Control method and system for biscuit blank pressing process
By combining machine vision with real-time feedback control, the problem of low manual inspection efficiency in the traditional biscuit dough pressing process has been solved, automated quality inspection and adjustment have been achieved, and production efficiency and product quality stability have been improved.
Patent Information
- Application Number
- CN202510828154.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-09-23
AI Technical Summary
The traditional biscuit dough pressing process relies on manual inspection, resulting in low production efficiency, unstable product quality, and the inability to achieve real-time feedback and adjustment.
A method combining machine vision and real-time feedback control is adopted. The camera is used to capture biscuit embryo images in real time, and a pre-trained classification model is used to classify them based on thickness, edge flatness and surface texture features. The pressure of the pressing device is adjusted through the control chip, and precise control is achieved by combining semiconductor strain gauges and motor encoders.
It realizes the automated quality inspection and adjustment of the biscuit dough pressing process, improves production efficiency, ensures product quality stability and pressing accuracy, and reduces manual intervention and raw material waste.
Smart Images
Figure CN120686748A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of biscuit manufacturing, and in particular to a control method and system for a biscuit embryo pressing process. Background Art
[0002] The traditional biscuit dough pressing process suffers from numerous technical bottlenecks, severely hindering improvements in production efficiency and product quality. In traditional biscuit production lines, quality inspection often relies on manual visual inspection or offline sampling. This method is not only inefficient but also lacks real-time feedback and adjustment. For example, after the pressing process, if the biscuit dough is found to be over-pressed or under-pressed, manual machine adjustments are required, resulting in production interruptions and material waste. Furthermore, the subjectivity of manual inspection can easily lead to inaccurate classification, impacting the stability of subsequent processes. Summary of the Invention
[0003] In order to help solve the above technical problems, the present application provides a control method and system for a biscuit dough pressing process.
[0004] In a first aspect, the present application provides a control method for a biscuit dough pressing process, which adopts the following technical solution: A control method for a biscuit dough pressing process, wherein the method comprises: S1: Use the camera to collect biscuit embryo images in real time. S2: Preliminary classification of the biscuit embryo image using a biscuit embryo standard template in a biscuit embryo database; S3: Classifying the biscuit embryo image based on thickness features, edge flatness features, and surface texture features using a pre-trained classification model to obtain an over-pressed biscuit embryo image, an under-pressed biscuit embryo image, and a normally pressed biscuit embryo image; S4: When the biscuit embryo image is the over-pressed biscuit embryo image or the under-pressed biscuit embryo image, the control chip controls the downward pressure of the pressing device until all the biscuit embryo images classified in S3 are normally pressed, and the pressure sensor at the bottom of the pressing device detects the value of the downward pressure and transmits the value of the downward pressure to the control chip; S5: storing the value of the downward pressure and the model of the biscuit embryo in a downward pressure database.
[0005] Preferably, S1 includes: preprocessing the biscuit embryo image by using an OpenCV algorithm, including denoising, edge enhancement and ROI cropping.
[0006] Preferably, S3 includes: extracting edge flatness features of the biscuit embryo image: Extract the edge of the biscuit embryo image by edge detection algorithm; Detect straight line segments in the edge of the biscuit embryo image by Hough transform; Calculate the angle between the detected straight line segment and the preset standard straight line, and calculate the variance of the angle.
[0007] Preferably, S3 includes: extracting surface texture features of the biscuit embryo image: Construct the gray-level co-occurrence matrix of the biscuit embryo image; Contrast texture features, energy texture features and entropy texture features are calculated based on the gray level co-occurrence matrix.
[0008] Preferably, S3 includes: inputting the classified over-pressed biscuit embryo images, under-pressed biscuit embryo images and normally pressed biscuit embryo images with classification marks into the biscuit embryo database.
[0009] Preferably, S3 includes: retraining the classification model based on the classified over-pressed biscuit embryo images, insufficiently pressed biscuit embryo images and normally pressed biscuit embryo images.
[0010] Preferably, S4 includes: when the biscuit embryo image is the over-pressed biscuit embryo image or the insufficiently pressed biscuit embryo image, the control chip controls the downward pressure of the pressing device based on the control motor until the images classified in S3 are all normally pressed biscuit embryo images, and an encoder is provided inside the control motor.
[0011] Preferably, S5 includes: storing the value of the downward pressure, the motor rotation angle detected in real time by the encoder, and the model of the biscuit embryo in a downward pressure database.
