Capacitor element stacking method and system

Abstract

The invention discloses a capacitor element stacking method and system, and relates to the technical field of metal spraying technology, and the method comprises the steps: obtaining an element image of a to-be-stacked capacitor, inputting the element image into a pre-trained capacitor detection model which comprises a feature extraction network and a detection network, performing feature extraction on the element image through a feature extraction network to obtain a capacitor feature map, performing identification positioning detection on the capacitor feature map by adopting a detection network to obtain an element model and an element coordinate corresponding to the to-be-stacked capacitor image, and controlling a feeding device corresponding to the element model to move to the element coordinate, and carrying out stacking operation on the to-be-stacked capacitors. According to an existing capacitor element stacking method, the capacitor elements are stacked in an element frame after being clamped by a feeding robot, but due to the fact that the capacitor elements are various in model, the error probability of the capacitor element stacking method clamped by the feeding robot is gradually increased, and the adaptability of the feeding robot to stack the capacitor elements is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of gold spraying technology, and in particular to a capacitor element stacking method and system. Background Art

[0002] With the development of flexible power transmission projects, civilian new energy devices, and high-energy-density, high-pulse-power military equipment, the application and demand of dry capacitors, especially high-end, high-voltage, and large-capacity capacitors, are also growing exponentially. Component spraying is an important process in the production of dry capacitors. In order to improve its processing efficiency, after the components are rolled, they are generally stacked on the component frame by hand, and then the entire frame of components is sprayed with gold. As labor costs continue to increase, there is an urgent need for a method to improve the efficiency of capacitor component stacking.

[0003] At present, the existing capacitor component stacking method mainly involves a loading robot clamping the component outer diameter and then placing it into a component frame. However, due to the diverse models of capacitor components and frequent production changes, the error probability of the loading robot clamping component stacking method is gradually increasing, reducing the adaptability of the loading robot to stack capacitor components. Summary of the invention

[0004] The present invention provides a capacitor element stacking method and system, which solves the technical problem that the existing capacitor element stacking method mainly relies on a loading robot to clamp the outer diameter of the element and then put it into a component frame. However, due to the diverse models of capacitor elements and frequent production changes, the error probability of the loading robot clamping the element stacking method is gradually increased, which reduces the adaptability of the loading robot to stack capacitor elements.

[0005] A first aspect of the present invention provides a capacitor element stacking method, comprising:

[0006] Acquire a component image of a capacitor to be stacked, and input the component image into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network;

[0007] Extracting features from the component image through the feature extraction network to obtain a capacitor feature map;

[0008] Using the detection network to perform identification and positioning detection on the capacitor characteristic image, and obtaining the component model and component coordinates corresponding to the capacitor image to be stacked;

[0009] The loading device corresponding to the component model is controlled to move to the component coordinate, and the capacitor to be stacked is stacked.

[0010] Optionally, the feature extraction network includes a 3×3 convolution layer, a 1×1 standard convolution layer, an extraction branch and a feature fusion layer, and the step of extracting features from the component image through the feature extraction network to obtain a capacitor feature map includes:

[0011] Extracting features from the component image through a 3×3 convolutional layer to obtain a first component feature map;

[0012] Performing feature extraction on the first component feature map through a 1×1 standard convolutional layer to obtain a second component feature map;

[0013] Performing feature extraction on the first component feature map through the extraction branch to obtain a third component feature map, wherein the extraction branch includes a first recognition module, a 2×2 maximum pooling layer, a first recognition module, a cascaded attention module, a second recognition module, and a 1×1 convolutional layer connected in sequence;

[0014] A feature fusion layer is used to perform feature fusion on the second element feature map and the third element feature map to obtain a capacitor feature map.

[0015] Optionally, the step of extracting features from the first component feature map by the extraction branch to obtain a third component feature map includes:

[0016] Extracting features from the first component feature map through a first recognition module, a 2×2 maximum pooling layer, and a first recognition module in sequence to obtain a first recognition feature map, wherein the first recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a Leaky-ReLU activation layer connected in sequence;

[0017] Using the cascaded attention module to extract features from the first recognition feature map to obtain a second recognition feature map, wherein the cascaded attention module includes a 1×1 convolution layer, a feature fusion layer, a feature fusion layer, a 3×3 convolution layer, and a feature fusion layer;

[0018] Extracting features from the second recognition feature map through a second recognition module to obtain a third recognition feature map, wherein the second recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a DY-ReLU activation layer connected in sequence;

[0019] The third identification feature map is subjected to feature extraction through a 1×1 convolutional layer to obtain a third component feature map.

