A weighing data processing method applied to a dynamic track scale of a pork slaughtering flow line

Abstract

This invention discloses a weighing data processing method for a dynamic rail scale applied to a pig slaughtering production line. The method includes: acquiring the original weight time-series data stream and visual data of pig swaying during carcass transport; constructing a weight data interference mask incorporating mechanical vibration and swaying interference through collaborative analysis; performing adaptive frequency domain notch filtering and time-series alignment correction on the original data stream based on this mask to obtain a denoised candidate weight data stream; calculating the transport process stability index by combining visual data, and performing sliding window clustering analysis on the candidate data stream to identify high-confidence weighing data envelopment intervals; and finally determining the carcass weighing data based on these intervals. This invention effectively improves the weighing accuracy and data stability of the dynamic rail scale in a high-speed production line environment by fusing visual and weight data to collaboratively suppress composite interference.

Description

Technical Field

[0001] This invention relates to the fields of automation in slaughtering and processing and weighing technology, and in particular to a weighing data processing method for a dynamic track scale applied to a pig slaughtering production line. Background Technology

[0002] In modern pig slaughtering production lines, dynamic rail scales are widely used for online weighing during carcass transport, and their data is directly linked to production statistics, settlement, and quality traceability. However, existing technology faces serious challenges in practical applications. Because pig carcasses are suspended on hooks and move at high speed along the conveyor chain, they inevitably experience violent shaking and swaying. Combined with the mechanical vibration of the conveyor track and fluctuations in chain tension, this results in severe signal distortion in the raw weight time-series data stream output by the dynamic rail scale. Specifically, this manifests as large data fluctuations, frequent jumps, and difficulty in distinguishing between gross weight and tare weight, leading to poor consistency between the data collected by the system and the data displayed on the rail scale instrument, causing statistical deviations and reconciliation disputes.

[0003] To address these issues, traditional methods typically rely on hardware upgrades (such as adding damping devices) or frequent manual parameter calibration. The former increases equipment costs and modification difficulties, while the latter is inefficient, slow, and difficult to adapt to the continuous and automated production rhythm of assembly lines. Although existing software filtering algorithms (such as mean filtering and Kalman filtering) can smooth data to some extent, they cannot effectively separate low-frequency, large-amplitude interference such as carcass swaying, resulting in significant deviations in weighing results. Therefore, there is an urgent need for a weighing data processing method that can integrate multi-dimensional data and adaptively suppress composite interference to significantly improve dynamic weighing accuracy and data stability without changing the existing hardware structure. Summary of the Invention

[0004] To address at least one of the aforementioned technical problems, this invention proposes a weighing data processing method for a dynamic track scale applied to a pig slaughtering production line.

[0005] The first aspect of this invention provides a weighing data processing method for a dynamic track scale applied to a pig slaughtering production line, comprising: The original weight time-series data stream and pig swaying visual data of the dynamic track scale during the transport of pig carcasses are acquired. The original weight time-series data stream and pig swaying visual data are analyzed collaboratively to construct a weight data interference mask, including a mechanical vibration interference mask and a swaying interference mask. Based on the weight data interference mask, the original weight time-series data stream is subjected to adaptive frequency domain notch filtering and time-series alignment correction to obtain the de-scrambled candidate weight data stream. Based on the candidate weight data stream and the shaking visual data, the stability index of the pig carcass transport process is calculated. Based on the stability index, a sliding window clustering analysis is performed on the candidate weight data stream to identify high-confidence weighing data envelopment intervals. The weight data of the pig carcass is determined based on the high-confidence weighing data envelopment interval.

[0006] In this solution, the acquisition of the original weight time-series data stream and visual data of pig swaying during the transport of pig carcasses using a dynamic track scale, and the collaborative analysis of the original weight time-series data stream and visual data of pig swaying to construct a weight data interference mask, including a mechanical vibration interference mask and a swaying interference mask, specifically: The original weight time-series data stream output by the dynamic track scale during the transportation of pig carcasses is obtained, and the visual image sequence including the hook and pig carcass during the transportation process is collected simultaneously as visual data of pig swaying. Perform time-domain difference operation on the original weight time-series data stream to extract the instantaneous fluctuation amplitude and fluctuation frequency of the weight data between adjacent sampling points, and identify periodic high-amplitude oscillation components based on the instantaneous fluctuation amplitude and fluctuation frequency; The periodic high-amplitude oscillation component is time-aligned with the original weight time-series data stream to determine the corresponding periodic weight change value of the periodic high-amplitude oscillation component in the original weight time-series data stream, and the periodic weight change value is used to construct a mechanical vibration interference mask. The visual data of the pig swaying is analyzed frame by frame to extract the centroid coordinates of the hook area and the carcass outline area. The displacement vector of the centroid coordinates between adjacent frames is calculated. Based on the displacement vector, the lateral sway amplitude and longitudinal sway period of the pig carcass during the transportation process are determined. The lateral sway amplitude and longitudinal sway period are mapped to low-frequency drift components in the original weight time-series data stream. The weight abrupt change intervals at the corresponding positions of the low-frequency drift components in the original weight time-series data stream are marked, and the weight abrupt change intervals are used to construct a swaying interference mask. A weight data interference mask is constructed from the mechanical vibration interference mask and the swaying interference mask to create a weight data interference mask for the original weight time-series data stream.

