A method and system for preventing fraud in unattended weighing and counting of pigs at slaughter, based on dual-visual cross-verification.
By analyzing the trajectories of live pigs and the sensor data of the weighing platform, a common trampling identifier is generated, which solves the data contradiction caused by the simultaneous return of live pigs and trampling by personnel, ensuring the accuracy of the weighing system and the continuity of the process, and avoiding misjudgment and fraud.
Patent Information
- Application Number
- CN202611346705.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-01
- Publication Date
- 2026-11-17
Smart Images

Figure CN123111438A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of pig weighing and counting technology, and in particular to a method and system for preventing fraud in unattended weighing and counting of pigs at slaughter, involving dual-visual cross-verification. Background Technology
[0002] Current conventional methods for counting pigs at slaughter typically involve a vision device recording the number of pigs and a weighing platform recording their weight. These devices operate independently, transmitting data to a backend system. The backend system simply compares the number of pigs with their weight, assuming that pigs will remain in the center of the weighing platform until weighing is complete, and that any increase in weight is solely due to the pig's own weight. However, in actual loading scenarios, there are overlapping interferences caused by pigs being startled and turning back, and by personnel unintentionally stepping on sensors. In such situations, when a pig reaches the middle of the weighing platform, it may be startled by external noise and turn back. The vision device at the gate side records this as the pig leaving the weighing platform and prepares to reduce the number of pigs in the count.
[0003] Meanwhile, the transport driver stepped on the sensor at the edge of the weighing platform to observe the situation. The pressure sensed by the weighing platform did not decrease as the pigs exited; instead, the weight reading jumped due to the driver's step. The conventional back-end system received a signal from the gate-side visual acquisition device indicating a decrease in the number of pigs, while simultaneously receiving a signal from the weighing platform indicating no weight reduction or even a sudden weight change. The conventional back-end system could not distinguish whether this data discrepancy stemmed from deliberate manipulation by adding foreign objects or from other heavy objects falling as the pigs ran out. Therefore, the conventional back-end system directly determined that fraud had occurred, triggered an alarm, and locked the exit process.
[0004] The aforementioned misjudgments forced the interruption of the normal loading and unloading process, causing pigs to be crushed in the aisle, transport drivers to be wrongly accused, and disrupting normal on-site operations. Conventional systems lack time correlation comparisons between turning-back actions and personnel trampling actions, lack distinction between the boundary of the force-bearing location and the normal standing area, and lack verification of the interference source, thus failing to accurately identify the aforementioned superimposed interference and causing the unloading process to be disrupted. Summary of the Invention
[0005] In view of the aforementioned problems, this application is hereby filed.
[0006] Therefore, this application provides a method and system for preventing fraud by unattended weighing and counting of pigs at slaughter, which can solve the problem of data discrepancy between the decrease in the number of pigs and the failure to decrease the weight when pigs turn back and personnel accidentally step on the weighing platform sensors at the same time, and avoid the problem of the system misjudging fraud and interrupting the slaughter process.
[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: In the first aspect, this application provides a method for preventing fraud in unattended weighing and counting of pigs at slaughter by dual visual cross-verification, including: acquiring data on the direction of the pig's trajectory through the gate line recorded by the visual acquisition device on the gate line side and the first force distribution data of multiple sensors at the bottom of the weighing platform; Based on the data of the direction of the pig's trajectory through the gate and the first force distribution data, when the pig's turning action and the human trampling action occur in the same time window, the off-center load force waveform and the time period of weight contradiction are extracted and combined into a trampling feature pair. The force position of the off-center load force waveform is compared with the pre-recorded edge force amplification feature to generate a homogeneous trampling identifier. Based on the same-source trampling mark, the force center position data is determined based on the force ratio relationship of each sensor at the bottom of the weighing platform. The force center position data is compared with the area boundary in the preset spatial mapping relationship to generate edge trampling mark, overlapping interference mark, or position uncertain mark. In response to the generation of the edge stepping mark or the position uncertainty mark, a process pause command and a back prompt command are generated, and the execution feedback signal is verified, and an on-site review command is generated. According to the on-site verification instruction, the second force distribution data is collected and compared with the trampling characteristics to generate an interference removal mark or an interference not removed mark.
[0008] Preferably, the extracted eccentric load waveform and the time period of weight inconsistency are combined into a trampling feature pair, and the stress position of the eccentric load waveform is compared with the edge stress increase feature to generate a homologous trampling identifier, including: By comparing the current total weight change sequence on the weighing platform with the smooth decrease waveform when the pigs leave normally, the abrupt change period where the weight increases instead of decreasing is identified as the period of weight inconsistency. By comparing the force distribution waveforms of each sensor with the central force waveform of a normal standing pig, the waveform in which the force distribution changes from being concentrated at the center to suddenly increasing at the edge is identified as the off-center load waveform. The combination of the eccentric load waveform and the weight contradiction period falling within the same time window is identified as the trampling characteristic pair; The stress location data of the off-center load waveform is compared with the pre-recorded edge stress amplification feature data. When the stress location ranges of the two are determined to correspond, a common-source trampling identifier is generated.
[0009] Preferably, the determination of the force center location data based on the force ratio relationship of each sensor at the bottom of the weighing platform includes: Obtain the current force values of the first, second, third, and fourth sensors at the bottom of the weighing platform; Based on the proportional relationship of the current force values of the first, second, third, and fourth sensors, determine the coordinate position of the current force concentration point on the weighing platform plane; The coordinate position is determined as the position data of the force center.
[0010] Preferably, the step of comparing the force center location data with the region boundary in a preset spatial mapping relationship to generate an edge trampling marker, an overlapping interference marker, or a position uncertainty marker includes: When the force center location data is determined to fall within the boundary of the first region but exceed the boundary of the second region, an edge trampling mark is generated. When the force center location data is determined to fall within the boundary of the second region, an overlap interference flag is generated; When the force center location data falls within the boundary between the first region boundary and the third region boundary, and the force ratio of the four sensors is within a preset critical range, a position uncertainty identifier is generated.
[0011] Preferably, the step of generating a process pause command and a rewind prompt command and verifying the execution feedback signal in response to generating the edge stepping mark or the position uncertainty mark includes: In response to the generation of the edge stepping mark or the position uncertainty mark, a process pause command is generated and sent to the unattended weighing management software, and a back prompt command is generated and sent to the voice intercom speaker. The process pause completion signal returned by the unattended weighing management software and the personnel position change signal returned by the voice intercom column are obtained as execution feedback signals. When the process pause completion signal and the personnel position change signal are received within the preset waiting time, the verification execution feedback signal is completed. If no personnel location change signal is received within the preset waiting time, the back prompt instruction is resent and a prompt execution error flag is generated. The personnel location change signal is then obtained again as an execution feedback signal to complete the verification.
[0012] Preferably, the step of collecting the second force distribution data according to the on-site verification instruction and comparing it with the trampling feature pair to generate an interference removal indicator or an interference not removed indicator includes: Collect the second force distribution data according to the on-site verification instructions; The second force distribution data from two consecutive detections are compared with the off-center load force waveform in the trampling feature pair; When it is confirmed that there is no off-center load waveform in the second force distribution data of two consecutive tests, and the total weight change time sequence returns to the normal state of decreasing as the pigs return, an interference cancellation flag is generated; If it is determined that there is still an off-center load waveform in the second force distribution data of any detection, or that the total weight change time sequence is still in the weight contradiction period, an interference not cleared flag is generated.
[0013] Preferably, comparing the second force distribution data detected in two consecutive tests with the off-center load force waveform in the trampling feature pair includes: The second force distribution data detected in the first test is compared with the off-center load waveform to obtain the first comparison result; The second force distribution data from the second detection is compared with the off-center load waveform to obtain the second comparison result; When the first comparison result indicates that there is no off-center load waveform and the second comparison result indicates that there is an off-center load waveform, a question mark is generated; When the first comparison result indicates the presence of an off-center load waveform and the second comparison result indicates the absence of an off-center load waveform, a question mark is generated. The suspicious identifier is sent to the central monitoring room backend, and the current process is kept paused to complete the comparison.
