An intelligent dispensing machine and dispensing path optimization system based on visual feedback

Abstract

The present application relates to the technical field of glue dispensing machines, and discloses an intelligent glue dispensing machine and glue dispensing path optimization system based on visual feedback, including an equipment frame, a main control computer installed inside the equipment frame, a glue dispensing path optimization system installed on the equipment frame, and electrically connected to the main control computer, the inner bottom wall of the equipment frame is fixedly connected to a base, a three-axis motion mechanism is installed on the top of the base, the three-axis motion mechanism includes an X-axis linear guide, a Y-axis linear guide and a Z-axis linear guide, and all are driven by a stepper motor, and a servo drive is installed on the outside of the three-axis motion mechanism. The present invention controls the positioning accuracy within ±0.01mm through the collaborative work of the laser positioning unit and the image recognition positioning unit in the high-precision positioning module, accurately determines the glue dispensing position, avoids problems such as glue dispensing offset and leakage caused by positioning deviation, greatly improves the reliability of glue dispensing, and significantly improves product quality.

Description

Technical Field

[0001] The present invention relates to the technical field of glue dispensing machines, and in particular to an intelligent glue dispensing machine and a glue dispensing path optimization system based on visual feedback. Background Art

[0002] With the increasing miniaturization and complexity of electronic products, traditional manual dispensing can no longer meet production demands. While the automated dispensing machines currently available have improved efficiency to a certain extent, they still suffer from issues such as inaccurate positioning and poor adaptability. Recent developments in machine vision technology have made it possible to enhance the intelligence of dispensing machines. By integrating vision sensors, they can detect target position and shape in real time, enabling more precise dispensing operations.

[0003] However, existing vision-guided dispensing systems still face many challenges, such as slow image processing speeds, low recognition rates, and misjudgments caused by changes in ambient lighting. Traditional manual dispensing relies on manual operation and has high flexibility, but is greatly affected by human factors, accuracy is difficult to guarantee, and the labor intensity is high. Ordinary automatic dispensing machines control the position and movement of the dispensing head through preset programs, which has high work efficiency, but adjustments are cumbersome when dealing with different products and have poor adaptability. Dispensing machines with simple vision systems: These integrate basic visual sensors and can assist in positioning to a certain extent, but their processing capabilities and algorithms are relatively simple and easily affected by external interference. They still have significant limitations in practical applications.

[0004] Drawbacks of the existing technology: Existing dispensing machines commonly suffer from inaccurate positioning and poor adaptability. This often requires frequent parameter adjustments, especially for complex and changing products, increasing production costs and time. Furthermore, simple visual systems are sensitive to changes in ambient light, making them prone to misjudgment and further reducing the reliability and stability of dispensing. Therefore, the present invention provides an intelligent dispensing machine and dispensing path optimization system based on visual feedback to address these shortcomings of the existing technology. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an intelligent dispensing machine and a dispensing path optimization system based on visual feedback, which solves the problems of poor dispensing accuracy and reliability of the existing dispensing machines and quality problems caused by inaccurate positioning.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent dispensing machine based on visual feedback, comprising an equipment frame, a main control computer is installed inside the equipment frame, the equipment frame is provided with a dispensing path optimization system, and is electrically connected to the main control computer, the inner bottom wall of the equipment frame is fixedly connected to a base, a three-axis motion mechanism is installed on the top of the base, the three-axis motion mechanism includes an X-axis linear guide, a Y-axis linear guide and a Z-axis linear guide, and are all driven by a stepper motor, a servo drive is installed on the outside of the three-axis motion mechanism, a dispensing mechanism is installed on the moving seat of the three-axis motion mechanism, and the dispensing mechanism is used for dispensing operations.

[0007] Preferably, the dispensing mechanism includes a dispensing head, the outer side of the dispensing head is mounted on a movable seat of the three-axis motion mechanism, a pressure regulating valve is provided on the outer side of the dispensing head, and a dispensing needle is provided at the bottom of the dispensing head.

