A flexible tea rolling process based on machine vision and association models

Abstract

This invention provides a flexible tea-rolling process based on machine vision and a correlation model, relating to the technical field of tea processing technology. The process includes: acquiring a tea image stream and calculating an average real-time tea strand tightness index and a standard deviation index based on the image stream; acquiring tea sensor data to obtain real-time surface temperature and real-time carried humidity; constructing a real-time process tomography state vector based on the average real-time tea strand tightness index, the standard deviation index, the real-time surface temperature, and the real-time carried humidity; generating adjustment instructions based on the real-time process tomography state vector, and executing the adjustment instructions. This invention solves the problem of low precision in tea processing quality control, thereby improving the precision of tea processing quality control.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of tea processing technology, and more specifically, to a flexible tea kneading process based on machine vision and association models. Background Technology

[0002] Rolling is a core process that determines the appearance and internal quality of tea leaves. Traditional rolling techniques rely mainly on fixed parameter settings, which cannot adapt to batch-to-batch differences in the physical characteristics of tea raw materials (such as moisture content and tenderness), resulting in unstable finished product quality. To address this issue, the introduction of machine vision for closed-loop feedback control of the tightness of tea leaves has become a research hotspot.

[0003] However, existing technical solutions suffer from two deeper technical bottlenecks: First, the control system only focuses on the average tightness of tea leaves, lacking effective monitoring and control over the uniformity of quality (i.e., whether there are large amounts of over- or under-rolled tea leaves in a batch), resulting in a limited yield of high-grade finished products. Second, the process model on which the control system relies is usually calibrated offline, with fixed parameters. This makes it unable to adapt to dynamic changes in key physical properties such as leaf flexibility caused by real-time temperature and humidity variations in tea leaves. Consequently, the model quickly becomes mismatched in actual production, leading to decreased control performance and preventing truly flexible processing. Summary of the Invention

[0004] This invention provides a flexible tea rolling process based on machine vision and correlation models, which at least solves the problem of low precision in tea processing quality control in related technologies.

[0005] According to one embodiment of the present invention, a flexible tea rolling process based on machine vision and association models is provided, comprising:

[0006] The tea leaf image stream was acquired, and the average real-time tea leaf tightness index and the standard deviation of tightness index were calculated based on the tea leaf image stream.

[0007] Collect sensor data of tea leaves to obtain real-time surface temperature and real-time humidity.

[0008] A real-time process tomography state vector is constructed based on the average real-time strand tightness index, the tightness standard deviation index, the real-time surface temperature, and the real-time carried humidity.

[0009] Adjustment instructions are generated based on the real-time process tomography state vector and executed. The adjustment instructions are used to indicate the adjustment of kneading pressure and kneading speed.

[0010] In one exemplary embodiment, the generation of adjustment instructions based on the real-time process tomography state vector includes:

[0011] Calculate the blade flexibility index based on the real-time surface temperature and the real-time carried humidity.

[0012] Based on the blade flexibility index, the parameters of the kneading model are adjusted online.

[0013] The physical conveying delay within the kneading machine is compensated based on the parameters of the kneading model, and the feedback signal after compensation is determined.

[0014] Based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index, the multi-objective control error is calculated.

[0015] The adjustment command is generated based on the multi-objective control error.

[0016] In an exemplary embodiment, calculating the multi-objective control error based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index includes:

[0017] The average tightness error and uniformity error are calculated based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index.

[0018] The multi-objective control error is determined based on the average tightness error and the uniformity error.

[0019] In one exemplary embodiment, the determination of the compensated feedback signal includes:

[0020] The virtual kneading process model is invoked, and based on historical adjustment instructions, the predicted tightness index without delay and the predicted tightness index including delay are calculated respectively.

[0021] The average real-time strand tightness index is compared with the predicted tightness index including delay to obtain the model prediction error;

[0022] The delayed predicted tightness index is combined with the model prediction error to obtain the compensated feedback signal.

[0023] In an exemplary embodiment, the calculation of the average real-time tea leaf tightness index and the standard deviation of tightness index based on the tea image stream includes:

[0024] The tightness index of tea strips is determined based on multiple independent tea strip samples within a single frame image in the image stream.