[0014] Preferably, the acquisition of the pig's trajectory direction data across the gate line recorded by the gate-side visual acquisition device and the first force distribution data from multiple sensors at the bottom of the weighing platform includes: Acquire truck approach status indicators and truck approach time data; Based on the approach time of the truck, the environmental vibration waveform within a preset time after the vehicle approaches is extracted as an environmental noise sample. The environmental noise sample is compared with the first force distribution data to determine the force frequency band corresponding to the continuous vibration component in the noise sample. The components of the force-bearing frequency band are removed from the first force distribution data to obtain a pure force waveform as the first force distribution data.
[0015] Preferably, the acquisition of the truck approach status identifier and the truck approach time data includes: Send a status query request to the proximity sensor switch in front of the weighbridge; When a response signal is received from the proximity sensor switch within a preset response time, the truck approach time data is obtained based on the response signal and a truck approach status identifier is generated. If no response signal is received from the proximity sensor switch within the preset response time, a truck location inquiry prompt is sent to the driver's mobile terminal applet. Obtain the confirmation button response signal returned by the driver's mobile terminal applet; The truck approach time data is determined based on the confirmation button response signal, and a truck approach status indicator is generated.
[0016] Secondly, this application also provides a dual-vision cross-verification unattended weighing and counting anti-fraud system for pig slaughter, including: a gate line data acquisition module, used to acquire the direction data of the pig's trajectory through the gate line recorded by the gate line side vision acquisition device and the first force distribution data of multiple sensors at the bottom of the weighing platform; The feature extraction module is used to determine, based on the data of the direction of the pig's trajectory through the gate and the first force distribution data, when the pig's turning action and the human trampling action occur in the same time window, extract the combination of the off-center load waveform and the time period of weight contradiction as trampling feature pair, and compare the force position of the off-center load waveform with the pre-recorded edge force amplification feature to generate a homogeneous trampling identifier. The position determination module is used to determine the force center position data based on the force ratio relationship of each sensor at the bottom of the weighing platform according to the same source trampling mark, compare the force center position data with the area boundary in the preset spatial mapping relationship, and generate edge trampling mark, overlapping interference mark or position uncertain mark. The interception module is used to respond to the generation of the edge stepping mark or the position uncertainty mark, generate a process pause command and a back prompt command, verify the execution feedback signal, and generate an on-site review command; The verification module is used to collect the second force distribution data according to the on-site verification instruction and compare it with the trampling feature pair to generate an interference removal mark or an interference not removal mark.
[0017] Implementing this application will have the following beneficial effects: This application provides a method and system for preventing fraud in unattended weighing and counting of pigs at slaughter, based on dual-vision cross-verification. The invention extracts the offset load waveform and the time period of weight discrepancy to form a trampling feature pair. It then compares the force position of the offset load waveform with pre-recorded edge force amplification features to accurately identify the superimposed interference state of pig turning actions and personnel trampling actions occurring within the same time window. By determining the force center location data and comparing it with the area boundary in a preset spatial mapping relationship, it can clearly distinguish whether personnel trampling is located at the edge of the weighing platform or overlaps with the pig's position, avoiding misjudging unintentional trampling as fraudulent behavior. By generating process pause commands and backtracking prompts and verifying execution feedback signals, and then collecting and comparing second force distribution data according to on-site verification commands, it can confirm whether the driver has left the edge of the weighing platform and whether the weight has returned to normal. This allows the process to be paused rather than directly locked when superimposed interference occurs, avoiding forced interruption of normal slaughtering and loading progress and the problem of pigs being squeezed in the aisle, ensuring the continuity and accuracy of slaughtering operations. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is an overall flowchart of a method for preventing fraud in unattended weighing and counting of live pigs at slaughter, involving dual-visual cross-verification. Figure 2 This is a schematic diagram of the data acquisition and noise removal process in the unattended weighing and counting anti-fraud method for pig slaughtering, which is a dual-visual cross-verification method for preventing fraud in pig slaughtering according to this application. Figure 3 This is a schematic diagram of the logical branches for the extraction of trampling features and homology comparison in a dual-visual cross-verification unattended weighing and counting anti-fraud method for pig slaughtering involved in this application; Figure 4 This is a schematic diagram of the area determination and identification generation process based on the force center position data of a dual-visual cross-verification unattended weighing and counting anti-fraud method for pig slaughtering involved in this application; Figure 5 This is a schematic diagram of the closed-loop verification logic of the process of issuing a pause command and the execution feedback signal in a method for preventing fraud by dual-visual cross-verification of unattended weighing and counting at the slaughter of pigs, which is involved in this application. Figure 6 This is a schematic diagram of the secondary data comparison and interference status confirmation mechanism in the on-site verification stage of a dual-visual cross-verification unattended weighing and counting anti-fraud method for pig slaughtering involved in this application. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0021] Example 1, referring to Figures 1-6 This is the first embodiment of the present application, which provides a method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual-visual cross-verification, including: This application provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to implement the unattended weighing and counting method for preventing fraud in pig slaughter by combining multiple embodiments. Figure 1A flowchart illustrating a method for preventing fraud in unattended weighing and counting of pigs at slaughter, involving dual-visual cross-verification, is shown, including: S1: Acquire the trajectory direction data of the pig passing through the gate line recorded by the vision acquisition device on the gate line side and the first force distribution data of multiple sensors at the bottom of the weighing platform; It should be noted that in the on-site environment of pigs being loaded onto trucks, pigs may turn back due to external noise, and transport drivers may move around near the weighing platform. Conventional single-vision acquisition devices or single-weighing platform sensors cannot accurately distinguish between data discrepancies caused by pigs turning back and people trampling on them. Therefore, it is necessary to simultaneously acquire data on the direction of the pig's trajectory across the gate line recorded by the vision acquisition device on the gate side and the first force distribution data from multiple sensors at the bottom of the weighing platform.
[0022] Specifically, the gate-side visual acquisition device is installed at the entrance of the weighing room and can capture the movement trajectory direction of pigs as they pass through the gate in real time. The gate-side visual acquisition device acquires continuous video frame images and compares the displacement direction of the pig outline pixels in adjacent frames. When the pig outline pixels move inward toward the weighing platform, it is recorded as an entry direction marker; when the pig outline pixels move outward toward the weighing platform, it is recorded as an exit direction marker. This direction marker data is output as the trajectory direction data of the pig as it passes through the gate.
[0023] Meanwhile, weighing sensors are typically installed at the four corners of the bottom of the weighing platform, designated as the first, second, third, and fourth sensors in a clockwise direction. These four sensors together form the force detection array of the weighing platform, used to collect the pressure values at various locations on the platform surface in real time. In this application, the force voltage signals collected by these four sensors at the same time are converted into digital weight values and combined in a fixed order from the first to the fourth sensor to form a data array containing four digital elements, thus constituting the first force distribution data. This data array not only contains the total weight information of the weighing platform surface but also the spatial distribution of the weight on the platform surface. By synchronously acquiring these two types of data and aligning them on the timestamp, a data foundation is provided for subsequent identification of the superimposed interference state when the pig's turning back action and the human trampling action occur within the same time window, thereby avoiding the risk of misjudgment caused by relying on only a single data source.
[0024] In some embodiments, acquiring the data on the direction of the pig's trajectory through the gate line recorded by the gate-side visual acquisition device and the first force distribution data from multiple sensors at the bottom of the weighing platform may further include the following steps: S1.1, Obtain truck approach status identifier and truck approach time data; S1.2, Based on the approach time of the truck, extract the environmental vibration waveform within a preset time after the vehicle approaches as an environmental noise sample. S1.3, compare the environmental noise sample with the first force distribution data to determine the force frequency band corresponding to the continuous vibration component in the noise sample; S1.4, Remove the component of the force-bearing frequency band from the first force distribution data to obtain a pure force waveform as the first force distribution data.