[0020] Optionally, the step of extracting features from the first recognition feature map using the cascaded attention module to obtain a second recognition feature map includes:

[0021] Performing feature extraction on the first recognition feature map through a 1×1 convolutional layer to obtain a first feature map;

[0022] Performing feature extraction on the first recognition feature map through a 3×3 convolutional layer to obtain a second feature map;

[0023] Using a feature fusion layer to perform feature fusion on the first feature map and the first recognition feature map to obtain a first fused feature map;

[0024] Using a feature fusion layer to perform feature fusion on the first fused feature map and the first recognition feature map to obtain a second fused feature map;

[0025] A feature fusion layer is used to perform feature fusion on the second feature map and the second fused feature map to obtain a second recognition feature map.

[0026] Optionally, the step of extracting features from the second recognition feature map by a second recognition module to obtain a third recognition feature map includes:

[0027] Performing feature extraction on the second recognition feature map through a 3×3 convolutional layer to obtain a third feature map;

[0028] Performing a batch normalization operation on the third feature map through a batch normalization layer to obtain a fourth feature map;

[0029] A DY-ReLU activation layer is used to perform a nonlinear operation on the fourth feature map to obtain a third recognition feature map.

[0030] Optionally, the detection network includes a fully connected layer and a Softmax layer, and the step of using the detection network to identify and locate the capacitor characteristic map to obtain the component model and component coordinates corresponding to the capacitor image to be stacked includes:

[0031] Performing feature mapping on the capacitor feature map through a fully connected layer to obtain a component mapping feature map;

[0032] The Softmax layer is used to perform identification and positioning operations on the component mapping feature map to obtain the component model and component coordinates corresponding to the capacitor image to be stacked.

[0033] Optionally, the step of controlling the loading device corresponding to the component model to move to the component coordinate and stacking the capacitors to be stacked includes:

[0034] Taking the loading device corresponding to the component model as the target device, and controlling the target device to move from the stacking position to the component coordinates;

[0035] The target device is controlled to place the capacitor to be stacked at the stacking position.

[0036] A second aspect of the present invention provides a capacitor element stacking system, comprising:

[0037] An acquisition module, used for acquiring component images of capacitors to be stacked, and inputting the component images into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network;

[0038] An extraction module, used for performing feature extraction on the component image through the feature extraction network to obtain a capacitor feature map;

[0039] An identification module, used to use the detection network to perform identification and positioning detection on the capacitor characteristic image, and obtain the component model and component coordinates corresponding to the capacitor image to be stacked;

[0040] The stacking module is used to control the loading equipment corresponding to the component model to move to the component coordinates and perform stacking operations on the capacitors to be stacked.

[0041] A third aspect of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of any one of the above-mentioned capacitor element stacking methods.

[0042] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the capacitor element stacking method as described in any one of the above items.

[0043] It can be seen from the above technical solutions that the present invention has the following advantages:

[0044] The present invention obtains the component image of the capacitor to be stacked, inputs the component image into a pre-trained capacitor detection model, obtains the component model and component coordinates of the capacitor to be stacked, and then controls the loading device corresponding to the component model to move to the component coordinates to perform stacking operations on the capacitor to be stacked, thereby overcoming the technical problem that the existing capacitor component stacking method mainly uses a loading robot to clamp the component outer diameter and then put it into the component frame, but due to the variety of capacitor component models and frequent production changes, the error probability of the loading robot clamping component stacking method is gradually increasing. Compared with the traditional component stacking method, the present invention uses a pre-trained capacitor detection model to accurately identify the component model of the capacitor to be stacked, and then calls the loading device corresponding to the component model to perform stacking operations on the capacitor to be stacked, thereby improving the efficiency and adaptability of the loading robot stacking capacitor components. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0046] Figure 1 A flowchart of a capacitor element stacking method provided in Embodiment 1 of the present invention;

[0047] Figure 2 A flowchart of a capacitor element stacking method provided in Embodiment 2 of the present invention;

[0048] Figure 3 A schematic diagram of the structure of a capacitor detection model provided in Embodiment 2 of the present invention;

[0049] Figure 4 A schematic diagram of the pin clamping of the feeding device provided in the second embodiment of the present invention;

