A collaborative control method and system for an acoustic module injection molding production line

Abstract

This invention relates to the field of collaborative control technology for injection molding production lines, and particularly to a collaborative control method and system for an acoustic module injection molding production line. The method includes the following steps: calculating the illumination unevenness index of the ROI region based on the standard deviation of pixel grayscale values ​​and the interquartile range within the ROI region; determining the gradient feature confidence level of each ROI region based on the illumination unevenness index; obtaining the texture pose result and gradient pose result of the ROI region; weightedly fusing the texture pose result and gradient pose result of the ROI region to obtain the final pose; calculating the difference between the final pose of the ROI region and the standard target pose to obtain the pose error vector of the square key in the ROI region; and responding to the pose error vector, controlling the robotic arm in the production line to perform the mold insertion operation in the next cycle, effectively improving the accuracy of insert pose detection, thereby effectively improving the collaborative control efficiency of the acoustic module injection molding production line.

Description

Technical Field

[0001] This invention relates to the field of collaborative control technology for injection molding production lines, and in particular to a collaborative control method and system for an acoustic module injection molding production line. Background Technology

[0002] As a core sound-generating unit in modern consumer electronics products, acoustic modules require extremely high dimensional accuracy and structural consistency in their manufacturing. Before injection molding, prefabricated square keys need to be implanted into specific cavities of the mold using a robotic arm; this process is called mold insertion. The accuracy of mold insertion directly determines the structural integrity of the subsequently injection-molded parts and the acoustic performance of the final product. To ensure the quality of mold insertion, the position and orientation of the insert within the mold must be confirmed in real time before mold closing.

[0003] Currently, in multi-cavity injection molding scenarios, vision inspection systems need to simultaneously image and locate multiple inserts within the mold. However, mold cavities are typically made of highly reflective metallic materials, and their complex curved surfaces produce strong, uneven specular reflections and shadows under industrial lighting. Traditional vision inspection algorithms, such as template matching methods based on normalized cross-correlation (NCC), heavily rely on the consistency of grayscale texture between the target and the template. When the lighting inside the mold cavity is uneven and reflections are strong, the texture features of the insert surface are washed away by highlight areas or submerged by shadow areas, leading to a significant decrease in template matching scores and large deviations or even failures in the positioning results. If the actual pose error of each insert cannot be accurately assessed, the subsequent collaborative control system will make decisions based on erroneous or unreliable measurement information, ultimately resulting in low product yield and reduced production efficiency.

[0004] Therefore, overcoming the interference of complex lighting environment in multi-cavity molds and achieving accurate detection of the position and orientation of implanted inserts is a technical problem that needs to be solved to achieve efficient collaborative control of acoustic module injection molding production lines. Summary of the Invention

[0005] To address the technical challenge of overcoming interference from the complex lighting environment within multi-cavity molds and achieving accurate detection of the position and orientation of implanted inserts, this invention provides a collaborative control method and system for an acoustic module injection molding production line.

[0006] In a first aspect, the present invention provides a collaborative control method for an acoustic module injection molding production line, employing the following technical solution:

[0007] A collaborative control method for an acoustic module injection molding production line, comprising the following steps:

[0008] Real-time images of a multi-cavity mold containing multiple square key inserts are acquired, and the Regions of Interest (ROIs) corresponding to each square key insert are extracted. The illumination unevenness index of the ROI is calculated; the illumination unevenness index is positively correlated with the standard deviation and interquartile range of the pixel grayscale values ​​within the corresponding ROI. Based on the illumination unevenness index, the gradient feature confidence score of each ROI is determined; the gradient feature confidence score is positively correlated with the illumination unevenness index. Texture pose and gradient pose results of the ROI are obtained. The texture pose and gradient pose results of the ROI are weighted and fused to obtain the final pose, where the gradient weight of the gradient pose result is positively correlated with the gradient feature confidence score of the ROI. The difference between the final pose of the ROI and the standard target pose is calculated to obtain the pose error vector of the square key in the ROI. In response to the pose error vector, the robot arm in the production line is controlled to perform the mold-planting operation in the next cycle.