[0025] In this embodiment, during the weighing of live pigs, the truck reversing towards the weighbridge causes ground vibration. This vibration is transmitted through the weighbridge foundation to the sensors at the bottom of the weighing platform, resulting in the force data output by the sensors being affected by environmental vibration interference. If the force data with vibration interference is directly used for subsequent analysis, the truck's vibration may be misinterpreted as force changes caused by people stepping on it or the pigs turning back. Therefore, it is necessary to first identify the truck's approach status. The truck approach status indicator can be generated by a ground inductive loop or an infrared beam switch installed in front of the weighbridge. When the truck blocks the infrared light or presses against the ground inductive loop, the relay inside the inductive switch closes, generating a high-level signal, which generates the truck approach status indicator. The exact time of this occurrence is recorded as the truck approach time data.
[0026] Furthermore, as the truck approaches the weighbridge while its engine is off or idling, engine vibration and wheel-ground friction vibration are transmitted to the weighing platform sensors. Based on the truck's approach time data, a sequence of force fluctuation data collected by the four sensors at the bottom of the weighbridge within a preset time period (e.g., 10 seconds) after the vehicle approaches is extracted; this force fluctuation data sequence is the environmental vibration waveform. This environmental vibration waveform is extracted as an environmental noise sample, used to subsequently remove interference components from the force data.
[0027] Furthermore, the environmental noise sample contains continuous low-frequency vibration components generated by truck engines and mid-to-high-frequency vibration components generated by vehicle wheels. The environmental noise sample is compared point-by-point with the currently collected first force distribution data. Specifically, the absolute value of the difference between the force value at each moment in the environmental noise sample and the corresponding force value in the current force data is calculated. If the absolute value of the difference is less than a preset difference threshold, the force fluctuation at that moment is determined to originate from the environmental noise. This identifies the continuous vibration components in the current force data that match the characteristics of the environmental noise sample and determines the corresponding force frequency bands. It is understandable that the vibration generated by a truck engine typically exhibits relatively stable periodic fluctuations, while the force changes caused by trampling or pigs turning back typically exhibit instantaneous, non-periodic abrupt changes. Through comparison, it is possible to accurately identify which frequency bands in the force data belong to environmental vibration interference.
[0028] Specifically, after determining the force frequency bands corresponding to environmental vibration interference, components of these force frequency bands are removed from the first force distribution data. The specific technique for this removal is to replace the force values in the first force distribution data that match the environmental noise samples with the average force values of adjacent time points, thereby smoothing out these vibration interferences. The processed force waveform is used as the new clean force waveform, i.e., the updated first force distribution data, for subsequent extraction of off-center load waveforms and generation of homogeneous trampling identifiers. It should be noted that the above-mentioned process of removing environmental noise ensures that the subsequently extracted off-center load waveforms and weight discrepancies accurately reflect the force changes on the weighing platform, eliminating physical interference caused by the approach of trucks. For example, when a truck approaches the weighbridge causing continuous vibration, without noise removal, the sensor may continuously detect force fluctuations of approximately 3 kg, which may mask the instantaneous force changes caused by personnel trampling; however, after noise removal, the instantaneous force surge caused by personnel trampling can be clearly displayed, thus improving the accuracy of superimposed interference identification. The preset difference threshold is obtained by statistically analyzing the force amplitude of the weighing platform caused by the idling vibration of the truck, and its preferred setting range is 1 kg to 5 kg. In the specific operation of this application, the geological parameters of the current weighing room foundation are obtained. If the geological parameters are soft soil foundation, the preset difference threshold is set to 3 kg. If the geological parameters are hard soil foundation, the preset difference threshold is set to 2 kg.
[0029] In some embodiments, obtaining the truck approach status identifier and truck approach time data may further include the following steps: S1.11, Send a status query request to the proximity sensor switch in front of the weighbridge; S1.12, when a response signal is received from the proximity sensor switch within a preset response time, the truck approach time data is obtained based on the response signal and a truck approach status identifier is generated. S1.13, if no response signal is received from the proximity sensor switch within the preset response time, send a truck location inquiry prompt to the driver's mobile terminal applet; S1.14, Obtain the confirmation button response signal returned by the driver's mobile terminal applet; S1.15, determine the truck approach time data based on the confirmation button response signal and generate a truck approach status indicator.
[0030] In this embodiment, in actual unattended weighbridges, inductive loop detectors or infrared beam switches are typically installed at the entrance in front of the weighbridge as proximity sensors to detect approaching vehicles. To obtain the approach status of a truck, a status query request needs to be actively sent to the proximity sensor in front of the weighbridge. The query request is sent in the form of a high-level pulse signal through the I / O interface of the weighbridge controller to inquire about the current on / off status of the sensor switch.
[0031] Furthermore, upon receiving a query request, the proximity sensor switch will immediately return a trigger signal as a response signal if it detects a vehicle obstructing or pressing against the coil. If this response signal is received within a preset response time, it indicates that the truck has approached the weighbridge. At this point, the moment the response signal is received is recorded as the truck approach time data, and a high-level truck approach status indicator is generated simultaneously, indicating that the truck is in the approaching state.
[0032] On the other hand, if the proximity sensor switch in front of the weighbridge malfunctions, or if the switch is not triggered because the truck is parked outside the sensing area, no response signal will be received within the preset response time. In this case, relying solely on the sensor switch would prevent the truck's proximity from being recognized, and the subsequent noise reduction steps would not be triggered. To solve this problem, when it is determined that no response signal has been received, a truck location inquiry prompt is sent to the driver's WeChat mini-program via wireless network, prompting the driver to manually confirm whether they have arrived near the weighbridge.
[0033] Furthermore, after seeing the prompt on the mini-program interface, if the driver confirms arrival, they will tap the confirmation button on the screen. The driver's mobile mini-program converts the driver's tap into a confirmation button response signal and sends it back to the weighbridge controller via the network. It should be noted that the confirmation button response signal includes the timestamp of the driver's tap.
[0034] Specifically, after receiving the confirmation button response signal, the weighbridge controller extracts the timestamp as the truck's approach time data and generates a truck approach status indicator. This way, even if the proximity sensor switch fails, the truck's approach time data can still be obtained through manual confirmation by the driver, ensuring the smooth execution of the environmental noise removal process.
[0035] Furthermore, the aforementioned dual verification mechanism ensures the robustness of truck approach detection. In practical applications, depending on the hardware configuration of the weighbridge, a process of primary sensor switch detection and backup app verification can be selected, or both can be performed in parallel to improve response speed.
[0036] Optionally, the preset response time setting needs to balance response speed and fault tolerance time. If the time is too short, it may lead to misjudgment due to network latency; if the time is too long, it will affect the timeliness of noise removal. For example, the preset response time can be set to 3 seconds, which can tolerate normal network communication latency and can switch to the mini-program query mode in time when the proximity switch fails, ensuring that the entire weighing anti-fraud process is not interrupted. The preset response time is obtained based on the statistical analysis of the normal communication latency between the weighbridge controller and the proximity switch, and its preferred setting range is 1 to 5 seconds. In the specific operation of this application, the latency parameter of the current network environment is obtained. If the latency parameter is less than 50 milliseconds, the preset response time is set to 2 seconds; if the latency parameter is greater than or equal to 50 milliseconds, the preset response time is set to 4 seconds.
[0037] S2: Based on the data of the direction of the pig's trajectory through the gate and the first force distribution data, when the pig's turning action and the human trampling action occur in the same time window, extract the combination of the off-center load waveform and the time period of weight contradiction as a trampling feature pair, and compare the force position of the off-center load waveform with the pre-recorded edge force amplification feature to generate a homogeneous trampling identifier. It should be noted that when a pig is startled and turns back, the gate-side visual acquisition device will record the trajectory direction of the pig as it exits the weighing platform. Simultaneously, if the transport driver steps on the sensors at the edge of the weighing platform, the sensors at the bottom of the platform will feel additional pressure, causing the total weight to not only fail to decrease due to the pig's exit but may even fluctuate. Therefore, it is necessary to analyze both the trajectory direction data and the force distribution data simultaneously.