[0012] In a second aspect, the present application provides a control system that adopts the following technical solution: A control system using the control method for a biscuit dough pressing process as described in any one of the first aspects, wherein the control system comprises: The camera is installed on the gantry behind the pressing process; A pressing device is provided in the pressing process, and a pressure sensor is provided at the bottom thereof, and is used to press the biscuit crust into biscuit embryo; A control motor is provided in the pressing process and is used to control the downward pressure of the pressing device; A spare stepper motor is installed in the pressing process to operate the biscuit dough production line; A control chip is used to execute the above S1, S2, S3, S4, and S5.
[0013] Preferably, the pressure sensor is a semiconductor strain gauge, and the control chip is an Esp32 chip.
[0014] In summary, the beneficial effects of this application are as follows: 1. Through the combination of machine vision and real-time feedback control, the biscuit dough pressing process is automated with quality inspection and adjustment, which greatly reduces manual intervention and improves production efficiency.
[0015] 2. The intelligent analysis layer uses template matching and contour analysis algorithms, combined with the SVM classification model, to quickly and accurately classify the biscuit dough status within 10ms, ensuring the stability of product quality.
[0016] 3. The use of semiconductor strain gauges for pressure detection achieves highly sensitive and linear pressure measurement, ensuring accurate pressure control. The motor's internal encoder reads the motor's angular position in real time. Combined with the strain gauge's high sampling rate, the system precisely controls the pressing process, improving product quality. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 This is a flow chart of an embodiment of a control method for a biscuit dough pressing process of the present application; Figure 2 for Figure 1 A detailed flow chart of the illustrated embodiment; Figure 3 This is a code diagram for the first part of this application; Figure 4 This is a code diagram for the second part of this application; Figure 5 This is a code diagram for the third part of this application. DETAILED DESCRIPTION
[0018] The present invention will be further described below with reference to the accompanying drawings. The structure and principle of the present invention will be very clear to those skilled in the art. It should be understood that the specific embodiments described herein are only intended to explain the present invention and are not intended to limit the present invention.
[0019] Figure 1 This is a flow chart of an embodiment of a control method for a biscuit dough pressing process of the present application. Figure 2 for Figure 1 A detailed flow chart of the embodiment shown, Figure 3 This is the code diagram for the first part of this application. Figure 4 This is the code diagram for the second part of this application. Figure 5 This is the code diagram for the third part of this application. Figures 3 to 5 It constitutes a complete code, so the "first part", "second part" and "third part" do not mean the separation of the technical solution, but are only set up in this way to facilitate the presentation of the code.
[0020] Combine Figures 1 to 5It is understood that the control method may include: S1: The biscuit embryo image is captured in real time by a camera. In step S1, the biscuit embryo image is preprocessed using the OpenCV algorithm, including denoising, edge enhancement, and ROI cropping.
[0021] S2: Preliminary classification of biscuit embryo images using the biscuit embryo standard template in the biscuit embryo database.
[0022] S3: Using the pre-trained classification model, the biscuit embryo images are classified based on thickness features, edge flatness features, and surface texture features to obtain images of over-pressed biscuit embryos, images of under-pressed biscuit embryos, and images of normally pressed biscuit embryos. In step S3, the aforementioned features need to be extracted: Extract edge flatness features of biscuit embryo images: Extract the edge of the biscuit embryo image by edge detection algorithm; Detect straight line segments in the edge of the biscuit embryo image by Hough transform; Calculate the angle between the detected straight line segment and the preset standard straight line, and calculate the variance of the angle.
[0023] Extract surface texture features of biscuit embryo images: Construct the gray-level co-occurrence matrix of the biscuit embryo image; Contrast texture features, energy texture features and entropy texture features are calculated based on the gray-level co-occurrence matrix.
[0024] Specifically, feature extraction based on biscuit embryo images includes at least: 1. Thickness feature extraction Image preparation: The collected cookie embryo images are converted from RGB format to 8-bit format to reduce the amount of calculation and increase the processing speed.
[0025] Apply Gaussian blur to the image to remove noise and unnecessary details, making the image smoother.
[0026] Image Segmentation: The image is segmented using a brightness threshold method to separate the cookie embryo region from the background. Since there is a significant difference in brightness between the cookie embryo and the background, accurate segmentation can be achieved by setting an appropriate threshold.
[0027] Object separation and optimization: The segmented binary image may contain some small patches or noise points. Use morphological operations (such as filling holes and removing small objects) to optimize the segmentation results and ensure that only the complete cookie embryo area is retained.