[0009] This invention enables collaborative control of an acoustic module injection molding production line by acquiring the pose of a robotic arm during the production line cycle for mold insertion. During pose acquisition, this invention considers that texture features may be damaged under complex lighting conditions such as high mold reflectivity and shadows, leading to reduced positioning accuracy or failure. Therefore, this invention constructs an illumination non-uniformity index to evaluate the imaging quality of each cavity. Based on this index, this invention dynamically determines the gradient feature confidence level and adaptively weights and fuses the two positioning results: texture pose and gradient pose. This dual-path adaptive fusion strategy relies on high-precision texture information under good lighting conditions and on more robust gradient information under poor lighting conditions, thus obtaining stable and high-precision insert poses under various complex lighting conditions. This solves the problem of poor positioning robustness in high-reflectivity environments in existing technologies, providing a reliable basis for subsequent precise control of the robotic arm and effectively improving the accuracy of mold insertion and production yield.

[0010] According to the collaborative control method for an acoustic module injection molding production line provided by the present invention, the method further includes, before acquiring real-time images of a multi-cavity mold containing multiple square key inserts and extracting the ROI regions corresponding to each square key insert, acquiring real-time images of the mold containing multiple square key inserts in the current loop using an industrial camera. The multi-cavity images of the key insert are obtained, and the multi-cavity images are preprocessed to obtain real-time images of the multi-cavity mold.

[0011] This invention uses an industrial camera to acquire and preprocess images in real time, providing a clear and standardized foundation of raw image data for subsequent calculation of illumination non-uniformity and pose analysis.

[0012] According to the collaborative control method for an acoustic module injection molding production line provided by the present invention, the calculation of the illumination non-uniformity index of the ROI region includes:

[0013] ;

[0014] For the first Light unevenness index for each ROI region For the first The third quartile of the gray value distribution of each ROI region For the first The first quartile of the grayscale value distribution of each ROI region For the first The standard deviation of grayscale values ​​for each ROI region grayscale value The maximum value in, This is a hyperparameter.

[0015] This invention provides a precise method for calculating illumination unevenness index. By combining interquartile range and standard deviation, which have high resistance to outliers, the accuracy of illumination unevenness assessment results is improved, thereby truly and accurately reflecting the degree of grayscale dispersion in the core area caused by highlights and shadows.

[0016] According to the collaborative control method for an acoustic module injection molding production line provided by the present invention, the step of determining the gradient feature confidence level of each ROI region includes:

[0017] ;

[0018] For the first Gradient feature confidence of each ROI region This is the system proportionality coefficient. For the first Light unevenness index for each ROI region It is the natural logarithm function.

[0019] According to the collaborative control method of an acoustic module injection molding production line provided by the present invention, the method for obtaining the gradient weight of the gradient pose result includes:

[0020] ;

[0021] For the first Gradient weights for each ROI region For the first Gradient feature confidence of each ROI region The confidence level of the gradient features in the ROI region. The minimum confidence level of the gradient feature across all ROI regions. This represents the maximum confidence level of the gradient features across all ROI regions.

[0022] This invention provides a precise gradient weight calculation method. By normalizing the confidence level of a single cavity within the confidence level range of all cavities in the current batch, the influence of dimensions is effectively eliminated, so that the weight can not only reflect the absolute illumination in a single cavity, but also its relative illumination level in the entire batch.

[0023] According to the collaborative control method of an acoustic module injection molding production line provided by the present invention, the step of obtaining the texture pose result and gradient pose result of the ROI region includes: performing normalized cross-correlation (NCC) matching between the standard square key template and the ROI region to obtain the texture pose result; performing edge detection on the ROI region to obtain the edge image; and matching the standard square key contour model with the edge image to obtain the gradient pose result.

[0024] According to the collaborative control method of an acoustic module injection molding production line provided by the present invention, the step of weightedly fusing the texture pose result and gradient pose result of the ROI region to obtain the final pose includes:

[0025] ;

[0026] For the first The final pose of each ROI region For the first Gradient weights for each ROI region For the first Texture pose results for each ROI region For the first Gradient pose results for each ROI region.

[0027] This invention provides a precise final pose calculation method. By analyzing the unevenness of illumination and dynamically adjusting the contribution ratio of texture pose and gradient pose, a smooth transition between the two complementary poses is achieved. This ensures that, regardless of the illumination conditions, the final output pose is dominated by the most reliable information, thereby improving the accuracy of positioning.