[0038] Specifically, by comparing the timestamps in the data on the direction of the pig's trajectory through the gate with the timestamps in the first force distribution data showing a sudden increase in force, the absolute value of the time difference between the two is calculated. If the absolute value of the time difference is less than a preset time window threshold, it is determined that the pig's turning back action and the person's trampling action occurred within the same time window. If the two times overlap, it indicates that superimposed interference has occurred. At this time, it is necessary to extract the waveform from the first force distribution data where the force distribution changes from a central concentration to a sudden increase at the edge as the off-center load force waveform, and extract the jump period where the weight does not decrease but increases from the total weight change time sequence as the weight contradiction period. Combining these two forms a trampling feature pair. In this embodiment, the force position of the extracted off-center load force waveform is compared with the pre-recorded edge force increase features. The pre-recorded edge force increase features refer to the distribution pattern array of the force increase of the four sensors when the person stands at different edge positions of the weighing platform in historical calibration. The specific operation of the same-source comparison is as follows: The absolute value of the difference between each item in the force increase ratio array of the current off-center load waveform and the pre-recorded force increase ratio array is calculated. If all absolute values of the difference are less than the preset error tolerance, it is determined to be of the same source. If the force position of the currently extracted off-center load waveform matches the pattern in the historical records, a same-source trampling indicator is generated, indicating that the current weight discrepancy is indeed caused by personnel trampling, rather than by other foreign objects falling. This step provides crucial evidence for subsequent precise location of the trampling position and differentiation between unintentional trampling and fraudulent behavior.
[0039] In this application, the step of extracting the contradictory time periods of the off-center load waveform and weight into a trampling feature pair, and comparing the stress position of the off-center load waveform with the edge stress amplification feature to generate a homologous trampling identifier, may also include the following steps: S2.1, compare the current total weight change sequence of the weighing platform with the smooth decline waveform when the normal pig leaves, and determine the jump period when the weight does not decrease but increases as the weight contradiction period; S2.2, compare the force distribution waveform of each sensor with the center force waveform when a normal pig is standing, and determine the waveform where the force distribution changes from being concentrated at the center to suddenly increasing at the edge as the off-center load force waveform; S2.3, the combination of the off-center load waveform and the weight contradiction period falling within the same time window is determined as the trampling characteristic pair; S2.4, compare the force position data of the off-center load waveform with the pre-recorded edge force amplification feature data, and when it is determined that the force position ranges of the two correspond, generate a homogeneous trampling identifier.
[0040] In this application, as the pigs leave the weighing platform normally, the total weight on the platform exhibits a smooth decreasing waveform as the pigs' weight gradually moves off. The current real-time data on the total weight change on the weighing platform is compared point-by-point with this smooth decreasing waveform. Specifically, the comparison method involves calculating the difference between the current weight change rate and the weight change rate during the normal decrease. If, during a certain period, the current weight change rate not only fails to decrease negatively but instead increases positively—that is, a numerical jump in weight occurs—and the jump exceeds a preset weight jump threshold, then this abnormal period is marked as a weight inconsistency period. Understandably, a jump in weight instead of decreasing usually means that additional weight is applied to the weighing platform while the pigs are leaving.
[0041] Furthermore, when a normal pig stands in the center of the weighing platform, the force on the four corner sensors is roughly uniform, exhibiting the waveform characteristics of central force. When the driver steps on the edge of the weighing platform, the force on the sensors near the stepping point increases sharply, while the force on other sensors remains relatively unchanged, causing the force distribution to shift from a central concentration to a sudden increase at the edges. By comparing the waveforms of the force distribution of each sensor with the waveform of the central force, the absolute value of the difference between the force value of the edge sensors and the average value of the central force is calculated. When the absolute value of the difference exceeds a preset off-center load threshold, this waveform of sudden increase at the edges is extracted as the off-center load force waveform. It should be noted that the extraction of the off-center load force waveform can accurately pinpoint the time and magnitude of the force position shift.
[0042] Specifically, when the pigs' return and withdrawal coincide with the driver's trampling, the periods of weight discrepancy and the off-center load waveform overlap on the time axis. These two features are extracted and combined into a trampling feature pair to characterize this specific superimposed interference state. Furthermore, to confirm whether this off-center load is caused by human trampling, pre-recorded edge force amplification feature data needs to be retrieved. The pre-recorded data is an array of force amplification distribution patterns collected during the calibration phase by having people of different weights stand at different edge positions on the weighing platform. The force location data of the current off-center load waveform is compared with historical data to calculate the similarity between the two.
[0043] For example, if the sensor location with the largest current force increase matches the increase pattern of a certain location in the historical record, and the force amplitude is within a reasonable range of a person's weight, then the force location ranges of the two are determined to correspond, generating a common-source trampling indicator. For instance, when the sensor detects a sudden increase of 15 kg in force on the first sensor and a sudden increase of 5 kg in force on the second sensor, while the forces on the third and fourth sensors remain unchanged, this increase ratio is highly consistent with the force distribution pattern when a person stands in the corner where the first sensor is located in the historical calibration, and can be determined as a common-source trampling. The pre-recorded edge force increase characteristic data is obtained by calibrating the device during the installation phase using standard weights to simulate human trampling. The preferred setting range is an annular area covering a width of 20 cm to 50 cm outside the edge of the weighing platform. In the specific operation of this application, the model parameters of the current weighing platform are obtained. If the platform size is less than 3 meters by 2 meters, the calibration area is set to a width of 20 cm outside the edge; if the platform size is greater than or equal to 3 meters by 2 meters, the calibration area is set to a width of 40 cm outside the edge.
[0044] S3: Based on the same-source trampling mark, determine the force center position data based on the force ratio relationship of each sensor at the bottom of the weighing platform, compare the force center position data with the area boundary in the preset spatial mapping relationship, and generate edge trampling mark, overlapping interference mark, or position uncertain mark. It should be noted that generating a common-source trampling indicator only confirms that the edge of the weighing platform has been disturbed by personnel trampling. However, determining the exact location where the person stepped, and whether that location is a high-risk area for fraud, requires further precise judgment. Specifically, the bottom of the weighing platform is usually equipped with four corner sensors. By analyzing the force ratio relationship of these four sensors, the coordinates of the current force concentration point, i.e., the force center location data, can be calculated. In this embodiment, the calculated force center location data is compared with the area boundaries in a pre-calibrated spatial mapping relationship. The spatial mapping relationship is a coordinate partitioning data table stored in the controller's memory. This table maps each set of coordinate ranges in the weighing platform's planar coordinate system to an area identifier, such as the normal standing area for pigs, the edge area of the weighing platform, etc. If the force center location falls within the edge area, it indicates that the person stepped on the edge of the weighing platform, which is likely unintentional. If the force center location falls within the normal standing area for pigs, there may be suspicion of intentional trampling of sensors for fraud. If the force center location is at the boundary of an area, there is uncertainty regarding its location. By comparing regional boundaries in this way, we can accurately distinguish different scenarios of trampling locations, thereby generating edge trampling markers, overlapping interference markers, or uncertain location markers, providing a basis for adopting different processing strategies in the future.
[0045] In this embodiment of the application, determining the force center location data based on the force ratio relationship of each sensor at the bottom of the weighing platform may further include the following steps: S3.1, Obtain the current force values of the first, second, third, and fourth sensors at the bottom of the weighing platform; S3.2, Based on the proportional relationship of the current force values of the first sensor, second sensor, third sensor, and fourth sensor, determine the coordinate position of the current force concentration point on the weighing platform plane; S3.3, the coordinate position is determined as the force center position data.
[0046] In this embodiment, a weighing sensor is typically installed at each of the four corners of the weighing platform, and these sensors are designated as the first sensor, second sensor, third sensor, and fourth sensor, respectively. The current force values of these four sensors are read in real time via a data acquisition bus, and these four force values collectively reflect the weight distribution on the weighing platform.