[0007] In this scheme, the step of performing adaptive frequency domain notch filtering and timing alignment correction on the original weight time-series data stream based on the weight data interference mask to obtain the dedisturbed candidate weight data stream is specifically as follows: The mechanical vibration interference mask is subjected to spectral scanning to extract the center frequency and frequency bandwidth values ​​corresponding to the periodic high-amplitude oscillation components in the mechanical vibration interference mask, and a digital band-stop filter with the center frequency and frequency bandwidth values ​​is constructed. The original weight time-series data stream is input into the digital band-stop filter for frequency domain notch filtering to filter out periodic high-frequency oscillation sampling points in the original weight time-series data stream that match the mechanical vibration interference mask, and outputs the weight time-series data stream after first-stage filtering. The swaying interference mask is analyzed by time interval analysis to identify the start and end time points of the marked weight change intervals in the swaying interference mask and the corresponding low-frequency drift amplitude values. Wavelet packet decomposition is performed on the weight time-series data stream after the first-level filtering to extract the low-frequency approximation coefficient components in the weight time-series data stream after the first-level filtering. Based on the start and end times of the weight mutation interval marked in the shaking interference mask, the corresponding abnormal coefficient segment in the low-frequency approximation coefficient component is located. The abnormal coefficient segment is reconstructed by linear interpolation using the data of the normal coefficient segments adjacent to the abnormal coefficient segment, and the low-frequency drift component in the low-frequency approximation coefficient component is suppressed to obtain the weight time-series data stream after secondary filtering. Extract the timestamp sequence of the original weight time-series data stream and the timestamp sequence of the weight time-series data stream after secondary filtering. Using the sampling points in the original weight time-series data stream that are not marked by the mask of weight data interference as the reference anchor points, perform interpolation processing on the weight time-series data stream after secondary filtering, and output the candidate weight data stream after descrambling.

[0008] In this scheme, the stability index of the pig carcass transport process is calculated based on the candidate weight data stream and the shaking visual data. Then, a sliding window clustering analysis is performed on the candidate weight data stream according to the stability index to identify high-confidence weighing data envelopment intervals. Specifically: The first derivative of the descrambled candidate weight data stream is calculated to obtain the weight change rate sequence of the candidate weight data stream at each sampling time. The centroid displacement of the carcass contour of each frame image in the shaking visual data of the pig is extracted, and a centroid displacement sequence aligned with the timestamp of the weight change rate sequence is constructed. The instantaneous stability value at each sampling time is obtained by weighted summation of the weight change rate sequence and the centroid displacement sequence at the same time. A fixed-time analysis window is set for the candidate weight data stream along the time axis. Within each analysis window, the mean and variance of the instantaneous stability value are calculated. The product of the mean and the reciprocal of the variance is then used to obtain the stability index for each analysis window. A stability index sequence is generated from the stability indices of all analysis windows. Using the stability index as a clustering feature, a density-based clustering algorithm is used to perform cluster analysis on all sampling points of the candidate weight data stream, and the sampling points are mapped to a feature space with the stability index as the coordinate axis. In the feature space, regions with sampling point density higher than the density threshold are identified. Sampling points belonging to the same high-density region and being temporally continuous are grouped into a cluster. The mean of the stability index corresponding to all sampling points in each cluster is calculated. Clusters with a mean stability index higher than the stability threshold are selected and mapped back to the original time interval in the candidate weight data stream. Extract the candidate weight data stream corresponding to the original time interval to obtain the high-confidence weighing data envelopment interval.

[0009] In this solution, determining the weight data of the pig carcass based on the high-confidence weighing data envelopment interval specifically involves: Extreme point distribution density analysis is performed on the candidate weight data stream within the high-confidence weighing data envelopment interval. Local maxima and local minima of the candidate weight data stream within the interval are extracted. The amplitude difference and time interval between adjacent extreme points are calculated, and an extreme point dynamic characteristic map is constructed. Obtain the swaying visual data of the pigs corresponding to the high-confidence weighing data envelopment interval, perform skeletal key point tracking on the swaying visual data, extract the rate of change of swaying angular momentum in the carcass trunk region, map the rate of change of swaying angular momentum to the extreme point dynamic feature map, and generate a spatiotemporal coupling feature matrix. Singular value decomposition is performed on the spatiotemporal coupling feature matrix to determine whether there are principal component vectors that satisfy the preset orthogonality condition. When there are principal component vectors that satisfy the preset orthogonality condition, the target extreme point cluster in the extreme point dynamic feature map corresponding to the principal component vector is determined. The candidate weight data stream amplitudes in the target extreme point cluster are weighted and averaged, and the weighted average value is determined as the weight data of the pig carcass. When there is no principal component vector that satisfies the preset orthogonality condition, the candidate weight data stream in the high-confidence weighing data envelopment interval is compensated and corrected, and the weighing data of the pig carcass is determined based on the corrected weight data stream.

[0010] In this scheme, when there is no principal component vector satisfying the preset orthogonality condition, the candidate weight data stream within the high-confidence weighing data envelopment interval is compensated and corrected, and the weighing data of the pig carcass is determined based on the corrected weight data stream. Specifically: When there is no principal component vector that satisfies the preset orthogonality condition, the start time and end time of the high-confidence weighing data envelopment interval are extracted, and the integral area and centroid time coordinate of the candidate weight data stream within the time interval of the start time and end time are calculated. The mechanical vibration frequency of the dynamic track scale is determined by the mechanical vibration interference mask, and the tension feedback data of the conveyor chain is obtained. Based on the mechanical vibration frequency and tension feedback data, an elastic deformation compensation model of the scale body is constructed. The integral area and the center of gravity time coordinate are input into the elastic deformation compensation model of the scale body to calculate the weight reading deviation caused by the mechanical hysteresis effect. Based on the weight reading deviation, the candidate weight data stream in the high confidence weighing data envelopment interval is compensated and corrected to obtain the corrected weight data stream. Gaussian kernel density estimation is performed on the corrected weight data stream to extract the amplitude interval with the highest probability density, and the median of the amplitude interval is determined as the weight data of the pig carcass.