[0025] Calculate the arithmetic mean of the strand tightness index, and use it as the average real-time strand tightness index;

[0026] Calculate the standard deviation of the multiple independent strand tightness indices, and use it as the tightness standard deviation index.

[0027] According to another embodiment of the present invention, a flexible tea-rolling system based on machine vision and correlation models is provided, comprising:

[0028] The image acquisition module is used to acquire tea image streams and calculate the average real-time tea strand tightness index and the standard deviation of tightness index based on the tea image streams;

[0029] The sensor data acquisition module is used to collect sensor data of tea leaves to obtain real-time surface temperature and real-time humidity.

[0030] The vector construction module is used to construct a real-time process chromatography state vector based on the average real-time strip tightness index, the tightness standard deviation index, the real-time surface temperature, and the real-time carried humidity.

[0031] The execution module is used to generate adjustment instructions based on the real-time process tomography state vector and execute the adjustment instructions, which are used to indicate the adjustment of kneading pressure and kneading speed.

[0032] In one exemplary embodiment, the generation of adjustment instructions based on the real-time process tomography state vector includes:

[0033] Calculate the blade flexibility index based on the real-time surface temperature and the real-time carried humidity.

[0034] Based on the blade flexibility index, the parameters of the kneading model are adjusted online.

[0035] The physical conveying delay within the kneading machine is compensated based on the parameters of the kneading model, and the feedback signal after compensation is determined.

[0036] Based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index, the multi-objective control error is calculated.

[0037] The adjustment command is generated based on the multi-objective control error.

[0038] In an exemplary embodiment, calculating the multi-objective control error based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index includes:

[0039] The average tightness error and uniformity error are calculated based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index.

[0040] The multi-objective control error is determined based on the average tightness error and the uniformity error.

[0041] According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0042] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0043] This invention constructs a multi-dimensional vector of tea images and sensor data, and analyzes it to make targeted real-time adjustments to the tea rolling process, effectively improving tea quality. Therefore, it can solve the problem of low precision in tea processing quality control and achieve the effect of improving the precision of tea processing quality control. Attached Figure Description

[0044] Figure 1 This is a structural block diagram of a flexible tea rolling system based on machine vision and association model according to an embodiment of the present invention;

[0045] Figure 2 This is a flowchart of a flexible tea-rolling method based on machine vision and association model according to an embodiment of the present invention.

[0046] Figure 3 This is a schematic diagram of a model relating leaf flexibility index to tea temperature and humidity, provided by an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0048] In the following description, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0049] Furthermore, in this application, directional terms such as "upper," "lower," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and may change accordingly depending on the orientation of the components in the accompanying drawings.

[0050] In this application, unless otherwise expressly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a direct connection or an indirect connection through an intermediate medium. Furthermore, the term "coupled" can refer to an electrical connection that enables signal transmission.

[0051] As used herein, “about,” “approximately,” or “approximately” includes the stated value and the average value within an acceptable range of deviation from the given value, wherein the acceptable range of deviation is determined by a person skilled in the art taking into account the measurement under discussion and the error associated with the measurement of the given quantity (i.e., the limitations of the measurement system).

[0052] This application provides a flexible tea rolling process based on machine vision and correlation models. This process upgrades the traditional single-objective, static model control paradigm to a multi-objective, dynamically adaptive "process tomography" control paradigm. It senses the mean and distribution of tea quality and the physical state of the tea material (temperature and humidity) in real time. Based on this multi-dimensional information, it adaptively adjusts its internal control model and control objectives online, thereby achieving precise and flexible closed-loop control of the rolling process while compensating for physical transport delays.

[0053] Example 1

[0054] Reference Figure 1 This embodiment provides an overall architecture for a flexible tea kneading system based on machine vision and correlation models. The system includes a multimodal online sensing module and an execution module 120.

