A Visual Intelligence-Based Dynamic Weighing Method and System for Logistics Parcels Using Three Collaborative Scales
By combining visual and photoelectric detection with a three-scale collaborative dynamic weighing method, the weighing deviation problem during the handover of ultra-long packages in multi-scale systems was solved. This enabled accurate calculation of the total weight of the packages and intelligent switching of working modes, improving the accuracy and reliability of the weighing system.
Patent Information
- Application Number
- CN202611279927.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-21
- Publication Date
- 2026-11-17
AI Technical Summary
In a dynamic continuous weighing system with multiple scales seamlessly connected, when an extra-long or large package crosses the junction area of adjacent scales, the real-time weight value output by the system experiences a momentary abnormal jump, resulting in a significant deviation between the measurement result and the actual value, which affects the reliability of logistics operations.
The package length and location status are collected by a visual photoelectric joint detection component. The original weighing sequence is obtained by combining the weighing sensors of three series-connected scales. The mechanical cross-section handover interval is located based on the running speed value. The total weight compensation value of the package is calculated by weighted fusion. The working mode is intelligently switched by combining the package length value and the motor drive control is adjusted to achieve accurate weighing.
It effectively suppressed the sudden jump in weighing output caused by mechanical cross-section handover, eliminated positive or negative deviations, realized accurate measurement and coordinated transportation of ultra-long parts, and improved the real-time performance and adaptability of the weighing system.
Smart Images

Figure CN123111434A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of weighing control, and in particular to a method and system for dynamic weighing of logistics parcels using three scales in a coordinated manner based on visual intelligent discrimination. Background Technology
[0002] In dynamic continuous weighing systems with multiple weighing scales seamlessly connected in series, multiple independent weighing scales are typically arranged sequentially along the conveyor line to collect the weight of packages throughout their movement. The sensors on each scale output weighing signals in a time-series fashion, and the system obtains the total weight of the package through data matching and accumulation. This technology enables continuous, non-stop weighing of extra-long or large packages and is a common solution in logistics sorting and billing processes.
[0003] In actual production, when an oversized package crosses the junction of adjacent weighing platforms, the real-time weight value output by the system may exhibit momentary abnormal fluctuations. For example, as a large package moves from the first weighing platform to the second, the weight reading on the first platform may suddenly and irregularly rise or fall sharply, while the signal from the second platform may show a momentary overshoot peak before quickly falling back. Several abnormal spikes can be observed in the original weighing data sequence during this junction. If these data are directly extracted and summed, operators will see a significant positive or negative deviation between the final displayed total package weight and the actual value, leading to incorrect waybill charges generated by the back-end billing system or triggering unnecessary heavy cargo alarms. This deviation renders the dynamic measurement results of large packages on the production line meaningless, affecting the reliability of logistics operations. Therefore, how to reduce the weight measurement deviation of packages crossing segments during multi-platform dynamic continuous weighing is a problem that urgently needs to be solved in this field. Summary of the Invention
[0004] To address the shortcomings of existing technologies, the present invention aims to provide a dynamic weighing method for logistics packages using three scales in collaboration based on visual intelligent discrimination. By combining visual and photoelectric joint detection with dynamic weighing using three scales in collaboration, the method achieves accurate weight calculation and intelligent switching of operating modes for logistics packages.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a dynamic weighing method for logistics parcels using three scales in a coordinated manner based on visual intelligent discrimination, comprising: The package length value and package position status mark are collected by the visual optoelectronic joint detection component, the original weighing sequence is obtained by the weighing sensors of three series-connected scales, and the running speed value is read from the drive motor. Based on the package location status marker and the running speed value, the mechanical cross-segment handover interval of the original weighing sequence is located. The extracted sensor signals are weighted and fused according to the dynamic coverage ratio to calculate and generate the total weight compensation value of the package. Based on the total weight compensation value and the length value of the package, a rejection determination is made and the working mode is intelligently switched. The corresponding motor drive control command is sent to the active roller drive motor to adjust the collaborative conveying state.
[0006] Preferably, reading the running speed value from the drive motor includes: reading the pulse frequency of the speed encoder built into the drive motor in real time; dividing the pulse frequency of the speed encoder by the number of pulses output per revolution of the speed encoder to obtain the number of revolutions per second of the drive roller; and then multiplying the number of revolutions per second of the drive roller by the circumference of the drive roller to calculate the running speed value of the corresponding belt.
[0007] Preferably, the acquisition of package length value and package position status mark by the visual photoelectric joint detection component includes: acquiring the physical duration of the visual recognition field of view being blocked by the package by the lateral visual contour sensor, multiplying the physical duration by the running speed value to calculate the package length value; and acquiring the hardware on / off timing sequence of the package pulling photoelectric sensor and the entry / exit scale photoelectric sensor deployed along the conveying direction in real time to generate a package position status mark that tracks the trajectory of the package.
[0008] Preferably, the process of generating a package location status marker for tracking the trajectory of the package includes: determining the scale segment where the head and tail of the package are located based on the occlusion trigger and conduction recovery events of each sensor in the hardware on / off timing sequence; outputting the package location status marker indexed by the sampling time, wherein the package location status marker includes the scale segment number currently occupied by the head of the package, the scale segment number currently occupied by the tail of the package, and the hardware timestamp of the photoelectric sensor at the feed end of the next level scale triggered by the head of the package.
[0009] Preferably, the process of obtaining the original weighing sequence through the weighing sensors of three scales connected in series includes: synchronously acquiring gravity load signals through the weighing sensors at the bottom of the first, second, and third weighing scales; converting the gravity load signals output by the three weighing sensors at each sampling moment into weighing values via analog-to-digital conversion, arranging them in ascending order according to hardware timestamps, and generating a multi-channel original weighing sequence.
[0010] Preferably, locating the mechanical segment handover interval of the original weighing sequence based on the package position status marker and the running speed value includes: determining the physical timestamps of the package head touching the next level scale body and the tail leaving the previous level scale body based on the package position status marker; using the running speed value and the physical timestamps to calibrate the impact time window crossing the mechanical splicing transition gap, and extracting the corresponding segment abnormal fluctuation fragment from the original weighing sequence.
[0011] Preferably, after extracting the corresponding cross-segment abnormal fluctuation segments from the original weighing sequence, the method further includes: performing a bandpass filtering algorithm on the cross-segment abnormal fluctuation segments; filtering out high-frequency transient spikes caused by mechanical impact and center of gravity drop, retaining the basic gravity load component, and generating a noise-reduced weighing sequence.
[0012] Preferably, the step of weighting and fusing the extracted sensor signals according to the dynamic coverage ratio to calculate the total weight compensation value of the package includes: calculating the real-time physical coverage length of the package distributed on adjacent weighing platforms by integrating the running speed value over time; dividing the real-time physical coverage length by the package length value to calculate the corresponding dynamic coverage weight coefficient; weighting and accumulating the noise-reduced weighing sequences output by two adjacent weighing platforms in real time according to the dynamic coverage weight coefficient, and superimposing the system's preset mechanical zero-point error compensation amount to calculate the total weight compensation value of the large package spanning multiple segments.
[0013] Preferably, the mechanical zero-point error compensation amount preset in the superposition system includes: the mechanical zero-point error compensation amount is composed of the sum of a zero-point compensation component, a speed compensation component, and a vibration compensation component; the noise-reduced weighing values of three weighing scales are collected under no-load conditions and the average value is taken, and the opposite of the average value is taken as the zero-point compensation component; the square of the belt running speed value is multiplied by the speed compensation coefficient and the opposite of the average value is taken as the speed compensation component; the opposite of the time average value of the residual sequence after bandpass filtering of the cross-segment abnormal fluctuation segment is taken as the vibration compensation component.
[0014] Preferably, the step of determining rejection and intelligently switching working modes based on the total package weight compensation value and the package length value includes: comparing the total package weight compensation value with a preset specification weight alarm threshold; if the total package weight compensation value exceeds the specification weight alarm threshold, generating an abnormal rejection control command and sending it to the side pneumatic sorting push rod to perform a physical rejection action; reading the package length value of the next package to be weighed, determining the preset physical size range to which the package length value belongs, and generating an adaptive switching logic status identifier.
[0015] Preferably, the step of determining the preset physical size range to which the package length value belongs and generating an adaptive switching logic state identifier includes: generating an adaptive switching logic state identifier that triggers the independent weighing mode of a single scale when the package length value is less than or equal to the effective conveying length of a single scale; generating an adaptive switching logic state identifier that triggers the linked weighing mode of two scales when the package length value is greater than the effective conveying length of a single scale and less than or equal to the total length of two consecutive scales; and generating an adaptive switching logic state identifier that triggers the combined weighing mode of three scales when the package length value is greater than the total length of two consecutive scales.
