Component information coding image generation and identification method based on neural network

Abstract

The invention provides a component information coding image generation and identification method based on a neural network, and belongs to the technical field of electronic component information processing, and the method comprises the steps: analyzing the model composition rule of a component development manufacturer, and obtaining and normalizing the component model data information of the manufacturer; converting the component model information into a character string for coding; converting the codes into an information matrix; converting the component coding information matrix into a two-dimensional coding image containing component model information, and constructing a CNN neural network model; and training a neural network model to perform deep learning on the construction characteristics of the component model information, and classifying the component model information from different development manufacturers by using the trained model, thereby obtaining the component information identification method. The information code identification and classification method which is more efficient and convenient and has certain feasibility is provided for identification and classification of component data information, and the efficiency of data information processing is improved.

Description

Technical Field

[0001] The present application relates to the field of electronic component information processing, and in particular to a component information coding image generation and recognition method based on a neural network. Background Art

[0002] Electronic components are an indispensable part of various industries and fields. Whether in the application scenarios of small and medium-sized electronic products such as mobile phones and computers or other large electronic products such as smart vehicles, their basic supporting role is crucial. Electronic components have many models, complex information, and scattered manufacturers. It is an inevitable trend of development to carry out information management on them. As an indispensable link in supporting the development, production, maintenance and service guarantee of aviation products, electronic component information management can not only ensure the smooth progress of development, but also stimulate the vitality of independent research and development of the downstream industry chain, and greatly promote the healthy development of the domestic electronic components industry.

[0003] However, due to the huge number of components used in electronic equipment and the fact that similar components can be produced by multiple manufacturers at the same time, it is a huge workload to directly match information by component model. It is also easy for the manufacturer's reported information to be inconsistent with the actual information, resulting in a high error rate, which reduces the accuracy of the information and slows down the overall progress of the project. Summary of the invention

[0004] In view of this, the present application provides a component information coding image generation method based on a neural network, which solves the problems in the prior art, provides a more efficient, convenient and feasible information coding identification and classification method for the identification and classification of component data information, and improves the efficiency of data information processing.

[0005] On the one hand, the present application provides a method for generating component information encoding images based on a neural network, which adopts the following technical solution:

[0006] A method for generating component information coding images based on a neural network, comprising:

[0007] Step 1: Analyze the model composition rules of component manufacturers, obtain and organize the manufacturer's component model data information;

[0008] Step 2, convert the component model information into a string for encoding;

[0009] Step 3, convert the code into an information matrix;

[0010] Step 4: Convert the component coding information matrix into a two-dimensional coded image containing component model information.

[0011] Optionally, in step 2, the component model information is converted into a binary string for encoding.

[0012] Optionally, in step 3, the binary code is first extended to a uniform length and then converted into an information matrix.

[0013] Optionally, in step 4, each element in the information matrix is ​​used as a pixel value of the component coded image, so as to draw and generate the component information coded image.

[0014] On the other hand, the present application provides a component information coding image recognition method based on a neural network, which adopts the following technical solution:

[0015] A component information coded image recognition method based on a neural network is used to recognize the two-dimensional coded image generated above, and the recognition method comprises the following steps: constructing a CNN neural network model;

[0016] The neural network model is trained to conduct deep learning on the structural characteristics of component model information, and the trained model is used to classify component model information from different research and production manufacturers, thereby obtaining a component information recognition method.

[0017] Optionally, the CNN neural network model includes a first convolutional layer, a second pooling layer, a third convolutional layer, a fourth pooling layer, a fifth fully connected layer and an output layer;

[0018] The first convolutional layer uses 32 convolution kernels for convolution operation, the convolution kernel size is 3*3, and the activation function is ReLU;

[0019] The second pooling layer uses the maximum pooling method with a window size of 2*2;

[0020] The third convolutional layer uses 64 convolution kernels of size 3*3 for convolution operation, and the activation function is the ReLU function;

[0021] The fourth pooling layer uses the maximum pooling method with a window size of 2*2;

[0022] The fifth fully connected layer contains 32 neurons, which are fully connected to the 256 neurons in the previous layer, and the activation function is the ReLU function;

[0023] The output layer contains 1 neuron, and the activation function is the sigmoid function.

[0024] In summary, this application includes the following beneficial technical effects:

[0025] The beneficial effects that can be achieved by at least one of the above-mentioned technical solutions adopted in the present application include at least the following: the present application utilizes the differences in the rules and elements of component model specifications composed by different component manufacturers to generate personalized coded images of component model information. The component information coded image generation and recognition method formed based on the local perception ability and feature extraction ability of the CNN neural network for image recognition effectively improves the efficiency and convenience of component information classification and recognition, and further reduces the labor cost caused by the complexity of component data information, so as to achieve efficient component information processing and more convenient classification and recognition effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 A flowchart of a component information coding image generation and recognition method based on a neural network provided in an embodiment of the present application;

[0028] Figure 2 An example diagram of a coded image generated for the component model information provided in an embodiment of the present application;

[0029] Figure 3 This is an example diagram of the CNN neural network structure provided for the embodiments of this application. DETAILED DESCRIPTION

[0030] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0031] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0032] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0033] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application. The illustrations only show the components related to the present application rather than being drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed at will, and the component layout may also be more complicated.

