Intelligent gold bonding wire length identification method and system

Abstract

The invention relates to the technical field of electronic manufacturing detection, and discloses a length identification method and system for an intelligent gold bonding wire, and the method comprises the following steps: collecting image sequence information in a dynamic deformation process of the gold bonding wire; constructing a feature extraction model based on a convolutional neural network; in combination with the bending degree and speed in the obtained motion parameters of the gold bonding wire, modeling is carried out on the dynamic change of the gold bonding wire by utilizing a long-short-term memory network; and inputting the feature vector set into a trained dynamic change model to obtain a hidden state representing a dynamic change rule of the gold bonding wire, and obtaining a length value of the gold bonding wire in each frame of image by using a regression algorithm. The dynamic length of the gold bonding wire can be accurately obtained through multi-angle collection, feature extraction, dynamic modeling and a regression algorithm in electronic manufacturing detection, and the production quality and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic manufacturing detection, and more specifically, to a length recognition method and system for an intelligent gold bonding wire. Background Art

[0002] In the current electronic manufacturing field, especially in the production and application scenarios of gold bonding wires in flexible electronic manufacturing, accurate identification of the length of gold bonding wires is crucial to ensure product quality and production efficiency. However, existing methods for identifying the length of gold bonding wires have many shortcomings. For example, traditional methods based on manual measurement are not only inefficient, but also easily affected by human factors, resulting in large measurement errors. Even if some automated measurement methods are used, in complex production environments, such as ultra-small production micro-environments, due to the presence of various interference factors such as temperature and humidity, the existing length recognition systems based on simple image processing or single parameter modeling are difficult to accurately capture the dynamic length changes of gold bonding wires under different conditions. This makes it impossible to obtain the length information of the gold bonding wires in a timely and accurate manner during the production process, which may lead to quality defects in the product and increase production costs. Summary of the invention

[0003] The present invention provides a method for identifying the length of an intelligent bonding wire, comprising: Image acquisition: using multiple sets of high-speed cameras to collect image sequence information from different angles during the dynamic deformation of the bonding wire; Feature extraction: construct a feature extraction model based on convolutional neural network, input the image sequence information into the feature extraction model, extract the morphological features of the bonding wire, and form a feature vector set; Dynamic modeling, combining the bending degree and speed of the obtained bonding wire motion parameters, using the long short-term memory network to model the dynamic changes of the bonding wire; taking the feature vector set as input, and integrating the bending degree and speed parameters into the long short-term memory network, allowing the long short-term memory network to learn the dynamic change law of the bonding wire, and obtain a trained dynamic change model; Length calculation: Input the feature vector set into the trained dynamic change model to obtain the hidden state representing the dynamic change law of the bonding wire, and use the regression algorithm to obtain the length value of the bonding wire in each frame of the image.

[0004] In a preferred embodiment, the image sequence information is recorded as an image sequence set in Indicates the number of frames of the image; for the acquired image sequence For each frame of the image, the Gaussian filter formula is used to perform convolution operation on the image to remove noise interference and obtain the preprocessed image sequence .

[0005] In a preferred embodiment, the feature extraction model input is the preprocessed image After multiple convolutional layers and pooling layer The convolution operation is expressed as: ; in, is the input feature map, Indicates that the convolution kernel is at position The weight value at is bias; and It is the coordinate offset during the sliding process of the convolution kernel; After the convolution operation, the coordinate The eigenvalue output at the output; the morphological features of the bonding wire are extracted through the feature extraction model to form a feature vector set It corresponds to Feature vectors of frame images; these feature vectors will serve as important input for subsequent dynamic modeling.

