extraction:

[0101] Peak pulse within the window Defined as:

[0102] ;

[0103] S5.3.C, Pulse Width Feature Extraction:

[0104] Set threshold Find the sampling point where the signal first exceeds the threshold. and the last sampling point that exceeded the threshold Then the pulse is wide. for:

[0105] .

[0106] In step S6, the time position, peak value, pulse width, frequency domain energy, and covariance feature deviation of the candidate over-seeding event corresponding to the k-th theoretical over-seeding event are extracted and compared with the normal single-seed over-seeding reference template to construct a normalized deviation:

[0107] ,

[0108] in, For time and location features; This represents the maximum allowable time deviation between the theoretical seeding time and the candidate seeding event time. Let be the absolute value of the time difference between the theoretical time of the k-th theoretical sorting event and the actual time of the corresponding candidate sorting event;

[0109] ,

[0110] in, Peak characteristics; This serves as the baseline value for peak characteristics;

[0111] ,

[0112] in, Pulse width characteristics; This serves as the baseline value for pulse width characteristics;

[0113] ,

[0114] in, Frequency domain energy characteristics; This serves as a benchmark value for the frequency domain energy characteristics;

[0115] ,

[0116] Wherein: d C This refers to multi-channel eigenvalue features; This serves as the baseline value for multi-channel feature values;

[0117] Based on this, a comprehensive criterion function is constructed:

[0118] ,

[0119] in, , , , , , All are weighting coefficients;

[0120] If there are no candidate events in the window, it is directly determined as a missed broadcast; if there are candidate events in the window but the comprehensive criterion function value is greater than the preset threshold, it is determined as a missed broadcast. When there are candidate events in the window and the value of the comprehensive criterion function is not greater than the threshold, it is determined to be a normal seeding event.

[0121] Final Missed Broadcast Detection Function Represented as:

[0122] ,

[0123] in, An empty set means no response signal was received; There is a set of candidate events within the window;

[0124] The time indicates a missed broadcast. The time indicates a normal process.

[0125] This application also proposes a detection system for implementing the above-mentioned peanut sowing missed sowing detection method, comprising:

[0126] Theoretical beat acquisition module, used to acquire the theoretical beat signal of the seeding mechanism;

[0127] The actual seed passage detection module is used to collect the actual passing signals of seeds in the seed dispensing channel, seed guiding channel, or seed landing area;

[0128] The interference and operating condition acquisition module is used to acquire vibration interference signals and operating status signals;

[0129] The time mapping module is used to establish the time mapping relationship between theoretical seeding events and actual seeding events and to construct an effective detection window;

[0130] The signal processing and feature extraction module is used for input and frequency domain analysis, time-frequency analysis, and extraction of time position, peak value, pulse width, and frequency domain energy features.

[0131] The missed broadcast detection module is used to determine whether a missed broadcast has occurred based on window constraints and a comprehensive criterion function.

[0132] The beneficial effects of this invention are:

[0133] (1) This application achieves precise tracking of a single seeding event during the sowing process by introducing a corresponding analysis mechanism between theoretical seeding rhythm signals and actual over-seeding signals, thereby improving the pertinence and accuracy of missed seeding identification;

[0134] (2) By constructing an effective detection time range, this application matches theoretical seeding events with actual seeding events, which can effectively reduce mismatches caused by transmission delay, speed fluctuations or random disturbances;

[0135] (3) This application comprehensively utilizes multi-dimensional features such as time position, peak value, pulse width, frequency band energy and time frequency distribution to judge candidate events, which can effectively distinguish normal signal, pseudo pulse signal and abnormal interference signal, and improve the detection reliability under complex working conditions;

[0136] (4) This application combines filtering, frequency domain analysis and time-frequency analysis, which can effectively suppress vibration noise, impact noise and random interference, and enhance the anti-interference capability of the system;

[0137] (5) This application can not only identify single missed broadcast events, but also analyze the degree of missed broadcast and the risk of missed broadcast, and output alarm, warning or control signals, which has good real-time performance and engineering application value.