[0011] This invention discloses a weighing data processing method for a dynamic rail scale applied to a pig slaughtering production line. The method includes: acquiring the original weight time-series data stream and visual data of pig swaying during carcass transport; constructing a weight data interference mask incorporating mechanical vibration and swaying interference through collaborative analysis; performing adaptive frequency domain notch filtering and time-series alignment correction on the original data stream based on this mask to obtain a denoised candidate weight data stream; calculating the transport process stability index by combining visual data, and performing sliding window clustering analysis on the candidate data stream to identify high-confidence weighing data envelopment intervals; and finally determining the carcass weighing data based on these intervals. This invention effectively improves the weighing accuracy and data stability of the dynamic rail scale in a high-speed production line environment by fusing visual and weight data to collaboratively suppress composite interference. Attached Figure Description

[0012] Figure 1 A flowchart of a weighing data processing method for a dynamic track scale applied to a pig slaughtering production line according to the present invention is shown; Figure 2 The flowchart illustrating the construction of the weight data interference mask according to the present invention is shown; Figure 3 The flowchart illustrating the present invention for descrambling the original weight time-series data stream is shown. Detailed Implementation

[0013] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0014] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0015] Figure 1 The flowchart illustrates a weighing data processing method for a dynamic track scale applied to a pig slaughtering production line according to the present invention.

[0016] like Figure 1 As shown, the first aspect of the present invention provides a weighing data processing method for a dynamic track scale applied to a pig slaughtering production line, comprising: The original weight time-series data stream and pig swaying visual data of the dynamic track scale during the transport of pig carcasses are acquired. The original weight time-series data stream and pig swaying visual data are analyzed collaboratively to construct a weight data interference mask, including a mechanical vibration interference mask and a swaying interference mask. Based on the weight data interference mask, the original weight time-series data stream is subjected to adaptive frequency domain notch filtering and time-series alignment correction to obtain the de-scrambled candidate weight data stream. Based on the candidate weight data stream and the shaking visual data, the stability index of the pig carcass transport process is calculated. Based on the stability index, a sliding window clustering analysis is performed on the candidate weight data stream to identify high-confidence weighing data envelopment intervals. The weight data of the pig carcass is determined based on the high-confidence weighing data envelopment interval.

[0017] It should be noted that by acquiring and collaboratively analyzing the original weight time-series data stream and visual data of pig swaying during the transport of pig carcasses using a dynamic track scale, a weight data interference mask was constructed, including a mechanical vibration interference mask and a swaying interference mask. This enabled the identification and modeling of the two main interference sources—high-frequency mechanical vibration and low-frequency carcass swaying—in both the time and frequency domains. Furthermore, based on this weight data interference mask, adaptive frequency-domain notch filtering and time-series alignment correction were applied to the original weight time-series data stream. This effectively filtered out high-frequency periodic noise caused by track mechanical resonance and suppressed low-frequency drift caused by large carcass swaying. Simultaneously, it corrected any time axis offset that might be introduced by the filtering process, resulting in a smooth waveform and precise time-series data. The system generates a precise candidate weight data stream. Based on this data stream and the visual data of shaking, a stability index for the transport process is calculated. This index guides a sliding window clustering analysis of the candidate weight data stream, intelligently identifying time periods of relatively stable carcass data with high reliability from complex dynamic signals. Invalid data from periods of violent shaking or transitional phases is successfully eliminated, accurately locating high-confidence weighing data envelope intervals. Finally, the weighing data of pig carcasses is determined based on these high-confidence weighing data envelope intervals, ensuring that the weighing samples used have extremely high stability and representativeness. This achieves near-static weighing accuracy in a high-speed production line environment, significantly improving the reliability and accuracy of dynamic weighing.

[0018] Figure 2 A flowchart illustrating the construction of the weight data interference mask according to the present invention is shown.

[0019] According to an embodiment of the present invention, the acquisition of the original weight time-series data stream and the visual data of pig swaying during the transport of pig carcasses using a dynamic track scale, and the collaborative analysis of the original weight time-series data stream and the visual data of pig swaying to construct a weight data interference mask, including a mechanical vibration interference mask and a swaying interference mask, specifically: The original weight time-series data stream output by the dynamic track scale during the transportation of pig carcasses is obtained, and the visual image sequence including the hook and pig carcass during the transportation process is collected simultaneously as visual data of pig swaying. Perform time-domain difference operation on the original weight time-series data stream to extract the instantaneous fluctuation amplitude and fluctuation frequency of the weight data between adjacent sampling points, and identify periodic high-amplitude oscillation components based on the instantaneous fluctuation amplitude and fluctuation frequency; The periodic high-amplitude oscillation component is time-aligned with the original weight time-series data stream to determine the corresponding periodic weight change value of the periodic high-amplitude oscillation component in the original weight time-series data stream, and the periodic weight change value is used to construct a mechanical vibration interference mask. The visual data of the pig swaying is analyzed frame by frame to extract the centroid coordinates of the hook area and the carcass outline area. The displacement vector of the centroid coordinates between adjacent frames is calculated. Based on the displacement vector, the lateral sway amplitude and longitudinal sway period of the pig carcass during the transportation process are determined. The lateral sway amplitude and longitudinal sway period are mapped to low-frequency drift components in the original weight time-series data stream. The weight abrupt change intervals at the corresponding positions of the low-frequency drift components in the original weight time-series data stream are marked, and the weight abrupt change intervals are used to construct a swaying interference mask. A weight data interference mask is constructed from the mechanical vibration interference mask and the swaying interference mask to create a weight data interference mask for the original weight time-series data stream.

