Intelligent dispensing machine based on visual feedback and dispensing path optimization system

Abstract

The invention relates to the technical field of dispensing machines, and discloses an intelligent dispensing machine based on visual feedback and a dispensing path optimization system.The intelligent dispensing machine comprises an equipment rack, a main control computer is installed in the equipment rack, and the dispensing path optimization system is erected on the equipment rack and electrically connected with the main control computer; the inner bottom wall of the equipment rack is fixedly connected with a base, the top of the base is provided with a three-axis movement mechanism, the three-axis movement mechanism comprises an X-axis linear guide rail, a Y-axis linear guide rail and a Z-axis linear guide rail which are all driven by a stepping motor, and the outer side of the three-axis movement mechanism is provided with a servo driver. Through cooperative work of a laser positioning unit and an image recognition positioning unit in the high-precision positioning module, the positioning precision is controlled within + / -0.01 mm, the dispensing position is accurately determined, the problems of dispensing deviation, point leakage and the like caused by positioning deviation are avoided, the dispensing reliability is greatly improved, and the product quality is remarkably improved.

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 miniaturization and increasing complexity of electronic products, traditional manual dispensing can no longer meet production needs. Although the common automatic dispensing machines on the market have improved efficiency to a certain extent, they still have problems such as inaccurate positioning and poor adaptability. In recent years, the development of machine vision technology has made it possible to improve the intelligence of dispensing machines. By integrating visual sensors, the target position and shape can be detected in real time, thereby achieving more accurate dispensing operations.

[0003] However, the existing vision-guided dispensing system still faces many challenges, such as slow image processing speed, low recognition rate, and misjudgment caused by changes in ambient light. 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, and have high work efficiency, but adjustments are cumbersome when dealing with different products, and the adaptability is poor. Dispensing machine with a simple vision system: It integrates basic visual sensors and can assist in positioning to a certain extent, but the processing power and algorithm are relatively simple, and it is easily affected by external interference, and there are still great limitations in practical applications.

[0004] Defects of the prior art: Existing dispensing machines generally have problems of inaccurate positioning and poor adaptability. Especially when facing complex and changeable products, they often need to frequently debug parameters, which increases production costs and time. In addition, simple visual systems are sensitive to changes in ambient light and are prone to misjudgment, further reducing the reliability and stability of dispensing. Therefore, the present invention provides an intelligent dispensing machine based on visual feedback and a dispensing path optimization system to solve the deficiencies in the prior art. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an intelligent dispensing machine and a dispensing path optimization system based on visual feedback, which solve the problems of poor dispensing accuracy and reliability of the prior art 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, a dispensing path optimization system is installed on the equipment frame, and the system is electrically connected to the main control computer, a base is fixedly connected to the inner bottom wall of the equipment frame, 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, 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 comprises 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 arranged on the outer side of the dispensing head, and a dispensing needle is arranged 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. An annular LED light is installed on the outside of the industrial camera, and the annular LED light is located on the periphery of the lens.

[0010] The dispensing path optimization system includes 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 optical sensors to extract and match product feature points; Parameter Adaptive 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.

[0011] Preferably, the high-precision positioning module comprises the following units: Laser positioning unit: using laser ranging technology to accurately measure the product position, with a positioning accuracy within ±0.01mm; Image recognition and positioning unit: It collects images through a high-resolution camera and combines it with an image recognition algorithm to identify the product outline and further improve positioning accuracy.

[0012] Preferably, the intelligent vision module comprises 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 the feature points of products and can accurately identify products under different lighting conditions.

[0013] Preferably, the parameter adaptation module comprises the following units: Product type identification unit: Automatically identify 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: Using laser scanning or image measurement technology, it can detect product size in real time and automatically adjust the dispensing path and parameters to adapt to products of different sizes.

[0014] Preferably, the system integration and control module comprises the following units: Central control unit: responsible for coordinating the work of each module to achieve the overall operation and control of the system.

[0015] 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.