[0138] (6) This application introduces a covariance matrix to jointly characterize the correlation and consistency of multi-channel detection signals, thus solving the technical deficiency of existing methods that cannot accurately identify missed seeding events based solely on a single signal amplitude or threshold under complex vibration environments. Compared with single-channel discrimination methods, the covariance matrix can not only reflect whether a signal response exists, but also further reflect whether the responses of each channel are synchronized and coordinated. Based on this feature, it can effectively distinguish between different states such as normal single-seed passage, missed seeding, pseudo-pulse interference, and abnormal multi-seed passage, thereby reducing misjudgments caused by vibration noise and local abnormal signals, and improving the robustness and engineering applicability of online missed seeding detection.

[0139] In summary, the method and system proposed in this application do not rely on a single detection principle, are compatible with multiple combinations of detection units, and are easy to apply to different types of peanut planting monitoring scenarios. Attached Figure Description

[0140] Figure 1 This is a general flowchart of the method described in this application;

[0141] Figure 2 This is a schematic diagram illustrating the time mapping relationship between the theoretical seeding events and the actual seeding events in this application;

[0142] Figure 3 A schematic diagram illustrating the construction of the effective detection window for this application;

[0143] Figure 4 This is a schematic diagram of the time-domain waveform of the detection signal in this application;

[0144] Figure 5 This is a schematic diagram of the spectrum of the detected signal after the Fast Fourier Transform (FFT) in this application;

[0145] Figure 6 This is a schematic diagram of the power spectral density analysis of this application. Detailed Implementation

[0146] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0147] Specific details are set forth in the following description to provide a full understanding of the invention. However, the invention can be practiced in many ways other than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0148] This application proposes a method for detecting missed planting in peanut cultivation, such as... Figure 1 As shown, the method includes the following steps.

[0149] The first step is to calculate the theoretical seeding rhythm signal of the seeding mechanism to obtain the theoretical seeding cycle and theoretical seeding time.

[0150] The theoretical seeding cycle signal of the seeding mechanism is acquired by a speed encoder. The speed encoder is mainly used to detect the real-time angular velocity of the seeding disc and is placed on the output shaft of the seeding disc.

[0151] The seed metering disc of the seeding mechanism has Each seeding station has a real-time angular velocity of the seeding disc. The theoretical seeding cycle is calculated using the following formula:

[0152] ,

[0153] in, This is the theoretical seeding cycle; For seeding workstations; This refers to the real-time angular velocity of the seeding disc.

[0154] Based on this, the theoretical timing of the k-th sorting event can be obtained:

[0155] ,

[0156] in, This represents the time of the k-th theoretical seeding event; This is the time of the (k-1)th theoretical seeding event.

[0157] The second step involves collecting the actual passing signal of the seeds through a detection unit installed in the seed dispensing or seed guiding channel.

[0158] The detection unit used in this application is a photoelectric detection unit, preferably located in the seed dispensing channel or seed guiding channel. The photoelectric detection unit includes a light-emitting device and a light-receiving device, with the light-emitting device emitting a detection beam towards the light-receiving device.

[0159] When the seed does not pass through the detection area, the light-receiving device receives the reference light intensity. When a seed passes through the detection area, the seed blocks, absorbs, or scatters the detection beam, causing a change in the received light intensity:

[0160] ,

[0161] in, The occlusion coefficient of the seed on the light beam; This represents the received light intensity in the seedless state.

[0162] The light-receiving device converts the received light intensity into photocurrent:

[0163] ,

[0164] in, The photocurrent output by the light-receiving device; This refers to the photoelectric response sensitivity of the light-receiving device.

[0165] And it is converted into output voltage by the signal conditioning circuit:

[0166] ,

[0167] in, To detect the output voltage of the unit; The equivalent gain of the conditioning circuit; This is the circuit bias voltage.

[0168] With preset threshold When comparing, When a seed actually passes through, it is determined that a seed has passed through, thereby obtaining a signal indicating that the seed has actually passed through the seeding channel or seed guiding channel.

[0169] The third step is to collect vibration interference signals during the seeding process.