[0016] Secondly, the present invention provides a three-scale collaborative dynamic weighing system for logistics parcels based on visual intelligent discrimination, used to implement the above-mentioned three-scale collaborative dynamic weighing method for logistics parcels based on visual intelligent discrimination, the system comprising: The data acquisition module is used to acquire the package length value and package position status mark through the visual photoelectric joint detection component, obtain the original weighing sequence through the weighing sensors of three series-connected scales, and read the running speed value from the drive motor; The handover positioning and weighted fusion module is used to locate the mechanical cross-segment handover interval of the original weighing sequence based on the package position status mark and the running speed value, and to perform weighted fusion on the extracted sensor signals according to the dynamic coverage ratio to calculate and generate the total weight compensation value of the package. The mode switching and drive control module is used to make rejection judgments and intelligently switch working modes based on the total weight compensation value and the length value of the package, and to send corresponding motor drive control commands to the active roller drive motor to adjust the collaborative conveying state.
[0017] Compared to existing technologies, the advantages of this invention are as follows: By synchronously acquiring the occlusion sequence of the visual-photoelectric joint detection component and the gravity load signals of the three weighing sensors, and uniformly binding them to the same hardware timestamp reference, this invention establishes a spatiotemporal correspondence between the package position, geometric dimensions, and instantaneous weight with strict alignment at the underlying physical acquisition stage, thereby achieving precise positioning of the impact zone at the mechanical cross-section transition. During the dynamic weighing process of the package cross-section, the dynamic coverage length of the package on adjacent weighing platforms is calculated in real time based on the integral of the running speed value over time, constructing a dynamic coverage weight coefficient. The square of the coefficient is used as the inverse variance weight to weight and fuse the estimated full weight of each weighing platform, effectively suppressing the sudden jump in weighing output caused by local suspension above the gap and mechanical impact, and eliminating the positive or negative deviation of the total weight of the package caused by the instantaneous impact interference in the transition zone. Simultaneously, by combining the feedforward package length value with real-time package position status markers, this method adaptively and asynchronously adjusts the speed of the three-stage active roller drive motor, intelligently switching between single-scale independent weighing, dual-scale linked weighing, and three-scale combined weighing modes. This achieves coordinated transport of small items for independent release and ultra-long items for synchronous and constant-speed linkage. From the underlying motor drive logic, it breaks the control dead zone where long scales must stop entirely to accommodate ultra-long items, significantly shortening the package dwell interval for high-density, multi-specification mixed items. The overall solution constructs an automatic control architecture that includes measurement feedback and closed-loop control, demonstrating the real-time performance, adaptability, and high-precision dynamic measurement capabilities of multi-scale collaborative weighing in industrial control systems. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a flowchart of the three-scale collaborative dynamic weighing method for logistics parcels based on visual intelligent discrimination in this invention; Figure 2 This is a schematic diagram of a three-scale collaborative dynamic weighing system for logistics parcels in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the physical duration of package length value acquisition in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the generation of package location status markers in an embodiment of the present invention; Figure 5 This is a schematic diagram of the impact time window extraction interval calibration in an embodiment of the present invention; Figure 6This is a schematic diagram of real-time physical coverage length segmentation in an embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the physical discarding action of the side pneumatic sorting pusher in an embodiment of the present invention; Figure 8 This is a schematic diagram of the adaptive switching logic state identifier determination interval in an embodiment of the present invention; Figure 9 This is a schematic diagram of the physical speed regulation control of the active roller drive motor in an embodiment of the present invention; Figure 10 This is a functional block diagram of the three-scale collaborative dynamic weighing system for logistics parcels based on visual intelligent discrimination in this invention. Detailed Implementation
[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] Example 1: Please see Figure 1 As shown, this embodiment provides a three-scale collaborative dynamic weighing method for logistics packages based on visual intelligent discrimination, including: This embodiment provides a method and system for dynamic weighing of logistics parcels using three scales based on visual intelligent discrimination, applicable to continuous dynamic measurement scenarios in express delivery, warehousing, and airport cargo and mail lines. The system, at the hardware level, includes a bottom frame, three weighing scales (first, second, and third) seamlessly connected in series along the conveying direction, a visual-photoelectric joint detection component, and a control unit. The bottom frame, composed of legs, feet, and wheels, supports the three independent weighing scales. Each weighing scale includes a driving roller, a driven roller, a drive motor, a conveyor belt, a side plate, sensor mounting hardware, a sensor junction box, and a weighing sensor mounted on the bottom of the scale. The drive motor has a built-in speed encoder and drives the driving roller to rotate the conveyor belt. The first, second, and third weighing scales are seamlessly connected end-to-end along the conveying direction, with physical transition gaps between adjacent weighing scales. The combined visual and photoelectric detection assembly includes a lateral visual contour sensor installed on the side plate of the upstream inlet section of the production line, a bag-pulling photoelectric sensor deployed at the bag-pulling station upstream of the production line, and an inlet / outlet photoelectric sensor deployed at the inlet and outlet ends of each weighing scale. The control unit is a programmable logic controller (PLC) main control system. The PLC main control system is connected to each drive motor, each weighing sensor, each photoelectric sensor, and the side pneumatic sorting push rod via an industrial communication bus, forming an automatic control architecture integrating measurement feedback and motor drive closed-loop control. The effective conveying length of a single weighing scale is a structural parameter of the equipment, determined by the center distance between the driving and driven rollers of each weighing scale. For example, the effective conveying length of a single weighing scale is set to 0.6 meters, the total length of two continuous weighing scales is 1.2 meters, the total length of three continuous weighing scales is 1.8 meters, and the belt running speed is set to 2.0 meters per second.
[0022] See Figure 2 This is a schematic diagram of a three-scale collaborative dynamic weighing system for logistics parcels provided in an embodiment of this application. Figure 2 As shown, the first, second, and third weighing scales are seamlessly connected in series along the conveying direction, with physical transition gaps marked between adjacent scales. A package-pulling photoelectric sensor is installed at the upstream inlet section of the assembly line, and inlet and outlet photoelectric sensors are installed at the feed and discharge ends of each weighing scale to track the trajectory of packages crossing each scale. This arrangement establishes a correspondence between the physical position of a package and its scale number, which can be captured by the on / off timing of the photoelectric sensor hardware, as the package continuously passes through three independent scales. This provides a hardware acquisition benchmark for subsequently locating the mechanical section handover area based on the package's position status and determining the scale section where the package's head and tail are located. It also lays the physical foundation for the PLC main control system to perform measurement feedback and closed-loop control of each drive motor, weighing sensor, photoelectric sensor, and side pneumatic sorting push rod via the industrial communication bus.
[0023] The following describes this embodiment in the order of steps.
[0024] S1: The package length value and package position status mark are collected by the visual photoelectric joint detection component, the original weighing sequence is obtained by the weighing sensors of three series scales, and the running speed value is read from the drive motor.
[0025] In a high-speed dynamic weighing scenario where multiple weighing scales are seamlessly connected, a package passes through three independent scales on a production line at a speed of 2.0 meters per second. Its physical position, geometric dimensions, and instantaneous weight change rapidly over time and are interconnected. If a unified spatiotemporal correspondence between the package's position, length, and weight signals is not established at the underlying physical acquisition stage, it will be impossible to accurately locate the mechanical transition zone in the original weighing sequence or determine the current scale segment the package is on. Therefore, this step simultaneously acquires four types of basic data at the physical acquisition level: belt speed, package length, package position status markers, and the multi-channel original weighing sequence. This provides a unified hardware acquisition benchmark for the subsequent accurate extraction of the cross-segment impact noise zone and the switching of operating modes.
[0026] Further, step S1 includes: S11: Reads the pulse frequency of the speed encoder built into the active roller drive motor in real time, and calculates the corresponding belt running speed value.
[0027] The speed encoder built into the drive roller motor outputs a pulse frequency that is proportional to the rotational angular velocity of the drive roller. The linear velocity of the conveyor belt is directly determined by the rotational linear velocity of the drive roller. Therefore, the belt running speed can be calculated from the pulse frequency. This physical relationship provides a direct basis for obtaining the running speed value without an external speed measuring device.
[0028] Specifically, the PLC main control system reads in real time the number of pulses output by the built-in speed encoder of the first weighing scale's active roller drive motor per unit time through the industrial communication bus, and obtains the speed encoder pulse frequency.
[0029] The calculation method for the operating speed value is as follows: divide the pulse frequency of the speed encoder by the number of pulses output per revolution of the speed encoder to obtain the rotational speed per second of the drive roller; then multiply the rotational speed per second of the drive roller by the circumference of the drive roller to obtain the belt running speed value. The circumference of the drive roller is equal to pi multiplied by the diameter of the drive roller. The number of pulses output per revolution of the speed encoder is an encoder selection parameter; for example, the number of pulses output per revolution of the speed encoder is set to 1024. The diameter of the drive roller is a structural parameter of the equipment; for example, the diameter of the drive roller is set to 0.12 meters, then the circumference of the drive roller is 0.377 meters.
[0030] For example, if the diameter of the drive roller is set to 0.12 meters, then the circumference of the drive roller is C = π × 0.12 ≈ 0.377 meters. For example, when the speed encoder pulse frequency is 5432 Hz, the drive roller rotates at 5432 / 1024 = 5.305 revolutions per second, and the belt running speed is 5.305 × 0.377 = 2.0 meters per second.