[0050] Figure 5 A structural block diagram of a capacitor element stacking system provided in Embodiment 3 of the present invention;

[0051] Figure 6 A structural block diagram of an electronic device provided in Embodiment 4 of the present invention;

[0052] exist Figure 4 middle:

[0053] 1. Capacitors to be stacked; 2. Mandrel hole; 3. Insert pins; 4. Robot arm. DETAILED DESCRIPTION

[0054] An embodiment of the present invention provides a capacitor element stacking method and system, which is used to solve the technical problem that the existing capacitor element stacking method mainly relies on a loading robot to clamp the outer diameter of the element and then put it into the element frame. However, due to the diverse models of capacitor elements and frequent production changes, the error probability of the loading robot clamping element stacking method is gradually increased, reducing the adaptability of the loading robot to stack capacitor elements.

[0055] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below 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.

[0056] See also Figure 1 , Figure 1 A flowchart of a capacitor element stacking method provided in Embodiment 1 of the present invention.

[0057] The present invention provides a capacitor element stacking method, comprising:

[0058] Step 101, obtaining a component image of a capacitor to be stacked, and inputting the component image into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network;

[0059] In an embodiment of the present invention, when the capacitors to be stacked on the conveyor belt are limited by blocking members, component images of the capacitors to be stacked are collected by a preset visual inspection system, and the component images are input into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network.

[0060] It should be noted that the training process of the capacitor detection model is specifically as follows: 1. Obtain multiple training component images, perform image preprocessing on the multiple training component images, and obtain a component feature set. 2. Use the component feature set to train the capacitor detection model to obtain training detection data. 3. Calculate the training loss function value of the component feature set based on the training detection data (that is, select standard data associated with the training detection data from the component feature set, calculate the deviation value between the training detection data and the standard data, and use the deviation value as the training loss function value). 4. When the training loss function value is greater than or equal to the preset standard loss function value, use a grid search method or a random search method to adjust the network parameters of the capacitor detection model until the training loss function value is less than the preset standard loss function value. 5. When the training loss function value is less than the preset standard loss function value, a trained capacitor detection model is generated.

[0061] Step 102, extracting features from the component image through a feature extraction network to obtain a capacitor feature map;

[0062] In an embodiment of the present invention, feature extraction is performed on a component image through a feature extraction network to obtain a capacitor feature map, wherein the feature extraction network includes a 3×3 convolution layer, a 1×1 standard convolution layer, an extraction branch and a feature fusion layer.

[0063] Step 103: Use a detection network to identify and locate the capacitor characteristic image to obtain the component model and component coordinates corresponding to the capacitor image to be stacked;

[0064] In an embodiment of the present invention, a detection network is used to identify and locate the capacitor feature map to obtain the component model and component coordinates corresponding to the capacitor image to be stacked, wherein the detection network includes a fully connected layer and a Softmax layer.

[0065] Step 104 , control the loading device corresponding to the component model to move to the component coordinate, and perform a stacking operation on the capacitors to be stacked.

[0066] In the embodiment of the present invention, after the component model and component coordinates corresponding to the image of the capacitor to be stacked are determined, the corresponding loading robot is called according to the component model and moved to the component coordinates to perform a stacking operation on the capacitor to be stacked.

[0067] It should be noted that each loading robot is associated with a unique component frame or stacking reference template. When the loading robot moves to the component coordinates, the robot's manipulator is used to insert and clamp the capacitors to be stacked. After the pin insertion and clamping are completed, the capacitors to be stacked are placed on the unique component frame or stacking reference template associated with the loading robot.

[0068] In an embodiment of the present invention, by acquiring the component image of the capacitor to be stacked, the component image is input into a pre-trained capacitor detection model to obtain the component model and component coordinates of the capacitor to be stacked, and then the loading device corresponding to the component model is controlled to move to the component coordinates to perform the stacking operation on the capacitor to be stacked, thereby overcoming the technical problem that the existing capacitor component stacking method mainly uses a loading robot to clamp the component outer diameter and then put it into the component frame, but due to the variety of capacitor component models and frequent production changes, the error probability of the loading robot clamping component stacking method is gradually increasing. Compared with the traditional component stacking method, the present invention uses a pre-trained capacitor detection model to accurately identify the component model of the capacitor to be stacked, and then calls the loading device corresponding to the component model to perform the stacking operation on the capacitor to be stacked, thereby improving the efficiency and adaptability of the loading robot stacking capacitor components.