[0006] In a preferred embodiment, the input gate of the long short-term memory network Forget Gate Output Gate and new memory cell candidate values Calculated by the following formulas: ; in, is the sigmoid activation function; is the hyperbolic tangent activation function; denote the first and second weight matrices respectively, denote the third and fourth weight matrices respectively; denote the fifth and sixth weight matrices respectively; denote the seventh and eighth weight matrices respectively; It corresponds to Feature vector of frame image; express Always hide the status, hour denote the first, second, third, and fourth bias vectors respectively; The memory cell state update formula is as follows: ; in, Indicates time The state of memory cells; express The state of memory cells at each moment; Indicates the previous moment The forget gate value of Indicates the previous moment The input gate value of Indicates time New memory cell candidate value; The hidden state is updated to: ; in, Indicates time The hidden state of Indicates time The state of memory cells; Indicates time The output gate value of represents the hyperbolic tangent activation function; The expression of the output layer is: ; in express Motion parameters of the bonding wire at each moment; Train the long short-term memory network, and then remove the output layer to obtain the dynamic change model of the hidden state of the dynamic change law of the bonding wire. The dynamic change model can capture the dynamic change characteristics of the bonding wire at different times.

[0007] In a preferred embodiment, in the length calculation, the length of the bonding wire in each frame of the image is calculated by a regression algorithm; the length calculation formula is: ; in, is the regression function, The length of the frame image is .

[0008] In a preferred embodiment, sensors are set on the moving path of the bonding wire to collect these parameters in real time, and angle sensors are used to obtain the degree of bending. Get speed using speed sensor .

[0009] In a preferred embodiment, a method for identifying the length of an intelligent bonding gold wire further includes: Fast image preprocessing, for the acquired image sequence Each frame of the image A simplified median filter method is used to replace Gaussian filter; Lightweight feature extraction model construction, build a lightweight convolutional neural network feature extraction model MobileNet; Dynamic modeling of fusion action information, combined with the acquired maintenance personnel operation action types and the characteristic vector of the bond wire The dynamic change model of the bonding wire is obtained by using the gated recurrent unit to model the dynamic change of the bonding wire and obtain the dynamic change model of the bonding wire that integrates the action information. ; Length calculation based on device parameters, dynamic change model of bonding wire based on trained fusion motion information Combined with the hardware parameters of the handheld detection device, the length of the bonding wire in each frame of the image is calculated using an optimized regression algorithm.

[0010] In a preferred embodiment, the gated recurrent unit resets the gate and update gate To control the flow of information; the input is the feature vector of the image and the type of operation action of the maintenance personnel and the previous hidden state The reset gate and update gate are: ; ; in, is the sigmoid activation function, Respectively represent the ninth, tenth, eleventh, and twelfth weight matrices, Respectively represent the fifth and sixth bias vectors, Indicates the corresponding Feature vector of frame image; The candidate hidden states are: ; in, is the hyperbolic tangent activation function, denote the thirteenth and fourteenth weight matrices respectively, represents the seventh bias vector; The hidden state is updated to: ; in, Indicates time The hidden state of Indicates time The update gate value of express The hidden state of the moment; Indicates time Candidate hidden states of ; The expression of the output layer is: ; in express Motion parameters of the bonding wire at each moment; Train the gated recurrent unit and then remove the output layer to get the dynamic change model of the bond wire that incorporates the action information. .

[0011] In a preferred embodiment, the length formula is calculated using an optimized regression algorithm as follows: ; in, is the optimized regression function, It is The length of the bonding wire in the frame image, It's time The hidden state of It is the type of operation action performed by maintenance personnel. is the resolution of the handheld detection device, is the frame rate of the handheld detection device.

[0012] In a preferred embodiment, a length identification system of an intelligent gold bonding wire comprises: An image acquisition module, using a handheld detection device tool to acquire an image sequence containing the bonded gold wire; Feature extraction module, which builds a lightweight convolutional neural network feature extraction model and processes images using deep separable convolution and pooling layers; The action information fusion and dynamic modeling module combines the operation action type of the maintenance personnel and the feature vector of the bonding wire, and uses the gated recurrent unit to model the dynamic changes of the bonding wire; The length calculation module calculates the length of the bonding wire in each frame of the image through an optimized regression algorithm based on the trained dynamic change model of the bonding wire that integrates the motion information and the feature vector set, combined with the hardware parameters of the handheld detection device.

[0013] The beneficial effects of the present invention are: Improve recognition accuracy: By building a complex and deeper convolutional neural network model and adding an attention mechanism module, the morphological characteristics of the bonding wire can be extracted more accurately. At the same time, combined with the improved long short-term memory network for dynamic modeling, the influence of micro-environmental parameters (temperature, humidity, production line speed, etc.) on the length of the bonding wire is fully considered, thereby achieving high-precision length recognition, which improves recognition accuracy compared to traditional methods.