[0028] According to the present invention, a collaborative control method for an acoustic module injection molding production line is provided, wherein the control of the molding operation of the robotic arm in the next cycle includes: taking the average pose error vector of the square keys in all ROI regions of the real-time image of the multi-cavity mold in the current cycle as the molding correction signal of the current cycle; substituting the molding correction signal into an iterative learning control algorithm to calculate the global pose correction vector of the robotic arm in the next cycle; and generating motion control commands for the end effector of the robotic arm based on the global pose correction vector and the initial ideal pose of the robotic arm.

[0029] According to the present invention, a collaborative control method for an acoustic module injection molding production line, wherein the iterative learning control algorithm includes:

[0030] ;

[0031] For the first The global pose correction vector of the robotic arm in the loop. For the first The global pose correction vector of the robotic arm in the loop. For the proportional learning matrix, For the first The implantation correction signal of the robotic arm in the loop, For the differential learning matrix, For the first The molding correction signal of the robotic arm in the loop.

[0032] Secondly, the present invention provides a collaborative control system for an acoustic module injection molding production line, which adopts the following technical solution:

[0033] A collaborative control system for an acoustic module injection molding production line includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned collaborative control method for an acoustic module injection molding production line is implemented.

[0034] By adopting the above technical solution, a computer program is generated from the above-mentioned collaborative control method for an acoustic module injection molding production line, and stored in a memory for loading and execution by a processor. This allows for the creation of terminal equipment based on the memory and processor, making it convenient to use.

[0035] The present invention has the following technical effects:

[0036] Based on the above technical solution, this invention provides a collaborative control method and system for an acoustic module injection molding production line. By acquiring the pose of the robotic arm during the production line cycle for mold placement, collaborative control of the acoustic module injection molding production line can be achieved. During the pose acquisition process, this invention considers that texture features may be damaged under complex lighting conditions such as high mold reflectivity and shadows, leading to reduced positioning accuracy or failure. Therefore, this invention constructs an illumination non-uniformity index to evaluate the imaging quality of each cavity. Based on this index, this invention dynamically determines the gradient feature confidence level and adaptively weights and fuses the two positioning results: texture pose and gradient pose. This dual-path adaptive fusion strategy can rely on high-precision texture information under good lighting conditions and rely more on robust gradient information under poor lighting conditions, thus obtaining stable and high-precision insert poses under various complex lighting conditions. This solves the problem of poor positioning robustness in high-reflectivity environments in existing technologies, providing a reliable basis for the subsequent precise control of the robotic arm and effectively improving the accuracy of mold placement and production yield. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating a collaborative control method for an acoustic module injection molding production line, as provided in an embodiment of the present invention. Detailed Implementation

[0038] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0039] It should be noted that the production process of the acoustic module injection molding production line can be realized by two types of implantation machines, which can be referred to as the first implantation machine and the second implantation machine in this embodiment of the invention. The first implantation machine produces 8 square keys per cycle, and the second implantation machine produces 32 square keys per cycle. The production process includes five main parts: pre-processing and feeding, material unloading and molding, visual inspection, injection molding and shell removal, and collection and inspection.

[0040] Specifically, during pre-processing, the material is first fed or cut, and the square key is transferred. In the material picking and molding cycle, the robotic arm of the implantation machine will send the picked square key to the loading platform and place the square key into the mold cavity for molding. After molding, the position of the square key is confirmed by visual inspection above the mold. Then, the mold is closed, injection is performed, and the shell is automatically removed. Finally, the collected square key products are inspected for quality by visual inspection. This cycle continues until the injection molding production of all components is completed.

[0041] Therefore, the accuracy of visual inspection during implantation is fundamental to the entire collaborative control system. If the errors caused by the highly reflective mold cavity in existing technologies are not addressed, the implantation machine's pose error may lead to the failure of the robotic arm's pose correction, or even worsen the overall error.

[0042] Based on this, this invention discloses a collaborative control method for an acoustic module injection molding production line. For details, please refer to [link / reference needed]. Figure 1 As shown, Figure 1 This is a flowchart illustrating a collaborative control method for an acoustic module injection molding production line, provided by an embodiment of the present invention. The method specifically includes the following steps:

[0043] S1: Obtain real-time images of a multi-cavity mold containing multiple square key inserts, and extract the ROI regions corresponding to each square key insert.

[0044] For example, in the current loop, images containing all... are captured by an industrial camera. After preprocessing the real-time images of the multi-cavity components of each square key embedding, a multi-cavity image is obtained; the ROI regions corresponding to each square key in the multi-cavity image are extracted.

