Molten iron torpedo ladle number visual identification method and molten iron torpedo ladle number visual identification device

Abstract

The invention provides a molten iron torpedo ladle number visual identification method. The method comprises the following steps: generating a two-dimensional code according to a serial number of a molten iron torpedo ladle; fixing the two-dimensional code at a preset position of a corresponding molten iron torpedo ladle; shooting the two-dimensional code by using a camera to form a two-dimensional code image; preprocessing the two-dimensional code image to obtain a preprocessed two-dimensional code image; setting a corresponding label for the preprocessed two-dimensional code image to form a data set; inputting the data set into a deep recognition model for training to obtain a molten iron torpedo ladle number visual recognition model; and the identification of the target molten iron torpedo ladle number is completed by using the molten iron torpedo ladle number visual identification model. According to the method, the deep learning model is adopted, so that the serial number identification process of the molten iron torpedo ladle is automatic, the working efficiency is improved, and the error rate of manual identification is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of tank number recognition of industrial torpedo tanks, and in particular to a tank number visual recognition method and a device for molten iron torpedo tanks. Background Art

[0002] QR code is a matrix graphic code that stores a large amount of information. Due to its high-density encoding, easy to read and easy to generate characteristics, it has been widely used in civil and industrial fields. In the field of commodity logistics, QR code can be used to trace the source of products, verify authenticity and provide product information. In the field of mobile payment, QR code can be used as a payment code, allowing users to perform convenient payment operations by scanning QR codes. In the field of industrial manufacturing, QR codes are used to identify products and parts, record production information, quality inspection results, supply chain information, etc. However, the reliability and robustness of QR code recognition in the industrial field is an important research issue. Due to interference from factors such as industrial environment, image contamination, and light changes, the quality of QR code images will deteriorate and cannot be accurately recognized. Therefore, it is of great significance to study robust QR code recognition methods.

[0003] QR code technology can be applied to the industrial scenario of identifying the tank number of molten iron torpedo tanks to realize the automatic identification and analysis of molten iron torpedo tanks. However, molten iron torpedo tanks are special containers for holding high-temperature liquid iron. The QR code on its surface is easily corroded by liquid metal, causing the boundary of the QR code and the information code block to deform, which puts higher requirements on the robustness of the QR code and the reliability of the recognition algorithm. The shortcomings of the existing common QR code recognition methods are as follows:

[0004] (1) The two-dimensional code recognition method based on traditional machine vision (such as Hough transform) has poor ability to extract image features, which makes it more sensitive to image noise interference and unable to achieve accurate recognition.

[0005] (2) Detection methods based on deep networks, such as the R-CNN algorithm, have low detection efficiency and lack real-time performance when performing QR code detection and recognition.

[0006] (3) The existing two-dimensional code recognition technology is not suitable for two-dimensional code recognition tasks in complex environments. It has poor robustness to environmental changes and cannot meet the high reliability requirements of industrial production. It is not conducive to the identification of molten iron torpedo tanks. Summary of the invention

[0007] In order to overcome the deficiencies of the prior art, the object of the present invention is to provide a method and device for visually identifying the tank number of a molten iron torpedo tank.

[0008] To achieve the above objectives, the present invention provides the following solutions:

[0009] A method for visually identifying a tank number of a molten iron torpedo tank, comprising:

[0010] Step 1: Generate a QR code according to the serial number of the molten iron torpedo tank;

[0011] Step 2: Fix the QR code at a preset position of the corresponding molten iron torpedo tank;

[0012] Step 3: Use a camera to capture the QR code to form a QR code image;

[0013] Step 4: preprocessing the two-dimensional code image to obtain a preprocessed two-dimensional code image;

[0014] Step 5: Set corresponding labels for the preprocessed QR code images to form a data set;

[0015] Step 6: Input the data set into the deep recognition model for training to obtain a molten iron torpedo tank number visual recognition model;

[0016] Step 7: Use the molten iron torpedo tank number visual recognition model to complete the recognition of the target molten iron torpedo tank number.