[0014] Adapt to different scenarios: For different application scenarios, such as portable wearable flexible electronic device maintenance scenarios, the algorithm is further optimized, and methods such as image fast preprocessing and lightweight feature extraction model construction are adopted to reduce the computational complexity while ensuring measurement accuracy, adapt to the limited resources of portable devices, and broaden the application scope of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of a method for identifying the length of an intelligent bonding gold wire of the present invention; Figure 2 is an example diagram of data collected by the present invention; Figure 3 is an example diagram of a convolutional layer of the present invention; Figure 4 is an example diagram of the pooling layer of the present invention. DETAILED DESCRIPTION

[0016] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0017] At least one embodiment of the present invention discloses a method for identifying the length of an intelligent bonding wire. Figure 1 As shown, including: Step 100, image preprocessing, using multiple sets of high-speed cameras to collect image sequence information during the dynamic deformation of the bonding wire from different angles; In one embodiment of the present invention, during the flexible electronics manufacturing process, the bonding gold wire may undergo dynamic deformation due to the operation of the production equipment, the material's own characteristics, and external environmental factors. For example, in a flexible electronics manufacturing scenario where a high-speed camera captures images, the bonding gold wire will be in a dynamic deformation state.

[0018] In one embodiment of the present invention, the image sequence information is recorded as an image sequence set in Representative Frame image. For the acquired image sequence Each frame of the image The Gaussian filter formula is used to perform a convolution operation on the image. The purpose of this operation is to remove noise interference and obtain a preprocessed image sequence. The Gaussian filter formula is: ; in, is the image coordinate, is the center of the Gaussian kernel function, is the standard deviation of the Gaussian kernel function, which is used to control the degree of smoothing. is the constant part related to pi, is a natural constant and is used as the base of the exponential function in this formula. Through this step, the image quality of subsequent processing is higher and the influence of noise on feature extraction is reduced.

[0019] Step 200, feature extraction, building a feature extraction model based on convolutional neural network (CNN); In one embodiment of the present invention, the CNN model is generally composed of a convolutional layer, a pooling layer, and a fully connected layer, and the input is a preprocessed image. After multiple convolutional layers and pooling layer The convolution operation can be expressed as: ; in, is the input feature map, Indicates that the convolution kernel is at position The weight value at is bias. and It is the coordinate offset during the sliding process of the convolution kernel, which is used to traverse the corresponding positions of the convolution kernel and the input feature map for multiplication. After the convolution operation, the coordinate The pooling layer mainly downsamples the convolutional feature map to reduce the amount of data while retaining the main features. The model extracts the morphological features of the bonding wire to form a feature vector set It corresponds to The feature vectors of the frame images will serve as important inputs for subsequent dynamic modeling.

[0020] Step 300, dynamic modeling, combined with the obtained bonding wire motion parameters and bending degree and speed Long short-term memory network (LSTM) is used to model the dynamic changes of bonding wires; In one embodiment of the present invention, these parameters can be collected in real time by setting sensors on the moving path of the bonding wire, for example, using an angle sensor to obtain the degree of bending. Get speed using speed sensor .

[0021] In one embodiment of the present invention, LSTM is a special recurrent neural network (RNN) that controls the flow of information through input gates, forget gates, and output gates. The input is a feature vector and the previous hidden state Memory cell status Input Gate Forget Gate Output Gate and new memory cell candidate values Calculated by the following formulas: ; in, is the sigmoid activation function; is the hyperbolic tangent activation function; Respectively represent the first and second weight matrices of the input gate, Respectively represent the third and fourth weight matrices of the forget gate; Respectively represent the fifth and sixth weight matrices of the output gate; Respectively represent the seventh and eighth weight matrices in the new memory cell candidate values; Represent the first, second, third, and fourth bias vectors respectively.

[0022] The memory cell state update formula is as follows: ; The hidden state is updated to: ; Finally, the dynamic change model of the bonding wire is obtained This model can capture the dynamic change characteristics of the bonding gold wire at different times and provide strong support for length calculation.