[0069] See also Figure 2 , Figure 2 A flowchart of a capacitor element stacking method provided in Embodiment 2 of the present invention.

[0070] The present invention provides a capacitor element stacking method, comprising:

[0071] Step 201, obtaining a component image of a capacitor to be stacked, and inputting the component image into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network;

[0072] It should be noted that the capacitor element stacking method provided by the present invention can be applied to a preset capacitor element stacking system, and the capacitor element stacking system includes a blocking member, a visual inspection system, a control system, and a loading robot. The blocking member is arranged at the end of the conveyor belt of the capacitor element. When the capacitor element is blocked and limited by the blocking member, the visual inspection system is triggered to shoot the capacitor element, and the model, size and position data of the capacitor element are obtained based on the captured image data. The control system determines the corresponding loading robot according to the model data of the capacitor element detected by the visual inspection system, and sends the model, size and position data to the loading robot. The manipulator of the loading robot uses pins to clamp the capacitor element and perform stacking operations.

[0073] It is worth mentioning that the device used to detect the presence of a capacitor element at the blocking member can be a pressure sensor arranged at the bottom of the end of the conveyor belt. When the pressure sensor detects pressure, it means that there is a capacitor element at the end of the conveyor belt; it can also be an infrared detection element arranged on the blocking member. When the emission signal of the infrared detection element is blocked, it means that there is a capacitor element at the end of the conveyor belt.

[0074] In an embodiment of the present invention, when a component image of a capacitor to be stacked is obtained, the component image is input into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network.

[0075] Step 202, extracting features from the component image through a feature extraction network to obtain a capacitor feature map;

[0076] Further, see Figure 3 As shown, the feature extraction network includes a 3×3 convolution layer, a 1×1 standard convolution layer, an extraction branch and a feature fusion layer, and step 202 includes the following sub-steps:

[0077] S11, extracting features from the component image through a 3×3 convolutional layer to obtain a first component feature map;

[0078] In the embodiment of the present invention, a 3×3 convolutional layer is used to extract features from the component image to obtain a first component feature map.

[0079] S12, extracting features from the first component feature map through a 1×1 standard convolutional layer to obtain a second component feature map;

[0080] In the embodiment of the present invention, a 1×1 standard convolutional layer is used to extract features from the first component feature map to obtain a second component feature map.

[0081] S13, extracting features from the first component feature map through an extraction branch to obtain a third component feature map, wherein the extraction branch includes a first recognition module, a 2×2 maximum pooling layer, a first recognition module, a cascaded attention module, a second recognition module, and a 1×1 convolutional layer connected in sequence;

[0082] Further, see Figure 3 As shown, S13 includes the following sub-steps:

[0083] S131, extracting features from the first component feature map through a first recognition module, a 2×2 maximum pooling layer, and a first recognition module in sequence to obtain a first recognition feature map, wherein the first recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a Leaky-ReLU activation layer connected in sequence;

[0084] In an embodiment of the present invention, the first component feature map is feature extracted by the first recognition model to obtain a first intermediate component feature map, and the first intermediate component feature map is downsampled by a 2×2 maximum pooling layer to obtain a second intermediate component feature map. The second intermediate component feature map is feature extracted by the first recognition model to obtain a first recognition feature map, wherein the first recognition module includes a 3×3 convolution layer, a batch normalization layer, and a Leaky-ReLU activation layer connected in sequence.

[0085] S132, using a cascaded attention module to extract features from the first recognition feature map to obtain a second recognition feature map, wherein the cascaded attention module includes a 1×1 convolution layer, a feature fusion layer, a feature fusion layer, a 3×3 convolution layer, and a feature fusion layer;

[0086] Further, S132 includes the following sub-steps:

[0087] S1321, extracting features from the first recognition feature map through a 1×1 convolutional layer to obtain a first feature map;

[0088] In the embodiment of the present invention, a convolution operation is performed on the first recognition feature map through a 1×1 convolution layer to obtain a first feature map.

[0089] It should be noted that the convolution operation extracts features from the data by sliding the convolution kernel (also called a filter) on the input data (such as an image or feature map) and calculating the dot product of the convolution kernel and the local data at each position.

[0090] It is worth mentioning that the 1×1 convolution layer is a special convolution layer in which the convolution kernel size is 1 × 1. This means that the convolution kernel only considers one pixel at each position on the input feature map, but operates on the channel dimension.