[0028] Thickness measurement: Determine the axis of the cookie dough, usually the center line of the image is selected as the axis.
[0029] Divide the axis into several short segments, draw a vertical line on each segment, and measure the thickness of the biscuit at the part where the vertical line overlaps with the edge of the biscuit embryo.
[0030] Repeat the above steps until the thickness of all segments is measured, and finally take the average value of all segments as the overall thickness of the biscuit embryo.
[0031] 2. Edge Flatness Feature Extraction Edge Detection: Use edge detection algorithms (such as the Canny or Sobel operator) to extract the edges of the cookie dough. These algorithms detect edges by calculating the gradient of pixels in the image, which can accurately locate the outline of the cookie dough.
[0032] Straight line detection: The straight line segments in the edge are detected by Hough transform. Hough transform is a method for converting between image space and parameter space, which can effectively detect straight lines and curves in the image.
[0033] When detecting straight line segments, an appropriate threshold can be set to filter out straight line segments with longer lengths and consistent directions, so as to reduce the impact of noise and short line segments on subsequent analysis.
[0034] Linear deformation analysis: Calculate the angle between the detected straight line segment and the standard template line, and calculate the variance of the angle. A larger variance indicates more severe edge deformation and poorer biscuit pressing quality.
[0035] The roundness coefficient of the straight line segment can also be calculated, that is, the relationship between the average distance from the pixel point on the contour to the center and the standard deviation to evaluate the regularity of the edge.
[0036] 3. Surface Texture Feature Extraction Gray-level co-occurrence matrix (GLCM) construction: Gray-level co-occurrence matrix is a statistical-based texture feature extraction method that describes texture features by statistically analyzing the gray-level distribution of pixel pairs with specific spatial position relationships in an image.
[0037] When constructing the GLCM, it is necessary to set the appropriate spatial distance and direction. For cookie embryo images, different directions and distances can be selected to capture different texture features.
[0038] Texture feature calculation: Based on the constructed GLCM, various texture features such as contrast, energy, entropy, etc. can be calculated.
[0039] Contrast: Measures the local changes in an image, reflecting the clarity of the image and the depth of the texture grooves.
[0040] Energy: It is a measure of the uniformity of image grayscale distribution and texture coarseness. The larger the energy value, the finer the texture.
[0041] Entropy: Measures the randomness of the amount of information contained in an image. A larger entropy value indicates a more complex image.
[0042] In biscuit dough images, GLCM can effectively capture surface texture features such as roughness and texture direction. By calculating the contrast, energy, entropy and other features of GLCM, the pressing quality of the biscuit dough can be evaluated. For example: Over-pressing: If the cookie dough is over-pressed, its surface texture may become too smooth, resulting in reduced contrast, increased energy, and reduced entropy.
[0043] Insufficient pressing: If the biscuit dough is not pressed sufficiently, its surface texture may appear rough, resulting in increased contrast, decreased energy, and increased entropy.
[0044] In step S3, the classified over-pressed biscuit embryo images, under-pressed biscuit embryo images and normally pressed biscuit embryo images with classification marks are input into the biscuit embryo database, thereby improving the accuracy of step S2.
[0045] In step S3, the classification model is also retrained based on the classified over-pressed biscuit embryo images, insufficiently pressed biscuit embryo images and normally pressed biscuit embryo images.
[0046] Step S3 intelligently outputs the training model for the day and re-injects the identified cookie data into the model training process, thereby achieving the goal of using production data to feed the high-precision classifier. The cookie style classification output by the high-precision classifier is manually judged and corrected daily to prevent possible classifier errors (this process helps prevent problems such as classifier errors).
[0047] S4: When the biscuit dough image is an over-pressed biscuit dough image or an under-pressed biscuit dough image, the control chip controls the downward pressure of the pressing device until all the biscuit dough images classified in S3 are normal pressed biscuit dough images. The pressure sensor at the bottom of the pressing device detects the value of the downward pressure and transmits the value of the downward pressure to the control chip. In step S4, when the biscuit dough image is an over-pressed biscuit dough image or an under-pressed biscuit dough image, the control chip controls the downward pressure of the pressing device based on the control motor until all the biscuit dough images classified in S3 are normal pressed biscuit dough images. The control motor is internally provided with an encoder.
[0048] S5: Storing the value of the pressing pressure and the model of the biscuit embryo in the pressing pressure database. In step S5, the value of the pressing pressure, the motor rotation angle detected in real time by the encoder, and the model of the biscuit embryo are stored in the pressing pressure database.