[0016] Preferably, 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 that the input image is , the convolution kernel is , the output feature map is , then the convolution operation formula is: , then the convolution operation formula is: ; 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; 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: ; in, is the position in the output feature map, and the pooling window size is ; 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: ; in, is the activation function, and the formula is ; 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: ; 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.

[0017] The present invention provides an intelligent glue dispensing machine and glue dispensing path optimization system based on visual feedback, which has the following beneficial effects: 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 also 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.

[0018] 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.

[0019] 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 best 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

[0020] Figure 1 A perspective view of the present invention; Figure 2 It is a schematic diagram of the internal structure of the equipment rack of the present invention; Figure 3 for Figure 2 The enlarged view of point A in the middle; Figure 4 This is a system architecture diagram for optimizing the dispensing path of the present invention; Figure 5 It is a schematic diagram of a high-precision positioning module of the present invention; Figure 6 It is a schematic diagram of the intelligent vision module of the present invention; Figure 7 Schematic diagram of the parameter adaptive module of the present invention; Figure 8 It is a schematic diagram of the system integration and control module of the present invention.

[0021] 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

[0022] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0023] Embodiment 1: Please see attached Figure 1-Figure 3The embodiment of the present invention provides an intelligent glue dispensing machine based on visual feedback, including an equipment frame 1, a main control computer 2 is installed inside the equipment frame 1, a glue dispensing path optimization system is set up on the equipment frame 1, and the system 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, and 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 all are driven by a stepper motor. A servo driver 5 is installed on the outer side of the three-axis motion mechanism 4. A glue dispensing mechanism is installed on the moving seat of the three-axis motion mechanism 4. The glue dispensing mechanism is used for glue dispensing operation. The glue dispensing mechanism includes a glue dispensing head 6. The outer side of the head 6 is installed on the moving seat of the three-axis motion mechanism 4, the outer side of the dispensing head 6 is provided with a pressure regulating valve 8, the bottom of the dispensing head 6 is provided with a dispensing needle 7, the inner side of the equipment frame 1 is provided with a glue storage tank 13, 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, the outer side of the dispensing head 6 is provided with a visual mechanism, 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, the bottom of the industrial camera 9 is provided with a lens 10, the outer side of the industrial camera 9 is provided with a ring-shaped LED light 11, and the ring-shaped LED light 11 is located on the periphery of the lens 10.

[0024] 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, and the servo driver 5 will accurately control 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, and providing 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 products to be processed. The captured images are quickly transmitted to the image recognition and positioning unit in the dispensing path optimization system. The unit uses advanced image recognition algorithms to analyze and process product contours, feature points, etc. in the image. By comparing with the product template data pre-stored in the main control computer 2, the product position, size, and key information such as the starting point and end point of the dispensing are accurately calculated, thereby generating the optimal dispensing path, and the path data is transmitted back to the main control computer 2. The main control computer 2 generates detailed control instructions based on the received dispensing path data and the preset dispensing process parameters, such as the dispensing amount and dispensing speed, and sends them to the servo driver 5. According to the instructions, the servo driver 5 accurately adjusts the speed and direction of the stepper motor, drives the three-axis motion mechanism 4 to drive the dispensing head 6 to move accurately along the planned dispensing path. In the dispensing process, the pressure regulating valve 8 plays a key role. It will adjust the glue pressure inside the dispensing head 6 in real time according to the pressure control instructions sent by the main control computer 2, ensuring that the glue extracted from the glue storage tank 13 through the conduit 12 can be squeezed out from the dispensing needle 7 at a stable and appropriate flow rate, ensuring the uniformity and accuracy of the dispensing. For example, for high-precision dispensing tasks, the pressure regulating valve 8 can keep the glue pressure control accuracy within the range of ±0.05MPa. At the same time, the industrial camera 9 will continue to capture images during the dispensing process and provide real-time feedback to the dispensing path optimization system. The system will compare and monitor the actual dispensing position with the preset path through image analysis. Once a deviation is found, the control instructions will be adjusted in time to ensure that the dispensing operation is always carried out along the optimal path, thereby improving the dispensing quality and production efficiency.