[0091] S1322, extracting features from the first recognition feature map through a 3×3 convolutional layer to obtain a second feature map;

[0092] In the embodiment of the present invention, a convolution operation is performed on the first recognition feature map through a 3×3 convolution layer to obtain a second feature map.

[0093] S1323, using a feature fusion layer to perform feature fusion on the first feature map and the first recognition feature map to obtain a first fused feature map;

[0094] In the embodiment of the present invention, the first feature map and the first recognition feature map are feature fused by a feature fusion layer to obtain a first fused feature map.

[0095] It should be noted that different feature maps may contain different types of information, which may have certain limitations when used alone, but feature fusion can complement their advantages and provide a more comprehensive and rich feature representation. Fusion of multiple feature maps can provide more supervision information for the model, which helps the model converge to the optimal solution faster.

[0096] S1324, using a feature fusion layer to perform feature fusion on the first fused feature map and the first recognition feature map to obtain a second fused feature map;

[0097] In the embodiment of the present invention, the first fused feature map and the first recognition feature map are input into the feature fusion layer for feature fusion to obtain a second fused feature map.

[0098] S1325. Use a feature fusion layer to perform feature fusion on the second feature map and the second fused feature map to obtain a second recognition feature map.

[0099] In the embodiment of the present invention, the second feature map and the second fused feature map are input into the feature fusion layer for feature fusion to obtain a second recognition feature map.

[0100] S133, extracting features from the second recognition feature map through a second recognition module to obtain a third recognition feature map, wherein the second recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a DY-ReLU activation layer connected in sequence;

[0101] Further, S133 includes the following sub-steps:

[0102] S1331, extracting features from the second recognition feature map through a 3×3 convolutional layer to obtain a third feature map;

[0103] In the embodiment of the present invention, a convolution operation is performed on the second recognition feature map through a 3×3 convolution layer to obtain a third feature map.

[0104] S1332. Perform a batch normalization operation on the third feature map through a batch normalization layer to obtain a fourth feature map;

[0105] In this embodiment of the present invention, the third feature map is input into a batch normalization layer for batch normalization operation to obtain a fourth feature map.

[0106] S1333. Use a DY-ReLU activation layer to perform a nonlinear operation on the fourth feature map to obtain a third recognition feature map.

[0107] In the embodiment of the present invention, the fourth feature map is input into the DY-ReLU activation layer for nonlinear operation to obtain the third recognition feature map.

[0108] It should be noted that the Leaky-ReLU activation layer: The parameters of the Leaky-ReLU (such as the slope of the negative semi-axis) are fixed and will not be automatically adjusted during training. This means that it may not adapt well to changes in data, and for some complex data distributions, fixed parameters may not achieve the best activation effect. Compared with the Leaky-ReLU activation layer, the DY-ReLU activation layer has the ability of adaptive learning and can dynamically adjust the parameters of the activation function according to the characteristics of the input data. This allows it to better adapt to different data distributions and task requirements, and show greater flexibility and adaptability when processing complex data. For example, in different data sets or different training stages, DY-ReLU can automatically adjust its parameters to optimize the performance of the model.

[0109] It is worth mentioning that replacing the Leaky-ReLU activation layer with the DY-ReLU activation layer helps the model converge to the optimal solution faster. During the training process, the DY-ReLU activation layer can adjust the parameters of the activation function in time according to the feedback of the data, so that the model can adapt to the distribution of the training data more quickly, thereby accelerating the convergence process. At the same time, this adaptive adjustment also helps to improve the stability of training and reduce the possibility of oscillation or falling into the local optimal solution during the training process.

[0110] S134. Perform feature extraction on the third recognition feature map through a 1×1 convolutional layer to obtain a third component feature map.

[0111] In the embodiment of the present invention, a convolution operation is performed on the third recognition feature map through a 1×1 convolution layer to obtain a third component feature map.

[0112] S14. Use a feature fusion layer to perform feature fusion on the second component feature map and the third component feature map to obtain a capacitor feature map.

[0113] In the embodiment of the present invention, the second element characteristic map and the third element characteristic map are input into the feature fusion layer for feature fusion to obtain the capacitor characteristic map.