[0020] It should be noted that during the weighing process of the dynamic track scale in the pig slaughtering line, because the carcass is in a continuous conveying state, the original weight time-series data stream output by the dynamic track scale not only contains the true weight information of the pig carcass, but also superimposed with dynamic interference signals generated by the vibration of the track mechanical structure and the swaying of the carcass itself. If the weighing value is directly extracted from the original weight time-series data stream, it is easy to cause abnormally amplified weight fluctuations, distortion of instantaneous peak values, and deviation of the weighing result from the true weight. Therefore, it is necessary to pre-determine a weight data interference mask to achieve explicit marking of the interference source, interference location, and interference intensity in the weight time-series data. By acquiring the raw weight time-series data stream output by the dynamic track scale and the synchronously collected visual data of pig swaying, time-domain difference operations are performed on the raw weight time-series data stream. By amplifying the change trend between adjacent sampling points, the instantaneous fluctuation amplitude and fluctuation frequency of the weight data in a short period of time are extracted. The instantaneous fluctuation amplitude is used to characterize the strength of weight change, and the fluctuation frequency is used to characterize the repetition period of weight fluctuation. Since mechanical vibration usually manifests as periodic weight fluctuations with stable frequency, repetition, and consistent amplitude variation, periodic high-amplitude oscillation components can be identified based on the instantaneous fluctuation amplitude and fluctuation frequency. Furthermore, by performing time-series alignment analysis with the raw weight time-series data stream, the specific location of mechanical vibration in the weight time-series data and the corresponding periodic weight change value are determined. The periodic weight change value is then used to construct a mechanical vibration interference mask to clearly mark the high-frequency sampling area affected by mechanical vibration. Meanwhile, frame-by-frame image analysis was performed on the visual data of pig swaying to extract the centroid coordinates of the hook area and the carcass contour area. The motion trajectory of the pig carcass during the conveying process was obtained by calculating the displacement vector between adjacent frames. The lateral sway amplitude was used to characterize the sway intensity of the carcass deviating from the conveying centerline, and the longitudinal sway period was used to characterize the temporal regularity of the sway. Since the swaying of the pig carcass causes changes in the direction of the weight force and forms a slow, undulating low-frequency drift in the weight time-series data stream, the lateral sway amplitude and the longitudinal sway period were mapped to the original weight time-series data stream to identify the low-frequency drift components caused by the sway. Furthermore, their corresponding weight abrupt change intervals were marked to form a swaying interference mask, thereby clearly identifying the low-frequency interference area affected by the carcass swaying.

[0021] Figure 3 The flowchart illustrating the present invention for descrambling the original weight time-series data stream is shown.

[0022] According to an embodiment of the present invention, the step of performing adaptive frequency domain notch filtering and timing alignment correction on the original weight time-series data stream based on the weight data interference mask to obtain the dedisturbed candidate weight data stream specifically includes: The mechanical vibration interference mask is subjected to spectral scanning to extract the center frequency and frequency bandwidth values ​​corresponding to the periodic high-amplitude oscillation components in the mechanical vibration interference mask, and a digital band-stop filter with the center frequency and frequency bandwidth values ​​is constructed. The original weight time-series data stream is input into the digital band-stop filter for frequency domain notch filtering to filter out periodic high-frequency oscillation sampling points in the original weight time-series data stream that match the mechanical vibration interference mask, and outputs the weight time-series data stream after first-stage filtering. The swaying interference mask is analyzed by time interval analysis to identify the start and end time points of the marked weight change intervals in the swaying interference mask and the corresponding low-frequency drift amplitude values. Wavelet packet decomposition is performed on the weight time-series data stream after the first-level filtering to extract the low-frequency approximation coefficient components in the weight time-series data stream after the first-level filtering. Based on the start and end times of the weight mutation interval marked in the shaking interference mask, the corresponding abnormal coefficient segment in the low-frequency approximation coefficient component is located. The abnormal coefficient segment is reconstructed by linear interpolation using the data of the normal coefficient segments adjacent to the abnormal coefficient segment, and the low-frequency drift component in the low-frequency approximation coefficient component is suppressed to obtain the weight time-series data stream after secondary filtering. Extract the timestamp sequence of the original weight time-series data stream and the timestamp sequence of the weight time-series data stream after secondary filtering. Using the sampling points in the original weight time-series data stream that are not marked by the mask of weight data interference as the reference anchor points, perform interpolation processing on the weight time-series data stream after secondary filtering, and output the candidate weight data stream after descrambling.