[0055] The multimodal online sensing module 110 is a highly integrated sensing unit deployed at the discharge port of the kneading machine. Internally, it includes an upgraded image acquisition module 111, a sensor data acquisition module 112, and a vector construction module 113, constructing and outputting a structured "process tomography state vector" to comprehensively describe the kneading process state; specifically:

[0056] Image acquisition module 111 is used to acquire tea image streams and calculate the average real-time tea strand tightness index and tightness standard deviation index based on the tea image streams;

[0057] The sensor data acquisition module 112 is used to collect sensor data of tea leaves to obtain real-time surface temperature and real-time carried humidity.

[0058] The vector construction module 113 is used to construct a real-time process chromatography state vector based on the average real-time strip tightness index, the tightness standard deviation index, the real-time surface temperature, and the real-time carried humidity.

[0059] The execution module 120 is used to receive the complete state vector from the multimodal online sensing module. It includes a blade flexibility index calculation unit for interpreting the material state, an adaptive virtual process model unit for updating its own model in real time, and a multi-objective error calculation unit for simultaneously optimizing multiple quality indicators. Based on the real-time insight into the material state, it dynamically adjusts its control strategy to ensure that the twisting degree meets the standard and the quality is uniform. Specifically, the execution module 120 is used to generate adjustment instructions based on the real-time process tomography state vector and execute the adjustment instructions, which are used to indicate the adjustment of the twisting pressure and twisting speed.

[0060] In an optional embodiment, the generation of adjustment instructions based on the real-time process tomography state vector includes:

[0061] Calculate the blade flexibility index based on the real-time surface temperature and the real-time carried humidity.

[0062] Based on the blade flexibility index, the parameters of the kneading model are adjusted online.

[0063] The physical conveying delay within the kneading machine is compensated based on the parameters of the kneading model, and the feedback signal after compensation is determined.

[0064] Based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index, the multi-objective control error is calculated.

[0065] The adjustment command is generated based on the multi-objective control error.

[0066] In an optional embodiment, calculating the multi-objective control error based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index includes:

[0067] The average tightness error and uniformity error are calculated based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index.

[0068] The multi-objective control error is determined based on the average tightness error and the uniformity error.

[0069] Example 2

[0070] like Figure 2 As shown, this embodiment also provides a flexible tea-rolling process based on machine vision and a correlation model, specifically including the following steps:

[0071] Step S100: Construct the real-time process tomography state vector.

[0072] In this embodiment, the system performs multi-dimensional online monitoring of the tea leaves at the discharge port to generate a data vector that comprehensively reflects the current process status and results. This is achieved through the following sub-steps:

[0073] S110: Obtain average compactness and uniformity indicators.

[0074] A stream of images of falling tea leaves was captured using a high-speed industrial camera. The images were then segmented to identify N independent tea leaf samples. The industrial camera used was a 20-megapixel CMOS camera at 30fps with a 45° shooting angle. Image segmentation employed a Mask R-CNN model, with a training set containing 100,000 tea leaf images and an mAP ≥ 92%. Subsequently, for each sample... (in Each algorithm independently calculates its own strand tightness index. After obtaining a set of independent TSI sample values ​​(e.g., N=50), their statistical characteristics are calculated. When the sample size is ≥50, the mean calculation error of the tightness is ≤1.0TSI and the standard deviation error is ≤0.5. When the sample size is <50, the error will exceed the allowable range of the process. Therefore, N=50 is set.

[0075] Among them, the average real-time strand tightness index This index, obtained by calculating the arithmetic mean of the samples, reflects the overall degree of rolling in the current batch of tea.

[0076]

[0077] Standard deviation of tightness The uniformity of kneading is then quantified by calculating the standard deviation of the sample.

[0078] .

[0079] This indicator The smaller the value, the more consistent the quality of the tea within a batch; the larger the value, the more polarized there is between over-rolling and under-rolling.

[0080] For example, if N=50 samples are successfully segmented from a frame of an image, their TSI values ​​are calculated and form a set. The calculated mean of this set is 82.5, and the standard deviation is 6.8. At this point, the controller acquires two morphological indices:

[0081] , .

[0082] S120: Obtain physical state indicators.