[0114] Step 203: using a detection network to perform identification and positioning detection on the capacitor characteristic image to obtain the component model and component coordinates corresponding to the capacitor image to be stacked;

[0115] Furthermore, the detection network includes a fully connected layer and a Softmax layer, and step 203 includes the following sub-steps:

[0116] S21, feature mapping the capacitor feature map through a fully connected layer to obtain a component mapping feature map;

[0117] In the embodiment of the present invention, a fully connected layer is used to perform nonlinear mapping on the capacitor characteristic map to obtain a component mapping characteristic map.

[0118] S22. Use the Softmax layer to perform identification and positioning operations on the component mapping feature map to obtain the component model and component coordinates corresponding to the capacitor image to be stacked.

[0119] In the embodiment of the present invention, the component mapping feature map is identified and located through the Softmax layer to obtain the component model and component coordinates corresponding to the capacitor image to be stacked.

[0120] Step 204: Use the loading device corresponding to the component model as the target device, and control the target device to move from the stacking position to the component coordinates;

[0121] In an embodiment of the present invention, a corresponding loading robot is determined according to the identified component model, and the loading robot is used as a target device to control the target device to move from the stacking position to the component coordinates.

[0122] It should be noted that when the target device is already loading the material of this model, wait until the target device is idle before controlling the target device to move to the component coordinates.

[0123] Step 205: Control the target device to place the capacitors to be stacked at the stacking position.

[0124] In the embodiment of the present invention, the mechanical arm of the target device is controlled to clamp the capacitor to be stacked, and move to the corresponding stacking position of the target device, and then place the capacitor to be stacked at the stacking position.

[0125] It should be noted that, see Figure 4 As shown, a core shaft hole 2 is opened on the capacitor 1 to be stacked, and the target device (i.e., a loading robot, one end of which is connected to a manipulator 4 to transmit power and the other end can be inserted into the core shaft hole 2 of the component) inserts the pin 3 on the manipulator 4 into the core shaft hole 2 of the capacitor 1 to be stacked, thereby picking up the capacitor 1 to be stacked for component clamping.

[0126] In an embodiment of the present invention, by acquiring the component image of the capacitor to be stacked, the component image is input into a pre-trained capacitor detection model to obtain the component model and component coordinates of the capacitor to be stacked, and then the loading device corresponding to the component model is controlled to move to the component coordinates to perform the stacking operation on the capacitor to be stacked, thereby overcoming the technical problem that the existing capacitor component stacking method mainly uses a loading robot to clamp the component outer diameter and then put it into the component frame, but due to the variety of capacitor component models and frequent production changes, the error probability of the loading robot clamping component stacking method is gradually increasing. Compared with the traditional component stacking method, the present invention uses a pre-trained capacitor detection model to accurately identify the component model of the capacitor to be stacked, and then calls the loading device corresponding to the component model to perform the stacking operation on the capacitor to be stacked, thereby improving the efficiency and adaptability of the loading robot stacking capacitor components.

[0127] See also Figure 5 , Figure 5 This is a structural block diagram of a capacitor element stacking system provided in Embodiment 3 of the present invention.

[0128] The present invention provides a capacitor element stacking system, comprising:

[0129] The acquisition module 301 is used to obtain component images of capacitors to be stacked, and input the component images into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network;

[0130] An extraction module 302 is used to extract features from the component image through a feature extraction network to obtain a capacitor feature map;

[0131] The identification module 303 is used to use the detection network to identify and locate the capacitor characteristic image to obtain the component model and component coordinates corresponding to the capacitor image to be stacked;

[0132] The stacking module 304 is used to control the loading equipment corresponding to the component model to move to the component coordinates and perform stacking operations on the capacitors to be stacked.

[0133] Furthermore, the feature extraction network includes a 3×3 convolution layer, a 1×1 standard convolution layer, an extraction branch and a feature fusion layer, and the extraction module 302 includes:

[0134] A first extraction submodule, used for extracting features from the component image through a 3×3 convolutional layer to obtain a first component feature map;

[0135] A second extraction submodule is used to extract features from the first component feature map through a 1×1 standard convolution layer to obtain a second component feature map;

[0136] A third extraction submodule is used to extract features from the first component feature map through an extraction branch to obtain a third component feature map, wherein the extraction branch includes a first recognition module, a 2×2 maximum pooling layer, a first recognition module, a cascaded attention module, a second recognition module, and a 1×1 convolutional layer connected in sequence;

[0137] The first fusion submodule is used to use a feature fusion layer to perform feature fusion on the second element feature map and the third element feature map to obtain a capacitor feature map.