[0045] Specifically, after the robotic arm completes the task... After the key insert is implanted into the multi-cavity mold, before mold closing, an industrial camera deployed above the mold takes a picture containing all the components. The multi-cavity real-time image of each cavity is processed to obtain the pixels in the multi-cavity image. The preprocessing includes: converting the real-time multi-cavity image from the camera to grayscale after transmission to the vision processing unit; verifying the integrity of the image data frames transmitted via the high-speed interface; and performing frame counting verification to ensure no missing or duplicate frames. By pre-calibrating and real-time segmenting the Region of Interest (ROI) in the multi-cavity image, the processing task of a single image can be decomposed into multiple parts. Each region of interest (ROI) is isolated. Random noise in the image is removed using a median filter, and illumination correction or contrast enhancement is performed on the ROI regions.

[0046] in, It can be 8 or 32. The preprocessing method can be selected according to actual needs. The preprocessing steps can be implemented by existing technology, and will not be described in detail here in the embodiments of the present invention.

[0047] For example, the ROI region can be extracted by using the preset coordinates of the center point of the cavity of the key and the fixed cavity size. This can be achieved using existing technologies, and will not be elaborated on in this embodiment of the invention.

[0048] Thus, by extracting the ROI region, the embodiments of the present invention enable subsequent processing of the cavity region of the square key insert to be detected, thereby reducing background interference and data processing volume.

[0049] S2: Calculate the illumination unevenness index of the ROI region. The illumination unevenness index is positively correlated with the standard deviation of the pixel grayscale value and the interquartile range of the corresponding ROI region.

[0050] It's important to note that in the injection molding production environment, the high-brightness reflections and shadows inside the cavity are the core defects that cause a decrease in pose estimation accuracy based on traditional texture matching. Without assessing the severity of lighting conditions, the reliability of texture features cannot be determined. Therefore, it is necessary to construct an illumination non-uniformity index to quantitatively describe the lighting environment. Furthermore, the coexistence of extremely bright highlights and extremely dark shadows in the ROI region leads to a highly discrete state of pixel grayscale values.

[0051] Furthermore, the outliers caused by extreme brightness and darkness are essentially due to uneven illumination. The interquartile range (IQR) of grayscale within the ROI region, obtained by subtracting the first IQR from the third IQR, exhibits higher noise resistance and resilience against individual highlight outliers, better reflecting the concentration of grayscale in the core area. When uneven illumination causes grayscale distribution stretching, the IQR increases.

[0052] Based on this, embodiments of the present invention can construct an illumination unevenness index by combining the normalized interquartile range and normalized standard deviation in the current ROI region.

[0053] Specifically, when constructing the illumination unevenness index for the current ROI region, the grayscale values ​​of the pixels in the current ROI region can be obtained first and sorted from smallest to largest. This yields the first and third quartiles of the grayscale value distribution in the current ROI region. The first quartile is located at the 25th percentile of the sort, and the third quartile is located at the 75th percentile. The third quartile is greater than the first quartile. The difference between the third and first quartiles is taken as the grayscale interquartile range of the current ROI region. The illumination unevenness index of the current ROI region is obtained by summing the normalized value of the grayscale interquartile range and the normalized value of the grayscale standard deviation of the current ROI region.

[0054] For example, in an embodiment of the present invention, the illumination unevenness index of the ROI region is calculated, and the specific formula can be found in the following relationship:

[0055] ;

[0056] For the first Light unevenness index for each ROI region For the first The third quartile of the gray value distribution of each ROI region For the first The first quartile of the grayscale value distribution of each ROI region For the first The standard deviation of grayscale values ​​for each ROI region grayscale value The maximum value in, This is a hyperparameter.

[0057] in, This represents the grayscale value, which ranges from 0 to 255. The maximum value is 255. The hyperparameter is used to avoid the case where the denominator is 0 when the first quartile and the third quartile are equal. As an example, the hyperparameter can be set to 0.01.

[0058] In the above relation, It is the first Gray-level interquartile range of each ROI region This is the interquartile skewness coefficient, which is the normalized value of the interquartile range of gray levels. This value is used to assess the relative dispersion of the core 50% of gray-level data. The larger the value, the more dispersed the gray-level data in the middle 50% of the ROI region, the stronger the contrast between highlights and shadows, and the more severe the unevenness of illumination. The denominator is used to offset the excessive influence of extreme gray-level values ​​on the indicator, making the indicators of different ROI regions comparable.