[0025] Embodiment 2: A binocular vision system is introduced to obtain stereo images through two cameras, enhance depth perception, and improve dispensing accuracy. A laser rangefinder is installed on the dispensing head to achieve real-time height measurement to avoid collision between the dispensing needle and the workpiece. A rotation axis (R axis) is added to enable the dispensing head to dispense glue 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.

[0026] Embodiment three: Please see attached Figure 4-Figure 8 , the embodiment of the present invention provides a dispensing path optimization system, which includes the following modules: high-precision positioning module: using laser ranging technology and high-resolution camera to collect images for precise positioning; intelligent vision module: equipped with light sensors to extract and match the feature points of the product; parameter adaptive module: the product type recognition unit uses deep learning algorithms to automatically identify the product type and quickly adjust the dispensing parameters according to the characteristics of various products; system integration and control module: let each module work together to ensure the overall operation and control of the system. The high-precision positioning module includes the following units: laser positioning unit: using laser ranging technology to accurately measure the product position, with a positioning accuracy of within ±0.01mm; image recognition positioning unit: through high-resolution camera to collect images, combined with image recognition algorithms, to identify the product contour, and further improve the positioning accuracy. The intelligent vision module includes the following units: ambient light adaptive unit: equipped with light sensors, real-time monitoring of ambient light intensity, and automatic adjustment of camera exposure parameters; feature extraction and matching unit: extract and match the feature points of the product, and accurately identify 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: Use laser scanning or image measurement technology to detect product size in real time, and 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. Assume that the input image is , the convolution kernel is , the output feature map is , then the convolution operation formula is: ; 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; 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: ; in, is the position in the output feature map, and the pooling window size is ; 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: ; in, is the activation function, and the formula is ; 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: ; 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.

[0027] Specifically, the laser positioning unit uses laser ranging technology to emit a laser beam to the product and receive reflected light. By accurately calculating the laser round-trip time, the distance between the product and the equipment is calculated, thereby accurately measuring the product position. The effect is to control the positioning accuracy within ±0.01mm, greatly reducing the deviation of the dispensing position, ensuring that the dispensing can accurately fall on the required position of the product, effectively avoiding quality problems such as dispensing offset and leakage caused by inaccurate positioning, and providing a stable and reliable position basis for subsequent dispensing operations. The image recognition positioning unit uses a high-resolution camera to collect product images in all directions, and uses advanced image recognition algorithms such as edge detection and template matching to identify the product contour. The algorithm compares and analyzes the collected image with the pre-stored standard contour template to accurately identify the actual position and posture of the product. This unit further improves the positioning accuracy and complements the laser positioning unit. It can adapt to the positioning needs of complex-shaped products, improve the stability and reliability of positioning, and ensure the accuracy of dispensing path planning. The light sensor equipped with the ambient light adaptive unit monitors the changes in the ambient light intensity in real time. Once the light intensity fluctuation is detected, it quickly feeds back the signal to the camera control system and automatically adjusts the camera exposure parameters, such as shutter speed, aperture size, etc. It can effectively avoid the camera imaging being too bright or too dark due to changes in ambient light, such as the opening and closing of workshop lights, interference from external natural light, etc., which leads to misjudgment, ensuring that the visual system can stably obtain clear and accurate product images under different lighting conditions, and provide 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 corners, holes, textures, etc. Under different lighting conditions, the extracted feature points are described by unique feature descriptions and matched with the product features stored in the database. The unit ensures that the product can be accurately identified. Even if there are slight stains, wear or uneven lighting on the product surface, the key parts of the product can be accurately located, the reliability of dispensing can be improved, and the dispensing position can be highly consistent with the actual needs of the product, thereby improving the dispensing quality of the product. As the core hub of the entire system, the central control unit is responsible for receiving data information transmitted by each module and issuing instructions to each module based on the preset control logic and algorithm. For example, according to the product position information fed back by the high-precision positioning module and the product recognition results provided by the intelligent vision module, the parameter adaptive module is coordinated to adjust the dispensing parameters, and the three-axis motion mechanism is controlled to drive the dispensing head to move along the planned path. It realizes the overall operation and control of the system, ensures the coordinated work of each module, makes the entire dispensing process proceed in an orderly manner, and improves the system operation efficiency and stability. The data storage and analysis unit stores various data in the dispensing process in real time, including dispensing parameters, product quality data, and the operating status data 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 best combination of dispensing process parameters; tracing the root cause of the problem based on product quality data to provide a scientific basis for optimizing the dispensing process, helping to continuously improve the dispensing process and enhance product quality and production efficiency.