[0023] It should be noted that a spectral scan is performed on the mechanical vibration interference mask to extract the center frequency and bandwidth values ​​corresponding to the periodic high-amplitude oscillation components in the mask. The center frequency value is used to determine the concentrated area of ​​mechanical vibration energy, and the bandwidth value is used to determine the range of vibration influence. Based on this, a digital band-stop filter is constructed so that the filter parameters can adaptively match the mechanical vibration characteristics. Subsequently, the original weight time-series data stream is input into the digital band-stop filter for frequency domain notch filtering. The periodic high-frequency oscillation sampling points that match the mechanical vibration interference mask are filtered out in a targeted manner to reduce the high-frequency noise caused by the resonance of the conveyor track, chain vibration, and mechanical structure oscillation of the scale. The first-stage filtered weight time-series data stream is then output. Furthermore, the swaying interference mask is analyzed over time intervals to identify the start and end times of weight abrupt change intervals and the corresponding low-frequency drift amplitude values. The start and end times are used to determine the range of swaying interference, and the low-frequency drift amplitude values ​​characterize the impact of carcass swaying on the weight data. Subsequently, wavelet packet decomposition is performed on the first-stage filtered weight time-series data stream to extract low-frequency approximation coefficient components. Based on the weight abrupt change intervals marked in the swaying interference mask, the corresponding abnormal coefficient segments are located in the low-frequency approximation coefficient components. Linear interpolation reconstruction is performed using data from adjacent normal coefficient segments before and after the abnormal coefficient segments, replacing the abnormal trend affected by swaying with the normal trend, suppressing the low-frequency drift component, and obtaining the second-stage filtered weight time-series data stream. This achieves layered descrambling of high-frequency mechanical vibration interference and low-frequency carcass swaying interference, effectively reducing noise impact during dynamic weighing while preserving the true weight change trend, and improving the stability, timing accuracy, and reliability of the candidate weight data stream. The interpolation processing completes the timing break sampling points caused by the filtering operation.

[0024] According to an embodiment of the present invention, the step of calculating the stability index of the pig carcass transport process based on the candidate weight data stream and the shaking visual data, and performing sliding window clustering analysis on the candidate weight data stream according to the stability index to identify high-confidence weighing data envelopment intervals, specifically includes: The first derivative of the descrambled candidate weight data stream is calculated to obtain the weight change rate sequence of the candidate weight data stream at each sampling time. The centroid displacement of the carcass contour of each frame image in the shaking visual data of the pig is extracted, and a centroid displacement sequence aligned with the timestamp of the weight change rate sequence is constructed. The instantaneous stability value at each sampling time is obtained by weighted summation of the weight change rate sequence and the centroid displacement sequence at the same time. A fixed-time analysis window is set for the candidate weight data stream along the time axis. Within each analysis window, the mean and variance of the instantaneous stability value are calculated. The product of the mean and the reciprocal of the variance is then used to obtain the stability index for each analysis window. A stability index sequence is generated from the stability indices of all analysis windows. Using the stability index as a clustering feature, a density-based clustering algorithm is used to perform cluster analysis on all sampling points of the candidate weight data stream, and the sampling points are mapped to a feature space with the stability index as the coordinate axis. In the feature space, regions with sampling point density higher than the density threshold are identified. Sampling points belonging to the same high-density region and being temporally continuous are grouped into a cluster. The mean of the stability index corresponding to all sampling points in each cluster is calculated. Clusters with a mean stability index higher than the stability threshold are selected and mapped back to the original time interval in the candidate weight data stream. Extract the candidate weight data stream corresponding to the original time interval to obtain the high-confidence weighing data envelopment interval.

[0025] It should be noted that during the dynamic conveying and weighing of pig carcasses, even after removing mechanical vibration and swaying interference, the candidate weight data stream still contains locally unstable sections caused by hanging inertia, conveying speed fluctuations, local swaying of the carcass, and changes in force. If the weighing result is determined directly based on the entire candidate weight data stream, data in transitional or highly volatile phases may be included in the calculation, causing the final weight value to deviate from the true weight. Therefore, by coupling the weight change rate in the candidate weight data stream with the lateral displacement of the carcass contour centroid in the swaying visual data, a stability index reflecting the degree of weight stability and posture stability is constructed. The persistence characteristics of stable states are identified using a sliding window approach, and then a density-based clustering algorithm is used to cluster the stability index, identifying high-density regions with high stability, consistent fluctuations, and temporal continuity from continuous time periods, ultimately determining the high-confidence weighing data envelopment interval. The high-confidence weighing data envelopment interval refers to the weight data interval in the candidate weight data stream that has a low rate of weight change, a small carcass swing amplitude, a consistently high stability index, and is continuous in time. Essentially, it is an effective weighing window in which the pig carcass is relatively stable and the weight output has high reliability during dynamic weighing.

[0026] According to an embodiment of the present invention, determining the weight data of the pig carcass based on the high-confidence weighing data envelopment interval specifically involves: Extreme point distribution density analysis is performed on the candidate weight data stream within the high-confidence weighing data envelopment interval. Local maxima and local minima of the candidate weight data stream within the interval are extracted. The amplitude difference and time interval between adjacent extreme points are calculated, and an extreme point dynamic characteristic map is constructed. Obtain the swaying visual data of the pigs corresponding to the high-confidence weighing data envelopment interval, perform skeletal key point tracking on the swaying visual data, extract the rate of change of swaying angular momentum in the carcass trunk region, map the rate of change of swaying angular momentum to the extreme point dynamic feature map, and generate a spatiotemporal coupling feature matrix. Singular value decomposition is performed on the spatiotemporal coupling feature matrix to determine whether there are principal component vectors that satisfy the preset orthogonality condition. When there are principal component vectors that satisfy the preset orthogonality condition, the target extreme point cluster in the extreme point dynamic feature map corresponding to the principal component vector is determined. The candidate weight data stream amplitudes in the target extreme point cluster are weighted and averaged, and the weighted average value is determined as the weight data of the pig carcass. When there is no principal component vector that satisfies the preset orthogonality condition, the candidate weight data stream in the high-confidence weighing data envelopment interval is compensated and corrected, and the weighing data of the pig carcass is determined based on the corrected weight data stream.