[0047] Furthermore, the principle of torque balance in physics and mechanics can be used to calculate the position of the force center based on the force values at the four corners. Consider the weighing platform as a rectangular plane, with four sensors located at the four vertices of the rectangle. Let the coordinates of the four sensors be the origin at the lower left corner, the horizontal coordinate of the lower right corner as the platform length, the vertical coordinate of the upper left corner as the platform width, and the horizontal and vertical coordinates of the upper right corner as the platform length and width, respectively. The lateral coordinate of the force center can be determined as follows: Multiply the force value of the first sensor by the horizontal coordinate of the origin, multiply the force value of the second sensor by the platform length, and add these two products to obtain the first lateral torque; multiply the force value of the third sensor by the horizontal coordinate of the origin, multiply the force value of the fourth sensor by the platform length, and add these two products to obtain the second lateral torque; add the first and second lateral torques to obtain the total lateral torque; add the force values of the four sensors to obtain the total force value; divide the total lateral torque by the total force value to obtain the lateral coordinate value of the force center point on the weighing platform plane. The longitudinal coordinate value is calculated in the same way by summing the products of the platform width and the longitudinal positions of each sensor, and dividing by the total force value. Through this arithmetic operation, the coordinate position of the current force concentration point on the weighing platform can be determined. This coordinate position intuitively reflects the concentration point of the weight on the platform.
[0048] Specifically, when the driver steps on the edge of the weighing platform, the sensor closest to the step experiences significantly greater force than other sensors, causing the calculated coordinate position to shift towards that sensor's location, thus falling within the edge area. When the pig stands in the center of the weighing platform, all four sensors experience uniform force, resulting in a calculated coordinate position closer to the platform's geometric center.
[0049] Furthermore, the calculated coordinates are directly output as the force center location data for subsequent area boundary comparison. For example, if the force value of the first sensor is 50 kg, the second sensor is 10 kg, the third sensor is 50 kg, and the fourth sensor is 10 kg, then the force center is clearly biased towards the side where the first and third sensors are located, and the calculated coordinates will fall on the edge of that side, thus providing data support for determining edge trampling.
[0050] In some embodiments, comparing the force center location data with the region boundary in a preset spatial mapping relationship to generate an edge trampling marker, an overlap interference marker, or a position uncertainty marker may further include the following steps: S3.4, when the force center position data falls within the boundary of the first region and exceeds the boundary of the second region, an edge trampling mark is generated; S3.5, when the force center location data is determined to fall within the boundary of the second region, an overlap interference identifier is generated; S3.6, when the force center location data falls within the boundary between the first region boundary and the third region boundary, and the force ratio of the four sensors is within a preset critical range, generate a position uncertainty identifier.
[0051] In some embodiments, a preset spatial mapping relationship divides the weighing platform plane into different regions. The boundary of the first region defines the entire platform surface, the boundary of the second region defines the core area where the pigs normally stand, and the boundary of the third region defines the transition zone between the edge region and the core region. When the calculated force center location data falls within the boundary of the first region but exceeds the boundary of the second region, it indicates that the force point is located at the edge of the weighing platform, and an edge stepping marker is generated. This indicates that a person has stepped on the edge of the weighing platform, which is likely an unintentional interference.
[0052] Furthermore, if the data on the location of the force center falls within the boundary of the second region, i.e., the core area where the pig is normally standing, it indicates that the force point overlaps with the pig's location. In this case, an overlap interference flag is generated, suggesting possible fraudulent activity such as people deliberately approaching the pig or stepping on the sensor, which requires close monitoring.
[0053] It should be noted that when the force center location data falls precisely within the boundary between the first and third regions, the force point is in the transition zone between the edge and the core, and its location is somewhat ambiguous. In this case, it is further determined whether the force distribution of the four sensors is within a preset critical range. The preset critical range refers to an intermediate state where the force distribution is neither completely biased towards the edge nor completely concentrated in the center. Specifically, this is determined by calculating the percentage of the force difference between adjacent edge sensors and the center sensor relative to the total weight; a percentage between 15% and 20% is considered an intermediate state. Specifically, if the force distribution meets this intermediate state, a position uncertainty flag is generated.
[0054] Furthermore, this three-way branching judgment logic can comprehensively cover various force position scenarios that may occur on the weighing platform. Optionally, the specific ranges of the first, second, and third region boundaries can be determined through on-site calibration and adjusted according to the size of the weighing platform and the size of the pigs. For example, if the weighing platform is 3 meters by 2 meters, the second region boundary can be set as a rectangular area of 1 meter by 0.5 meters with the center as the origin; the third region boundary can be set as a ring-shaped area extending outward by 0.2 meters. When the force center coordinates fall within 0.2 meters of the edge, it is determined to be edge trampling; when it falls within 0.5 meters of the center, it is determined to be overlapping interference; when it falls in between, it is determined to be positional uncertainty. The preset critical interval is obtained by statistically analyzing the fluctuation range of the force difference between the four corner sensors of the weighing platform in the transition area between the edge and the center. The preferred setting range is 10% to 20% of the total weight of the force difference between adjacent sensors. In the specific operation of this application, the average weight parameter of the current pig is obtained. If the average weight is less than 100 kg, the preset critical interval is set to 15% of the force difference. If the average weight is greater than or equal to 100 kg, the preset critical interval is set to 20% of the force difference.
[0055] S4: In response to generating the edge stepping mark or the position uncertainty mark, generate a process pause command and a back prompt command and verify the execution feedback signal, and generate an on-site review command; It should be noted that when an edge trampling indicator or an uncertain position indicator is generated, it indicates that there is superimposed interference from personnel trampling on the weighing platform. If normal weighing and counting continue, it will lead to data inconsistencies. Therefore, it is necessary to immediately pause the slaughtering process and prompt the driver to back up. Specifically, in response to the generation of the above indicator, the controller generates a process pause command and sends it to the unattended weighing management software, pausing the data upload and weighbridge generation for the current weighing. At the same time, a back-up prompt command is generated and sent to the voice intercom outside the weighbridge, broadcasting a voice prompt to the driver to move away from the edge of the weighing platform. In this application, after sending the command, it is also necessary to verify the execution feedback signal, that is, to confirm whether the management software has successfully paused and whether the driver has backed up. After verification, a field verification command with a timestamp and the current indicator status is generated and placed in the task queue for delayed execution, which is used to trigger the subsequent secondary confirmation process to check whether the interference has been truly eliminated. This step, through physical pause and voice prompt, can effectively prevent the generation of erroneous weighing data when the interference has not been eliminated, ensuring the rigor of the slaughtering process.
[0056] In this embodiment, the step of generating the edge stepping mark or the position uncertainty mark, generating the process pause command and the rewind prompt command, and verifying the execution feedback signal may further include the following steps: S4.1, in response to generating the edge stepping mark or the position uncertainty mark, a process pause command is generated and sent to the unattended weighing management software, and a back prompt command is generated and sent to the voice intercom speaker. S4.2, Obtain the process pause completion signal returned by the unattended weighing management software and the personnel position change signal returned by the voice intercom column as execution feedback signals; S4.3, when the process pause completion signal and the personnel position change signal are received within the preset waiting time, the verification execution feedback signal is completed; S4.4 If the personnel position change signal is not received within the preset waiting time, the back prompt instruction is resent and a prompt execution error flag is generated. The personnel position change signal is obtained again as an execution feedback signal to complete the verification.
[0057] In this embodiment, when the controller determines that the current center of force falls in the edge area or an area of uncertain location, it indicates that the driver may have unintentionally stepped on the weighing platform sensor. To prevent such unintentional stepping from causing data discrepancies and being misjudged as fraudulent behavior, immediate intervention is required. At this time, the weighbridge controller generates a process pause command and sends it to the unattended weighing management software via the internal bus. Upon receiving the command, the management software immediately suspends the current weighing and counting task, ceasing to read and upload weight and head count data to prevent erroneous data from entering the back-end settlement system.
[0058] Furthermore, a reverse prompt command is simultaneously generated and sent via audio cable to the voice intercom column outside the weighbridge. Upon receiving the command, the intercom column plays a pre-recorded voice prompt, such as "Weighing error, please move back to the safe area," reminding the driver that their current position is incorrect and they need to move away from the edge of the weighing platform. Understandably, this dual intervention of physical pause and voice prompts can promptly prevent the generation of erroneous data and guide the driver to correct their stance.