[0008] Preferably, a glue storage tank is provided on the inner side of the equipment frame, a conduit is fixedly connected to the top of the glue storage tank, one end of the conduit is connected to the outer side of the glue dispensing head, and a visual mechanism is installed on the outer side of the glue dispensing head, and the visual mechanism is used to assist the glue dispensing operation.

[0009] Preferably, the visual mechanism includes an industrial camera, which is installed on the outside of the dispensing head. A lens is provided at the bottom of the industrial camera, and a ring-shaped LED light is installed on the outside of the industrial camera, and the ring-shaped LED light is located on the periphery of the lens.

[0010] The dispensing path optimization system includes the following modules:

[0011] High-precision positioning module: uses laser ranging technology and high-resolution camera to collect images for precise positioning;

[0012] Intelligent vision module: equipped with light sensors to extract and match product feature points;

[0013] Parameter Adaptation Module: The product type recognition unit uses a deep learning algorithm to automatically identify product types and quickly adjust dispensing parameters according to the characteristics of each product;

[0014] System integration and control module: enables each module to work together to ensure the overall operation and control of the system.

[0015] Preferably, the high-precision positioning module includes the following units:

[0016] Laser positioning unit: uses laser ranging technology to accurately measure product position with a positioning accuracy within ±0.01mm;

[0017] Image recognition and positioning unit: It uses a high-resolution camera to capture images and combines them with image recognition algorithms to identify product contours, further improving positioning accuracy.

[0018] Preferably, the intelligent vision module comprises the following units:

[0019] Ambient light adaptive unit: equipped with a light sensor to monitor the ambient light intensity in real time and automatically adjust the camera exposure parameters;

[0020] Feature extraction and matching unit: Extracts and matches product feature points, enabling accurate product identification under different lighting conditions.

[0021] Preferably, the parameter adaptation module includes the following units:

[0022] Product type recognition unit: Automatically identify product types through deep learning algorithms and quickly adjust dispensing parameters based on the characteristics of different types of products;

[0023] Size detection and adjustment unit: Utilizes laser scanning or image measurement technology to detect product dimensions in real time and automatically adjust dispensing paths and parameters to accommodate products of different sizes.

[0024] Preferably, the system integration and control module includes the following units:

[0025] Central control unit: responsible for coordinating the work of each module to achieve the overall operation and control of the system.

[0026] Data storage and analysis unit: stores dispensing parameters and product quality data during the dispensing process and performs analysis to provide a basis for optimizing the dispensing process.

[0027] Preferably, in the product type recognition unit, the convolution layer: performs a convolution operation by sliding the convolution kernel on the image to extract the local features of the image. Assuming that the input image is , the convolution kernel is , the output feature map is , then the convolution operation formula is:

[0028] , then the convolution operation formula is: ;

[0029] in, is the position in the output feature map, is the position in the convolution kernel, and the convolution kernel size is or , step length Set to 1 or 2;

[0030] Pooling layer: downsamples the feature map to reduce the amount of data while retaining the features. Assume that the input feature map is , the output feature map is, and the pooling window size is , then the maximum pooling formula is:

[0031] ;

[0032] in, is the position in the output feature map, and the pooling window size is ;

[0033] Fully connected layer: The feature map after convolution and pooling is expanded into a one-dimensional vector, and then classified through the fully connected layer. Assume that the input vector is , the weight matrix is , the bias vector is , the output vector is , then the calculation formula of the fully connected layer is:

[0034] ;

[0035] in, is the activation function, the formula is ;

[0036] After the fully connected layer, there will be a The classifier converts the output vector into the probability distribution of each category. Assume that the output vector of the fully connected layer is , the number of categories is ,but The output probability of the classifier The calculation formula is:

[0037] ;

[0038] in, Indicates that the sample belongs to the category Probability, compare these probability values, and select the category with the largest probability as the recognition result.

[0039] The present invention provides an intelligent glue dispensing machine and glue dispensing path optimization system based on visual feedback. It has the following beneficial effects:

[0040] 1. The present invention controls the positioning accuracy within ±0.01mm through the coordinated work of the laser positioning unit and the image recognition positioning unit in the high-precision positioning module, accurately determines the dispensing position, and avoids problems such as dispensing offset and leakage caused by positioning deviation. The feature extraction and matching unit of the intelligent vision module can accurately identify the key parts of the product under different lighting conditions, ensuring that the dispensing position is highly consistent with the actual needs of the product, greatly improving the reliability of dispensing and significantly improving product quality.