[0170] During seeding, factors such as the rotation of the seeding mechanism, the impact of the machine's movement, and the vibration of the frame can disturb the output of the photoelectric detection unit, introducing vibration interference components unrelated to seed passage into the actual detection signal. To improve the accuracy and stability of identifying actual seed over-seeding events, vibration interference signals generated during seeding are collected in this step. These vibration interference signals are acquired by an accelerometer. The vibration interference signals characterize the intensity of mechanical disturbance within the current detection period and further participate in detection threshold correction, effective detection window adjustment, and auxiliary determination of actual seed over-seeding events.

[0171] Let the output voltages of the accelerometer in the x, y, and z directions be respectively... , , The zero bias voltages of the accelerometer in the x, y, and z directions are respectively , , The sensitivity coefficients of the accelerometer in the x, y, and z directions are respectively , , Then the corresponding triaxial acceleration can be expressed as:

[0172] ,

[0173] ,

[0174] ,

[0175] in: The vibration acceleration is in the x-axis direction; The y-axis is the vibration acceleration. The vibration acceleration is in the z-axis direction; This is the real-time output voltage signal of the accelerometer in the x-axis measurement channel; This is the real-time output voltage signal of the accelerometer in the y-axis measurement channel; This is the real-time output voltage signal of the accelerometer in the z-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the x-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the y-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the z-axis measurement channel; This represents the sensitivity coefficient of the accelerometer in the x-axis measurement channel. This represents the sensitivity coefficient of the accelerometer in the y-axis measurement channel. This represents the sensitivity coefficient of the accelerometer in the z-axis measurement channel.

[0176] The above three-axis accelerations are combined into a comprehensive vibration signal, which is the vibration interference signal:

[0177] .

[0178] The fourth step is to establish a time mapping relationship between theoretical seeding events and actual seeding events, and to construct an effective detection window.

[0179] Considering the propagation delay τ from the seeding position to the detection position, construct the effective detection window corresponding to the k-th theoretical seeding event:

[0180] ,

[0181] in: τ represents the effective detection window; τ represents the propagation delay. To allow for time deviation.

[0182] The purpose of constructing an effective detection window is to limit the continuously collected detection signals to the target time interval in which the seeds may actually pass through, based on the theoretical seeding sequence and the actual movement of seeds in the seeding or seed-guiding channel. Through the effective detection window, a temporal correspondence between theoretical seeding events and actual seeding events can be established, providing a basis for subsequent signal feature extraction, effective seeding determination, and missed seeding identification, while reducing the amount of data for signal processing and improving the real-time performance of online detection.

[0183] like Figure 2 and Figure 3 As shown, for the k-th theoretical sorting event, according to ,τ, Construct an effective detection window .

[0184] Search for candidate events within this window. If no candidate event is found, it is directly determined that the k-th theoretical seeding event has been missed. If a candidate event exists within the window, further feature extraction and discrimination are performed on the candidate event.

[0185] The fifth step involves setting up a multi-channel structure based on the actual transmitted signals to perform frequency and time domain analysis on both the transmitted and interference signals, extracting missed seeding identification features to achieve real-time monitoring of the seeding status. The specific implementation process of this step is described below.

[0186] First, set up a multi-channel structure for actual signal transmission.

[0187] like Figure 4 As shown, a signal segment is extracted within the valid detection window of the k-th seeding event. ,in , This represents the number of sampling points within the window.

[0188] An array of photoelectric detection units is arranged in the seed dispensing channel, seed guiding channel, or seed landing area. The array of photoelectric detection units forms a multi-channel structure, and each channel synchronously collects the signal of the seed passing through.

[0189] ,

[0190] Where m represents the total number of detection channels in the multi-channel structure, or the number of channels formed by the array photoelectric detection unit; n is the sampling point number of the time series signal of each channel.

[0191] Arrange the signals from each channel in rows to construct a seed-through signal matrix:

[0192] ,

[0193] Each row represents the time-series signal of one detection channel; each column represents the response at different spatial locations at the same time. To eliminate DC bias and amplitude differences between channels, the signal matrix undergoes mean-reduction processing. If the zero-mean signal matrix is ​​a column vector of all 1s, then... :

[0194] ,

[0195] Among them, the mean vector for:

[0196] ,

[0197] Based on zero-mean signal matrix Construct the covariance matrix :

[0198] ,

[0199] Based on covariance matrix Further features such as total energy, principal eigenvalues, energy concentration, and channel correlation can be extracted. The specific feature extraction process is described below.