[0031] S12: The physical duration of the visual recognition field of view being blocked by the package is collected by the lateral visual contour sensor. The physical duration is multiplied by the running speed value to calculate the package length value.
[0032] The subsequent step S32 requires the package length value as the basis for determining the physical size range to which the package belongs, thereby generating an adaptive switching logic state identifier. However, the lateral vision contour sensor can only output the time information of the opening and closing of the field of view, and does not yet include the spatial length information of the package. Therefore, the time information needs to be combined with the belt running speed value output by S11 to convert it into a package length value with spatial length dimensions.
[0033] Specifically, the lateral vision contour sensor is installed on the side plate of the upstream inlet section of the production line, and its vision scanning light curtain is projected horizontally along a direction perpendicular to the conveying direction. When a package enters the vision scanning field of view, the PLC main control system records the start time stamp when the field of view changes from open to closed; when the tail of the package leaves the field of view, the PLC main control system records the end time stamp when the field of view changes from closed to open.
[0034] The physical duration is calculated by subtracting the start timestamp from the end timestamp to obtain the physical duration during which the visual recognition field of view is covered and obstructed.
[0035] The method for calculating the package length value is as follows: multiply the physical duration by the belt running speed value output by S11 to obtain the package length value along the conveying direction.
[0036] For example, when the physical duration is 0.75 seconds and the belt speed is 2.0 meters per second, the package length is 0.75 × 2.0 = 1.5 meters.
[0037] See Figure 3 This is a schematic diagram illustrating the physical duration of package length value acquisition provided in an embodiment of this application. For example... Figure 3As shown, a lateral visual contour sensor is installed on one side of the production line. Its scanning field of view is projected horizontally along a direction perpendicular to the conveyor direction. Packages enter and pass through this field of view. The diagram marks the direction of the package's movement at the belt speed, the start time stamp when the field of view changes from open to closed, and the end time stamp when it returns to open. The physical projection interval corresponding to the start and end timestamps is indicated for the package length. This schematic diagram intuitively illustrates the conversion relationship between the physical duration and the package length: subtracting the start time stamp from the end time stamp yields the physical duration, and multiplying the physical duration by the belt speed gives the package length. This allows the lateral visual contour sensor, which can only output the on / off time information of the field of view, to obtain a package length value with spatial length dimensions by combining it with the belt speed value, providing a basis for subsequently determining the physical size range to which the package belongs.
[0038] S13: Real-time acquisition of the hardware on / off timing sequence of the package pulling photoelectric sensor and the entry / exit scale photoelectric sensor deployed along the conveying direction, generating package position status markers to track the trajectory of the package.
[0039] The belt speed value output by S11 and the package length value output by S12 only describe the speed and geometric dimensions of the package's movement. They do not yet include information about which weighing scale segment the package is currently on or the position of the package's head and tail relative to the mechanical joint transition gap. If the weighing fusion process proceeds directly, it will be impossible to determine which segment of the original weighing sequence corresponds to the cross-segment handover process. Therefore, this step collects the hardware on / off timing sequence of multi-level photoelectric sensors to generate a package position status marker describing the package's trajectory.
[0040] Specifically, the bag-pulling photoelectric sensor is deployed at the upstream bag-pulling station of the production line, and the inlet and outlet photoelectric sensors are deployed at the inlet and outlet ends of the first, second, and third weighing scales, respectively. The PLC main control system collects the on / off status of the bag-pulling photoelectric sensor and each inlet / outlet scale photoelectric sensor in real time, and records the corresponding hardware timestamp for each state flip, forming a hardware on / off timing sequence. The data structure of the hardware on / off timing sequence is a key-value pair sequence arranged in ascending order of timestamps. Each key-value pair uses the sensor number as the key and a tuple consisting of the state flip type and the hardware timestamp as the value. The state flip type can be either occlusion trigger or on / off recovery.
[0041] Furthermore, the package pulling photoelectric sensor is used to complete the first occlusion trigger at the upstream position of the package entering the assembly line, establish a global tracking start timestamp for the package, and serve as the identity matching benchmark for the package in the subsequent entry and exit photoelectric sensor sequence, preventing erroneous association of position status markers when adjacent packages are too close; the package pulling photoelectric sensor itself does not participate in the determination of the scale body number, which is still determined by the on / off events of each entry and exit photoelectric sensor.
[0042] The method for generating the package location status marker is as follows: based on the occlusion trigger and conduction recovery events of each sensor in the hardware on / off timing sequence, the scale segment where the package head and tail are located is determined, and a package location status marker indexed by the sampling time is output. The data structure of the package location status marker is a tuple, including the scale number currently occupied by the package head, the scale number currently occupied by the package tail, and the hardware timestamp of the package head triggering the photoelectric sensor at the feed end of the next-level scale. For example, when the package head has triggered the photoelectric sensor at the feed end of the second weighing scale while the package tail is still blocking the photoelectric sensor at the discharge end of the first weighing scale, the package location status marker records the scale number occupied by the package head as 2 and the scale number occupied by the package tail as 1, indicating that the package is in a transitional state between the first and second weighing scales.
[0043] See Figure 4 This is a schematic diagram illustrating the generation of package location status markers provided in an embodiment of this application. For example... Figure 4 As shown, the package spans both the first and second weighing scales. The diagram indicates that the inlet / outlet photoelectric sensors are located at the junction of the two scales, and shows the positional relationship of the package's head and tail relative to the sensors, along with their corresponding hardware timestamps. This diagram corresponds to a scenario where the package head has triggered the inlet / outlet photoelectric sensor at the second weighing scale's feed end, while the package tail still obstructs the inlet / outlet photoelectric sensor at the first weighing scale's discharge end. In this case, the package position status is marked as scale number 2 for the package head and scale number 1 for the package tail, indicating that the package is in a transitional state between the first and second weighing scales. This data structure allows the PLC main control system to accurately determine the scale segment where the package head and tail are located based on the hardware on / off timing sequence, providing a direct basis for subsequently locating the mechanical transition zone in the original weighing sequence.
[0044] S14: Gravity load signals are synchronously collected by the weighing sensors at the bottom of the first, second, and third weighing scales and converted into a multi-channel raw weighing sequence.
[0045] The weighing sensors at the bottom of the first, second, and third weighing scales each bear the component force of the package pressing on the corresponding scale belt. When the package passes through a section, the gravity load of the same package is distributed to two or three adjacent scales. Only by synchronously acquiring the gravity load signals of the three weighing sensors under the same sampling clock reference can the three channels of data be strictly aligned on the time axis, thereby reflecting the complete weight distribution of the package at the instant of the segment.
[0046] Specifically, the load cells at the bottom of the first, second, and third weighing scales are all resistance strain gauge load cells, installed between the corresponding scale belt support frame and the machine frame, directly collecting the gravity load signal wrapped around and pressing on the corresponding scale belt. Synchronous acquisition means that the three load cells use the same sampling clock signal provided by the PLC main control system as a reference, completing analog-to-digital conversion and recording the same hardware timestamp at the same sampling time. The sampling clock frequency is a device parameter; for example, the sampling clock frequency is set to 1000 Hz.
[0047] The method for generating the multi-channel original weighing sequence is as follows: the gravity load signals output by the three weighing sensors at each sampling time are converted into weighing values via analog-to-digital conversion, and arranged in ascending order according to the hardware timestamps to form a three-channel parallel time series. The data structure of the multi-channel original weighing sequence is a matrix indexed by the hardware timestamp, with each index corresponding to three components: the weighing values of the first, second, and third weighing scales. The number of rows equals the number of sampling points, and the number of columns equals 3.
[0048] In the underlying physical acquisition stage, the multi-level photoelectric sensor hardware on / off timing sequence output by S13 is bound to the gravity load signals of the three weighing sensors to the same hardware timestamp reference, establishing a strict spatiotemporal correspondence between the physical location of the package segment and its instantaneous weight performance. In continuous weighing scenarios where multiple scales are seamlessly connected, the moment when the package crosses the mechanical splicing transition gap lasts only on the order of milliseconds. If the photoelectric on / off event and the weighing sampling each use independent clocks, the timestamps of the two types of data will have an unquantifiable relative offset, causing the subsequent step S21 to be unable to locate the sampling segment that truly corresponds to the impact event in the multi-channel original weighing sequence. The synchronous acquisition structure based on the same sampling clock ensures that any photoelectric sensor occlusion trigger event can correspond to a determined sampling point in the multi-channel original weighing sequence, providing a hardware acquisition reference that does not rely on post-processing alignment for the accurate extraction of the subsequent impact noise range.
[0049] S2: Based on the package location status marker and running speed value, locate the mechanical cross-segment handover interval of the original weighing sequence, perform weighted fusion of the extracted sensor signals according to the dynamic coverage ratio, and calculate the total weight compensation value of the package.