[0034] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0035] The embodiment of the present application provides a method for generating a component information coding image based on a neural network.

[0036] like Figure 1 As shown, a method for generating component information coding images based on a neural network includes:

[0037] Step 1: Analyze the model composition rules of component manufacturers, and obtain and organize the manufacturer's component model data information.

[0038] Step 2: Convert the component model information into a string for encoding; convert the component model information into a binary string for encoding.

[0039] For example, if the component model information is “TLE9015QU”, its binary code can be converted into “010101000100110001000101001110010011000000110001001101010101000101010101”.

[0040] Step 3, convert the code into an information matrix; first expand the binary code to a uniform length, and then convert it into an information matrix.

[0041] Specifically including: S31. Calculate the longest binary code length of the component model specification, and expand the binary code string to the same length to generate a consistent matrix size. The unified extension length provided in the embodiment of the present application is 256 bits. Convert its code from string form to sequence form.

[0042] S32. Re-divide the expanded binary coding sequence of component information to generate a coding matrix of component model information. The matrix provided in the embodiment of the present application has a size of 16*16.

[0043] Step 4, converting the component coding information matrix into a two-dimensional coding image containing component model information;

[0044] According to the component information coding matrix generated in the previous step, each element in the matrix is ​​used as the pixel value of the component coding image, so as to draw and generate the component information coding image. In the embodiment of the present application, the model information of the SAF-XC886CM-8FFI 5V AC microcontroller is taken as an example, and its information coding image is as follows: Figure 2 As shown in . Each element in the information matrix is ​​used as the pixel value of the component coding image, so as to draw and generate the component information coding image.

[0045] The embodiment of the present application also discloses a component information coded image recognition method based on a neural network, which is used to recognize the two-dimensional coded image generated by the above generation method, and the recognition method includes:

[0046] Construct a CNN neural network model; the CNN neural network model includes an input layer, a hidden layer, and an output layer. Among them, the hidden layer includes a convolutional layer, a pooling layer, and a fully connected layer;

[0047] In the input layer, the image is composed of a three-dimensional matrix, including its length, width and RGB color channels. In this layer, the basic information of the image to be recognized is clarified.

[0048] The convolution layer is used to extract image features. Each element in the convolution kernel corresponds to a weight coefficient and a bias vector. Each neuron in the convolution layer is connected to several neurons in the perception area near the input image of the previous layer. The size of the area depends on the size of the convolution kernel. When the convolution layer is running, it scans the input image features, multiplies and sums the matrix elements of the input features in the perception area, and superimposes the bias vector:

[0049]

[0050] In the above formula, w is the weight, b is the bias vector, and I l and I l+1 Represents the convolution input and output in layer l+1, i.e., the feature map, L l+1Refers to the characteristic graph I l+1 I(i, j) is the pixel corresponding to the feature map, K refers to the number of channels in the feature map, f, d0, p are the convolution kernel size, convolution stride, and the number of pixels of the padding edge around the convolution layer input image.

[0051] The pooling layer is used in the convolution layer to extract features, and the output feature map will be sent to the pooling layer for feature selection and information filtering. Pooling layers are usually divided into two types, maximum pooling and average pooling. Maximum pooling divides the input image into rectangular areas, and each area outputs the maximum value to enter the next layer. This pooling method only extracts the strongest part of the feature map, discards a lot of redundant information, and can retain more texture information of the image while losing detail information. After average pooling divides the image into rectangular areas, each area outputs the average value of all elements and passes it to the next layer. This pooling method can retain more background information of the image.

[0052] The fully connected layer is the last layer in the CNN hidden layer. The feature map will be expanded into a vector in the fully connected layer and calculated through the activation function.

[0053] The choice of activation function is usually based on ReLU function, softmax function and siigmoid function. Among them, ReLU activation function has sparseness, high efficiency and biological rationality, which can alleviate the problem of gradient disappearance to a certain extent and accelerate gradient convergence. Its calculation formula is as follows:

[0054] f(x)=max(0,x) (2);

[0055] The Softmax function has a normalization characteristic that can enhance the significance of input features and is widely used in multi-classification problems. Its calculation formula is as follows:

[0056]

[0057] Sigmoid function gradient smoothing is mostly used for binary classification problems. Its calculation formula is as follows:

[0058]

[0059] The fully connected layer contains 32 neurons, which are fully connected to the 256 neurons in the previous layer, and the activation function is the ReLU function;

[0060] The output layer contains 1 neuron, and the activation function is the sigmoid function.