[0059] The standard deviation of grayscale reflects the overall dispersion of grayscale values ​​within the ROI region. If the... When the illumination is uniform within each ROI region, the grayscale values ​​of all pixels tend to be consistent, close to the mean, thus the standard deviation is small, the overall dispersion is smaller, and the illumination unevenness index is smaller; conversely, if the illumination is uneven within the ROI region, the standard deviation is smaller, the overall dispersion is smaller, and the illumination unevenness index is smaller. When there are strong reflective highlights and deep shadows within a ROI region, the range of pixel grayscale values ​​will widen, resulting in a larger standard deviation, a greater overall dispersion, and a larger illumination unevenness index.

[0060] The above steps can be used to obtain the illumination unevenness index for each ROI region. The larger the illumination unevenness index, the more severe the illumination unevenness and the less reliable the texture features.

[0061] S3: Based on the illumination unevenness index, determine the gradient feature confidence level of each ROI region. The gradient feature confidence level is positively correlated with the illumination unevenness index.

[0062] It should be noted that the above steps assess the severity of the lighting environment by obtaining the illumination unevenness index of the ROI region. Texture matching is highly accurate under uniform lighting but fails under uneven lighting; gradient matching is more reliable under poor lighting but suffers from blurred edges and insufficient accuracy under uniform lighting. Therefore, when the illumination unevenness is high, features based on pixel grayscale textures become unreliable, while edges formed by highlight or shadow boundaries—i.e., the gradient features of the image—become clearer and more stable. The higher the illumination unevenness, the lower the reliability of texture features, and the more reliant on gradient features becomes. Therefore, an index needs to be constructed to characterize the confidence level of gradient features under the current lighting conditions.

[0063] Based on this, embodiments of the present invention can determine the confidence level of gradient features according to the illumination unevenness index of the ROI region. The higher the illumination unevenness, the more reliable the gradient features and the larger the confidence level value.

[0064] For example, in an embodiment of the present invention, the gradient feature confidence of the ROI region is calculated, and the specific relationship can be found in the following formula:

[0065] ;

[0066] For the first Gradient feature confidence of each ROI region This is the system proportionality coefficient. For the first Light unevenness index for each ROI region It is the natural logarithm function.

[0067] The system scaling factor is used to adjust the sensitivity of the system scaling factor. It can be fixed by the dynamic range and image bit depth of the industrial camera. The specifics can be obtained through existing technologies, which will not be elaborated here.

[0068] In the above relationship, the gradient feature confidence is a monotonically increasing function of the illumination unevenness index. When the illumination is uniform and the illumination unevenness index approaches 0, the gradient feature confidence also approaches 0, indicating that the confidence of the gradient feature is low at this time. As the illumination unevenness index increases, the more uneven the illumination, the greater the gradient feature confidence, indicating that the gradient feature becomes increasingly reliable.

[0069] The logarithmic function is used to reflect a diminishing marginal effect: when the lighting conditions are already extremely poor, further deterioration will reduce its contribution to improving the relative reliability of the gradient features.

[0070] The gradient feature confidence level of each ROI region can be obtained by following the steps above.

[0071] S4: Obtain the texture pose and gradient pose results of the ROI region; perform weighted fusion of the texture pose and gradient pose results of the ROI region to obtain the final pose, wherein the gradient weight of the gradient pose result is positively correlated with the confidence of the gradient feature of the ROI region.

[0072] It should be noted that after obtaining the gradient feature confidence score, the ultimate goal is to calculate the precise pose of the square key. This invention does not employ a single detection algorithm, but rather designs an adaptive fusion strategy. First, by calculating the texture pose and gradient pose results in parallel, the gradient feature confidence score is transformed into a fusion weight, thereby dynamically weighting and fusing the two pose estimation results.

[0073] For example, texture pose results can be obtained by matching a standard square bond template with the region of interest using the traditional normalized cross-correlation (NCC) algorithm, resulting in a texture-based pose estimation. This method has high accuracy under uniform illumination.

[0074] For example, when obtaining gradient pose results, edge detection can be performed on the region of interest, and then an edge contour-based matching algorithm can be used to match the standard contour model of the key with the detected edge image to obtain a gradient-based pose estimation result. This method has high accuracy when the edges are clear.