[0049] In the embodiment of the present application, the pressure sensor is a semiconductor strain gauge, and the control chip is an Esp32 chip. Strain gauge theory: Strain gauges are based on the piezoresistive effect of materials: when a conductor or semiconductor material undergoes mechanical deformation (strain), its resistance value changes. The change in resistance (ΔR) is proportional to the strain (ε), and the mathematical relationship is: ΔR / R=K×ε, where R is the initial resistance; ΔR is the change in resistance; K is the gauge factor of the strain gauge, and ε is the strain (dimensionless, usually expressed in microstrain με, 1με= 1×10 -6 This project uses a semiconductor strain gauge with a K value of 181.7. The electrical connection uses a full-bridge connection, which offers high sensitivity and linearity.
[0050] After receiving the signal from the ESP32, the motor rotates clockwise and counterclockwise to increase or decrease pressure. The specific pressure is read in real time by a resistance strain gauge located at the bottom of the pressure frame and displayed and fed back to the ESP32 chip. During this process, the ESP32 chip records the motor's real-time angular position (the angular position is read by an internal encoder). Simultaneously, due to the strain gauge's fast response time and extremely high local pressure sampling rate, the ESP32 chip reads both the motor's real-time angular position and the pressure value, generates a corresponding table, and feeds this data to a backend intelligent database for recording. Simultaneously, the appropriate pressure for each biscuit product is calibrated (determined by the strain gauge readings). This database will also comprehensively record the hopper's product moisture, dryness, material discharge conditions, and other daily environmental conditions, including humidity, temperature, and other daily conditions, to achieve the goal of establishing a database and prepare for future production efficiency improvements.
[0051] When over-pressing is detected, the ESP32 uses PWM signals to control the stepper motor in the preceding process, rotating the screw counterclockwise (with an accuracy of ±0.1mm) and reducing the downforce by 5% to 10%. When under-pressing is detected, the ESP32 triggers the motor to increase downforce, using feedback from the pressure sensor to form a closed-loop control loop. Under normal conditions, the system automatically records the current process parameters (downforce, speed, and temperature) in an SQLite database (downforce database) for optimizing initial parameter settings for subsequent products of the same model.
[0052] The present application also provides a control system using the above control method, the control system comprising: The camera is installed on the gantry behind the pressing process; A pressing device is provided in the pressing process, and a pressure sensor is provided at the bottom thereof, and is used to press the biscuit crust into biscuit embryo; A control motor is provided in the pressing process and is used to control the downward pressure of the pressing device; A spare stepper motor is installed in the pressing process to operate the biscuit dough production line; Control chip, used to execute S1, S2, S3, S4, and S5.
[0053] In the embodiment of the present application, the pressure sensor is a semiconductor strain gauge, and the control chip is an Esp32 chip.
[0054] The current production model is upgrading from the traditional method of adjusting downforce based on manual experience to an automated torque adjustment system based on dual-motor collaborative control. The system adopts a master-slave motor redundant design. The main servo motor is responsible for adjusting the pressing parameters in real time based on visual inspection feedback. The backup stepper motor ensures the continuous operation of the production line, while retaining the upper-level manual adjustment interface as the ultimate guarantee. This design not only achieves a torque control accuracy of ±0.5%, but also forms a closed-loop control through a built-in torque sensor and PID algorithm. While improving stability, it also reserves space for subsequent intelligent upgrades. The system is specially set up with a manual intervention learning mode, so that engineers' debugging experience can be converted into digital parameters, providing high-quality samples for AI model training.
[0055] As the system continues to operate, a process parameter database covering the entire production process will be built. This database utilizes a time-series storage architecture to record multidimensional data, including torque, speed, ambient temperature and humidity, in real time. Edge computing nodes are used for data cleaning and feature extraction. Based on this data, the system initially uses machine learning algorithms such as random forests for parameter recommendations, gradually transitioning to an online learning model that enables the model to dynamically adapt to fluctuations in raw material properties. This data-driven decision-making approach effectively reduces quality fluctuations caused by human factors, reducing the trial-and-error costs of traditional production by over 60%, while also accumulating valuable industrial datasets for deep neural network training.