[0083] The sensor data acquisition module includes a module for measuring the average surface temperature of tea leaves as they leave the rolling drum. A non-contact infrared temperature array sensor; and a sensor for measuring the average humidity carried by tea leaves. High-precision capacitive or near-infrared spectral humidity sensors, with an infrared temperature measurement range of 20-80℃ and a humidity sensor accuracy of ±1% RH.

[0084] For example, at the same time point in acquiring the aforementioned morphological indicators, the sensor data acquisition module measured... Celsius (°C) Relative humidity (%RH).

[0085] S130: Combine and output the state vector.

[0086] All the acquired data is combined into a structured data packet to obtain the process tomography state vector. And send it to the execution module 120 via industrial Ethernet; for example, at the current time t, the output state vector is .

[0087] Step S200: Perform model adaptation and delay compensation.

[0088] In this embodiment, the execution module 120 interprets and processes the received state vector to dynamically adjust its internal model, and performs delay compensation accordingly. Specifically:

[0089] S210: Calculate the blade flexibility index.

[0090] The blade flexibility index calculation unit first extracts the temperature from the state vector. and humidity Based on a pre-defined physical correlation model, these two measurable external physical quantities are transformed into a leaf flexibility index that can characterize the internal mechanical properties of tea leaves. The higher the index, the more humid and flexible the tea leaves are, and the easier they are to deform under kneading. In simpler terms, the higher the flexibility of the tea leaves, the easier they are to deform, and the greater the increase in compactness under the same kneading pressure. Therefore, the process gain increases linearly with the increase of the flexibility index; the time constant decreases with the increase of flexibility because flexible leaves respond to shaping faster. Preferably, this model can be a linear weighted model, whose mathematical form is:

[0091]

[0092] in, and These are preset weighting coefficients, and ; It is a normalized range of temperature and humidity set based on process knowledge.

[0093] To more intuitively understand the relationship between the blade flexibility index λ(t) and physical state, please refer to [link to relevant documentation]. Figure 3 . Figure 3 This diagram illustrates a model where the leaf flexibility index λ(t) is a bivariate function of the tea leaf surface temperature T_leaf and the carried humidity H_leaf. The horizontal axis represents the real-time surface temperature of the tea leaf, and the vertical axis represents the real-time carried humidity. The values ​​of the contour lines or the shades of the colors represent the magnitude of the flexibility index λ(t) calculated using the aforementioned formula.

[0094] As can be clearly seen from the figure, within a given normalization range, the leaf flexibility index increases monotonically with the increase of temperature and humidity, meaning that the tea leaves become more humid and flexible.

[0095] For example, assuming weights The normalization range is ℃, %RH; calculated based on the aforementioned data:

[0096]

[0097] This value represents the current flexibility of the tea leaves.

[0098] S220: Parameters of the online scheduling virtual process model.

[0099] The adaptive virtual process model unit receives the calculated blade flexibility index. Then, the key parameters of its internal virtual kneading process model are immediately adjusted (or "scheduled") online. Structurally, this virtual model can still be described using a first-order inertial element with pure time delay, but its process gain is adjusted. and time constant No longer a fixed value, but The function is a function whose relationship can be obtained by fitting experimental data.

[0100] For example, suppose the parameter scheduling function is set as follows:

[0101] Process gain:

[0102] Time constant:

[0103] in These are baseline parameters. It is the scheduling coefficient.

[0104] Assumption Using the aforementioned calculations The controller immediately updates the parameters of the internal model to:

[0105] (Unit: TSI / kPa)

[0106] Second

[0107] Through this step, the control model adapts to the current physical state of the tea leaves in real time.

[0108] S230: Perform Smith predictor calculations.

[0109] Based on the updated model parameters, standard Smith predictor logic is executed to compensate for the physical transport delay within the kneading chamber; that is, the updated virtual model is invoked to calculate the delay-free prediction value. and delayed prediction values Then calculate the model prediction error. (Note that the average tightness is used here), and finally the compensated feedback signal is obtained. .

[0110] Step S300: Generate and execute multi-target control instructions.

[0111] In this embodiment, the following sub-steps are specifically included:

[0112] S310: Calculate the multi-objective control error.