[0017] Preferably, in step 1, the QR code is a robust QR code; the code area of ​​the robust QR code is divided into four parts, each part is in the shape of a square, each square is evenly divided into 4 blocks, a hollow circle is used in each block to represent "1" in the binary code, and a blank space represents "0" in the binary code, and the converted binary code of each block is arranged in a "Z" shape sequence to identify the 4-digit binary number of different molten iron torpedo tanks.

[0018] Preferably, in step 6, the deep recognition model is a YOLOv5lite network; the YOLOv5lite network is constructed by replacing the DarkNet53 network in the YOLOv5 network with the MobileNet network.

[0019] Preferably, the step 4: preprocessing the two-dimensional code image to obtain a preprocessed two-dimensional code image includes:

[0020] Convert the two-dimensional code image into RGB color space to form a first image;

[0021] Performing feature enhancement processing on the first image based on the values ​​of each color channel in the first image to form a feature-enhanced first image;

[0022] Gaussian noise is added to the feature-enhanced first image and the original two-dimensional code image to expand the data samples to form a preprocessed two-dimensional code image.

[0023] Preferably, performing feature enhancement processing on the first image based on the values ​​of each color channel in the first image to form the feature-enhanced first image includes:

[0024] Calculate the mean of each color channel in the first image;

[0025] Calculating a pixel mean of the first image according to the mean of each color channel;

[0026] Among them, the pixel mean calculation formula of the first image is:

[0027]

[0028] Among them, R ave represents the mean value of the red color channel, G ave represents the mean value of the green color channel, B ave represents the mean value of the blue color channel, M represents the length of the first image, N represents the width of the first image, and f R (i, j) represents the value of the red channel at point (i, j), f G (i, j) represents the value of the green channel at point (i, j), f B (i, j) represents the value of the blue color channel at point (i, j), and K represents the pixel mean of the first image;

[0029] Calculating an enhancement coefficient according to a pixel mean of the first image;

[0030] The enhancement coefficient is used to perform feature enhancement on the first image to form a feature-enhanced first image.

[0031] Preferably, using the enhancement coefficient to perform feature enhancement on the first image to form a feature-enhanced first image includes:

[0032] Using the formula:

[0033]

[0034] The first image is feature enhanced to form a first image after feature enhancement; wherein the first enhancement coefficient The second enhancement factor The third enhancement factor represents the first image, Represents the first image after feature enhancement.

[0035] The present invention also provides a molten iron torpedo tank number visual recognition device, comprising:

[0036] A molten iron torpedo tank, on which a QR code is fixed;

[0037] A camera is used to take a picture of the QR code and upload the QR code image to a host computer;

[0038] The host computer is used to identify the serial number of the molten iron torpedo tank according to the QR code image.

[0039] Preferably, it also includes:

[0040] The Bluetooth module is connected to the host computer for communication and is used for receiving the two-dimensional code image sent by the camera and uploading the two-dimensional code image to the host computer.

[0041] Preferably, it further comprises: a target molten iron torpedo tank reminder module, the target molten iron torpedo tank reminder module comprises:

[0042] A current limiting resistor R1, one end of which is connected to a host computer;

[0043] A transistor Q, wherein the base of the transistor Q is connected to the other end of the current limiting resistor R1, and the emitter of the transistor Q is grounded;

[0044] A shunt resistor R2, one end of the shunt resistor R2 is connected to the base of the transistor, and the other end of the shunt resistor R2 is grounded;

[0045] A voltage regulator tube D, wherein the anode of the voltage regulator tube D is connected to the collector of the transistor Q, the cathode of the voltage regulator tube D is connected to the positive electrode of the buzzer, and the negative electrode of the buzzer is connected to the collector of the transistor Q.

[0046] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0047] The present invention provides a method and device for visually identifying tank numbers of molten iron torpedo tanks. Compared with the prior art, the present invention adopts a deep learning model to automate the number recognition process of molten iron torpedo tanks, improves work efficiency, and reduces the error rate of manual recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] 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 will be briefly introduced below. Obviously, the drawings described below are 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.