[0023] Step 400, length calculation, input the feature vector set into the trained dynamic change model to obtain the hidden state representing the dynamic change law of the bonding wire, and use the regression algorithm to obtain the length value of the bonding wire in each frame image; the length calculation formula is: ; in, is the regression function, The length of the frame image is .

[0024] Through the above steps, the real-time length sequence of the bonding wire during dynamic deformation is finally generated. Further modeling can be used to obtain the actual length or total length L of the bonding wire. Indicates the calculation The model can be a multi-layer perceptron or a linear regression model.

[0025] In one embodiment of the present invention, an example of the aforementioned method for identifying the length of an intelligent bonding wire is provided: Image preprocessing: like Figure 2 As shown in Figure 1, in a flexible electronics manufacturing plant, 5 sets of high-speed cameras are used to collect image sequences of the dynamic deformation process of bonding wires from different angles. For example, in one acquisition task, a total of 100 frames of images are acquired to form an image sequence set. ; For each frame The Gaussian filter formula is used to calculate and perform convolution operations, where is the center of the image. After processing, the preprocessed image sequence is obtained .

[0026] Feature extraction: Construct a feature extraction model based on convolutional neural network (CNN). Take the preprocessed image of the 10th frame For example, enter the model. Figure 3 As shown, the convolutional layer in the model The convolution kernel size is 3x3, the step size is 1, and the feature map size obtained after the convolution layer is the original image size minus the convolution kernel size plus 1. Figure 4 As shown, the pooling layer The maximum pooling is used, the pooling kernel size is 2x2, and the step size is 2. After multiple convolutional layers and pooling layers, the morphological features of the frame image are extracted to form a feature vector .

[0027] Dynamic Modeling: Combined with the obtained bonding wire motion parameters bending degree and speed The dynamic changes of the bonding wire are modeled using a long short-term memory network (LSTM). For example, Input Gate: ; in, is the input gate value at the 10th moment; is the sigmoid activation function; Respectively represent the first and second weight matrices of the input gate; is the feature vector corresponding to the 10th frame image; is the hidden state at the 9th moment; Represents the first bias vector.

[0028] Forget Gate: ; in, is the forget gate value at the 10th moment; is the sigmoid activation function; Respectively represent the third and fourth weight matrices of the input gate, Used to adjust the hidden state of the previous moment The degree of influence on the calculation results of the forget gate; represents the second bias vector.

[0029] Output Gate: ; in, is the output gate value at the 10th moment; is the sigmoid activation function; Respectively represent the fifth and sixth weight matrices of the input gate, Used to adjust the hidden state of the previous moment The degree of influence on the calculation results of the output gate; represents the third bias vector.

[0030] New memory cell candidate values: ; in, is the new candidate value of memory cell at the 10th moment; is the hyperbolic tangent activation function; is the seventh weight matrix in the new memory cell candidate value, used to adjust the feature vector The degree of influence on the calculation results of new memory cell candidate values; It is the eighth weight matrix in the new memory cell candidate value, which is used to adjust the hidden state at the previous moment. The degree of influence on the calculation results of new memory cell candidate values; is the fourth bias vector in the new memory cell candidate value.

[0031] The memory cell status is updated as follows: ; in, is the memory cell state after being updated at the 10th moment; is the state of memory cells at the 9th moment; is the forget gate value at the 10th moment, controlling the state of the memory cell at the 9th moment Retained to the extent of the 10th moment; is the input gate value at the 10th moment, controlling the new memory cell candidate value The degree of entry into the memory cell state; It is the new candidate value of memory cell at the 10th moment.

[0032] The hidden state is updated to: ; in, is the hidden state after updating at the 10th moment; is the hyperbolic tangent activation function; is the output gate value at the 10th moment, controlling the state of the memory cell The degree to which output is sent to the hidden state; This is the memory cell status after being updated at the 10th moment.

[0033] Length calculation: The length of the bonding wire in each frame of the image is calculated by a regression algorithm; The length calculation formula is: ; in, is the regression function, The length of the frame image is After calculation, we get .