[0050] Given the unavoidable physical transition gaps between adjacent independent weighing sections, when an extra-long package's head touches the next high-speed conveyor belt or its tail detaches from the previous belt, instantaneous mechanical impact and localized center of gravity imbalance occur, causing it to fall. This physical impact force couples with the belt's vibration and is directly transmitted to the weighing sensor, superimposing high-frequency interference noise onto the actual gravity load signal. If data from each weighing section is simply extracted and directly accumulated and fused based on a time window, impact errors will be introduced into the final result, causing a positive or negative deviation from the actual total weight of the package. Therefore, this step first locates the mechanical section transition zone in the multi-channel original weighing sequence output by S14 based on the package position status marker output by S13 and the belt speed value output by S11. Then, the signals within the transition zone are denoised and dynamically weighted by coverage ratio to generate a package total weight compensation value.
[0051] Further, step S2 includes: S21: Based on the package position status marker, determine the physical timestamps of the package head touching the next level of the weighing platform and the tail leaving the previous level of the weighing platform. Use the running speed value and the physical timestamps to mark the impact time window that crosses the mechanical splicing gap, and extract the corresponding cross-segment abnormal fluctuation segments from the original weighing sequence.
[0052] The original multi-channel weighing sequence contains both a stable load segment where the package is smoothly positioned on a single scale and an impact disturbance segment where the package crosses a mechanical joint gap. Subsequent noise reduction and fusion only need to process the cross-segment impact disturbance segment. If the impact time window is not identified first and the entire sequence is processed uniformly, the true weight information of the stable load segment will be destroyed. Therefore, this step uses the physical timestamp of the package crossing the gap to mark the impact time window.
[0053] Specifically, the hardware timestamp of the package head triggering the photoelectric sensor at the feed end of the next-level scale body is read from the package position status mark output by S13, and used as the physical timestamp of the package head reaching the next-level scale body; the hardware timestamp of the package tail leaving the photoelectric sensor at the discharge end of the previous-level scale body is read from the package position status mark, and used as the physical timestamp of the package tail leaving the previous-level scale body.
[0054] The method for calibrating the impact time window is as follows: Using the physical timestamp of the package head touching the next level of the weighing platform as the center of the window, extend forward and backward by one half-width of impact duration to obtain the impact time window for the package head; using the physical timestamp of the package tail leaving the previous level of the weighing platform as the center of the window, extend forward and backward by one half-width of impact duration to obtain the impact time window for the package tail. The method for determining the impact duration half-width is as follows: Divide the width of the mechanical splicing transition gap by the belt running speed value output by S11 to obtain the physical passage time of the package edge crossing the gap, and take the sum of the physical passage time and the sensor mechanical response time as the impact duration half-width. The width of the mechanical splicing transition gap is a structural parameter of the equipment; for example, the width of the mechanical splicing transition gap is set to 0.02 meters; the sensor mechanical response time is a selection parameter of the weighing sensor; for example, the sensor mechanical response time is set to 0.005 seconds. For example, when the belt speed is 2.0 meters per second, the physical passage time is 0.02 / 2.0 = 0.01 seconds, and the impact duration is 0.01 + 0.005 = 0.015 seconds.
[0055] The method for extracting the cross-segment abnormal fluctuation segment is as follows: from the multi-channel original weighing sequence output by S14, sampling points whose hardware timestamps fall within the impact time window of the package head and the impact time window of the package tail are extracted to form the cross-segment abnormal fluctuation segment. The data structure of the cross-segment abnormal fluctuation segment is consistent with the multi-channel original weighing sequence, which is a three-channel weighing value sequence indexed by hardware timestamp.
[0056] See Figure 5 This is a schematic diagram of the impact time window extraction interval calibration provided in an embodiment of this application. For example... Figure 5 As shown, the physical timestamp is marked along the hardware timestamp axis as the center of the window, extending forward and backward by half the impact duration to form the impact time window. The sampling points captured within the window constitute the cross-segment abnormal fluctuation segment. This figure illustrates the calibration method for the impact time window: using the physical timestamp of the package head touching the next level of the weighing platform or the package tail detaching from the previous level of the weighing platform as the center, the window extends outwards along the impact duration half-width, capturing sampling points within the corresponding timestamp range from the original multi-channel weighing sequence. This calibration method ensures that subsequent noise reduction and fusion processing only targets the cross-segment impact disturbance segment, avoiding damage to the stable load segment where the package is smoothly positioned on a single weighing platform. This provides a quantifiable window boundary basis for accurately capturing cross-segment abnormal fluctuation segments.
[0057] S22: Perform a bandpass filtering algorithm on the cross-segment abnormal fluctuation segment to filter out high-frequency transient spikes caused by mechanical impact and center of gravity drop, retain the basic gravity load component, and generate a noise-reduced weighing sequence.
[0058] The abnormal fluctuation segment output by S21 consists of three superimposed components in the frequency domain: an ultra-low frequency baseline drift component caused by sensor temperature drift and installation stress; a transition band load component constituted by the gravity load itself surrounding the segment; and a high-frequency transient spike component caused by the coupling of mechanical impact and belt vibration. These three components occupy different frequency bands on the frequency axis. This frequency domain separability determines that using a bandpass filtering algorithm can suppress the other two types of interference components while preserving the transition band load component.
[0059] To ensure the bandpass filter fully establishes a steady-state response and avoids boundary effects affecting filtering accuracy, the data segments actually fed into the bandpass filter are extended forward and backward by a buffer period no less than the filter establishment time, based on the impact time window calibrated in S21. For example, the buffer duration is set to 3 seconds. The data within the buffer is only used for filter state convergence and does not participate in the subsequent denoising and weighing sequence extraction. After filtering, the corresponding denoising and weighing values are still extracted according to the impact time window boundaries calibrated in S21 to form the denoising and weighing sequence.
[0060] Specifically, the bandpass filtering algorithm employs a second-order Butterworth bandpass filter. It takes the weighing value sequence of each channel of the expanded, buffered data segment as input and outputs a noise-reduced weighing value sequence for the corresponding channel. The lower cutoff frequency of the bandpass filter is used to filter out ultra-low frequency baseline drift components. The lower cutoff frequency is determined as follows: the reciprocal of the time required for a package to smoothly pass through the effective transport length of a single weighing platform is taken as the lower cutoff frequency. For example, the time required for a package to pass through a 0.6-meter effective transport length at 2.0 meters per second is 0.3 seconds, and its reciprocal is 3.33 Hz. Therefore, the lower cutoff frequency is set to 0.33 Hz. The upper cutoff frequency of the bandpass filter is used to filter out the high-frequency transient spike components generated by the coupling of mechanical impact and belt vibration. The upper cutoff frequency is determined by taking half of the main frequency of mechanical impact excitation as the upper cutoff frequency. The main frequency of mechanical impact excitation is calibrated by taking the frequency at the maximum amplitude of the output signal of the weighing sensor when the belt is running at high speed under no-load conditions through a fast Fourier transform. For example, the main frequency of mechanical impact excitation is calibrated to 40 Hz, and the upper cutoff frequency is set to 20 Hz.
[0061] The method for generating the noise-reduced weighing sequence is as follows: the three-channel weighing value sequences of the first, second, and third weighing scales are independently processed through bandpass filters to obtain a three-channel noise-reduced weighing value sequence. After being truncated according to the impact time window boundary specified in S21, the sequences are combined according to the hardware timestamp index to form the noise-reduced weighing sequence. The data structure of the noise-reduced weighing sequence is consistent with that of the original multi-channel weighing sequence.
[0062] S23: Calculate the real-time physical coverage length of the package distributed on adjacent weighing platforms by integrating the running speed value over time, divide the real-time physical coverage length by the package length value, and calculate the corresponding dynamic coverage weight coefficient.
[0063] Although the noise-reduced weighing sequence output by S22 has eliminated high-frequency transient spikes, when an extra-long package crosses a mechanical splicing transition gap, the bottom of the package is partially suspended above the gap, causing uneven gravity distribution. The component force borne by each of the two adjacent scales changes in real time as the center of gravity of the package shifts across the gap. If the readings of the two scales are directly added together according to a fixed time window, it will not be able to follow the trajectory of the center of gravity shift and produce a sudden jump in the weighing output. Therefore, this step uses the integral of the belt running speed value over time to calculate the physical coverage length of the package on the adjacent scales in real time, generating a dynamic coverage weight coefficient that follows the movement of the center of gravity.
[0064] Specifically, the physical timestamp when the package head touches the next level scale is defined as the start time of the coverage integration. The start time of the coverage integration is taken from the hardware timestamp of the package head triggering the photoelectric sensor at the feed end of the next level scale in the package position status mark output by S13.
[0065] The calculation method for the real-time physical coverage length is as follows: For any sampling moment during the cross-segment handover process, the belt running speed value output by S11 is integrated from the start time of coverage integration to the current sampling moment to obtain the real-time physical coverage length of the package that has entered the next level weighing platform. Taking the next level weighing platform as the second weighing platform as an example, it is recorded as the real-time physical coverage length of the second weighing platform. The package length value output by S12 is subtracted from the real-time physical coverage length of the second weighing platform to obtain the real-time physical coverage length of the first weighing platform where the package is still distributed on the previous level first weighing platform.