[0113] The multi-target error calculation unit receives the compensated feedback signal. And extract the standard deviation of tightness from the state vector. Simultaneously, the target strand tightness index is obtained from the process parameter library. Standard deviation of target tightness index (A small, expected value representing high uniformity), then, two independent error components are calculated:

[0114] Average tightness error:

[0115] Uniformity error:

[0116] S320: Calculate the weighted total error.

[0117] To drive a single-input PID controller, these two error components need to be combined into a total error signal:

[0118]

[0119] Among them, the weighting coefficient and These are adjustable process parameters. For example, in the initial stage of kneading, if it is necessary to quickly increase the average compactness, then the parameters can be set as follows: In the later stages of kneading, to improve the uniformity of quality, settings can be adjusted. .

[0120] For example, suppose Weight ; Assuming the result And the current measurement .but:

[0121]

[0122]

[0123]

[0124] It's easy to see that the average value is slightly insufficient, but the uniformity is far worse than the target. Based on this overall error, the controller will generate stable control actions, at which point the weighting coefficients... and It is not fixed, but is adjusted in real time through a dynamic scheduling strategy with weighting coefficients. This strategy aims to intelligently allocate limited control resources (i.e., the total control output generated by PID) to different control objectives (i.e., correcting the average value or optimizing uniformity) at different processing stages.

[0125] Specifically, the dynamic scheduling strategy for weighting coefficients is based on the duration of the kneading process. The kneading process can usually be divided into a clear initial, middle and later stage, with each stage having a different focus on the process objectives.

[0126] For example, the time-based scheduling strategy could be:

[0127] Rule 1 (Initial - Rapid Prototyping Stage): When the duration of the kneading process... At the minute mark, the system's primary task is to quickly curl the loose tea leaves into shape, that is, to rapidly increase the average tightness of the tea leaves. To bring the value closer to the target value; therefore, the controller sets the weighting coefficient as follows: , This makes the total error Its composition mainly consists of average tightness error This leads to the PID controller generating powerful control actions primarily aimed at correcting the mean.

[0128] Rule 2 (Mid-to-late stage - shaping and homogenization): the duration of kneading processing At the 10-minute mark, assuming the average tightness has stabilized near the target value, the focus of the process shifts to refining the strand morphology and improving the consistency of the entire batch's product quality, i.e., reducing the standard deviation of tightness. Therefore, the controller automatically switches the weighting coefficient as follows: , This increases the uniformity error. The proportion of the total error means that subsequent adjustments by the PID controller will focus more on suppressing quality fluctuations, such as "breaking in" the tea leaves through a smoother combination of pressure and speed, rather than drastically changing their average tightness. Of course, the dynamic scheduling strategy for weighting coefficients can also be adaptively adjusted based on the system's current real-time state, rather than relying on a fixed time division.

[0129] For example, this state-based scheduling strategy can be defined as follows:

[0130] Rule 1 (Mean Priority Mode): The controller continuously monitors the magnitude of the average tightness error; if If the TSI (Total Segment Indicator) deviates significantly from the target value, the system will be forced into "mean-first mode" regardless of the current processing time. In this mode, the weights are set to... , This allows for the mobilization of the majority of control resources to quickly correct mean deviations.

[0131] Rule 2 (Equalization Control Mode): Once the controller detects that the average tightness error has converged to the allowable range, i.e. In TSI mode, the system automatically switches to "Balanced Control Mode". In this mode, the weights are set to... , This allows the controller to give equal attention to mean deviation and uniformity deviation.

[0132] S330: Generates and executes adjustment instructions.

[0133] The calculated As input to a standard PID controller, the total control output is calculated. Subsequently, a dynamic association model will be used to... Decoupling The instruction (ΔP=0.6×u(t)-0.2×λ(t), ΔS=0.4×u(t)+0.3×λ(t)) is executed. For example, the kneading pressure adjustment step is ±0.1kPa, the speed adjustment step is ±5r / min, and a 2-second interval is required after each adjustment before the next test to avoid parameter fluctuations. This process is repeated.