[0034] In one embodiment of the present invention, in view of the limited resources of portable devices in the wearable flexible electronic device maintenance scenario, the relevant technology is further optimized and expanded; a method for identifying the length of an intelligent bonding gold wire also includes: Step 101, fast image preprocessing. Due to the limited resources of handheld devices, the acquired image sequence Each frame of the image A simplified median filter method is used to replace Gaussian filter.

[0035] In one embodiment of the present invention, median filtering is to replace the value of a pixel in an image with the median value of the pixel values ​​in the neighborhood of the pixel. Assume that the neighborhood size is For the pixels in the image The pixel value after median filtering It is the middle value of all pixel values ​​in the neighborhood after sorting. Represents the side length of the median filter neighborhood. This step can reduce noise interference while reducing computational complexity, adapting to the computing power of portable devices, and obtaining a preprocessed image sequence .

[0036] Step 102, lightweight feature extraction model construction, build a lightweight convolutional neural network (CNN) feature extraction model MobileNet. MobileNet uses deep separable convolution to decompose the standard convolution into deep convolution and point-by-point convolution, greatly reducing model parameters and calculations. Assume the input is the preprocessed image After multiple layers of depth-wise separable convolutional layers and pooling layer The depthwise convolution operation can be expressed as: ; in, is the depth convolution kernel, and is the convolution kernel in and Upward index, is the input feature map at position The value at is the output feature map at position The point-by-point convolution operation can be expressed as: ; in, is the point-wise convolution kernel, is the bias, and The same convolution kernel is and The index in the direction, is the input feature map at position The value at is the output feature map at position The pooling layer also downsamples the convolutional feature map. The lightweight model is used to extract the morphological features of the bonding wire to form a feature vector set. It corresponds to Feature vector of the frame image.

[0037] Step 103: Dynamic modeling of fusion action information, combined with the acquired operation action types of maintenance personnel and the characteristic vector of the bond wire The dynamic changes of the bonding wire are modeled using a gated recurrent unit (GRU). GRU is a simplified RNN that and update gate To control the flow of information. Let the input be the feature vector of the image and the previous hidden state The reset gate and update gate are: ; ; in, is the Sigmoid activation function, Respectively represent the ninth, tenth, eleventh, and twelfth weight matrices, Represent the fifth and sixth bias vectors respectively.

[0038] The candidate hidden states are: ; in, is the hyperbolic tangent activation function, denote the thirteenth and fourteenth weight matrices respectively, Represents the seventh bias vector.

[0039] The hidden state is updated to: ; in, Indicates time The hidden state of Indicates time The update gate value of express The hidden state of the moment; Indicates time Candidate hidden states of ; The expression of the output layer is: ; in express Motion parameters of the bonding wire at each moment; Train the gated recurrent unit and then remove the output layer to get the dynamic change model of the bond wire that incorporates the action information. .

[0040] Step 104, length calculation based on device parameters. According to the trained dynamic change model And the set of feature vectors Combined with the hardware parameters of the handheld detection device (such as resolution Frame rate The length of the bonding wire in each frame is calculated using an optimized regression algorithm. The formula for calculating the length using the optimized regression algorithm is as follows: ; in, is the optimized regression function, It is The length of the bonding wire in the frame image, It's time The hidden state of It is the type of operation action performed by maintenance personnel. is the resolution of the handheld detection device, is the frame rate of the handheld detection device.

[0041] In one embodiment of the present invention, based on the field of flexible electronic education practice, a method for identifying the length of an intelligent bonding wire further includes: Step 201, data preprocessing, obtaining data sets Preprocess the feature vector data in the data set; In one embodiment of the present invention, the data set It is a data set obtained in the flexible electronics education practice scenario, which contains feature vector data related to the dynamic change of the bonding wire length; Perform normalization and set the eigenvector The normalization formula is: ; in, Represents a set of feature vectors The minimum value in Represents a set of feature vectors The maximum value in . After calculation by this formula, the normalized feature vector set is obtained .

[0042] Step 202, model construction and training, build a deep learning model based on long short-term memory network (LSTM). Take the normalized feature vector set And the corresponding length set As training data. The model input is the feature vector and the previous hidden state Memory cell status .