[0059] Specifically, after sending a command, the controller needs to confirm whether the command has been executed. It acquires the process pause completion signal returned by the unattended weighing management software; this signal is a status indicator generated by the management software after successfully suspending the task. Simultaneously, it acquires the personnel position change signal returned by the voice intercom column. The personnel position change signal is generated by analyzing whether the driver's outline pixels have left the edge area of the weighing platform using the vision acquisition device on the gate side. A personnel position change signal is generated when the overlap area between the driver's outline pixels and the edge area of the weighing platform in the video frame becomes zero. These two signals are used together as execution feedback signals to verify whether on-site intervention measures have been implemented effectively.
[0060] On the other hand, if the process pause completion signal and the personnel position change signal are successfully received within the preset waiting time, it indicates that the management software has paused, the driver has reversed, the verification execution feedback signal is complete, and on-site interference has been initially eliminated. It should be noted that if the driver does not hear the prompt or does not move for other reasons, the personnel position change signal will not be received within the preset waiting time. In this case, the prompt execution is determined to be abnormal, a prompt execution abnormality flag is generated, and the reverse prompt instruction is resent, with the volume increased or the prompt message changed and broadcast again. Furthermore, the personnel position change signal is acquired again as the execution feedback signal to complete the verification.
[0061] Specifically, the aforementioned verification and retry logic ensures that even if the driver does not respond promptly, multiple prompts can guide the driver to retreat, rather than directly judging it as fraud and locking the process. For example, the preset waiting time can be set to 8 seconds, which gives the driver enough time to react and move, while avoiding excessive waiting that could cause the process to stall. If no person is detected leaving within 8 seconds after the first prompt, the controller will broadcast the prompt again and record an execution anomaly. If no personnel position change signal is received three times consecutively, no retry will be made, and the process will directly switch to manual intervention alarm procedure to prevent a dead loop. Through this intervention mechanism with feedback verification and retry limit, false alarms caused by unintentional trampling by the driver are minimized, improving fault tolerance and on-site adaptability.
[0062] The preset waiting time is obtained by statistically analyzing the average time it takes for drivers to walk from the outer edge of the weighbridge to the safe area, with a preferred setting range of 5 to 15 seconds. In the specific operation of this application, the length parameter of the passage outside the weighbridge is obtained. If the passage length is less than 5 meters, the preset waiting time is set to 8 seconds. If the passage length is greater than or equal to 5 meters, the preset waiting time is set to 12 seconds.
[0063] As a preferred embodiment, the specific implementation of the verification execution feedback signal includes: A1. The process pause completion signal and the personnel position change signal are input into a preset feedback confirmation logic table for comparison. The feedback confirmation logic table is stored in the controller's memory and includes the mapping relationship between various signal state combinations and execution branches. Proceed to step A2.
[0064] A2. Determine whether both the process pause completion signal and the personnel position change signal are in a received state. If yes, it indicates that the on-site intervention measures have been implemented, and proceed to step A3; if no, it indicates that there are unimplemented intervention measures, and proceed to step A4. Here, "received state" means that the controller has captured the corresponding high-level transition signal within a preset waiting time.
[0065] A3, directly complete the verification execution feedback signal, generate a review trigger identifier, write the review trigger identifier into the on-site review instruction, and end the process.
[0066] Step A4 determines whether only the process pause completion signal was received without the personnel position change signal. If yes, it means the software suspension was successful but the driver did not move, and proceeds to step A5; if no, it means the software suspension failed, and proceeds to step A7. This step is used to distinguish between software communication failure and on-site personnel not following instructions, so as to take different retry strategies.
[0067] A5 generates an error message and accumulates an error count. It then determines whether the error count exceeds a preset retry limit. If yes, proceed to step A6; otherwise, proceed to step A8. The preset retry limit is obtained based on a comprehensive assessment of the weighbridge noise environment and the driver's reaction time, with a preferred range of 2 to 4 times. In the specific operation of this application, the current ambient noise level outside the weighbridge is obtained. If the ambient noise level is greater than 70 decibels, the preset retry limit is set to 3 times; if the ambient noise level is less than or equal to 70 decibels, the preset retry limit is set to 2 times.
[0068] A6: Stop resending the reversal prompt command, generate a manual intervention alarm flag and send it to the central monitoring room backend, maintain the current process paused state, and end the process. If the retry limit is reached and no personnel are detected leaving, it indicates that the driver may be trapped, there may be a conflict, or equipment malfunction, requiring manual on-site handling to avoid system deadlock.
[0069] A7 generates a software communication anomaly flag, resends the process pause command to the unattended weighing management software, and returns to step A2 to re-evaluate the signal status. The software communication anomaly may be caused by transient bus interference; resending can recover from most transient faults.
[0070] A8, resend the back prompt command to the voice intercom speaker, and increase the prompt volume by a preset decibel increment, then return to step A2 to re-evaluate the signal status. The preset decibel increment is obtained based on experimental data on the impact of environmental noise gradients on speech intelligibility, with a preferred setting range of 3 to 5 decibels. In the specific operation of this application, the current environmental noise decibel value is obtained. If the environmental noise decibel value is between 60 and 70 decibels, the preset decibel increment is set to 3 decibels; if the environmental noise decibel value is greater than 70 decibels, the preset decibel increment is set to 5 decibels.
[0071] S5: Collect the second force distribution data according to the on-site verification instruction and compare it with the trampling feature pair to generate an interference removal mark or an interference not removed mark.
[0072] It should be noted that after the process pause and reverse prompts are executed, the driver may have already left the edge of the weighing platform, or may still be standing in place because they did not hear the prompts. In this case, on-site verification is required to confirm whether the interference has truly been resolved. Specifically, according to the on-site verification instruction, after waiting for a short stabilization delay, the controller collects the force distribution data at the bottom of the weighing platform again to obtain the second force distribution data. The second force distribution data is compared with the previously extracted trampling features. In this embodiment, the specific comparison method is to check whether there is still an edge surge waveform in the second force distribution data that exceeds the preset off-center load threshold, and whether the total weight change sequence has returned to the normal decreasing waveform as the pigs return. If the comparison finds that the off-center load force waveform has disappeared and the total weight change has returned to normal, it indicates that the interference has been resolved, and an interference resolution indicator is generated; if the comparison finds that the off-center load force waveform still exists, it indicates that the driver has not left or there is still a heavy object, the interference has not been resolved, and an interference not resolved indicator is generated. Through this secondary verification, the on-site status can be accurately determined, providing a basis for whether to resume the process or upgrade the lock.
[0073] In some preferred embodiments, the step of comparing the second force distribution data collected according to the on-site verification instruction with the trampling feature pair to generate an interference removal indicator or an interference not removed indicator may further include the following steps: S5.1, Collect the second force distribution data according to the on-site verification instruction; S5.2, compare the second force distribution data detected in two consecutive tests with the off-center load force waveform in the trampling feature pair; S5.3, when it is determined that there is no off-center load waveform in the second force distribution data of two consecutive detections, and the total weight change time sequence is restored to the normal state of decreasing as the pigs return, an interference removal flag is generated; S5.4 If it is determined that there is still an off-center load waveform in the second force distribution data of any detection, or that the total weight change time sequence is still in the weight contradiction period, an interference not cleared flag is generated.
[0074] In some preferred embodiments, after the controller issues a reverse prompt and waits for a period of time, it is necessary to collect the force data at the bottom of the weighing platform again, i.e., the second force distribution data. The collection method is the same as that for the first force distribution data, using the four corner sensors to obtain the current force value in real time.
[0075] Furthermore, to avoid misjudgments caused by force fluctuations due to driver movement, a double-check method is used. After the first check, a second check is performed after a 2-second interval. The second force distribution data obtained from these two checks are compared with the off-center load force waveform of the pedal feature to check whether the waveform with sudden increases at the edge has disappeared.
[0076] Understandably, if the driver has left, the off-center load waveform should disappear, and the total weight should show a normal decreasing trend as the pigs return to their origin. It should be noted that if no off-center load waveform is detected in two consecutive tests (i.e., the increase in edge force is below the preset off-center load threshold), and the total weight change sequence returns to the normal state of decreasing as the pigs return, then the driver has left, the superimposed interference has been eliminated, and an interference cancellation flag is generated. The controller can then lift the process pause and resume normal slaughtering.