[0138] Furthermore, the third extraction submodule includes:

[0139] A first extraction unit is used to extract features from the first component feature map through a first recognition module, a 2×2 maximum pooling layer, and a first recognition module in sequence to obtain a first recognition feature map, wherein the first recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a Leaky-ReLU activation layer connected in sequence;

[0140] A second extraction unit is used to extract features from the first recognition feature map by using a cascaded attention module to obtain a second recognition feature map, wherein the cascaded attention module includes a 1×1 convolution layer, a feature fusion layer, a feature fusion layer, a 3×3 convolution layer, and a feature fusion layer;

[0141] a third extraction unit, configured to extract features from the second recognition feature map through a second recognition module to obtain a third recognition feature map, wherein the second recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a DY-ReLU activation layer connected in sequence;

[0142] The fourth extraction unit is used to extract features from the third recognition feature map through a 1×1 convolution layer to obtain a third component feature map.

[0143] Furthermore, the second extraction unit comprises:

[0144] A first extraction subunit, configured to extract features from the first recognition feature map through a 1×1 convolutional layer to obtain a first feature map;

[0145] A second extraction subunit is used to extract features from the first recognition feature map through a 3×3 convolution layer to obtain a second feature map;

[0146] A first fusion subunit is used to perform feature fusion on the first feature map and the first recognition feature map by using a feature fusion layer to obtain a first fused feature map;

[0147] A second fusion subunit is used to use a feature fusion layer to perform feature fusion on the first fusion feature map and the first recognition feature map to obtain a second fusion feature map;

[0148] The third fusion subunit is used to use the feature fusion layer to perform feature fusion on the second feature map and the second fused feature map to obtain a second recognition feature map.

[0149] Further, the third extraction unit comprises:

[0150] A third extraction subunit is used to extract features from the second recognition feature map through a 3×3 convolution layer to obtain a third feature map;

[0151] a batch normalization subunit, configured to perform a batch normalization operation on the third feature map through a batch normalization layer to obtain a fourth feature map;

[0152] The nonlinear operation subunit is used to perform a nonlinear operation on the fourth feature map by using a DY-ReLU activation layer to obtain a third recognition feature map.

[0153] Furthermore, the detection network includes a fully connected layer and a Softmax layer, and the recognition module 303 includes:

[0154] A mapping submodule, used for performing feature mapping on the capacitor feature map through a fully connected layer to obtain a component mapping feature map;

[0155] The identification and positioning submodule is used to use the Softmax layer to perform identification and positioning operations on the component mapping feature map to obtain the component model and component coordinates corresponding to the capacitor image to be stacked.

[0156] Furthermore, the stacking module 304 includes:

[0157] The matching submodule is used to take the loading device corresponding to the component model as the target device and control the target device to move from the stacking position to the component coordinates;

[0158] The stacking submodule is used to control the target device to place the capacitors to be stacked at the stacking position.

[0159] See also Figure 6 , Figure 6 This is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention.

[0160] An electronic device according to an embodiment of the present invention includes: a memory 401 and a processor 402, wherein the memory 402 stores a computer program; when the computer program is executed by the processor 402, the processor 402 executes the capacitor element stacking method according to any of the above embodiments.

[0161] The memory 401 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk or a ROM. The memory 401 has a storage space 403 for a program code 413 for executing any method step in the above method. For example, the storage space 403 for the program code may include individual program codes 413 for implementing the various steps in the above method, respectively. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as a hard disk, a compact disk (CD), a memory card or a floppy disk. The program code may be compressed, for example, in an appropriate form. When these codes are run by a computing and processing device, the computing and processing device is caused to execute the various steps in the above-described method.

[0162] The fifth embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the capacitor element stacking method as described in any of the above embodiments is implemented.

[0163] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0164] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0165] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0166] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0167] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0168] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A capacitor element stacking method, characterized in that: include: Acquire a component image of a capacitor to be stacked, and input the component image into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network; Extracting features from the component image through the feature extraction network to obtain a capacitor feature map; Using the detection network to perform identification and positioning detection on the capacitor characteristic image, and obtaining the component model and component coordinates corresponding to the capacitor image to be stacked; The loading device corresponding to the component model is controlled to move to the component coordinate, and the capacitor to be stacked is stacked.