[0027] It should be noted that in determining the weight data of pig carcasses based on the high-confidence weighing data envelope interval, residual fluctuations caused by suspension inertia, local swaying, and track micro-disturbances may still exist within the envelope interval. Therefore, the weight data within the interval cannot be directly averaged. First, extreme point distribution density analysis is performed on the candidate weight data stream within the high-confidence weighing data envelope interval to extract local maxima and local minima. Local maxima represent instantaneous high force, and local minima represent instantaneous low force. The amplitude difference between adjacent extreme points is used to characterize the intensity of weight fluctuations, and the time interval is used to characterize the duration of fluctuations. Based on this, a dynamic feature map of extreme points is constructed to describe the dynamic stability of weight fluctuations. Further, the corresponding visual data of pig swaying is obtained, and the rate of change of angular momentum of the carcass trunk region is extracted through skeletal key point tracking. The rate of change of angular momentum of the carcass reflects the speed of carcass posture change and the degree of inertial disturbance. The lower the rate of change of angular momentum, the closer the carcass is to a stable suspension state. Subsequently, the rate of change of angular momentum is mapped onto the dynamic characteristic map of extreme points to generate a spatiotemporal coupling characteristic matrix, enabling the correlation analysis between weight fluctuations and attitude changes. Singular value decomposition is then performed on the spatiotemporal coupling characteristic matrix, decomposing the coupling relationship into multiple principal component vectors. It is then determined whether any principal component vectors satisfy a preset orthogonality condition exist. The orthogonality condition is required because orthogonal principal components have low correlation with other disturbance components, indicating that they have largely escaped the influence of mechanical vibration, residual oscillation, and random noise, and can independently characterize the true stress state of the pig carcass. When a principal component vector satisfying the preset orthogonality condition exists, it indicates that a stable dominant weight pattern has formed within the current envelope interval. Based on this, the corresponding target extreme point cluster is determined, and the weight amplitude within the target extreme point cluster is weighted and averaged according to the degree of correlation to enhance the weight of stable data and weaken the influence of disturbance data. The weighted average value is then used as the pig carcass weighing data.

[0028] According to an embodiment of the present invention, when there is no principal component vector satisfying the preset orthogonality condition, the candidate weight data stream within the high-confidence weighing data envelopment interval is compensated and corrected, and the weighing data of the pig carcass is determined based on the corrected weight data stream, specifically as follows: When there is no principal component vector that satisfies the preset orthogonality condition, the start time and end time of the high-confidence weighing data envelopment interval are extracted, and the integral area and centroid time coordinate of the candidate weight data stream within the time interval of the start time and end time are calculated. The mechanical vibration frequency of the dynamic track scale is determined by the mechanical vibration interference mask, and the tension feedback data of the conveyor chain is obtained. Based on the mechanical vibration frequency and tension feedback data, an elastic deformation compensation model of the scale body is constructed. The integral area and the center of gravity time coordinate are input into the elastic deformation compensation model of the scale body to calculate the weight reading deviation caused by the mechanical hysteresis effect. Based on the weight reading deviation, the candidate weight data stream in the high confidence weighing data envelopment interval is compensated and corrected to obtain the corrected weight data stream. Gaussian kernel density estimation is performed on the corrected weight data stream to extract the amplitude interval with the highest probability density, and the median of the amplitude interval is determined as the weight data of the pig carcass.

[0029] It should be noted that when no principal component vectors satisfy the preset orthogonality condition exist, it indicates that the weight change pattern within the high-confidence weighing data envelopment interval is still subject to the coupling effects of multiple perturbation factors. That is, weight fluctuations and carcass swaying, track vibration, and the structural inertia of the scale have not yet formed mutually independent and stable dominant patterns. At this time, the principal components obtained by singular value decomposition still have strong correlations, indicating that there are still insufficiently decoupled mechanical hysteresis effects and dynamic transmission biases in the current weight data. Therefore, by extracting the start and end times of the high-confidence weighing data envelopment interval and calculating the integral area and centroid time coordinate of the candidate weight data stream within this time interval, the integral area is used to characterize the overall cumulative effect of weight action within the envelope interval, which can reflect the overall force level of the pig carcass on the scale. The centroid time coordinate is used to describe the concentrated position of weight action on the time axis, to characterize the direction of dynamic force offset and the degree of weight response hysteresis. Furthermore, the mechanical structure vibration frequency of the dynamic rail scale is determined based on the mechanical vibration interference mask, and the tension feedback data of the conveyor chain is obtained. The mechanical structure vibration frequency characterizes the periodic vibration intensity and inertial response characteristics of the scale body during operation, while the tension feedback data reflects the changes in the traction load applied by the conveyor chain to the suspended body, which alters the force transmission path and local elastic deformation of the scale body. The reason for constructing a compensation model based on both the mechanical structure vibration frequency and the tension feedback data is that the weight error of the dynamic rail scale does not solely originate from the vibration frequency itself. High-frequency mechanical vibration causes periodic deformation of the scale body's elastic structure, while changes in chain tension alter the stress state and structural rebound process of the scale body. The combined effect of these two factors creates a mechanical hysteresis effect, resulting in a time delay and amplitude deviation between the actual weight and the output weight. Based on this, when constructing the elastic deformation compensation model of the weighing platform, the vibration frequency of the mechanical structure is used as the excitation parameter, and the tension feedback data is used as the load constraint parameter. A dynamic mapping relationship between the elastic deformation of the weighing platform and the time response is established. Combining the actual stress state represented by the integral area and the time coordinate of the center of gravity, the weight reading deviation caused by the mechanical hysteresis effect is calculated. Then, based on this weight reading deviation, the candidate weight data stream in the high-confidence weighing data envelopment interval is compensated and corrected to obtain the corrected weight data stream. Finally, Gaussian kernel density estimation is performed on the corrected weight data stream to construct a probability distribution model of the weight amplitude. The amplitude interval with the highest probability density is extracted. The region with the highest probability density represents the weight range with the most stable weight data, the highest frequency of occurrence, and the least impact from random disturbances. The median of this amplitude interval is determined as the weighing data of the pig carcass, thereby improving the stability and accuracy of the dynamic weighing results.