[0028] Working principle: First, start the main control computer 2 and calibrate the positions of each component. Then, the ring LED lamp 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 to move the dispensing head 6 according to the control instruction, and draws glue from the glue storage tank 13 through the conduit 12 to complete the dispensing action by the dispensing needle 7.

[0029] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent glue dispensing machine based on visual feedback, comprising a device frame (1), characterized in that: A main control computer (2) is installed inside the equipment frame (1); a dispensing path optimization system is installed on the equipment frame (1) and is electrically connected to the main control computer (2); a base (3) is fixedly connected to the inner bottom wall of the equipment frame (1); a three-axis motion mechanism (4) is installed on the top of the base (3); the three-axis motion mechanism (4) comprises an X-axis linear guide rail, a Y-axis linear guide rail and a Z-axis linear guide rail, and all are 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.

2. The intelligent glue dispensing machine based on visual feedback according to claim 1 is characterized in that: The glue dispensing mechanism comprises a glue dispensing head (6), the outer side of the glue dispensing head (6) being mounted on a movable seat of the three-axis motion mechanism (4), the outer side of the glue dispensing head (6) being provided with a pressure regulating valve (8), and the bottom of the glue dispensing head (6) being provided with a glue dispensing needle (7).

3. The intelligent glue dispensing machine based on visual feedback according to claim 2 is characterized in that: A glue storage tank (13) is arranged on the inner side of the equipment frame (1), a conduit (12) is fixedly connected to the top of the glue storage tank (13), one end of the conduit (12) is connected to the outer side of the glue dispensing head (6), and a visual mechanism is installed on the outer side of the glue dispensing head (6), and the visual mechanism is used to assist the glue dispensing operation.

4. The intelligent glue dispensing machine based on visual feedback according to claim 3 is characterized in that: The visual mechanism comprises an industrial camera (9), the industrial camera (9) being mounted on the outside of the dispensing head (6), a lens (10) being arranged at the bottom of the industrial camera (9), and an annular LED light (11) being mounted on the outside of the industrial camera (9), and the annular LED light (11) being located on the periphery of the lens (10).

5. The dispensing path optimization system is 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 optical sensors to extract and match product feature points; Parameter Adaptive 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.

6. The dispensing path optimization system according to claim 5, characterized in that: The high-precision positioning module includes the following units: Laser positioning unit: using laser ranging technology to accurately measure the product position, with a positioning accuracy within ±0.01mm; Image recognition and positioning unit: It collects images through a high-resolution camera and combines it with an image recognition algorithm to identify the product outline and further improve positioning accuracy.

7. The dispensing path optimization system according to claim 5, characterized in that: 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 the feature points of products and can accurately identify products under different lighting conditions.

8. The dispensing path optimization system according to claim 5, characterized in that: The parameter adaptation module includes the following units: Product type identification unit: Automatically identify 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: Using laser scanning or image measurement technology, it can detect product size in real time and automatically adjust the dispensing path and parameters to adapt to products of different sizes.

9. The dispensing path optimization system according to claim 5, 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 analyzes them to provide a basis for optimizing the dispensing process.

10. The dispensing path optimization system according to claim 8, characterized in that: In the product type recognition unit, the convolution layer: performs convolution operation by sliding the convolution kernel on the image to extract the local features of the image. Assume that the input image is , the convolution kernel is , the output feature map is , then the convolution operation formula is: ; 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; 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: ； in, is the position in the output feature map, and the pooling window size is ; 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: ; in, is the activation function, and the formula is ; 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: ； 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.

Images