[0066] See Figure 6 This is a schematic diagram of real-time physical coverage length segmentation provided in an embodiment of this application. For example... Figure 6 As shown, the package length is marked above the package. The package body spans a mechanical splicing transition gap, with the real-time physical coverage lengths of the first and second weighing scales marked on the left and right sides of the gap, respectively. This diagram illustrates the physical relationship where the package length is divided into two parts by the mechanical splicing transition gap: the real-time physical coverage length of the second weighing scale is obtained by integrating the belt speed value from the start of the coverage integration to the current sampling time; the real-time physical coverage length of the first weighing scale is obtained by subtracting the real-time physical coverage length of the second weighing scale from the package length. This division provides a length benchmark for calculating the dynamic coverage weight coefficient, allowing the dynamic coverage weight coefficient to change in real time as the package's center of gravity shifts across the gap. This provides a basis for subsequent inverse variance weighted fusion based on the square of the dynamic coverage weight coefficient.
[0067] The calculation method of the dynamic coverage weight coefficient is as follows: divide the real-time physical coverage length of the second weighing scale body by the package length value output by S12 to obtain the dynamic coverage weight coefficient of the second weighing scale body; divide the real-time physical coverage length of the first weighing scale body by the package length value to obtain the dynamic coverage weight coefficient of the first weighing scale body; the sum of the dynamic coverage weight coefficient of the first weighing scale body and the dynamic coverage weight coefficient of the second weighing scale body is always 1.
[0068] For example, when the package length is 1.5 meters, the belt speed is 2.0 meters per second, and the current sampling time is 0.2 seconds from the start time of coverage integration, the real-time physical coverage length of the second weighing scale is 2.0 × 0.2 = 0.4 meters, and the dynamic coverage weight coefficient of the second weighing scale is 0.4 / 1.5 ≈ 0.267. The real-time physical coverage length of the first weighing scale is 1.5 - 0.4 = 1.1 meters, and the dynamic coverage weight coefficient of the first weighing scale is 1.1 / 1.5 ≈ 0.733.
[0069] The dynamic coverage length of a package on adjacent weighing platforms is calculated in real time by integrating the belt speed over time. This directly converts traditional static time-window data into a dynamic coverage weighting coefficient that follows the movement of the package's center of gravity. The design goal is to quantify the real-time mass distribution ratio of packages on two adjacent weighing platforms, thus providing a continuously changing weighting benchmark for the weighted fusion of S24. However, since the dynamic coverage weighting coefficient is physically equal to the real-time ratio of the length of the package that has crossed the gap to the total length of the package, its mathematical structure corresponds precisely to the real-time position of the package's center of gravity relative to the mechanical splicing transition gap: when the package's center of gravity has not yet crossed the gap, the dynamic coverage weighting coefficient of the first weighing platform is greater than that of the second weighing platform; when the package's center of gravity crosses the gap, the relationship reverses. This positional correspondence allows the low-coverage scale readings, which suffer from significant edge errors due to localized overhangs above the gaps, to be automatically assigned a smaller weight when the square of the dynamic coverage weight coefficient is used as the inverse variance weight in S24 for fusion. This naturally suppresses sudden jumps in weighing output caused by uneven gravity distribution due to localized bottom overhangs, without requiring an independent center-of-gravity sensor or additional overhang compensation module. This suppression capability is not the original design goal of the dynamic coverage weight coefficient, but rather arises naturally from its mathematical structure of being equal to the length percentage in cross-segment overhang scenarios. It overcomes the inherent limitation of existing multi-scale direct accumulation methods that produce positive or negative deviations when ultra-long pieces cross junction gaps due to failure to track the center-of-gravity translation trajectory.
[0070] S24: The noise-reduced weighing sequences output by two adjacent scales are weighted and accumulated in real time according to the dynamic coverage weight coefficient, and the system's preset mechanical zero-point error compensation is added to calculate the total weight compensation value of the large package across segments.
[0071] The subsequent step S31 requires the total weight compensation value of the package to be compared with the specified weight alarm threshold. If the total weight is simply summed from the noise reduction weighing sequences of the two scales, the component force error caused by the suspension of the span will be completely retained. Therefore, this step performs weighted fusion of the noise reduction weighing sequence output by S22 according to the dynamic coverage weight coefficient output by S23, and adds mechanical zero-point error compensation to eliminate the inherent offset of the system.
[0072] Specifically, for any sampling moment during the cross-segment handover process, taking the adjacent first and second weighing scales as an example, the denoised weighing value of the first weighing scale in the denoised weighing sequence output by S22 is divided by the dynamic coverage weight coefficient of the first weighing scale output by S23 to obtain the estimated full weight of the first weighing scale; the denoised weighing value of the second weighing scale is divided by the dynamic coverage weight coefficient of the second weighing scale to obtain the estimated full weight of the second weighing scale. The physical meaning of the estimated full weight is the overall weight estimate of the package obtained by back-calculating the component force of a single scale under the assumption that the package mass is uniformly distributed along the length direction.
[0073] The method for calculating the total weight compensation value of the package is to perform inverse variance weighted fusion based on the square of the dynamic coverage weight coefficient, and the calculation expression is as follows: in The dynamic coverage weighting coefficient for the first weighing scale body. This is the dynamic coverage weighting coefficient for the second weighing scale. This is the estimated full weight of the first weighing scale. This is the estimated full weight of the second weighing scale. This is the mechanical zero-point error compensation amount. This is the compensation value for the total weight of the package.
[0074] The design principle of the calculation expression is as follows: the full weight estimate is obtained by dividing the component force of a single scale body by the corresponding dynamic coverage weight coefficient. When the dynamic coverage weight coefficient is smaller, the amplification factor of the residual error in the component force by the division operation is larger. Therefore, the variance of the full weight estimate of a scale body with low coverage is inversely proportional to the square of its dynamic coverage weight coefficient. The square of the dynamic coverage weight coefficient is used as the weighting factor for the full weight estimate of each scale body in the numerator. It is exactly equal to the reciprocal of the variance of the corresponding full weight estimate, which conforms to the principle of weighting by inverse variance in the minimum variance unbiased estimation. This makes the scale body readings with large coverage, small edge overhang, and low estimation variance dominate in the fusion result, while the scale body readings with small coverage, located in the gap edge overhang area, and high estimation variance are suppressed accordingly. The denominator is normalized by the sum of the squares of the two dynamic coverage weight coefficients to ensure that the fusion result converges to the true weight of the package under the ideal conditions of uniform package quality and no impact error. The mechanical zero point error compensation is superimposed in the form of additive terms to eliminate the inherent bias of the system that is unrelated to the package weight. The overall design goal is to make the total weight compensation value of the package converge to the reading of the high-coverage scale with a lower estimated variance when the component force error is caused by the suspension of the cross section, thereby suppressing the instantaneous weighing deviation in the handover transition zone.
[0075] The mechanical zero-point error compensation is composed of the sum of the zero-point compensation component, the speed compensation component, and the vibration compensation component. The method for determining the zero-point compensation component is as follows: Under no-load conditions, the reduced weighing values of three weighing scales are collected and averaged. The negative of this average is taken as the zero-point compensation component; for example, the zero-point compensation component is set to -0.05 kg. In this embodiment, the three weighing scales are selected and installed using the same process, and the difference in zero-point drift is negligible within the allowable range of equipment accuracy. Therefore, the average of the three scales is used for unified compensation. If the difference in zero-point drift of each scale exceeds the allowable range in actual working conditions, the zero-point compensation component of each scale can be calibrated independently, and the zero-point compensation component of the corresponding scale is deducted before weighted fusion before calculating the estimated full-weight value. The speed compensation component is used to compensate for the dynamic additional load offset caused by the belt during high-speed operation. Its determination method is as follows: the square of the belt running speed value is multiplied by the speed compensation coefficient, and the negative value is taken as the speed compensation component. The speed compensation coefficient is determined by taking the measured dynamic weighing deviation at each speed point in historical calibration data as the target variable, the square of the corresponding belt running speed value as the independent variable, and the slope estimate when the sum of the squares of the residuals is minimum. For example, the speed compensation component is set to -0.02 kg at 2.0 m / s. The vibration compensation component is used to compensate for the rise in the mean value caused by the low-frequency vibration component remaining after the S22 bandpass filter. Its determination method is as follows: the negative value of the time average of the residual sequence after the cross-segment abnormal fluctuation segment is bandpass filtered is taken as the vibration compensation component.