[0056] The final intelligent stage will introduce locally deployed convolutional neural networks to achieve end-to-end optimization, from visual inspection to process adjustments. Pre-trained models directly process production line images, increasing recognition accuracy to over 99% and establishing real-time linkage with the control system. This system not only completes the transition from experience-based production to data-based production, but also continuously optimizes process parameters through continuous learning, increasing production efficiency by over 15%. In the future, a federated learning architecture will enable multi-factory data sharing and model co-evolution, building a truly self-optimizing intelligent production system.
[0057] This application has the following advantages: Quality improvement: The defective rate of pressing has been reduced from 8% in traditional manual inspection to below 0.5%.
[0058] Efficiency optimization: Through automatic parameter calling, the production change and debugging time is shortened by 70%.
[0059] Cost savings: Reduce raw material waste due to over-pressing by approximately 12 tons per year (based on an annual production of 5,000 tons of biscuits), and simultaneously reduce energy waste by 3% (based on pressing alone).
[0060] In the future, high-speed cameras and terminal processors can be introduced to enable simultaneous inspection of multiple doughs and multiple production lines, making this system suitable for high-speed production lines. At the same time, a comprehensive database will be established to expand its application to production lines such as egg rolls. Integrating more sensor modules, such as thermal imaging, will allow for the establishment of a central production database and twin factories, optimizing production efficiency and adaptability, contributing to improved production efficiency and streamlined scheduling.
Claims
1. A control method for a biscuit embryo pressing process, characterized in that: The method comprises: S1: Use the camera to collect biscuit embryo images in real time. S2: Preliminary classification of the biscuit embryo image using a biscuit embryo standard template in a biscuit embryo database; S3: Classifying the biscuit embryo image based on thickness features, edge flatness features, and surface texture features using a pre-trained classification model to obtain an over-pressed biscuit embryo image, an under-pressed biscuit embryo image, and a normally pressed biscuit embryo image; S4: When the biscuit embryo image is the over-pressed biscuit embryo image or the under-pressed biscuit embryo image, the control chip controls the downward pressure of the pressing device until all the biscuit embryo images classified in S3 are normally pressed, and the pressure sensor at the bottom of the pressing device detects the value of the downward pressure and transmits the value of the downward pressure to the control chip; S5: storing the value of the downward pressure and the model of the biscuit embryo in a downward pressure database.
2. The method according to claim 1, characterized in that S1 includes: The biscuit embryo image is preprocessed using the OpenCV algorithm, including denoising, edge enhancement, and ROI cropping.
3. The method according to claim 1, characterized in that S3 includes: extracting edge flatness features of biscuit embryo images: Extract the edge of the biscuit embryo image by edge detection algorithm; Detect straight line segments in the edge of the biscuit embryo image by Hough transform; Calculate the angle between the detected straight line segment and the preset standard straight line, and calculate the variance of the angle.
4. The method according to claim 1, wherein S3 includes: extracting surface texture features of biscuit embryo images: Construct the gray-level co-occurrence matrix of the biscuit embryo image; Contrast texture features, energy texture features and entropy texture features are calculated based on the gray level co-occurrence matrix.
5. The method according to claim 1, wherein S3 includes: inputting the classified over-pressed biscuit embryo images, under-pressed biscuit embryo images and normally pressed biscuit embryo images with classification marks into the biscuit embryo database.
6. The method according to claim 1, wherein S3 includes: retraining the classification model based on the classified over-pressed biscuit embryo images, insufficiently pressed biscuit embryo images, and normally pressed biscuit embryo images.
7. The method according to claim 1, characterized in that S4 includes: When the biscuit embryo image is the over-pressed biscuit embryo image or the under-pressed biscuit embryo image, the control chip controls the downward pressure of the pressing device based on the control motor until the images classified in S3 are all normally pressed biscuit embryo images, and an encoder is provided inside the control motor.
8. The method according to claim 7, characterized in that S5 includes: storing the value of the pressing pressure, the motor rotation angle detected in real time by the encoder, and the model of the biscuit embryo in a pressing pressure database.
9. A control system using the control method for a biscuit dough pressing process according to any one of claims 1 to 8, characterized in that: The control system includes: The camera is installed on the gantry behind the pressing process; A pressing device is provided in the pressing process, and a pressure sensor is provided at the bottom thereof, and is used to press the biscuit crust into biscuit embryo; A control motor is provided in the pressing process and is used to control the downward pressure of the pressing device; A spare stepper motor is installed in the pressing process to operate the biscuit dough production line; A control chip is used to execute the above S1, S2, S3, S4, and S5.
10. The control system according to claim 9, characterized in that: The pressure sensor is a semiconductor strain gauge, and the control chip is an Esp32 chip.