[0200] I. Extracting total energy characteristics.

[0201] Total energy features are extracted based on the covariance matrix. The total energy feature is defined as the sum of the diagonal elements of the covariance matrix, or expressed as the trace of the covariance matrix, and its expression is:

[0202] ,

[0203] Among them: E k This represents the total energy of all channels.

[0204] The total energy feature is used to characterize the overall fluctuation intensity of the multi-channel signal within the current detection window. When a seed actually passes through the detection area, the overall change in the detection signal is enhanced, corresponding to an increase in the total energy feature; when only background noise or weak interference exists, the total energy feature is relatively small. Therefore, introducing the total energy feature is beneficial for improving the ability to identify real seed passing events from the perspective of overall response intensity.

[0205] II. Extract the main feature values.

[0206] Based on the covariance matrix, eigenvalue decomposition is performed, and the largest eigenvalue is extracted as the principal eigenvalue. Its expression is as follows:

[0207] ,

[0208] in, For feature vectors; These are the eigenvalues.

[0209] The principal eigenvalues ​​are obtained using the above formula:

[0210] ,

[0211] The largest eigenvalue Characterizes the main energy direction.

[0212] The principal eigenvalue is used to characterize the intensity of the dominant change pattern of multi-channel signals within the current detection window. When a seed actually passes through the detection area, multiple channels typically produce relatively consistent response changes, increasing the fluctuation of the covariance matrix in the principal direction and correspondingly increasing the principal eigenvalue. When only random noise or scattered interference exists, the signal energy distribution is more discrete, and the principal eigenvalue is relatively small. Therefore, introducing the principal eigenvalue is beneficial for improving the ability to identify real seed passing events from the perspective of dominant structure.

[0213] III. Extracting energy concentration.

[0214] Energy Concentration The calculation formula is:

[0215] ,

[0216] Energy concentration is used to characterize the degree of concentration of signal energy in the dominant feature direction within the current detection window. A larger energy concentration value indicates that the signal energy is more concentrated in a few dominant components, suggesting that the candidate event has strong structural consistency and dominant features, and is more likely to correspond to the actual event. Conversely, a smaller energy concentration value indicates that the signal energy distribution is more dispersed, and is more likely to correspond to random noise, mechanical vibration, or other invalid disturbances. Therefore, introducing the energy concentration feature to measure whether the signal is concentrated in a certain main channel or direction is beneficial to improving the accuracy of candidate event validity judgment and missed event identification.

[0217] IV. Extract channel correlation features.

[0218] The mean of the off-diagonal elements of the covariance matrix can be selected. As a channel correlation feature:

[0219] ,

[0220] Channel correlation features are used to characterize the consistency and degree of coordinated change in responses among multiple detection channels. When a seed actually passes through the detection area, adjacent or multiple detection channels typically produce temporally correlated response changes, enhancing the correlation between channels; however, random noise, mechanical vibration, or local abnormal disturbances usually make it difficult to establish stable correlations among multiple channels. Therefore, introducing channel correlation features to measure the synchronicity among multiple channels is beneficial for improving the reliability of identifying actual seed passage events and determining missed seeding from the perspective of multi-channel consistency.

[0221] Second, frequency domain analysis.

[0222] like Figure 5 and Figure 6 As shown, for signals within the window Perform a Fast Fourier Transform:

[0223] ,

[0224] Establish the relationship between the discrete frequency point number d and the actual physical frequency:

[0225] ,

[0226] in, Where L is the sampling frequency and L is the window length. This represents the actual frequency corresponding to the d-th discrete frequency point.

[0227] Obtain the power spectrum of the corresponding d-th discrete frequency point :

[0228] ,

[0229] Let the target frequency band be Then the frequency domain energy characteristics Defined as:

[0230] ,

[0231] Frequency domain energy characteristics can reflect the degree of energy concentration of candidate events within the target frequency range, and are used to distinguish normal signals from vibration interference signals.