[0076] For example, assuming the actual weight of the package is 20.0 kg, the dynamic coverage weighting coefficient of the first weighing scale is 0.733, and the dynamic coverage weighting coefficient of the second weighing scale is 0.267. Ideally, without error, the noise-reduced weighing value of the first weighing scale should be 0.733 × 20.0 = 14.66 kg, and the noise-reduced weighing value of the second weighing scale should be 0.267 × 20.0 = 5.34 kg. If the noise-reduced weighing value of the second weighing scale increases to 6.0 kg due to a localized impact above the gap, then the estimated full weight of the first weighing scale is 14.66 / 0.733 = 20.0 kg. The estimated full weight of the second weighing scale is 6.0 / 0.267≈22.5 kg. Substituting into the calculation expression, the numerator is 0.7332×20.0+0.2672×22.5=0.537×20.0+0.0713×22.5=10.74+1.60=12.34, and the denominator is 0.537+0.0713=0.608. The combined result is 12.34 / 0.608≈20.29 kg. After adding the zero-point compensation component of -0.05 kg, the total weight compensation value of the package is 20.29-0.05=20.24 kg. Compared to the sum of the noise-reduced weighing values of the first and second weighing scales in the same scenario, which is 14.66 + 6.0 = 20.66 kg, the total weight compensation value of the package is closer to the actual weight of the package, 20.0 kg, and the cross-segment suspended impact error is suppressed.
[0077] S3: Based on the total weight compensation value and the length value of the package, it makes rejection judgments and intelligently switches the working mode, and sends the corresponding motor drive control command to the active roller drive motor to adjust the collaborative conveying status.
[0078] In scenarios involving high-volume, multi-sized parcels passing continuously on logistics conveyors, it is necessary to physically reject out-of-specification parcels to prevent them from flowing into lower-level conveyor lines and jamming the sorting turntables. Simultaneously, it is crucial to intelligently switch between single-scale, dual-scale, and triple-scale weighing and conveying modes based on parcel length to avoid entire lines stopping to accommodate oversized items. Therefore, this step uses the total parcel weight compensation value output by S24 and the parcel length value output by S12 as decision-making criteria to generate rejection control commands and adaptive switching logic status indicators. Independent motor drive control commands are then issued to the three active roller drive motors, forming a closed-loop control system integrating measurement feedback and motor drive.
[0079] Further, step S3 includes: S31: Compare the total weight compensation value of the package with the preset standard weight alarm threshold. If the total weight compensation value of the package exceeds the standard weight alarm threshold, generate an abnormal rejection control command and send it to the side pneumatic sorting push rod to perform a physical rejection action.
[0080] The total package weight compensation value output in the preceding step S24 has eliminated the cross-segment suspension impact error and has the accuracy to serve as the overweight judgment benchmark. If overweight packages are not physically rejected at the end of the weighing section of the assembly line, oversized heavy goods will continue to flow into the next level of the conveyor line and jam the sorting turntable. Therefore, this step compares the total package weight compensation value with the specified weight alarm threshold to trigger the pneumatic sorting pusher action.
[0081] Specifically, the PLC main control system compares the total package weight compensation value output by S24 with the preset specification weight alarm threshold. The specification weight alarm threshold is determined by taking the maximum single-item load-bearing weight allowed to pass through the downstream sorting turntable of the production line as the specification weight alarm threshold. For example, the specification weight alarm threshold is set to 50 kg.
[0082] The method for generating and executing the abnormal rejection control command is as follows: When the total weight compensation value of the package is greater than the specification weight alarm threshold, the PLC main control system generates an abnormal rejection control command and sends it to the side pneumatic sorting push rod deployed on the side of the assembly line via the industrial communication bus. When the package moves to the push rod station, the side pneumatic sorting push rod extends to push the oversized package laterally out of the main conveyor line, completing the physical rejection action; when the total weight compensation value of the package is not greater than the specification weight alarm threshold, the PLC main control system does not generate an abnormal rejection control command, and the package continues to flow along the main conveyor line.
[0083] See Figure 7 This is a schematic diagram illustrating the physical discarding action of the side pneumatic sorting pusher provided in an embodiment of this application. For example... Figure 7 As shown, oversized packages are marked on the main conveyor line. Side pneumatic sorting pushers are positioned on the side of the conveyor line. The execution path of the physical rejection action is indicated by a parabolic arrow in the diagram. This diagram corresponds to the handling process when the total weight compensation value of the package exceeds the specified weight alarm threshold: The PLC main control system generates an abnormal rejection control command and sends it to the side pneumatic sorting pushers. When the package moves to the pusher station, the side pneumatic sorting pushers extend, pushing the oversized package laterally out of the main conveyor line, completing the physical rejection action. Because the total weight compensation value of the package has already eliminated the cross-segment suspension impact error through dynamic coverage weight coefficient inverse variance weighted fusion in the previous steps, the triggering timing of this physical rejection action strictly corresponds to the actual weight status of the package, avoiding the situation where overweight packages mistakenly flow into the next level conveyor line due to weighing distortion.
[0084] The total package weight compensation value output by S24 is directly used as the pre-physical trigger condition for the side pneumatic sorting pusher action. Its effect is realized in the physical connection between the weighing section of the assembly line and the lower-level sorting turntable. In the multi-weighing cross-section dynamic weighing scenario, if the directly accumulated weight without cross-section suspension impact error suppression is used as the rejection criterion, the positive deviation generated in the handover transition zone will cause some compliant packages to be misjudged as overweight and incorrectly rejected, while the negative deviation will cause truly overweight packages to be missed and flow into the lower-level sorting turntable. Since the total package weight compensation value compared in this step has been eliminated by cross-section suspension impact error through dynamic coverage weight coefficient inverse variance weighted fusion in S24, the comparison benchmark for overweight judgment is consistent with the actual weight of the package, thereby eliminating the risk of overweight packages being incorrectly flowed into the lower-level conveyor line and jamming the sorting turntable due to weighing distortion, and ensuring that the triggering time of the physical rejection action strictly corresponds to the actual weight status.
[0085] S32: Read the length value of the next package to be weighed, determine the preset physical size range to which the package length value belongs, and generate an adaptive switching logic status indicator.
[0086] The subsequent step S33 requires issuing differentiated motor drive control commands to the three active roller drive motors based on the adaptive switching logic state identifier. However, the package length value output by S12 is a continuous value and has not yet been mapped to discrete identifiers for single-scale, dual-scale, or triple-scale modes, making direct mode switching impossible. Therefore, this step maps the package length value of the next package to be weighed to the adaptive switching logic state identifier according to its preset physical size range. The boundary of the preset physical size range is determined by two equipment structural parameters: the effective conveying length of a single scale and the total length of two consecutive scales. Since the lateral visual contour sensor is installed in the upstream inlet section of the production line, the package length value is calculated before the package reaches the feeding end of the first weighing scale. The PLC main control system can complete the mode determination in advance and issue preset speed adjustment commands to the corresponding drive motors before the package enters the scale, ensuring that the motor response timing strictly matches the package arrival timing.
[0087] Further, step S32 includes: S321: When the package length is less than or equal to the effective conveying length of a single scale, generate an adaptive switching logic status flag that triggers the independent weighing mode of a single scale.
[0088] When the length of the package is less than or equal to the effective conveying length of a single scale, the package can fall completely onto the belt of a single scale at any moment during the process of passing through the assembly line. Its gravity load will not cross the mechanical splicing transition gap. At this time, the weight of the package can be completely collected by using the single scale independent weighing mode, without the participation of adjacent scales.
[0089] Specifically, the PLC main control system reads the package length value output by S12 for the next package to be weighed and compares it with the effective conveying length of a single scale. The adaptive switching logic status flag for the single-scale independent weighing mode is an integer flag value of 1. When the package length value is less than or equal to the effective conveying length of 0.6 meters for a single scale, the PLC main control system generates an adaptive switching logic status flag with a value of 1. For example, when the package length value is 0.45 meters, since 0.45 meters is less than the effective conveying length of 0.6 meters for a single scale, an adaptive switching logic status flag with a value of 1 is generated.
[0090] S322: When the length of the package is greater than the effective conveying length of a single scale and less than or equal to the total length of two consecutive scales, an adaptive switching logic status flag is generated to trigger the dual-scale linkage weighing mode.
[0091] When the length of a package is greater than the effective conveying length of a single weighing unit but less than or equal to the total length of two consecutive weighing units, the package will simultaneously press against two adjacent weighing units and cross the mechanical splicing transition gap between them during the process of passing through the conveyor line. A single weighing unit cannot collect the complete weight of the package, so the two adjacent weighing units need to work together and calculate the total weight compensation value of the package according to the weighted fusion method of S24.
[0092] Specifically, the PLC main control system compares the length of the next package to be weighed with the effective conveying length of a single weighing unit and the total length of the two consecutive weighing units. The adaptive switching logic status identifier for the dual-weighing linkage mode is an integer identifier value of 2. When the package length is greater than the effective conveying length of a single weighing unit (0.6 meters) and less than or equal to the total length of the two consecutive weighing units (1.2 meters), the PLC main control system generates an adaptive switching logic status identifier with a value of 2. For example, when the package length is 1.5 meters, it does not fall into this range; when the package length is 1.0 meter, the package length value of 1.0 meter is greater than 0.6 meters and less than 1.2 meters, generating an adaptive switching logic status identifier with a value of 2.
[0093] S323: When the length of the package is greater than the total length of the two continuous weighing scales, an adaptive switching logic status flag is generated to trigger the three-scale combined weighing mode.