[0043] Input Gate Forget Gate Output Gate and new memory cell candidate values Calculated by the following formulas: ; ; in, is the Sigmoid activation function; They represent the fifteenth, sixteenth, seventeenth, eighteenth, nineteenth, twentieth, twenty-first and twenty-second weight matrices respectively, and are used to adjust the influence of input information on each calculation result; They are the eighth, ninth, tenth and eleventh bias vectors, which are used to increase the nonlinear expression ability of the model.

[0044] The memory cell status is updated as follows: ; in Represents element-by-element multiplication, i.e., forget gate The memory cell state at the previous moment Element-wise multiplication, plus input gate With the new memory cell candidate value The result of element-by-element multiplication is the updated memory cell state. .

[0045] The hidden state is updated to: Output Gate and through the hyperbolic tangent function Processed memory cell state Multiply element by element to get the updated hidden state .

[0046] After multiple iterations of training, the model learns the mapping relationship between feature vectors and lengths, and obtains a trained model. .

[0047] Step 203, virtual simulation interaction is realized, in the virtual simulation environment, the degree of bending of the parameters set by the students is obtained speed Convert it to a format that matches the feature vector in the training dataset and record it as the converted feature vector

[0048] Will Input the trained model Predict the length of the bonding wire based on the model output Through graphics rendering technology, the dynamic change of the length of the bonding wire as the bending degree and speed change in the setting are displayed in real time in the virtual environment. This step realizes the virtual simulation interaction function and provides students with an intuitive learning experience.

[0049] Through the above steps, a virtual simulation environment was built, and the dynamic simulation of the length of the bonding wire was realized based on the deep learning algorithm. Students can intuitively observe the length changes of the bonding wire under different bending degrees and speeds. This interactive learning method can significantly improve students' understanding of the dynamic change process of the bonding wire in flexible electronic manufacturing.

[0050] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A method for identifying the length of an intelligent bonding wire, characterized in that: include: Image acquisition: using multiple sets of high-speed cameras to collect image sequence information from different angles during the dynamic deformation of the bonding wire; Feature extraction: construct a feature extraction model based on convolutional neural network, input the image sequence information into the feature extraction model, extract the morphological features of the bonding wire, and form a feature vector set; Dynamic modeling, combining the bending degree and speed of the bonding wire motion parameters obtained, and using the long short-term memory network to model the dynamic changes of the bonding wire; The feature vector set is used as input, and the bending degree and speed parameters are integrated into the long short-term memory network, so that the long short-term memory network can learn the dynamic change law of the bonding wire and obtain a trained dynamic change model; Length calculation: Input the feature vector set into the trained dynamic change model to obtain the hidden state representing the dynamic change law of the bonding wire, and use the regression algorithm to obtain the length value of the bonding wire in each frame of the image.

2. The method for identifying the length of an intelligent bonding wire according to claim 1, characterized in that: Record the image sequence information as an image sequence set , Represent the 1st, 2nd, and kth frame images respectively; for the acquired image sequence For each frame of the image, the Gaussian filter formula is used to perform convolution operation on the image to remove noise interference and obtain the preprocessed image sequence .

3. The method for identifying the length of an intelligent bonding wire according to claim 2, characterized in that: The feature extraction model input is the preprocessed image After multiple convolutional layers and pooling layer The convolution operation is expressed as: ； in, is the input feature map, Indicates that the convolution kernel is at position The weight value at is bias; and It is the coordinate offset during the sliding process of the convolution kernel; After the convolution operation, the coordinate The eigenvalue output at the output; the morphological features of the bonding wire are extracted through the feature extraction model to form a feature vector set Respectively represent the 1st, 2nd, Feature vectors of frame images; these feature vectors will serve as important input for subsequent dynamic modeling.