[0030] This invention discloses a weighing data processing method for a dynamic rail scale applied to a pig slaughtering production line. The method includes: acquiring the original weight time-series data stream and visual data of pig swaying during carcass transport; constructing a weight data interference mask incorporating mechanical vibration and swaying interference through collaborative analysis; performing adaptive frequency domain notch filtering and time-series alignment correction on the original data stream based on this mask to obtain a denoised candidate weight data stream; calculating the transport process stability index by combining visual data, and performing sliding window clustering analysis on the candidate data stream to identify high-confidence weighing data envelopment intervals; and finally determining the carcass weighing data based on these intervals. This invention effectively improves the weighing accuracy and data stability of the dynamic rail scale in a high-speed production line environment by fusing visual and weight data to collaboratively suppress composite interference.

[0031] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0032] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0033] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for processing weighing data using a dynamic track scale applied to a pig slaughtering production line, characterized in that, Includes the following steps: The original weight time-series data stream and pig swaying visual data of the dynamic track scale during the transport of pig carcasses are acquired. The original weight time-series data stream and pig swaying visual data are analyzed collaboratively to construct a weight data interference mask, including a mechanical vibration interference mask and a swaying interference mask. Based on the weight data interference mask, the original weight time-series data stream is subjected to adaptive frequency domain notch filtering and time-series alignment correction to obtain the de-scrambled candidate weight data stream. Based on the candidate weight data stream and the shaking visual data, the stability index of the pig carcass transport process is calculated. Based on the stability index, a sliding window clustering analysis is performed on the candidate weight data stream to identify high-confidence weighing data envelopment intervals. The weight data of the pig carcass is determined based on the high-confidence weighing data envelopment interval.

2. The weighing data processing method for a dynamic track scale applied to a pig slaughtering production line according to claim 1, characterized in that, The process involves acquiring the original weight time-series data stream and visual data of pig swaying during the transport of pig carcasses using a dynamic track scale. A collaborative analysis of the original weight time-series data stream and the visual data of pig swaying is then performed to construct a weight data interference mask, including a mechanical vibration interference mask and a swaying interference mask. Specifically: The original weight time-series data stream output by the dynamic track scale during the transportation of pig carcasses is obtained, and the visual image sequence including the hook and pig carcass during the transportation process is collected simultaneously as visual data of pig swaying. Perform time-domain difference operation on the original weight time-series data stream to extract the instantaneous fluctuation amplitude and fluctuation frequency of the weight data between adjacent sampling points, and identify periodic high-amplitude oscillation components based on the instantaneous fluctuation amplitude and fluctuation frequency; The periodic high-amplitude oscillation component is time-aligned with the original weight time-series data stream to determine the corresponding periodic weight change value of the periodic high-amplitude oscillation component in the original weight time-series data stream, and the periodic weight change value is used to construct a mechanical vibration interference mask. The visual data of the pig swaying is analyzed frame by frame to extract the centroid coordinates of the hook area and the carcass outline area. The displacement vector of the centroid coordinates between adjacent frames is calculated. Based on the displacement vector, the lateral sway amplitude and longitudinal sway period of the pig carcass during the transportation process are determined. The lateral sway amplitude and longitudinal sway period are mapped to low-frequency drift components in the original weight time-series data stream. The weight abrupt change intervals at the corresponding positions of the low-frequency drift components in the original weight time-series data stream are marked, and the weight abrupt change intervals are used to construct a swaying interference mask. A weight data interference mask is constructed from the mechanical vibration interference mask and the swaying interference mask to create a weight data interference mask for the original weight time-series data stream.

3. The weighing data processing method for a dynamic track scale applied to a pig slaughtering production line according to claim 1, characterized in that, The step of performing adaptive frequency domain notch filtering and timing alignment correction on the original weight time-series data stream based on the weight data interference mask to obtain the scrambled candidate weight data stream is as follows: The mechanical vibration interference mask is subjected to spectral scanning to extract the center frequency and frequency bandwidth values ​​corresponding to the periodic high-amplitude oscillation components in the mechanical vibration interference mask, and a digital band-stop filter with the center frequency and frequency bandwidth values ​​is constructed. The original weight time-series data stream is input into the digital band-stop filter for frequency domain notch filtering to filter out periodic high-frequency oscillation sampling points in the original weight time-series data stream that match the mechanical vibration interference mask, and outputs the weight time-series data stream after first-stage filtering. The swaying interference mask is analyzed by time interval analysis to identify the start and end time points of the marked weight change intervals in the swaying interference mask and the corresponding low-frequency drift amplitude values. Wavelet packet decomposition is performed on the weight time-series data stream after the first-level filtering to extract the low-frequency approximation coefficient components in the weight time-series data stream after the first-level filtering. Based on the start and end times of the weight mutation interval marked in the shaking interference mask, the corresponding abnormal coefficient segment in the low-frequency approximation coefficient component is located. The abnormal coefficient segment is reconstructed by linear interpolation using the data of the normal coefficient segments adjacent to the abnormal coefficient segment, and the low-frequency drift component in the low-frequency approximation coefficient component is suppressed to obtain the weight time-series data stream after secondary filtering. Extract the timestamp sequence of the original weight time-series data stream and the timestamp sequence of the weight time-series data stream after secondary filtering. Using the sampling points in the original weight time-series data stream that are not marked by the mask of weight data interference as the reference anchor points, perform interpolation processing on the weight time-series data stream after secondary filtering, and output the candidate weight data stream after descrambling.