[0094] When the package length exceeds the total length of the two continuous weighing units, the package will simultaneously press against all three weighing units and cross two mechanically joined transition gaps during its passage through the conveyor line. Therefore, the complete package weight must be collected by combining the first, second, and third weighing units and weighting the three-channel noise-reduced weighing sequence according to a dynamic coverage weighting coefficient. This range corresponds to the scenario where this solution can handle weighing of extra-long items with a length greater than the total length of the two continuous weighing units but not exceeding 1.8 meters of the total length of the three continuous weighing units. If the package length exceeds 1.8 meters of the total length of the three continuous weighing units, it is considered to exceed the system's design capacity and will be intercepted by the pre-sorting stage or the S31 abnormal rejection process.
[0095] Specifically, the PLC main control system compares the length of the next package to be weighed with the total length of the two continuous weighing platforms. The adaptive switching logic state identifier for the three-scale combined weighing mode is an integer identifier value of 3. When the package length is greater than 1.2 meters (the total length of the two continuous weighing platforms), the PLC main control system generates an adaptive switching logic state identifier with a value of 3. For example, when the package length is 1.5 meters, it is greater than 1.2 meters (the total length of the two continuous weighing platforms), and an adaptive switching logic state identifier with a value of 3 is generated. At this time, the calculation and weighted fusion of the dynamic coverage weight coefficients in S23 and S24 are carried out simultaneously on the three channels of the first, second, and third weighing platforms. The total weight compensation value of the package is obtained by weighted fusion of the full weight estimates of the three channels according to the square inverse variance of their respective dynamic coverage weight coefficients. The calculation expression is as follows: in The dynamic coverage weighting coefficient for the third weighing scale body. This is the estimated full weight of the third weighing scale. The sum of the dynamic coverage weighting coefficients of the three scales is always 1, and the meanings of the other symbols are consistent with those in S24. This expression is a natural extension of the weighted fusion expression for dual scales in S24 in a three-channel scenario, and the inverse variance weighting criterion and design principle are completely consistent with those in S24.
[0096] See Figure 8 This is a schematic diagram of the adaptive switching logic state identifier determination range provided in the embodiments of this application. For example... Figure 8As shown, the number axis is marked with package length values, and three intervals are divided by three boundary points: the effective conveying length of a single weighing unit, the total length of two consecutive weighing units, and the total length of three consecutive weighing units. These intervals correspond to adaptive switching logic status identifiers with values of 1, 2, and 3, respectively. This diagram illustrates the mapping relationship between the physical size interval of the package length value and the adaptive switching logic status identifiers: when the package length value is less than or equal to the effective conveying length of a single weighing unit, status identifier 1 is generated, triggering the single-weighing independent weighing mode; when the package length value is greater than the effective conveying length of a single weighing unit but less than or equal to the total length of two consecutive weighing units, status identifier 2 is generated, triggering the dual-weighing linked weighing mode; and when the package length value is greater than the total length of two consecutive weighing units, status identifier 3 is generated, triggering the three-weighing combined weighing mode. This mapping relationship allows the PLC main control system to complete the mode determination before the package enters the weighing system and issue preset speed adjustment commands to the corresponding drive motors.
[0097] S33: Based on the adaptive switching logic status identifier and the package position status mark, send independent motor drive control commands to the active roller drive motors of the first weighing scale, the second weighing scale, and the third weighing scale, and control each level of belt to perform physical speed adjustment actions such as independent conveying and releasing of small items or synchronous and constant speed linkage of extra-long items.
[0098] In scenarios where multiple package sizes are mixed and continuously passed through, a long weighing platform stopping entirely to accommodate extra-long items creates a control dead zone. This prevents the weighing platform from immediately releasing subsequent packages after processing smaller items, resulting in prolonged package dwell times and conveyor line congestion. Therefore, this step, based on the adaptive switching logic status flag output by S32 and the package position status flag output by S13, issues independent motor drive control commands to the three active roller drive motors.
[0099] Specifically, the PLC main control system determines the number of weighing platforms required for the current package based on the adaptive switching logic status flag output by S32, and determines the weighing platform numbers occupied by the package head and tail in real time based on the package position status flag output by S13. Based on this, it generates independent motor drive control commands for the active roller drive motors of the first, second, and third weighing scales. These motor drive control commands are target operating frequency commands sent to the corresponding active roller drive motor frequency converters via the industrial communication bus.
[0100] When the adaptive switching logic status flag is 1, the PLC main control system controls the active roller drive motor of the scale currently carrying small packages to perform independent conveying and releasing actions for small packages. That is, after the tail of the package leaves the photoelectric sensor at the discharge end of the scale, the system immediately sends a rated conveying frequency command to the active roller drive motor of the scale to release subsequent packages into circulation. At the same time, the downstream active roller drive motors of the scale that are not carrying the current package maintain the rated conveying frequency and operate independently without waiting for the current small package.
[0101] When the adaptive switching logic status is marked as 2 or 3, the PLC main control system controls the adjacent two or three scale drive rollers that are being wrapped and pressed to perform synchronous and constant-speed linkage actions for the extra-long items. That is, it sends the same target operating frequency command to the adjacent scale drive rollers that are being pressed, so that the linear speed of the belts of the adjacent scales is strictly consistent, and avoids the additional pulling force on the package caused by the speed difference between adjacent belts at the junction of the sections. The scale drive rollers that are not being pressed by the current extra-long items resume their rated conveying frequency and operate independently according to the package position status mark, in order to receive subsequent packages.
[0102] See Figure 9 This is a schematic diagram of the physical speed regulation control of the active roller drive motor provided in an embodiment of this application. Figure 9 As shown in the diagram, two types of packages are labeled: small items and extra-long items. Small items correspond to a single first weighing scale and its drive roller motor. Extra-long items span two adjacent weighing scales that are simultaneously covered by the package and their respective drive roller motors. The dashed lines indicate two speed control actions: independent conveying and releasing of small items and synchronous constant-speed linkage of extra-long items. This diagram illustrates the process by which the PLC main control system issues independent motor drive control commands to each drive roller motor based on adaptive switching logic status indicators and package position status markers: When the tail of the package carrying the small item leaves the inlet / outlet photoelectric sensor at the discharge end, the drive roller motor of that scale immediately performs the independent conveying and releasing action of the small item; when adjacent weighing scales are simultaneously covered by extra-long items, the drive roller motors of the covered weighing scales are issued the same target operating frequency command to perform synchronous constant-speed linkage action of the extra-long items, ensuring that the belt linear speeds of adjacent weighing scales are strictly consistent, and avoiding additional pulling force on the packages at the junction of adjacent belt speeds.
[0103] Based on the preceding package length information carried by the adaptive switching logic status flag output by S32 and the real-time segment position information carried by the package position status flag output by S13, independent motor drive control commands are issued to the active roller drive motors of the three independent weighing platforms, forming a closed-loop automatic control logic that combines feedforward and feedback. The adaptive switching logic status flag originates from the feedforward judgment of the next package length value to be weighed by S32, and the package position status flag originates from the real-time feedback of the on / off timing of the multi-level photoelectric sensor hardware by S13. If the adaptive switching logic status indicator is used only and the package position status indicator is missing, the PLC main control system can determine the number of weighing platforms required for the current package, but cannot determine the weighing platform numbers occupied by the head and tail of the package in real time. Therefore, it cannot determine at what moment the weighing platform that has finished processing the small package is ready to immediately and independently release the subsequent packages. Independent speed adjustment will degenerate into the entire line operating uniformly according to the longest package cycle. If the package position status indicator is used only and the adaptive switching logic status indicator is missing, the PLC main control system can know the occupancy status of each weighing platform in real time, but cannot predict whether the package about to enter the weighing platform needs to be linked by one, two, or three weighing platforms. It cannot adjust the adjacent weighing platforms to a synchronous and equal speed linkage state in advance. The junction of the long package will generate additional pulling force due to the speed difference between adjacent belts and further aggravate the segment impact located by S21. The combination of these two types of information allows the first weighing scale to independently release subsequent packages into circulation immediately after processing small items, based on the position feedback of the package's tail. This eliminates the need to wait for the second and third weighing scales to complete their full cycle. Simultaneously, it ensures that adjacent scales maintain synchronized speeds when extra-long packages cross sections. The roles of these two types of information in closed-loop control are irreplaceable; the absence of either will render either independent speed regulation or synchronous linkage ineffective. This breaks the control dead zone where long scales must stop entirely to accommodate extra-long packages at the underlying motor drive logic, shortening the package dwell interval for high-density, multi-specification mixed packages without increasing floor space.
[0104] This embodiment, through the above steps, is based on the law of center of gravity imbalance and mechanical impact caused by the transfer of ultra-long packages between multiple physical weighing bodies. In S23, the dynamic coverage length of the bottom of the package is mapped to the dynamic coverage weight coefficient of cross-segment weight fusion. In S24, the instantaneous impact interference in the handover transition zone is eliminated by weighted fusion of the square inverse variance of the dynamic coverage weight coefficient. In S32 and S33, the feedforward package length value and the real-time package position status mark are used to perform asynchronous speed adjustment of the three-stage active roller drive motor. An automatic control architecture including measurement feedback and closed-loop control is constructed, realizing the coordination of cross-segment accurate weighing and multi-specification continuous non-speed conveying on the underlying physical mechanical surface.