[0041] 2. The present invention uses deep learning algorithms and measurement technology through the product type recognition unit and size detection and adjustment unit in the parameter adaptation module to quickly identify product types and sizes and automatically adjust dispensing parameters and paths without human intervention. This enables the system to quickly adapt to products of different types and sizes, reduce debugging time during product switching, greatly improve production efficiency, and meet diversified production needs.

[0042] 3. The present invention uses the data storage and analysis unit of the system integration and control module to analyze a large amount of production data to dig out the optimal dispensing process parameter combination, avoid glue waste and defective products caused by unreasonable parameters, and reduce production costs. At the same time, the automated operation and collaborative work of each module reduce the workload of manual parameter debugging, save labor costs and production time, and improve the economic benefits of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A perspective view of the present invention;

[0044] Figure 2 A schematic diagram of the internal structure of the equipment rack of the present invention;

[0045] Figure 3 for Figure 2 Enlarged view of point A in the middle;

[0046] Figure 4 This is a diagram of the dispensing path optimization system architecture of the present invention;

[0047] Figure 5 This is a schematic diagram of a high-precision positioning module of the present invention;

[0048] Figure 6 Schematic diagram of the intelligent vision module of the present invention;

[0049] Figure 7 Schematic diagram of the parameter adaptation module of the present invention;

[0050] Figure 8 It is a schematic diagram of the system integration and control module of the present invention.

[0051] Among them, 1. Equipment rack; 2. Main control computer; 3. Base; 4. Three-axis motion mechanism; 5. Servo drive; 6. Dispensing head; 7. Dispensing needle; 8. Pressure regulating valve; 9. Industrial camera; 10. Lens; 11. Ring LED light; 12. Catheter; 13. Glue storage tank. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] Example 1:

[0054] Please see the attached Figure 1-Figure 3 The embodiment of the present invention provides an intelligent dispensing machine based on visual feedback, including an equipment frame 1, a main control computer 2 is installed inside the equipment frame 1, a dispensing path optimization system is set up on the equipment frame 1, and is electrically connected to the main control computer 2, the inner bottom wall of the equipment frame 1 is fixedly connected to a base 3, a three-axis motion mechanism 4 is installed on the top of the base 3, the three-axis motion mechanism 4 includes an X-axis linear guide, a Y-axis linear guide and a Z-axis linear guide, and are all driven by a stepper motor, a servo driver 5 is installed on the outside of the three-axis motion mechanism 4, a dispensing mechanism is installed on the moving seat of the three-axis motion mechanism 4, the dispensing mechanism is used for dispensing operation, the dispensing mechanism includes a dispensing head 6, and the dispensing The outer side of the head 6 is installed on the moving seat of the three-axis motion mechanism 4. A pressure regulating valve 8 is provided on the outer side of the dispensing head 6. A dispensing needle 7 is provided at the bottom of the dispensing head 6. A glue storage tank 13 is provided on the inner side of the equipment frame 1. The top of the glue storage tank 13 is fixedly connected with a conduit 12. One end of the conduit 12 is connected to the outer side of the dispensing head 6. A visual mechanism is installed on the outer side of the dispensing head 6. The visual mechanism is used to assist the dispensing operation. The visual mechanism includes an industrial camera 9. The industrial camera 9 is installed on the outer side of the dispensing head 6. A lens 10 is provided at the bottom of the industrial camera 9. A ring-shaped LED light 11 is installed on the outer side of the industrial camera 9, and the ring-shaped LED light 11 is located on the periphery of the lens 10.