4. The weighing data processing method for a dynamic track scale applied to a pig slaughtering production line according to claim 1, characterized in that, The process involves calculating a stability index for the pig carcass transport process based on the candidate weight data stream and the shaking visual data, and then performing sliding window clustering analysis on the candidate weight data stream according to the stability index to identify high-confidence weighing data envelopment intervals. Specifically: The first derivative of the descrambled candidate weight data stream is calculated to obtain the weight change rate sequence of the candidate weight data stream at each sampling time. The centroid displacement of the carcass contour of each frame image in the shaking visual data of the pig is extracted, and a centroid displacement sequence aligned with the timestamp of the weight change rate sequence is constructed. The weight change rate sequence and the centroid displacement sequence are weighted and summed at the same time to obtain the instantaneous stability value at each sampling time. A fixed-time analysis window is set for the candidate weight data stream along the time axis. Within each analysis window, the mean and variance of the instantaneous stability value are calculated. The product of the mean and the reciprocal of the variance is then used to obtain the stability index for each analysis window. A stability index sequence is generated from the stability indices of all analysis windows. Using the stability index as a clustering feature, a density-based clustering algorithm is used to perform cluster analysis on all sampling points of the candidate weight data stream, and the sampling points are mapped to a feature space with the stability index as the coordinate axis. In the feature space, regions with sampling point density higher than the density threshold are identified. Sampling points belonging to the same high-density region and being temporally continuous are grouped into a cluster. The mean of the stability index corresponding to all sampling points in each cluster is calculated. Clusters with a mean stability index higher than the stability threshold are selected and mapped back to the original time interval in the candidate weight data stream. Extract the candidate weight data stream corresponding to the original time interval to obtain the high-confidence weighing data envelopment interval.

5. The weighing data processing method for a dynamic track scale applied to a pig slaughtering production line according to claim 1, characterized in that, The determination of the weight data of the pig carcass based on the high-confidence weighing data envelopment interval specifically involves: Extreme point distribution density analysis is performed on the candidate weight data stream within the high-confidence weighing data envelopment interval. Local maxima and local minima of the candidate weight data stream within the interval are extracted. The amplitude difference and time interval between adjacent extreme points are calculated, and an extreme point dynamic characteristic map is constructed. Obtain the swaying visual data of the pigs corresponding to the high-confidence weighing data envelopment interval, perform skeletal key point tracking on the swaying visual data, extract the rate of change of swaying angular momentum in the carcass trunk region, map the rate of change of swaying angular momentum to the extreme point dynamic feature map, and generate a spatiotemporal coupling feature matrix. Singular value decomposition is performed on the spatiotemporal coupling feature matrix to determine whether there are principal component vectors that satisfy the preset orthogonality condition. When there are principal component vectors that satisfy the preset orthogonality condition, the target extreme point cluster in the extreme point dynamic feature map corresponding to the principal component vector is determined. The candidate weight data stream amplitudes in the target extreme point cluster are weighted and averaged, and the weighted average value is determined as the weight data of the pig carcass. When there is no principal component vector that satisfies the preset orthogonality condition, the candidate weight data stream in the high-confidence weighing data envelopment interval is compensated and corrected, and the weighing data of the pig carcass is determined based on the corrected weight data stream.

6. The weighing data processing method for a dynamic track scale applied to a pig slaughtering production line according to claim 5, characterized in that, When no principal component vector satisfies the preset orthogonality condition, the candidate weight data stream within the high-confidence weighing data envelopment interval is compensated and corrected, and the weighing data of the pig carcass is determined based on the corrected weight data stream, specifically as follows: When there is no principal component vector that satisfies the preset orthogonality condition, the start time and end time of the high-confidence weighing data envelopment interval are extracted, and the integral area and centroid time coordinate of the candidate weight data stream within the time interval of the start time and end time are calculated. The mechanical vibration frequency of the dynamic track scale is determined by the mechanical vibration interference mask, and the tension feedback data of the conveyor chain is obtained. Based on the mechanical vibration frequency and tension feedback data, an elastic deformation compensation model of the scale body is constructed. The integral area and the center of gravity time coordinate are input into the elastic deformation compensation model of the scale body to calculate the weight reading deviation caused by the mechanical hysteresis effect. Based on the weight reading deviation, the candidate weight data stream in the high confidence weighing data envelopment interval is compensated and corrected to obtain the corrected weight data stream. Gaussian kernel density estimation is performed on the corrected weight data stream to extract the amplitude interval with the highest probability density, and the median of the amplitude interval is determined as the weight data of the pig carcass.

Images