[0075] The texture pose and gradient pose results described above are both pose vectors containing position and angle. Edge detection can be performed using edge detection algorithms. Preferably, in this embodiment of the invention, the Canny operator can be used for edge detection, and the Chamfer Matching algorithm can be used for edge contour matching. The specific steps for obtaining the above two pose estimation results can be implemented using existing technologies, and will not be elaborated here.

[0076] Specifically, when obtaining weights, one can perform a weighting of the current batch. The gradient feature confidence of each ROI region is normalized to obtain the gradient weight of each ROI region.

[0077] It should be understood that gradient feature confidence only reflects the reliability of gradient features in a single ROI region. However, pose fusion requires a clear understanding of the proportion of gradient pose results to texture pose results. Therefore, the gradient weights are calculated through the following steps to determine the contribution of the two pose results. Furthermore, in the same production cycle, the lighting conditions differ in different ROI regions of a multi-cavity mold, resulting in variations in their corresponding gradient feature confidence.

[0078] Based on this, when calculating the gradient weight of each ROI region, the embodiments of the present invention can use the extreme value of the confidence of the gradient features in the same batch as a benchmark to convert a single confidence into a relative proportion. While preserving the relative differences of each ROI region in the same batch, it can also eliminate the interference of dimensions.

[0079] For example, in an embodiment of the present invention, the gradient weights of the ROI region are calculated, and the specific relationship can be found in the following formula:

[0080] ;

[0081] For the first Gradient weights for each ROI region For the first Gradient feature confidence of each ROI region The confidence level of the gradient features in the ROI region. The minimum confidence level of the gradient feature across all ROI regions. This represents the maximum confidence level of the gradient features across all ROI regions.

[0082] In the gradient weight formula, the confidence level is converted into a weight ratio through a normalization operation. For the ROI region with the worst lighting conditions in the current batch, the gradient feature confidence level is the highest, and the lighting unevenness index is the highest; for the square bond cavity with the best lighting conditions in the current batch, the gradient feature confidence level is the lowest, and the lighting unevenness index is the lowest.

[0083] Right now The larger the value, the more dispersed the weight distribution of each ROI region in the same batch; The smaller the value, the more concentrated the weight distribution, indicating that the differences in lighting conditions among the ROI regions within the same batch are smaller. equal In this case, you can directly set the gradient weight to 0.5.

[0084] Finally, by applying this weight to the calculation of the final pose, the final pose of the ROI region can be obtained.

[0085] For example, in an embodiment of the present invention, the final pose of the ROI region is calculated, specifically from the following relationship:

[0086] ;

[0087] For the first The final pose of each ROI region For the first Gradient weights for each ROI region For the first Texture pose results for each ROI region For the first Gradient pose results for each ROI region.

[0088] The final pose relation is a dynamic weighted average process. When the... When the illumination is uniform in each ROI region, the illumination non-uniformity index With a relatively small gradient weight approaching 0, the final pose is primarily determined by the high-precision texture pose; when there is uneven illumination and severe reflection, the uneven illumination index... With a larger gradient, the gradient weights approach 1, and the final pose is mainly determined by the more reliable gradient pose. For lighting conditions between these two, this formula can achieve a smooth transition and complementary advantages between the two methods.

[0089] The final pose of each ROI region can be obtained by following the above steps, that is, the pose of the square key in the cavity. By comparing the final pose of the ROI region with the standard target pose, the pose error vector of the square key in each ROI region can be obtained.

[0090] For example, in an embodiment of the present invention, obtaining the pose error vector of the square key in the ROI region includes: using the difference between the final pose of the ROI region and the standard target pose as the pose error vector of the square key in the ROI region.

[0091] The pose error vector includes position error and angle error. The standard target pose is the ideal pose of the ROI region. After this set of high-precision error vectors is sent to the central collaborative controller of the production line, the controller can control the robotic arm based on the error information to achieve module production collaboration, that is, continue to execute the following steps.

[0092] S5: Calculate the difference between the final pose of the ROI region and the standard target pose to obtain the pose error vector of the square key in the ROI region; in response to the pose error vector, control the robot arm in the production line to perform the mold planting operation in the next cycle.

[0093] It should be noted that the above steps obtain the actual pose of the key ROI region in each cavity. The ultimate goal is to control the robot to accurately place the mold. If this error is not fed back to the robot, the robot will continue to place the mold in the standard pose, resulting in a decrease in yield and accuracy.