2. The capacitor element stacking method according to claim 1, characterized in that: The feature extraction network includes a 3×3 convolution layer, a 1×1 standard convolution layer, an extraction branch and a feature fusion layer. The step of extracting features from the component image through the feature extraction network to obtain a capacitor feature map includes: Extracting features from the component image through a 3×3 convolutional layer to obtain a first component feature map; Extract features from the first component feature map through a 1×1 standard convolutional layer to obtain a second component feature map; Performing feature extraction on the first component feature map through the extraction branch to obtain a third component feature map, wherein the extraction branch includes a first recognition module, a 2×2 maximum pooling layer, a first recognition module, a cascaded attention module, a second recognition module, and a 1×1 convolutional layer connected in sequence; A feature fusion layer is used to perform feature fusion on the second element feature map and the third element feature map to obtain a capacitor feature map.

3. The capacitor element stacking method according to claim 2, characterized in that: The step of extracting features from the first component feature map by the extraction branch to obtain a third component feature map comprises: Extracting features from the first component feature map through a first recognition module, a 2×2 maximum pooling layer, and a first recognition module in sequence to obtain a first recognition feature map, wherein the first recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a Leaky-ReLU activation layer connected in sequence; Using the cascaded attention module to extract features from the first recognition feature map to obtain a second recognition feature map, wherein the cascaded attention module includes a 1×1 convolution layer, a feature fusion layer, a feature fusion layer, a 3×3 convolution layer, and a feature fusion layer; Extracting features from the second recognition feature map through a second recognition module to obtain a third recognition feature map, wherein the second recognition module includes a 3×3 convolutional layer, a batch normalization layer, and a DY-ReLU activation layer connected in sequence; The third identification feature map is subjected to feature extraction through a 1×1 convolutional layer to obtain a third component feature map.

4. The capacitor element stacking method according to claim 3, characterized in that: The step of using the cascaded attention module to extract features from the first recognition feature map to obtain a second recognition feature map comprises: Performing feature extraction on the first recognition feature map through a 1×1 convolutional layer to obtain a first feature map; Performing feature extraction on the first recognition feature map through a 3×3 convolutional layer to obtain a second feature map; Using a feature fusion layer to perform feature fusion on the first feature map and the first recognition feature map to obtain a first fused feature map; Using a feature fusion layer to perform feature fusion on the first fused feature map and the first recognition feature map to obtain a second fused feature map; A feature fusion layer is used to perform feature fusion on the second feature map and the second fused feature map to obtain a second recognition feature map.

5. The capacitor element stacking method according to claim 3, characterized in that: The step of extracting features from the second recognition feature map by a second recognition module to obtain a third recognition feature map comprises: Performing feature extraction on the second recognition feature map through a 3×3 convolutional layer to obtain a third feature map; Performing a batch normalization operation on the third feature map through a batch normalization layer to obtain a fourth feature map; A DY-ReLU activation layer is used to perform a nonlinear operation on the fourth feature map to obtain a third recognition feature map.

6. The capacitor element stacking method according to claim 1, characterized in that: The detection network includes a fully connected layer and a Softmax layer. The step of using the detection network to identify and locate the capacitor feature map to obtain the component model and component coordinates corresponding to the capacitor image to be stacked includes: Performing feature mapping on the capacitor feature map through a fully connected layer to obtain a component mapping feature map; The Softmax layer is used to perform identification and positioning operations on the component mapping feature map to obtain the component model and component coordinates corresponding to the capacitor image to be stacked.

7. The capacitor element stacking method according to claim 1, characterized in that: The step of controlling the loading device corresponding to the component model to move to the component coordinate and stacking the capacitors to be stacked comprises: Taking the loading device corresponding to the component model as the target device, and controlling the target device to move from the stacking position to the component coordinates; The target device is controlled to place the capacitor to be stacked at the stacking position.

8. A capacitor element stacking system, characterized in that: include: An acquisition module, used for acquiring component images of capacitors to be stacked, and inputting the component images into a pre-trained capacitor detection model, wherein the capacitor detection model includes a feature extraction network and a detection network; An extraction module, used for performing feature extraction on the component image through the feature extraction network to obtain a capacitor feature map; An identification module, used to use the detection network to perform identification and positioning detection on the capacitor characteristic image, and obtain the component model and component coordinates corresponding to the capacitor image to be stacked; The stacking module is used to control the loading equipment corresponding to the component model to move to the component coordinates and perform stacking operations on the capacitors to be stacked.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the capacitor element stacking method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the capacitor element stacking method according to any one of claims 1 to 7 is implemented.

Images