[0055] Specifically, after starting the main control computer 2, the system will automatically enter the initialization calibration program. At this time, the main control computer 2 will send a calibration instruction to the servo driver 5. The servo driver 5 accurately controls the stepper motor to drive the X-axis linear guide, Y-axis linear guide and Z-axis linear guide in the three-axis motion mechanism 4, so that the dispensing head 6, industrial camera 9 and other components return to the preset initial position, ensuring that the position of each component is accurate and correct, providing a basic guarantee for subsequent dispensing operations. After the ring LED light 11 provides a stable light source, the industrial camera 9 starts working. The industrial camera 9 has high resolution and high frame rate characteristics, for example, the resolution can reach 2000×2000 pixels and the frame rate is 30 frames / second, which can quickly and clearly capture images of the product to be processed. The captured image is quickly transmitted to the image recognition and positioning unit in the dispensing path optimization system. This unit uses advanced image recognition algorithms to analyze and process the product contours, feature points, etc. in the image. By comparing the product template data pre-stored in the main control computer 2, the system accurately calculates key information such as the product's position, dimensions, and the dispensing start and end points. This generates an optimal dispensing path and transmits the path data back to the main control computer 2. Based on the received dispensing path data and pre-set dispensing process parameters, such as dispensing volume and dispensing speed, the main control computer 2 generates detailed control instructions and sends them to the servo driver 5. Following these instructions, the servo driver 5 precisely adjusts the speed and direction of the stepper motor, driving the three-axis motion mechanism 4 to precisely move the dispensing head 6 along the planned dispensing path. During the dispensing process, the pressure regulating valve 8 plays a key role. Based on the pressure control instructions sent by the main control computer 2, it adjusts the glue pressure inside the dispensing head 6 in real time, ensuring that the glue drawn from the glue storage tank 13 through the conduit 12 is extruded from the dispensing needle 7 at a stable and appropriate flow rate, ensuring uniform and accurate dispensing. For example, for high-precision dispensing tasks, the pressure regulating valve 8 can maintain glue pressure control accuracy within ±0.05 MPa. At the same time, industrial camera 9 continuously captures images of the dispensing process and provides real-time feedback to the dispensing path optimization system. The system analyzes the images and compares the actual dispensing position with the preset path. If any deviation is detected, the control instructions are adjusted promptly to ensure that the dispensing operation always follows the optimal path, thereby improving dispensing quality and production efficiency.

[0056] Example 2:

[0057] A binocular vision system is introduced to acquire stereo images through two cameras, thereby enhancing depth perception and improving dispensing accuracy. A laser rangefinder is installed on the dispensing head to achieve real-time height measurement and avoid collision between the dispensing needle and the workpiece. A rotation axis (R-axis) is added to enable the dispensing head to dispense at more angles and adapt to more complex workpiece geometries. This embodiment retains the contents of the first embodiment and adds other functions to improve the accuracy and reliability of dispensing.

[0058] Example 3:

[0059] Please see the attached Figure 4-Figure 8 , an embodiment of the present invention provides a dispensing path optimization system, which includes the following modules: a high-precision positioning module: using laser ranging technology and high-resolution camera to collect images for precise positioning; an intelligent vision module: equipped with a light sensor to extract and match the feature points of the product; a parameter adaptation module: the product type recognition unit uses a deep learning algorithm to automatically identify the product type and quickly adjust the dispensing parameters according to the characteristics of each type of product; a system integration and control module: allows each module to work together to ensure the overall operation and control of the system. The high-precision positioning module includes the following units: a laser positioning unit: using laser ranging technology to accurately measure the position of the product, with a positioning accuracy within ±0.01mm; an image recognition positioning unit: using a high-resolution camera to collect images, combined with an image recognition algorithm, to identify the contour of the product, and further improve the positioning accuracy. The intelligent vision module includes the following units: an ambient light adaptation unit: equipped with a light sensor, real-time monitoring of the ambient light intensity, and automatic adjustment of the camera exposure parameters; a feature extraction and matching unit: extracting and matching the feature points of the product, and accurately identifying the product under different lighting conditions. The parameter adaptation module includes the following units: Product type recognition unit: Automatically recognize product types through deep learning algorithms, and quickly adjust dispensing parameters according to the characteristics of different types of products; Size detection and adjustment unit: Utilize laser scanning or image measurement technology to detect product size in real time, automatically adjust dispensing paths and parameters to adapt to products of different sizes. The system integration and control module includes the following units: Central control unit: Responsible for coordinating the work of each module to achieve the overall operation and control of the system. Data storage and analysis unit: Stores dispensing parameters and product quality data during the dispensing process, and analyzes them to provide a basis for optimizing the dispensing process. In the product type recognition unit, the convolution layer: Performs a convolution operation by sliding the convolution kernel on the image to extract local features of the image. Assuming the input image is , the convolution kernel is , the output feature map is , then the convolution operation formula is:

[0060] ;

[0061] in, is the position in the output feature map, is the position in the convolution kernel, and the convolution kernel size is or , step length Set to 1 or 2;

[0062] Pooling layer: downsamples the feature map to reduce the amount of data while retaining the features. Assume that the input feature map is , the output feature map is , the pooling window size is , then the maximum pooling formula is:

[0063] ;

[0064] in, is the position in the output feature map, and the pooling window size is ;

[0065] Fully connected layer: The feature map after convolution and pooling is expanded into a one-dimensional vector, and then classified through the fully connected layer. Assume that the input vector is , the weight matrix is , the bias vector is , the output vector is , then the calculation formula of the fully connected layer is:

[0066] ;

[0067] in, is the activation function, the formula is ;

[0068] After the fully connected layer, there will be a The classifier converts the output vector into the probability distribution of each category. Assume that the output vector of the fully connected layer is , the number of categories is ,but The output probability of the classifier The calculation formula is:

[0069] ;

[0070] in, Indicates that the sample belongs to the category Probability, compare these probability values, and select the category with the largest probability as the recognition result.

[0071] Specifically, the laser positioning unit utilizes laser ranging technology to emit a laser beam toward the product and receive the reflected light. By precisely calculating the laser's round-trip time, it calculates the distance between the product and the device, thereby precisely measuring the product's position. This results in positioning accuracy within ±0.01mm, significantly reducing dispensing position deviation and ensuring that dispensing lands precisely where the product requires it. This effectively avoids quality issues such as dispensing offset and missed spots caused by inaccurate positioning, providing a stable and reliable positioning foundation for subsequent dispensing operations. The image recognition positioning unit utilizes a high-resolution camera to capture comprehensive product images and employs advanced image recognition algorithms, such as edge detection and template matching, to identify the product's contours. The algorithm compares and analyzes the captured image with pre-stored standard contour templates to accurately identify the product's actual position and posture. This unit further enhances positioning accuracy and complements the laser positioning unit, adapting to the positioning needs of complex-shaped products, enhancing positioning stability and reliability, and ensuring the accuracy of dispensing path planning. The light sensor equipped with the Ambient Light Adaptation Unit monitors changes in ambient light intensity in real time. Upon detecting fluctuations in light intensity, it rapidly feeds a signal to the camera control system, automatically adjusting camera exposure parameters such as shutter speed and aperture. This effectively prevents false positives caused by changes in ambient light, such as when workshop lights are turned on and off or by interference from natural light. This ensures that the vision system can consistently capture clear and accurate product images under varying lighting conditions, providing reliable visual data for subsequent feature extraction and dispensing operations. The Feature Extraction and Matching Unit uses digital image processing and machine learning algorithms to extract key feature points in product images, such as edges, holes, and textures. These extracted feature points are described using unique feature descriptions under varying lighting conditions and matched against product features stored in a database. This unit ensures accurate product recognition, even with minor surface stains, wear, or uneven lighting. It can precisely locate key product locations, improving dispensing reliability and ensuring that dispensing locations closely align with actual product requirements, ultimately enhancing dispensing quality. The central control unit, the core hub of the entire system, is responsible for receiving data transmitted by each module and issuing instructions to each module based on pre-set control logic and algorithms. For example, based on product position information provided by the high-precision positioning module and product recognition results provided by the intelligent vision module, it coordinates the parameter adaptation module to adjust dispensing parameters and controls the three-axis motion mechanism to drive the dispensing head along the planned path. It implements the overall operation and control of the system, ensuring the coordinated operation of each module, ensuring the orderly dispensing process, and improving system efficiency and stability. The data storage and analysis unit stores various data during the dispensing process in real time, including dispensing parameters, product quality data, and the operating status of each module.Through data analysis algorithms, these data are deeply mined, such as analyzing the relationship between different product types and dispensing parameters to find the optimal dispensing process parameter combination; tracing the root cause of the problem based on product quality data, providing a scientific basis for optimizing the dispensing process, and helping to continuously improve the dispensing process and enhance product quality and production efficiency.