[0105] Example 2: This embodiment, based on Embodiment 1, provides a dynamic three-scale collaborative weighing system for logistics parcels based on visual intelligent discrimination, such as... Figure 10 As shown, it includes: The data acquisition module is used to acquire the package length value and package position status mark through the visual photoelectric joint detection component, obtain the original weighing sequence through the weighing sensors of three series-connected scales, and read the running speed value from the drive motor; The handover positioning and weighted fusion module is used to locate the mechanical cross-segment handover interval of the original weighing sequence based on the package position status mark and the running speed value, and to perform weighted fusion on the extracted sensor signals according to the dynamic coverage ratio to calculate and generate the total weight compensation value of the package. The mode switching and drive control module is used to make rejection judgments and intelligently switch working modes based on the total weight compensation value and the length value of the package, and to send corresponding motor drive control commands to the active roller drive motor to adjust the collaborative conveying state.
Claims
1. A dynamic weighing method for logistics parcels using three scales in collaboration based on visual intelligence, characterized in that: include: The package length value and package position status mark are collected by the visual optoelectronic joint detection component, the original weighing sequence is obtained by the weighing sensors of three series-connected scales, and the running speed value is read from the drive motor. Based on the package location status marker and the running speed value, the mechanical cross-segment handover interval of the original weighing sequence is located. The extracted sensor signals are weighted and fused according to the dynamic coverage ratio to calculate and generate the total weight compensation value of the package. Based on the total weight compensation value and the length value of the package, the rejection judgment and intelligent switching of the working mode are performed. According to the relationship between the package length value, the effective conveying length of a single scale and the total length of two continuous scales, the mode is adaptively switched between single scale independent weighing mode, dual scale linkage weighing mode and three scale combination weighing mode, and the corresponding motor drive control command is sent to the active roller drive motor to adjust the collaborative conveying state.
2. The method for dynamic weighing of logistics parcels using three scales in collaboration based on visual intelligent discrimination according to claim 1, characterized in that, The process of reading the operating speed value from the drive motor includes: Real-time reading of the pulse frequency of the speed encoder built into the active roller drive motor; Divide the pulse frequency of the speed encoder by the number of pulses output per revolution of the speed encoder to obtain the rotational speed of the drive drum per second. Then multiply the rotational speed of the drive drum per second by the circumference of the drive drum to obtain the corresponding belt running speed value.
3. The method for dynamic weighing of logistics parcels using three scales based on visual intelligent discrimination according to claim 1, characterized in that, The acquisition of package length value and package position status marker through the visual-photoelectric joint detection component includes: The physical duration of the visual recognition field of view being obstructed by the package is collected by a lateral visual contour sensor. The physical duration is multiplied by the running speed value to calculate the package length value. The system collects the on / off timing sequence of the hardware on / off of the package-pulling photoelectric sensor and the in / out weighing photoelectric sensor deployed along the conveying direction in real time, and generates package position status markers to track the trajectory of the package.
4. The three-scale collaborative dynamic weighing method for logistics parcels based on visual intelligent discrimination according to claim 3, characterized in that, The generated package location status markers for tracking the trajectory of the package include: Based on the occlusion trigger and conduction recovery events of each sensor in the hardware on / off timing sequence, the scale segment where the head and tail of the package are located is determined. Output the package location status marker indexed by the sampling time. The package location status marker includes the scale number currently occupied by the package head, the scale number currently occupied by the package tail, and the hardware timestamp of the photoelectric sensor at the feed end of the next level scale triggered by the package head.
5. The method for dynamic weighing of logistics parcels using three scales based on visual intelligent discrimination according to claim 1, characterized in that, The process of obtaining the original weighing sequence through the weighing sensors of three scales connected in series includes: Gravity load signals are collected synchronously by the weighing sensors at the bottom of the first, second, and third weighing scales. The gravity load signals output by the three weighing sensors at each sampling time are converted into weighing values via analog-to-digital conversion and arranged in ascending order according to hardware timestamps to generate a multi-channel original weighing sequence.
6. The method for dynamic weighing of logistics parcels using three scales based on visual intelligent discrimination according to claim 1, characterized in that, The step of locating the mechanical segment handover interval of the original weighing sequence based on the package location status marker and the running speed value includes: Based on the package position status marker, determine the physical timestamps of when the package head touches the next level scale and when the package tail leaves the previous level scale. The impact time window across the mechanical splicing transition gap is determined by using the running speed value and the physical timestamp, and the corresponding cross-segment abnormal fluctuation segment is extracted from the original weighing sequence.
7. The method for dynamic weighing of logistics parcels using three scales based on visual intelligent discrimination according to claim 6, characterized in that, After extracting the corresponding cross-segment abnormal fluctuation segments from the original weighing sequence, the method further includes: Perform a bandpass filtering algorithm on the aforementioned cross-segment abnormal fluctuation segment; High-frequency transient spikes caused by mechanical impact and center of gravity drop are filtered out, while the basic gravity load component is retained to generate a noise-reduced weighing sequence.
8. The method for dynamic weighing of logistics parcels using three scales based on visual intelligent discrimination according to claim 7, characterized in that, The step of weighting and fusing the extracted sensor signals according to the dynamic coverage ratio to calculate the total package weight compensation value includes: The real-time physical coverage length of the package distributed on adjacent weighing platforms is calculated by integrating the running speed value over time. Divide the real-time physical coverage length by the package length value to calculate and generate the corresponding dynamic coverage weight coefficient; The noise-reduced weighing sequences output by two adjacent scales are weighted and accumulated in real time according to the dynamic coverage weight coefficient, and the mechanical zero-point error compensation amount preset by the system is added to calculate the total weight compensation value of the large package across segments.
9. The method for dynamic weighing of logistics parcels using three scales in collaboration based on visual intelligent discrimination according to claim 8, characterized in that, The preset mechanical zero-point error compensation amount of the superposition system includes: The mechanical zero-point error compensation amount is composed of the sum of the zero-point compensation component, the velocity compensation component, and the vibration compensation component; Under no-load conditions, the noise-reduced weighing values of three weighing scales are collected and the average value is taken. The opposite of the average value is taken as the zero-point compensation component. The square of the belt running speed is multiplied by the speed compensation coefficient and the opposite number is taken as the speed compensation component. The inverse of the time average of the residual sequence after bandpass filtering of the cross-segment abnormal fluctuation segment is taken as the vibration compensation component.
10. The method for dynamic weighing of logistics parcels using three scales in collaboration based on visual intelligent discrimination according to claim 1, characterized in that, The adaptive switching between single-scale independent weighing mode, dual-scale linked weighing mode, and three-scale combined weighing mode based on the relationship between the package length value, the effective conveying length of a single scale, and the total length of two continuous scales includes: When the length of the package is less than or equal to the effective conveying length of a single scale, an adaptive switching logic status flag is generated to trigger the independent weighing mode of a single scale, and the active roller drive motor of the scale currently carrying the package is controlled to immediately perform an independent conveying and releasing action after the tail of the package leaves the discharge end of the scale. When the length of the package is greater than the effective conveying length of the single scale body and less than or equal to the total length of the two consecutive scale bodies, an adaptive switching logic status flag is generated to trigger the dual-scale linkage weighing mode, and the active roller drive motors of the two adjacent scale bodies that are simultaneously wrapped and pressed are controlled to perform synchronous and constant speed linkage actions at the same target operating frequency. When the length of the package is greater than the total length of the two continuous weighing platforms, an adaptive switching logic status flag is generated to trigger the three-scale combined weighing mode, and the active roller drive motors of the three weighing platforms that are simultaneously wrapped and pressed are controlled to perform synchronous and constant speed linkage actions at the same target operating frequency.
11. A three-scale collaborative dynamic weighing system for logistics parcels based on visual intelligent discrimination, used to implement the three-scale collaborative dynamic weighing method for logistics parcels based on visual intelligent discrimination as described in any one of claims 1 to 10, characterized in that, The system includes: The data acquisition module is used to acquire the package length value and package position status mark through the visual photoelectric joint detection component, obtain the original weighing sequence through the weighing sensors of three series-connected scales, and read the running speed value from the drive motor; The handover positioning and weighted fusion module is used to locate the mechanical cross-segment handover interval of the original weighing sequence based on the package position status mark and the running speed value, and to perform weighted fusion on the extracted sensor signals according to the dynamic coverage ratio to calculate and generate the total weight compensation value of the package. The mode switching and drive control module is used to make rejection judgments and intelligently switch working modes based on the total weight compensation value and the length value of the package. According to the relationship between the length value of the package and the effective conveying length of a single scale and the total length of two continuous scales, it adaptively switches between single scale independent weighing mode, dual scale linkage weighing mode and three scale combination weighing mode, and sends corresponding motor drive control commands to the active roller drive motor to adjust the collaborative conveying state.