[0232] For non-stationary signals, short-time Fourier transform can be used to extract frequency distribution features at different time intervals to improve the recognition effect under variable speed and vibration conditions.

[0233] Third, time-domain analysis.

[0234] The time-domain analysis process includes time location feature extraction, peak value feature extraction, and pulse width feature extraction. The specific feature extraction process is described below.

[0235] V. Extraction of temporal and location features.

[0236] The role of time location feature extraction is to determine the specific time when the signal occurs, and to achieve time sequence matching between theoretical seeding events and actual detection events, thereby providing a basis for candidate event screening, window construction, and missed seeding determination.

[0237] By finding the sampling point location corresponding to the largest pulse within the window. :

[0238] ,

[0239] Obtain the actual time of passing :

[0240] ,

[0241] in: ΔT represents the start time of the window; ΔT represents the sampling period.

[0242] VI. Peak feature extraction.

[0243] Peak features are used to characterize the maximum response amplitude of a candidate event within the detection window, reflecting the signal strength of the candidate event.

[0244] When a seed actually passes through the detection area, the detection signal typically exhibits a significant amplitude change, corresponding to a large peak feature; while the signal changes caused by background noise, mechanical micro-vibration, or weak disturbances are usually smaller. Therefore, by extracting the peak feature and comparing it with a preset threshold, candidate events can be effectively screened and used as one of the amplitude constraints for determining actual seed passage events, thereby improving the accuracy of missed seed identification.

[0245] Peak pulse within the window Defined as:

[0246] .

[0247] VII. Pulse width feature extraction.

[0248] Pulse width features are used to characterize the duration of candidate events on the time axis to reflect the continuous process characteristics of the seed passing through the detection region.

[0249] When a seed actually passes through the detection area, the detection signal typically changes continuously within a certain time range, corresponding to a pulse width within a reasonable range. In contrast, spurious pulses caused by transient noise, mechanical shock, or stray interference usually have narrower pulse widths, while abnormal signals caused by seed retention, multiple seed stacking, or abnormal occlusion typically have wider pulse widths. Therefore, by extracting pulse width features and comparing them with a preset pulse width range, candidate events can be screened for authenticity, serving as an important time constraint for determining genuine seed passage events.

[0250] Set threshold Find the sampling point where the signal first exceeds the threshold. and the last sampling point that exceeded the threshold Then the pulse is wide. for:

[0251] .

[0252] The sixth step is to integrate theoretical seeding events, actual over-seeding events, and identification features to construct a comprehensive criterion function for missed seeding and output the missed seeding judgment result.

[0253] For the candidate over-seeding event corresponding to the k-th theoretical seeding event, extract the time position, peak value, pulse width, frequency domain energy, and covariance features, and compare them with the normal single-seed over-seeding reference template to construct a normalized bias:

[0254] ,

[0255] in, For time and location features; This represents the maximum allowable time deviation between the theoretical seeding time and the candidate seeding event time. Let be the absolute value of the time difference between the theoretical time of the k-th theoretical sorting event and the actual time of the corresponding candidate sorting event;

[0256] ,

[0257] in, Peak characteristics; This is the reference value for the peak characteristic, i.e., the peak value of the signal under normal reference conditions;

[0258] ,

[0259] in, Pulse width characteristics; This is the reference value for pulse width characteristics, i.e., the signal pulse width under normal reference conditions;

[0260] ,

[0261] in, Frequency domain energy characteristics; This is the reference value for the frequency domain energy characteristics, that is, the energy of the signal in the target frequency band under normal reference conditions;

[0262] ,

[0263] Wherein: d C This refers to multi-channel eigenvalue features; It serves as the reference value for the multi-channel characteristic values, i.e., the reference value for the correlation or fusion characteristics of multi-channel signals under normal reference conditions;

[0264] Based on this, a comprehensive criterion function is constructed:

[0265] ,

[0266] in, , , , , , All are weighting coefficients.