[0134] By constructing a multi-dimensional vector of tea images and sensor data and analyzing it, the tea rolling process can be adjusted in real time to effectively improve tea quality and solve the problem of low precision in tea processing quality control.

[0135] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0136] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0137] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0138] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0139] Embodiments of the present invention also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0140] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0141] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0142] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0143] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0144] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0145] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0146] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A flexible tea-rolling process based on machine vision and association models, characterized in that, include: The tea leaf image stream was acquired, and the average real-time tea leaf tightness index and the standard deviation of tightness index were calculated based on the tea leaf image stream. Collect sensor data of tea leaves to obtain real-time surface temperature and real-time humidity. A real-time process tomography state vector is constructed based on the average real-time strand tightness index, the tightness standard deviation index, the real-time surface temperature, and the real-time carried humidity. Adjustment instructions are generated based on the real-time process tomography state vector and executed. The adjustment instructions are used to indicate the adjustment of kneading pressure and kneading speed.

2. The method according to claim 1, characterized in that, The adjustment instruction generated based on the real-time process tomography state vector includes: Calculate the blade flexibility index based on the real-time surface temperature and the real-time carried humidity. Based on the blade flexibility index, the parameters of the kneading model are adjusted online. The physical conveying delay within the kneading machine is compensated based on the parameters of the kneading model, and the feedback signal after compensation is determined. Based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index, the multi-objective control error is calculated. The adjustment command is generated based on the multi-objective control error.

3. The method according to claim 2, characterized in that, The calculation of multi-objective control error based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index includes: The average tightness error and uniformity error are calculated based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index. The multi-objective control error is determined based on the average tightness error and the uniformity error.

4. The method according to claim 1, characterized in that, The determined feedback signal after compensation includes: The virtual kneading process model is invoked, and based on historical adjustment instructions, the predicted tightness index without delay and the predicted tightness index including delay are calculated respectively. The average real-time strand tightness index is compared with the predicted tightness index including delay to obtain the model prediction error; The delayed predicted tightness index is combined with the model prediction error to obtain the compensated feedback signal.

5. The method according to claim 1, characterized in that, The average real-time tea leaf tightness index and the standard deviation of tightness index calculated based on the tea image stream include: The tightness index of tea strips is determined based on multiple independent tea strip samples within a single frame image in the image stream. Calculate the arithmetic mean of the strand tightness index, and use it as the average real-time strand tightness index; Calculate the standard deviation of the multiple independent strand tightness indices, and use it as the tightness standard deviation index.

6. A flexible tea-rolling system based on machine vision and association models, characterized in that, include: The image acquisition module is used to acquire tea image streams and calculate the average real-time tea strand tightness index and the standard deviation of tightness index based on the tea image streams; The sensor data acquisition module is used to collect sensor data of tea leaves to obtain real-time surface temperature and real-time humidity. The vector construction module is used to construct a real-time process chromatography state vector based on the average real-time strip tightness index, the tightness standard deviation index, the real-time surface temperature, and the real-time carried humidity. The execution module is used to generate adjustment instructions based on the real-time process tomography state vector and execute the adjustment instructions, which are used to indicate the adjustment of kneading pressure and kneading speed.

7. The system according to claim 6, characterized in that, The adjustment instruction generated based on the real-time process tomography state vector includes: Calculate the blade flexibility index based on the real-time surface temperature and the real-time carried humidity. Based on the blade flexibility index, the parameters of the kneading model are adjusted online. The physical conveying delay within the kneading machine is compensated based on the parameters of the kneading model, and the feedback signal after compensation is determined. Based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index, the multi-objective control error is calculated. The adjustment command is generated based on the multi-objective control error.

8. The system according to claim 6, characterized in that, The calculation of multi-objective control error based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index includes: The average tightness error and uniformity error are calculated based on the compensated feedback signal, the target strand tightness index, the tightness standard deviation index, and the target tightness standard deviation index. The multi-objective control error is determined based on the average tightness error and the uniformity error.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is configured to perform the method described in any one of claims 1 to 5 when executed.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1 to 5.

Images