[0077] Furthermore, if the driver has not left or there is still a heavy object on the weighing platform, the second force distribution data will still show an off-center load waveform in any detection, or the total weight change sequence will still be in a period of weight inconsistency. In this case, it is determined that the interference has not been eliminated, and an interference not eliminated flag is generated. For example, if the off-center load waveform disappears in the first detection but reappears in the second detection, it is directly determined that the interference has not been eliminated, using a conservative strategy to avoid missed detection. Through this logic of two consecutive comparisons, misjudgments caused by transitional states such as single-leg exertion during the driver's departure process can be effectively eliminated, ensuring the accuracy of the verification results. The interval between the two detections is obtained based on the statistical analysis of the conventional time taken for a person to complete a center of gravity shift or leg lifting action, and its preferred setting range is 1 to 3 seconds. In the specific operation of this application, the current driver's action amplitude parameter is obtained. If the action amplitude is less than the preset amplitude, the interval is set to 2 seconds; if the action amplitude is greater than or equal to the preset amplitude, the interval is set to 3 seconds.
[0078] In this application, comparing the second force distribution data detected in two consecutive tests with the off-center load force waveform in the trampling feature pair may further include the following steps: S5.21, compare the second force distribution data detected in the first test with the off-center load waveform to obtain the first comparison result; S5.22, compare the second force distribution data detected in the second test with the off-center load waveform to obtain the second comparison result; S5.23, when it is determined that the first comparison result is that there is no off-center load waveform and the second comparison result is that there is an off-center load waveform, a doubtful identifier is generated; S5.24, when it is determined that the first comparison result is that there is an off-center load waveform and the second comparison result is that there is no off-center load waveform, a doubtful identifier is generated; S5.25, the suspicious identifier is sent to the central monitoring room backend and the current process is kept paused to complete the comparison.
[0079] In this application, during two consecutive detection processes, the second force distribution data from the first detection is matched with the off-center load waveform to determine whether the edge abrupt increase waveform has disappeared, thus obtaining the first comparison result. The first comparison result records whether off-center load interference still exists on the weighing platform at that moment.
[0080] Furthermore, a second test is conducted after a short interval. The second force distribution data is then compared with the off-center load waveform to determine if the waveform has disappeared, thus obtaining a second comparison result. This second comparison result reflects the changes in the disturbance state over a short period of time.
[0081] On the other hand, if the first comparison result shows no off-center load waveform, meaning the driver appears to have left, but the second comparison result shows an off-center load waveform, meaning the driver has stepped back, this contradictory situation indicates the driver is in a hesitant or exploratory state, and the interference is unstable. It should be noted that a doubt flag is generated at this point, indicating that it cannot be definitively determined whether the interference has been completely resolved.
[0082] Specifically, if the first comparison result shows the presence of an off-center load waveform, but the second comparison result shows its absence, it also indicates that the interference state is unstable, and a question mark is generated accordingly.
[0083] Furthermore, upon generating a suspicious flag, the controller does not automatically lift the process pause or directly determine it as fraud. Instead, it sends the flag to the centralized monitoring room, alerting remote supervisors for intervention. Supervisors can then manually review the footage from the site. Optionally, the controller maintains the current paused state, awaiting a manual review instruction to complete the comparison loop. For example, if the interval between two checks is 2 seconds, the driver might be moving their feet during those 2 seconds, causing the force application to be intermittent. Generating a suspicious flag and requesting manual intervention can effectively handle such boundary situations, preventing the controller from getting stuck in an infinite loop or making misjudgments.
[0084] Figure 2 This document illustrates a segment of the data acquisition and noise removal process in a method for preventing fraud during pig slaughter weighing. The visual acquisition device at the gate and the sensor array at the bottom of the weighing platform are activated synchronously to collect data on the trajectory direction of the pigs passing through the gate and the initial force distribution data on the platform. To eliminate ground vibration interference caused by trucks approaching the weighbridge, a status query request is sent to the proximity sensor switch. If a trigger signal is received within a preset response time, the truck's approach time is directly extracted and a status identifier is generated; if no response signal is received, a backup mechanism is triggered, sending a location query prompt to the driver's mobile terminal app. The time data is obtained based on the returned confirmation button response signal. Environmental vibration waveforms are extracted from this time data as noise samples. Continuous vibration components are removed through frequency band comparison, and a clean force waveform is output for subsequent feature extraction.
[0085] Figure 3The logical branches for extracting trampling feature pairs and comparing them with similar patterns are shown. The time sequence of the current total weight change on the weighing platform is compared with the smooth decrease waveform when a normal pig leaves, identifying the abrupt increase in weight as the period of weight inconsistency. Simultaneously, the force distribution waveforms of each sensor are analyzed, extracting the off-center load force waveform that changes from a concentrated center to a sudden increase at the edges. When these two types of features are within the same time window, they are combined into a trampling feature pair. Further, pre-calibrated edge force increase feature data are retrieved, and the similarity between the force location range of the current off-center load waveform and historical patterns is calculated. If the two correspond, it is confirmed that the weight inconsistency originates from human trampling rather than foreign object falling, generating a similar trampling identifier.
[0086] Figure 4 This demonstrates the process of region determination and marker generation based on force center location data. The current force values of the four corner sensors at the bottom of the weighing platform are acquired, and the coordinates of the force concentration point on the platform plane are calculated based on the torque balance principle. This coordinate position is compared with the region boundary in a preset spatial mapping relationship: if it falls within the core area where the pig is normally standing, an overlap interference marker is generated, indicating potential intentional fraud; if it falls within the edge area of the platform, an edge trampling marker is generated, indicating unintentional interference; if it is exactly within the boundary between the edge and the core, and the force ratio of the four corner sensors is within a preset critical range, an uncertain position marker is generated. Edge trampling and uncertain position markers will jointly trigger subsequent process intervention mechanisms.
[0087] Figure 5 The diagram illustrates the closed-loop verification logic for issuing and executing a process pause command and its feedback signal. In response to edge trampling or uncertain position markers, a process pause command is sent to the unattended weighing management software, and a backtracking prompt command is sent to the voice intercom. The system then verifies whether both the software's suspension completion signal and the personnel position change signal captured by the vision device are received simultaneously. If both are received normally, the verification is complete and a field review command is generated; if only the software pause signal is received without detecting personnel backtracking, an anomaly count is accumulated. When the anomaly count is within the limit, the prompt volume is increased and the backtracking command is resent; if the retry limit is reached, automatic intervention stops, a manual intervention alarm is generated and sent to the central monitoring room, maintaining the process pause state to avoid deadlock.
[0088] Figure 6This demonstrates the secondary data comparison and interference status confirmation mechanism during the on-site verification phase. Based on the on-site verification instruction, the second force distribution data at the bottom of the weighing platform is collected again after a short delay. A two-step detection method is used, and the data acquired in both steps are matched with the previously extracted off-center load waveform. If the first detection shows the off-center load waveform disappearing, but it reappears in the second detection, or vice versa, the two comparison results are considered contradictory, a doubtful flag is generated, and the system remains paused pending manual re-inspection. If no off-center load waveform is found in either of the two consecutive detections, and the total weight change sequence returns to the normal decreasing state as the pigs return, an interference clearance flag is generated to resume the slaughtering process; otherwise, an interference not clearance flag is generated, and the process is locked.
[0089] Example 2, this example also provides a dual-visual cross-verification unattended weighing and counting anti-fraud system for pig slaughter, including: a gate line data acquisition module, used to acquire the direction data of the pig's trajectory through the gate line recorded by the gate line side visual acquisition device and the first force distribution data of multiple sensors at the bottom of the weighing platform; The feature extraction module is used to determine, based on the data of the direction of the pig's trajectory through the gate and the first force distribution data, when the pig's turning action and the human trampling action occur in the same time window, extract the combination of the off-center load waveform and the time period of weight contradiction as trampling feature pair, and compare the force position of the off-center load waveform with the pre-recorded edge force amplification feature to generate a homogeneous trampling identifier. The position determination module is used to determine the force center position data based on the force ratio relationship of each sensor at the bottom of the weighing platform according to the same source trampling mark, compare the force center position data with the area boundary in the preset spatial mapping relationship, and generate edge trampling mark, overlapping interference mark or position uncertain mark. The interception module is used to respond to the generation of the edge stepping mark or the position uncertainty mark, generate a process pause command and a back prompt command, verify the execution feedback signal, and generate an on-site review command; The verification module is used to collect the second force distribution data according to the on-site verification instruction and compare it with the trampling feature pair to generate an interference removal mark or an interference not removal mark.