[0267] The aforementioned weighting coefficients are determined based on the contribution of each feature to distinguishing different seeding states. Specifically, the time position features, peak features, pulse width features, frequency domain energy features, multi-channel feature values, and acceleration auxiliary features are first normalized. Then, the inter-class separation degree of each feature is calculated using calibration samples among normal seeding, missed seeding, and interference states. The normalized separation degree of each feature is then used as the corresponding weighting coefficient. The greater the feature separation degree, the higher the contribution of that feature to state discrimination, and the larger its corresponding weighting coefficient. In practical applications, initial weights can be obtained first using offline calibration samples, and then the weights can be iteratively corrected by combining detection accuracy, false alarm rate, and false negative rate.

[0268] If there are no candidate events in the window, it is directly determined as a missed broadcast; if there are candidate events in the window but the comprehensive criterion function value is greater than the preset threshold, it is determined as a missed broadcast. When there is a candidate event in the window and the value of the comprehensive criterion function is not greater than the threshold, it is determined to be a normal seeding event.

[0269] Final Missed Broadcast Detection Function It can be represented as:

[0270] ,

[0271] in, This indicates a missed broadcast. This indicates a normal seeding process; An empty set means that no change signal was received; This is a set of candidate events that exist within the window.

[0272] This application also includes a detection system capable of implementing the above-mentioned peanut sowing missed detection method, the system comprising:

[0273] Theoretical beat acquisition module, used to acquire the theoretical beat signal of the seeding mechanism;

[0274] The actual seed passage detection module is used to collect the actual passing signals of seeds in the seed dispensing channel, seed guiding channel, or seed landing area;

[0275] The interference and operating condition acquisition module is used to acquire vibration interference signals and operating status signals;

[0276] The time mapping module is used to establish the time mapping relationship between theoretical seeding events and actual seeding events and to construct an effective detection window;

[0277] The signal processing and feature extraction module is used for input and frequency domain analysis, time-frequency analysis, and extraction of time position, peak value, pulse width, and frequency domain energy features.

[0278] The missed broadcast detection module is used to determine whether a missed broadcast has occurred based on window constraints and a comprehensive criterion function.

[0279] The above provides a detailed description of a method and system for detecting missed planting in peanut cultivation, as provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of these embodiments are merely for the purpose of helping to understand the method and core ideas of this invention. It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this invention. The above description of the disclosed embodiments enables those skilled in the art to implement or use this invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of this invention. Therefore, this invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting missed planting in peanut sowing, characterized in that, Includes the following steps: S1. Calculate the theoretical seeding cycle signal of the seeding mechanism to obtain the theoretical seeding cycle and theoretical seeding time; S2. Collect the actual passing signal of the seeds by a detection unit set in the seed dispensing channel or seed guiding channel; S3. Collect vibration interference signals during the seeding process; S4. Establish the time mapping relationship between theoretical seeding events and actual seeding events, and construct an effective detection window; S5. Set up a multi-channel structure based on the actual passing signal, perform frequency domain analysis and time domain analysis on the actual passing signal and interference signal, extract the missed seeding identification features, so as to realize real-time monitoring of the seeding status; S6. Integrate theoretical seeding events, actual seeding events, and identification features to construct a comprehensive criterion function for missed seeding and output the missed seeding judgment result.

2. The method for detecting missed planting in peanut sowing according to claim 1, characterized in that, In step S1, the seed metering disc of the seeding mechanism has Each seeding station has a real-time angular velocity of the seeding disc. The theoretical seeding cycle Calculate using the following formula: ， The kth theoretical seeding event time : ， in, This is the time of the (k-1)th theoretical seeding event.

3. The method for detecting missed planting in peanut sowing according to claim 1, characterized in that, In step S2, When the seed does not pass through the detection area, the light-receiving device of the detection unit receives the reference light intensity. ; When a seed passes through the detection area, the seed blocks, absorbs, or scatters the detection beam, causing the received light intensity to change as follows: ， in, The occlusion coefficient of the seed on the light beam; The light-receiving device converts the received light intensity into photocurrent: ， in, The photocurrent output by the light-receiving device; The photoelectric response sensitivity of the light-receiving device; And it is converted into output voltage by the signal conditioning circuit: ， in, To detect the output voltage of the unit; The equivalent gain of the conditioning circuit; This is the circuit bias voltage; With preset threshold When comparing, When a seed actually passes through, it is determined that a seed has passed through, thereby obtaining a signal indicating that the seed has actually passed through the seeding channel or seed guiding channel.