[0094] Based on this, embodiments of the present invention can adjust the pose of the robotic arm according to the pose error vector of each ROI region, thereby improving the injection molding production efficiency of acoustic modules.

[0095] For example, when adjusting the robotic arm based on the pose error vector, the average pose error vector of the square keys in all ROI regions can be used as the modeling correction signal for the robotic arm in the current loop; the pose of the robotic arm is adjusted according to the modeling correction signal to achieve control of the robotic arm.

[0096] Among them, the implantation correction signal is the expected pose error vector of the end effector of the robot arm in the world coordinate system in the current cycle. By obtaining the average pose error of all ROI regions, the systematic deviation of the implantation robot arm in the current cycle can be obtained, so as to make correction in the next cycle.

[0097] Specifically, when adjusting the pose of the robotic arm based on the implantation correction signal, the implantation correction signal of the robotic arm in the current cycle can be obtained. The implantation correction signal is then substituted into the iterative learning control algorithm to obtain the global pose correction vector that the implantation robotic arm needs to apply in the next cycle.

[0098] Preferably, according to the current number Calculate the next iteration in a loop. The global pose correction vector for the robotic arm can be found in the following formula:

[0099] ;

[0100] For the first The global pose correction vector of the robotic arm in the loop. For the first The global pose correction vector of the robotic arm in the loop. For the proportional learning matrix, For the first The implantation correction signal of the robotic arm in the loop, For the differential learning matrix, For the first The molding correction signal of the robotic arm in the loop.

[0101] Among them, the The global pose correction vector of the robotic arm in the loop is the first vector calculated according to the above formula. The global pose correction vector obtained by iterative calculation will not be described in detail in this embodiment of the invention.

[0102] Both the proportional learning matrix and the differential learning matrix are 3×3 diagonal matrices. The proportional learning matrix is ​​used to correct the current error, and its x and y axis values ​​can be set to 0.6. The x-axis can be set to 0.4; the differential learning matrix is ​​used to correct for error trends and enhance convergence speed, and its x and y axis values ​​can be set to 0.25. The axis can be set to 0.15, and the specific setting can be adjusted according to actual needs.

[0103] Used to inherit the first The system performs corrections in a loop and makes proportional corrections based on the current error to continuously approximate the constant error of the system.

[0104] It is the error differential term, used to predict and suppress the trend of error change.

[0105] Thus, according to the above steps, the global pose correction vector of the robotic arm in the next cycle can be obtained in the embodiment of the present invention. The global pose correction vector is the final command value finally applied to the end effector of the robotic arm. By obtaining the difference between the initial ideal pose of the robotic arm at the center of the mold and the global pose correction vector, the control signal of the final motion pose of the end effector of the robotic arm in the world coordinate system can be generated.

[0106] For example, in an embodiment of the present invention, implementing robotic arm control includes: obtaining the initial ideal pose of the robotic arm, and using the difference between the initial ideal pose and the global pose correction vector as a control signal to implement robotic arm control.

[0107] The initial ideal pose of the robotic arm can be obtained through hand-eye calibration. The specific steps can be implemented using existing technologies, and will not be elaborated here in the embodiments of the present invention.

[0108] As can be seen, in this embodiment of the invention, when realizing the collaborative control of the acoustic module injection molding production line, real-time images of a multi-cavity mold containing multiple square key inserts can be acquired, and the ROI regions corresponding to each square key insert can be extracted; the illumination unevenness index of the ROI region can be calculated, and the illumination unevenness index is positively correlated with the standard deviation and interquartile range of the pixel grayscale values ​​in the corresponding ROI region; based on the illumination unevenness index, the gradient feature confidence of each ROI region can be determined, and the gradient feature confidence is positively correlated with the illumination unevenness index; the texture pose results of the ROI region can be obtained and Gradient pose results; The texture pose results and gradient pose results of the ROI region are weighted and fused to obtain the final pose, where the gradient weight of the gradient pose result is positively correlated with the gradient feature confidence of the ROI region; The difference between the final pose of the ROI region and the standard target pose is calculated to obtain the pose error vector of the square key in the ROI region; In response to the pose error vector, the robot arm in the production line is controlled to perform the mold insertion operation in the next cycle, which effectively improves the accuracy of insert pose detection, thereby effectively improving the collaborative control efficiency of the acoustic module injection molding production line.