[0090] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.
[0091] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0092] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for preventing fraud in unattended weighing and counting of pigs at slaughter, characterized in that: include: Acquire data on the direction of the pig's trajectory through the gate line, recorded by the visual acquisition device on the gate line side, and the first force distribution data from multiple sensors at the bottom of the weighing platform; Based on the data of the direction of the pig's trajectory through the gate and the first force distribution data, when the pig's turning action and the human trampling action occur in the same time window, the off-center load force waveform and the time period of weight contradiction are extracted and combined into a trampling feature pair. The force position of the off-center load force waveform is compared with the pre-recorded edge force amplification feature to generate a homogeneous trampling identifier. Based on the same-source trampling mark, the force center position data is determined based on the force ratio relationship of each sensor at the bottom of the weighing platform. The force center position data is compared with the area boundary in the preset spatial mapping relationship to generate edge trampling mark, overlapping interference mark, or position uncertain mark. In response to the generation of the edge stepping mark or the position uncertainty mark, a process pause command and a back prompt command are generated, and the execution feedback signal is verified, and an on-site review command is generated. According to the on-site verification instruction, the second force distribution data is collected and compared with the trampling characteristics to generate an interference removal mark or an interference not removed mark.
2. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 1, characterized in that: The extracted eccentric load waveform and the time period of weight inconsistency are combined into a trampling feature pair. The stress position of the eccentric load waveform is compared with the edge stress increase feature to generate a homologous trampling identifier, including: By comparing the current total weight change sequence on the weighing platform with the smooth decrease waveform when the pigs leave normally, the abrupt change period where the weight increases instead of decreasing is identified as the period of weight inconsistency. By comparing the force distribution waveforms of each sensor with the central force waveform of a normal standing pig, the waveform in which the force distribution changes from being concentrated at the center to suddenly increasing at the edge is identified as the off-center load waveform. The combination of the eccentric load waveform and the weight contradiction period falling within the same time window is identified as the trampling characteristic pair; The stress location data of the off-center load waveform is compared with the pre-recorded edge stress amplification feature data. When the stress location ranges of the two are determined to correspond, a common-source trampling identifier is generated.
3. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 1, characterized in that: The determination of the force center location data based on the force ratio relationship of each sensor at the bottom of the weighing platform includes: Obtain the current force values of the first, second, third, and fourth sensors at the bottom of the weighing platform; Based on the proportional relationship of the current force values of the first, second, third, and fourth sensors, determine the coordinate position of the current force concentration point on the weighing platform plane; The coordinate position is determined as the position data of the force center.
4. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 1, characterized in that: The step of comparing the force center location data with the region boundary in the preset spatial mapping relationship to generate edge trampling markers, overlapping interference markers, or position uncertainty markers includes: When the force center location data is determined to fall within the boundary of the first region but exceed the boundary of the second region, an edge trampling mark is generated. When the force center location data is determined to fall within the boundary of the second region, an overlap interference flag is generated; When the force center location data falls within the boundary between the first region boundary and the third region boundary, and the force ratio of the four sensors is within a preset critical range, a position uncertainty identifier is generated.
5. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 1, characterized in that: The response to generating the edge stepping marker or the position uncertainty marker, generating a process pause command and a back prompt command, and verifying the execution feedback signal includes: In response to the generation of the edge stepping mark or the position uncertainty mark, a process pause command is generated and sent to the unattended weighing management software, and a back prompt command is generated and sent to the voice intercom speaker. The process pause completion signal returned by the unattended weighing management software and the personnel position change signal returned by the voice intercom column are obtained as execution feedback signals. When the process pause completion signal and the personnel position change signal are received within the preset waiting time, the verification execution feedback signal is completed. If no personnel location change signal is received within the preset waiting time, the back prompt instruction is resent and a prompt execution error flag is generated. The personnel location change signal is then obtained again as an execution feedback signal to complete the verification.
6. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 1, characterized in that: The step of collecting the second force distribution data according to the on-site verification instruction and comparing it with the trampling characteristics to generate an interference removal indicator or an interference not removal indicator includes: Collect the second force distribution data according to the on-site verification instructions; The second force distribution data from two consecutive detections are compared with the off-center load force waveform in the trampling feature pair; When it is confirmed that there is no off-center load waveform in the second force distribution data of two consecutive tests, and the total weight change time sequence returns to the normal state of decreasing as the pigs return, an interference cancellation flag is generated; If it is determined that there is still an off-center load waveform in the second force distribution data of any detection, or that the total weight change time sequence is still in the weight contradiction period, an interference not cleared flag is generated.
7. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 6, characterized in that: The step of comparing the second force distribution data detected in two consecutive tests with the off-center load force waveform in the trampling feature pair includes: The second force distribution data detected in the first test is compared with the off-center load waveform to obtain the first comparison result; The second force distribution data from the second detection is compared with the off-center load waveform to obtain the second comparison result; When the first comparison result indicates that there is no off-center load waveform and the second comparison result indicates that there is an off-center load waveform, a question mark is generated; When the first comparison result indicates the presence of an off-center load waveform and the second comparison result indicates the absence of an off-center load waveform, a question mark is generated. The suspicious identifier is sent to the central monitoring room backend, and the current process is kept paused to complete the comparison.
8. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 1, characterized in that: The acquisition of data on the direction of the pig's trajectory through the gate line recorded by the gate-side visual acquisition device and the first force distribution data from multiple sensors at the bottom of the weighing platform includes: Acquire truck approach status indicators and truck approach time data; Based on the approach time of the truck, the environmental vibration waveform within a preset time after the vehicle approaches is extracted as an environmental noise sample. The environmental noise sample is compared with the first force distribution data to determine the force frequency band corresponding to the continuous vibration component in the noise sample. The components of the force-bearing frequency band are removed from the first force distribution data to obtain a pure force waveform as the first force distribution data.
9. The method for preventing fraud in unattended weighing and counting of pigs at slaughter through dual visual cross-verification as described in claim 8, characterized in that: The acquisition of truck approach status identifier and truck approach time data includes: Send a status query request to the proximity sensor switch in front of the weighbridge; When a response signal is received from the proximity sensor switch within a preset response time, the truck approach time data is obtained based on the response signal and a truck approach status identifier is generated. If no response signal is received from the proximity sensor switch within the preset response time, a truck location inquiry prompt is sent to the driver's mobile terminal applet. Obtain the confirmation button response signal returned by the driver's mobile terminal applet; The truck approach time data is determined based on the confirmation button response signal, and a truck approach status indicator is generated.
10. A dual-visual cross-verification unattended weighing and counting anti-fraud system for pig slaughter, characterized in that, include: The gate line data acquisition module is used to acquire the trajectory direction data of the pig passing through the gate line recorded by the vision acquisition device on the gate line side and the first force distribution data of multiple sensors at the bottom of the weighing platform. The feature extraction module is used to determine, based on the data of the direction of the pig's trajectory through the gate and the first force distribution data, when the pig's turning action and the human trampling action occur in the same time window, extract the combination of the off-center load waveform and the time period of weight contradiction as trampling feature pair, and compare the force position of the off-center load waveform with the pre-recorded edge force amplification feature to generate a homogeneous trampling identifier. The position determination module is used to determine the force center position data based on the force ratio relationship of each sensor at the bottom of the weighing platform according to the same source trampling mark, compare the force center position data with the area boundary in the preset spatial mapping relationship, and generate edge trampling mark, overlapping interference mark or position uncertain mark. The interception module is used to respond to the generation of the edge stepping mark or the position uncertainty mark, generate a process pause command and a back prompt command, verify the execution feedback signal, and generate an on-site review command; The verification module is used to collect the second force distribution data according to the on-site verification instruction and compare it with the trampling feature pair to generate an interference removal mark or an interference not removal mark.