4. The method for detecting missed planting in peanut sowing according to claim 1, characterized in that, In step S3, vibration interference signals are acquired using an accelerometer. Let the output voltages of the accelerometer in the x, y, and z directions be respectively... , , The zero bias voltages of the accelerometer in the x, y, and z directions are respectively , , The sensitivity coefficients of the accelerometer in the x, y, and z directions are respectively , , Then the corresponding triaxial acceleration can be expressed as: ， ， ， in, The vibration acceleration is in the x-axis direction; The y-axis is the vibration acceleration. The vibration acceleration is in the z-axis direction; This is the real-time output voltage signal of the accelerometer in the x-axis measurement channel; This is the real-time output voltage signal of the accelerometer in the y-axis measurement channel; This is the real-time output voltage signal of the accelerometer in the z-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the x-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the y-axis measurement channel; This is the zero-acceleration bias voltage signal of the accelerometer in the z-axis measurement channel; This represents the sensitivity coefficient of the accelerometer in the x-axis measurement channel. This represents the sensitivity coefficient of the accelerometer in the y-axis measurement channel. This represents the sensitivity coefficient of the accelerometer in the z-axis measurement channel. The above three-axis accelerations are combined into a comprehensive vibration signal, which is the vibration interference signal: 。 5. The method for detecting missed planting in peanut sowing according to claim 1, characterized in that, In step S4, There is a propagation delay τ from the seeding position to the detection position. Construct an effective detection window corresponding to the k-th theoretical seeding event: ， in, For effective detection window; To allow for time deviation; Search for candidate events within the window: if there are no candidate events in the window, it is directly determined that the k-th theoretical seeding event has been missed; if there are candidate events in the window, further feature extraction and discrimination are performed on the candidate events.

6. The method for detecting missed planting in peanut sowing according to claim 1, characterized in that, The specific implementation process of step S5 is as follows: S5.1 Set up the actual multi-channel structure of the signal and extract the total energy characteristics, main characteristic value characteristics, energy concentration, and channel correlation characteristics; S5.2 Frequency domain analysis to obtain frequency domain energy characteristics; S5.3 Time domain analysis: extract time location features, peak features, and pulse width features.

7. The method for detecting missed planting in peanut sowing according to claim 6, characterized in that, In step S5.1, a signal segment is extracted within the valid detection window of the k-th theoretical priming event. ,in , This represents the number of sampling points within the window. An array of photoelectric detection units is arranged in the seed dispensing channel, seed guiding channel, or seed landing area. These array of photoelectric detection units form a multi-channel structure, and each channel synchronously collects the seed passing signal. ， Where m is the total number of detection channels in the multi-channel structure; n is the sampling point number of the time series signal of each channel; Arrange the signals from each channel in rows to construct a seed-through signal matrix: ， To eliminate DC bias and amplitude differences between channels, the signal matrix is ​​subjected to mean-reduction processing. If the zero-mean signal matrix is ​​a column vector of all 1s, then... : ， Among them, the mean vector for: ， Based on zero-mean signal matrix Construct the covariance matrix : ， Based on covariance matrix Extract total energy features, principal eigenvalue features, energy concentration, and channel correlation features: S5.1.A. Extracting Total Energy Characteristics: The expression for the total energy characteristic is: ， Among them, E k This represents the total energy of all channels; Total energy characteristics are used to characterize the overall fluctuation intensity of multi-channel signals within the current detection window; S5.1.B. Extracting principal feature values: Based on the covariance matrix, eigenvalue decomposition is performed, and the largest eigenvalue is extracted as the principal eigenvalue. Its expression is as follows: ， in, For feature vectors; For eigenvalues; The principal eigenvalues ​​are obtained using the above formula: ， The largest eigenvalue Characterizes the main energy direction; The principal eigenvalue is used to characterize the intensity of the dominant change pattern of the multi-channel signal within the current detection window; S5.1.C, Energy Concentration Extraction: Energy Concentration The calculation formula is: ， Energy concentration is used to characterize the degree of concentration of signal energy in the dominant feature direction within the current detection window; S5.1.D. Extracting channel correlation features: Choose the average of the off-diagonal elements of the covariance matrix As a channel correlation feature: ， Channel correlation features are used to characterize the consistency and degree of coordinated change in response among multiple detection channels.