[0109] This invention also discloses a collaborative control system for an acoustic module injection molding production line, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a collaborative control method for an acoustic module injection molding production line provided by this invention.

[0110] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0111] In this invention, the aforementioned memory can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0112] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A collaborative control method for an acoustic module injection molding production line, characterized in that, include: Acquire real-time images of a multi-cavity mold containing multiple square key inserts, and extract the ROI regions corresponding to each square key insert; The illumination unevenness index of the ROI region is calculated. The illumination unevenness index is positively correlated with the standard deviation of the pixel gray value and the interquartile range of the corresponding ROI region. Calculate the illumination non-uniformity index of the ROI region, including: ； For the first Light unevenness index for each ROI region For the first The third quartile of the gray value distribution of each ROI region For the first The first quartile of the grayscale value distribution of each ROI region For the first The standard deviation of grayscale values ​​for each ROI region grayscale value The maximum value in, For hyperparameters; Based on the illumination unevenness index, the gradient feature confidence level of each ROI region is determined. The gradient feature confidence level is positively correlated with the illumination unevenness index. The determination of the gradient feature confidence level of each ROI region includes: ； For the first Gradient feature confidence of each ROI region This is the system proportionality coefficient. For the first Light unevenness index for each ROI region It is the natural logarithm function; Obtain the texture pose and gradient pose results for the Region of Interest (ROI); perform weighted fusion of the texture pose and gradient pose results for the ROI to obtain the final pose, where the gradient weights of the gradient pose results are positively correlated with the confidence level of the gradient features of the ROI region; the methods for obtaining the gradient weights of the gradient pose results include: ； For the first Gradient weights for each ROI region For the first Gradient feature confidence of each ROI region The confidence level of the gradient features in the ROI region. The minimum confidence level of the gradient feature across all ROI regions. The maximum confidence level of the gradient feature across all ROI regions; Calculate the difference between the final pose of the ROI region and the standard target pose to obtain the pose error vector of the square key in the ROI region; respond to the pose error vector, control the robot arm in the production line to perform the molding operation in the next cycle; weightedly fuse the texture pose result and gradient pose result of the ROI region to obtain the final pose, including: ； For the first The final pose of each ROI region For the first Gradient weights for each ROI region For the first Texture pose results for each ROI region For the first Gradient pose results for each ROI region.

2. The collaborative control method for an acoustic module injection molding production line according to claim 1, characterized in that, The process of acquiring real-time images of a multi-cavity mold containing multiple square key inserts and extracting the ROI regions corresponding to each square key insert includes, prior to: In the current cycle, data is collected in real time via an industrial camera. The multi-cavity images of the key insert are obtained, and the multi-cavity images are preprocessed to obtain real-time images of the multi-cavity mold.

3. The collaborative control method for an acoustic module injection molding production line according to claim 1, characterized in that, The acquisition of texture pose and gradient pose results for the ROI region includes: Normalized cross-correlation (NCC) matching is performed between the standard square key template and the ROI region to obtain the texture pose result; edge detection is performed on the ROI region to obtain the edge image; and the standard square key contour model is matched with the edge image to obtain the gradient pose result.

4. The collaborative control method for an acoustic module injection molding production line according to claim 1, characterized in that, The control of the robotic arm in the production line for the molding operation in the next cycle includes: The mean of the pose error vectors of the square keys in all ROI regions of the real-time image of the multi-cavity mold in the current loop is used as the mold installation correction signal for the current loop. The mold installation correction signal is substituted into the iterative learning control algorithm to calculate the global pose correction vector of the robot arm in the next loop. Based on the global pose correction vector and the initial ideal pose of the robot arm, motion control commands for the end effector of the robot arm are generated.

5. The collaborative control method for an acoustic module injection molding production line according to claim 4, characterized in that, The iterative learning control algorithm includes: ； For the first The global pose correction vector of the robotic arm in the loop. For the first The global pose correction vector of the robotic arm in the loop. For the proportional learning matrix, For the first The implantation correction signal of the robotic arm in the loop, For the differential learning matrix, For the first The molding correction signal of the robotic arm in the loop.

6. A collaborative control system for an acoustic module injection molding production line, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a collaborative control method for an acoustic module injection molding production line according to any one of claims 1-5.

Images