[0061] like Figure 3 As shown, in the embodiment of the present application, the convolution layer and the pooling layer are respectively the first convolution layer, the second pooling layer, the third convolution layer, and the fourth pooling layer.

[0062] The first convolutional layer uses 32 convolution kernels for convolution operation, the convolution kernel size is 3*3, and the activation function is ReLU.

[0063] The second pooling layer uses the maximum pooling method with a window size of 2*2.

[0064] The third convolutional layer uses 64 convolution kernels of size 3*3 for convolution operation, and the activation function is the ReLU function;

[0065] The fourth pooling layer uses the maximum pooling method with a window size of 2*2.

[0066] The neural network model is trained to conduct deep learning on the structural characteristics of component model information, and the trained model is used to classify component model information from different research and production manufacturers, thereby obtaining a component information recognition method.

[0067] The specific training process includes:

[0068] The generated component information encoding images are divided into positive samples and negative samples. The positive samples are the component information encoding images of the manufacturing manufacturer to be identified, and the negative samples are the component information encoding images of the other manufacturing manufacturers.

[0069] The positive and negative sample data sets of component information encoding images are divided into training set, validation set and test set, among which the training set accounts for 70% of the data volume of each sample set, and the validation set and test set each account for 15% of the data volume of each sample set.

[0070] Based on the constructed CNN neural network basic model, model training is performed on the divided training set. After training, the model parameters are optimized and adjusted through the validation set. Finally, the accuracy of the model is evaluated through the test set to obtain the final result.

[0071] In summary, the present application converts component model information into binary form to initially form information coding, builds the basic model structure of the CNN neural network, further expands the component information coding to a uniform length and processes it, constructs a two-dimensional coded image containing the component model information, trains and adjusts the constructed CNN neural network model based on the component information coding image, and tests and evaluates the component information recognition and classification effect of the optimized and adjusted CNN neural network model.

[0072] This application utilizes the construction rules and specificity of the constituent elements of component models and specifications of different manufacturers, converts the differences in their characters into image features, retains the construction characteristics of the component models and specifications of each manufacturer, and utilizes the high efficiency and flexibility of CNN neural network for image recognition to perform feature learning of component information encoding images, thereby improving the efficiency of component information recognition.

[0073] The component information coding image generation and recognition method based on CNN neural network provided in the present application is efficient and flexible. Compared with the traditional method, it further improves the efficiency and accuracy of manual identification of component information. When the other auxiliary information of the component is lost, it guarantees the accuracy and stability of component model information identification to a certain extent, thereby achieving better security and accuracy of information recognition effect.

[0074] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for generating component information coding images based on a neural network, characterized in that: include: Step 1: Analyze the model composition rules of component manufacturers, obtain and organize the manufacturer's component model data information; Step 2, convert the component model information into a string for encoding; Step 3, convert the code into an information matrix; Step 4: Convert the component coding information matrix into a two-dimensional coded image containing component model information.

2. The method for generating component information coded images based on a neural network according to claim 1, characterized in that: In step 2, the component model information is converted into a binary string for encoding.

3. The method for generating component information coded images based on a neural network according to claim 2, characterized in that: In step 3, the binary code is first extended to a uniform length and then converted into an information matrix.

4. The method for generating component information coded images based on a neural network according to claim 3, characterized in that: In step 4, each element in the information matrix is ​​used as a pixel value of the component coded image, so as to draw and generate the component information coded image.

5. A component information coding image recognition method based on neural network, characterized in that: For identifying the two-dimensional coded image generated as claimed in claims 2 to 4, the recognition method comprises the following steps: constructing a CNN neural network model; The neural network model is trained to conduct deep learning on the structural characteristics of component model information, and the trained model is used to classify component model information from different research and production manufacturers, thereby obtaining a component information recognition method.

6. The component information coding image recognition method based on neural network according to claim 5 is characterized in that: The CNN neural network model includes a first convolutional layer, a second pooling layer, a third convolutional layer, a fourth pooling layer, a fifth fully connected layer and an output layer; The first convolutional layer uses 32 convolution kernels for convolution operation, the convolution kernel size is 3*3, and the activation function is ReLU; The second pooling layer uses the maximum pooling method with a window size of 2*2; The third convolutional layer uses 64 convolution kernels of size 3*3 for convolution operation, and the activation function is the ReLU function; The fourth pooling layer uses the maximum pooling method with a window size of 2*2; The fifth fully connected layer contains 32 neurons, which are fully connected to the 256 neurons in the previous layer, and the activation function is the ReLU function; The output layer contains 1 neuron, and the activation function is the sigmoid function.

Images