8. The method for detecting missed planting in peanut sowing according to claim 6, characterized in that, In step S5.2, a signal segment is extracted from the window. Perform a Fast Fourier Transform: ， Establish the relationship between the discrete frequency point number d and the actual physical frequency: ， in, Where is the sampling frequency; L is the window length; This is the actual frequency corresponding to the d-th discrete frequency point; Obtain the power spectrum of the corresponding d-th discrete frequency point : ， Let the target frequency band be Then the frequency domain energy characteristics Defined as: ， Frequency domain energy characteristics reflect the degree of energy concentration of candidate events within the target frequency range, and are used to distinguish normal signals from vibration interference signals. In step S5.3, the specific feature extraction process is as follows: S5.3.A. Extraction of temporal and location features: By finding the sampling point location corresponding to the largest pulse within the window. : ， Obtain the actual time of passing : ， in: ΔT is the start time of the window; ΔT is the sampling period. S5.3.B. Peak Feature Extraction: Pulse peak within window Defined as: ； S5.3.C, Pulse Width Feature Extraction: Set threshold Find the sampling point where the signal first exceeds the threshold. and the last sampling point that exceeded the threshold Then the pulse is wide for: 。 9. The method for detecting missed planting in peanut sowing according to claim 1, characterized in that, In step S6, the time position, peak value, pulse width, frequency domain energy, and covariance feature deviation of the candidate over-seeding event corresponding to the k-th theoretical over-seeding event are extracted and compared with the normal single-seed over-seeding reference template to construct a normalized deviation: ， in, For time and location features; This represents the maximum allowable time deviation between the theoretical seeding time and the candidate seeding event time. Let be the absolute value of the time difference between the theoretical time of the k-th theoretical sorting event and the actual time of the corresponding candidate sorting event; ， in, Peak characteristics; This serves as the baseline value for peak characteristics; ， in, Pulse width characteristics; This serves as the baseline value for pulse width characteristics; ， in, Frequency domain energy characteristics; This serves as a benchmark value for the frequency domain energy characteristics; ， Wherein: d C This refers to multi-channel eigenvalue features; This serves as the baseline value for multi-channel feature values; Based on this, a comprehensive criterion function is constructed: ， in, , , , , , All are weighting coefficients; If there are no candidate events in the window, it is directly determined as a missed broadcast; if there are candidate events in the window but the comprehensive criterion function value is greater than the preset threshold, it is determined as a missed broadcast. When there are candidate events in the window and the value of the comprehensive criterion function is not greater than the threshold, it is determined to be a normal seeding event. Final Missed Broadcast Detection Function Represented as: ， in, An empty set means no response signal was received; There is a set of candidate events within the window; The time indicates a missed broadcast. The time indicates a normal process.

10. A detection system for implementing the peanut sowing missed detection method according to any one of claims 1-9, characterized in that, include: Theoretical beat acquisition module, used to acquire the theoretical beat signal of the seeding mechanism; The actual seed passage detection module is used to collect the actual passing signals of seeds in the seed dispensing channel, seed guiding channel, or seed landing area; The interference and operating condition acquisition module is used to acquire vibration interference signals and operating status signals; The time mapping module is used to establish the time mapping relationship between theoretical seeding events and actual seeding events and to construct an effective detection window; The signal processing and feature extraction module is used for input and frequency domain analysis, time-frequency analysis, and extraction of time position, peak value, pulse width, and frequency domain energy features. The missed broadcast detection module is used to determine whether a missed broadcast has occurred based on window constraints and a comprehensive criterion function.

Images