[0072] Working principle: First, start the main control computer 2 and calibrate the positions of each component. Then, the ring LED light 11 provides a stable light source, and the industrial camera 9 takes an image of the product to be processed. After analysis by the image recognition and positioning unit, the optimal dispensing path is generated; finally, the servo driver 5 drives the three-axis motion mechanism 4 according to the control instructions to drive the dispensing head 6 to move, and glue is extracted from the glue storage tank 13 through the conduit 12, and the dispensing action is completed by the dispensing needle 7.

[0073] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Glue dispensing path optimization system, characterized by: Contains the following modules: High-precision positioning module: uses laser ranging technology and high-resolution camera to collect images for precise positioning; Intelligent vision module: equipped with light sensors to extract and match product feature points; Parameter Adaptation Module: The product type recognition unit uses deep learning algorithms to automatically identify product types and quickly adjust dispensing parameters according to the characteristics of various products; System integration and control module: enables each module to work together to ensure the overall operation and control of the system; The high-precision positioning module includes the following units: Laser positioning unit: uses laser ranging technology to accurately measure product position with a positioning accuracy within ±0.01mm; Image recognition and positioning unit: uses a high-resolution camera to capture images and combines them with image recognition algorithms to identify product contours, further improving positioning accuracy. The parameter adaptation module includes the following units: Product type recognition unit: Automatically identify product types through deep learning algorithms and quickly adjust dispensing parameters based on the characteristics of different types of products; Size detection and adjustment unit: uses laser scanning or image measurement technology to detect product dimensions in real time and automatically adjust dispensing paths and parameters to accommodate products of different sizes; In the product type recognition unit, the convolution layer performs a convolution operation by sliding the convolution kernel on the image to extract the local features of the image. Assuming that the input image is I, the convolution kernel is K, and the output feature map is F, the convolution operation formula is: Where (x, y) is the position in the output feature map, (m, n) is the position in the convolution kernel, the convolution kernel size is 3×3 or 5×5, and the stride is set to 1 or 2; Pooling layer: downsamples the feature map to reduce the amount of data while retaining the main features. Assuming the input feature map is F, the output feature map is P, and the pooling window size is s×s, the maximum pooling formula is: Where (i, j) is the position in the output feature map, and the pooling window size is 2×2; Fully connected layer: The feature map after convolution and pooling is expanded into a one-dimensional vector, and then classified through the fully connected layer. Assuming that the input vector is X, the weight matrix is W, the bias vector is b, and the output vector is Y, the calculation formula of the fully connected layer is: Y = ReLU(WX+b) Among them, ReLU is the activation function, and the formula is ReLU(x)=max(0,x); After the fully connected layer, a Softmax classifier is connected to convert the output vector into the probability distribution of each category. Assuming that the output vector of the fully connected layer is Y and the number of categories is C, the output probability p of the Softmax classifier is calculated as follows: Among them, p(c) represents the probability that the sample belongs to category c. Compare these probability values and select the category with the largest probability as the recognition result; The intelligent vision module includes the following units: Ambient light adaptive unit: equipped with a light sensor to monitor the ambient light intensity in real time and automatically adjust the camera exposure parameters; Feature extraction and matching unit: Extracts and matches product feature points, enabling accurate product identification under different lighting conditions.

2. The dispensing path optimization system according to claim 1, characterized in that: The system integration and control module includes the following units: Central control unit: responsible for coordinating the work of each module to achieve the overall operation and control of the system; Data storage and analysis unit: stores dispensing parameters and product quality data during the dispensing process and performs analysis to provide a basis for optimizing the dispensing process.

Images