4. The method for identifying the length of an intelligent bonding wire according to claim 1, characterized in that: Input gate of LSTM network Forget Gate Output Gate and new memory cell candidate values Calculated by the following formulas: ； in, is the sigmoid activation function; is the hyperbolic tangent activation function; denote the first and second weight matrices respectively, denote the third and fourth weight matrices respectively; denote the fifth and sixth weight matrices respectively; denote the seventh and eighth weight matrices respectively; It corresponds to Feature vector of frame image; express Always hide the status, hour denote the first, second, third, and fourth bias vectors respectively; The memory cell state update formula is as follows: ； in, Indicates time The state of memory cells; express The state of memory cells at each moment; Indicates the previous moment The forget gate value of Indicates the previous moment The input gate value of Indicates time New memory cell candidate value; The hidden state is updated to: ； in, Indicates time The hidden state of Indicates time The state of memory cells; Indicates time The output gate value of represents the hyperbolic tangent activation function; The expression of the output layer is: ； in express Motion parameters of the bonding wire at each moment; Train the long short-term memory network, and then remove the output layer to obtain the dynamic change model of the hidden state of the dynamic change law of the bonding wire. The dynamic change model can capture the dynamic change characteristics of the bonding wire at different times.

5. The method for identifying the length of an intelligent bonding wire according to claim 1, characterized in that: In the length calculation, the length of the bonding wire in each frame of the image is calculated by a regression algorithm; the length calculation formula is: ； in, is the regression function, The length of the frame image is .

6. The method for identifying the length of an intelligent bonding wire according to claim 1, characterized in that: Sensors are set up on the movement path of the bonding wire to collect these parameters in real time, and angle sensors are used to obtain the degree of bending. , use the speed sensor to get the speed .

7. The method for identifying the length of an intelligent bonding wire according to claim 1, characterized in that: Also includes: Fast image preprocessing, for the acquired image sequence Each frame of the image A simplified median filter method is used to replace Gaussian filter; Lightweight feature extraction model construction, build a lightweight convolutional neural network feature extraction model MobileNet; Dynamic modeling of fusion action information, combined with the acquired maintenance personnel operation action types and the characteristic vector of the bond wire The dynamic change model of the bonding wire is obtained by using the gated recurrent unit to model the dynamic change of the bonding wire and obtain the dynamic change model of the bonding wire that integrates the action information. Length calculation based on device parameters, dynamic change model of bonding wire based on trained fusion motion information Combined with the hardware parameters of the handheld detection device, the length of the bonding wire in each frame of the image is calculated using an optimized regression algorithm.

8. The method for identifying the length of an intelligent bonding wire according to claim 7, characterized in that: The gated recurrent unit is reset by and update gate To control the flow of information; the input is the feature vector of the image and the type of operation action of the maintenance personnel and the hidden state at the previous moment The reset gate and update gate are: ； ； in, is the sigmoid activation function, Respectively represent the ninth, tenth, eleventh, and twelfth weight matrices, Respectively represent the fifth and sixth bias vectors, Indicates the corresponding Feature vector of frame image; The candidate hidden states are: ； in, is the hyperbolic tangent activation function, denote the thirteenth and fourteenth weight matrices respectively, represents the seventh bias vector; The hidden state is updated to: ； in, Indicates time The hidden state of Indicates time The update gate value of express The hidden state of the moment; Indicates time Candidate hidden states of ; The expression of the output layer is: ； in express Motion parameters of the bonding wire at each moment; Train the gated recurrent unit and then remove the output layer to get the dynamic change model of the bond wire that incorporates the action information. .

9. The method for identifying the length of an intelligent bonding wire according to claim 7, characterized in that: The formula for calculating the length using the optimized regression algorithm is as follows: ； in, is the optimized regression function, It is The length of the bonding wire in the frame image, It's time The hidden state of It is the type of operation action performed by maintenance personnel. is the resolution of the handheld detection device, is the frame rate of the handheld detection device.

10. An intelligent bonding wire length identification system, characterized in that: A method for identifying the length of an intelligent bonding wire according to any one of claims 7 to 9, comprising: An image acquisition module, using a handheld detection device tool to acquire an image sequence containing the bonded gold wire; Feature extraction module, which builds a lightweight convolutional neural network feature extraction model and processes images using deep separable convolution and pooling layers; The action information fusion and dynamic modeling module combines the operation action type of the maintenance personnel and the feature vector of the bonding wire, and uses the gated recurrent unit to model the dynamic changes of the bonding wire; The length calculation module calculates the length of the bonding wire in each frame of the image through an optimized regression algorithm based on the trained dynamic change model of the bonding wire that integrates the motion information and the feature vector set, combined with the hardware parameters of the handheld detection device.

Images