[0049] Figure 1 The schematic diagram of the target molten iron torpedo tank reminder module provided by the present invention;

[0050] Figure 2 A flow chart of a method for visually identifying a tank number of a molten iron torpedo tank provided by the present invention;

[0051] Figure 3 A schematic diagram of a method for visually identifying tank numbers of molten iron torpedo tanks provided by the present invention;

[0052] Figure 4 A schematic diagram of the robust two-dimensional code construction provided by the present invention;

[0053] Figure 5 This is a first robust two-dimensional code recognition effect diagram provided by the present invention;

[0054] Figure 6 This is a second robust two-dimensional code recognition effect diagram provided by the present invention;

[0055] Figure 7 This is a third robust two-dimensional code recognition effect diagram provided by the present invention;

[0056] Figure 8 This is a fourth robust two-dimensional code recognition effect diagram provided by the present invention;

[0057] Fig. 9 This is a fifth robust two-dimensional code recognition effect diagram provided by the present invention;

[0058] Fig.10 This is a sixth robust two-dimensional code recognition effect diagram provided by the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0061] The terms "first", "second", "third" and "fourth" in the specification and claims of the present application and the drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a series of steps, processes, methods, etc. are not limited to the listed steps, but may optionally include steps that are not listed, or may optionally include other step elements inherent to these processes, methods, products or devices.

[0062] In order to make the above-mentioned objectives, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0063] The present invention provides a molten iron torpedo tank number visual recognition device, comprising:

[0064] A molten iron torpedo tank, on which a QR code is fixed;

[0065] A camera is used to take a picture of the QR code and upload the QR code image to a host computer;

[0066] The host computer is used to identify the serial number of the molten iron torpedo tank according to the QR code image.

[0067] Furthermore, it also includes:

[0068] The Bluetooth module is connected to the host computer for communication and is used for receiving the two-dimensional code image sent by the camera and uploading the two-dimensional code image to the host computer.

[0069] The target molten iron torpedo tank reminder module is connected to the host computer for communication and is used to issue a reminder when the host computer recognizes the marked molten iron torpedo tank number.

[0070] For further information, see Figure 1 The target molten iron torpedo tank reminder module includes: a current limiting resistor, a triode, a voltage regulator tube, a shunt resistor and a buzzer.

[0071] A current limiting resistor R1, one end of which is connected to a host computer; a transistor Q, a base of which is connected to the other end of the current limiting resistor R1, and an emitter of which is grounded; one end of a shunt resistor R2 is connected to the base of the transistor, and the other end of which is grounded; an anode of a voltage-stabilizing tube D is connected to the collector of the transistor Q, a cathode of the voltage-stabilizing tube D is connected to the positive electrode of a buzzer, and a negative electrode of the buzzer is connected to the collector of the transistor Q. When the host computer recognizes the number of the marked molten iron torpedo tank, the host computer will send a high level to the current limiting resistor R1, turn on the transistor Q, and then power on the buzzer to issue a reminder.

[0072] The present invention fixes the QR code on the molten iron torpedo tank and can automatically and quickly and accurately identify the serial number of the corresponding molten iron torpedo tank according to the QR code. Compared with manual search and recording, the automated identification method of the present invention can significantly shorten the identification time and improve work efficiency.

[0073] See also Figure 2-3 The present invention also provides a method for visually identifying the tank number of a molten iron torpedo tank, the visual recognition method is applied to the above-mentioned molten iron torpedo tank number visual recognition device, comprising:

[0074] Step 1: Generate a QR code according to the serial number of the molten iron torpedo tank;

[0075] The amount of information stored in the traditional two-dimensional code based on QR coding is large, and most of them adopt complex coding methods, and the pixel block matrix size of the two-dimensional code is more than 22*22. However, complex coding can cause the recognition effect to decline, and is easily affected by the environment, and the robustness is poor. Considering that the molten iron torpedo tank is numbered as a fixed four-digit number in the molten iron torpedo tank tank number recognition task, the present invention draws on traditional QR coding and makes some simplifications, designs a set of special coding rules (SQR coding), and generates a more robust two-dimensional code for each molten iron torpedo tank number based on the SQR coding rule, which is called robust two-dimensional code. Robust two-dimensional code is composed of 4*4 pixels, which is used to identify different molten iron torpedo tanks. Compared with traditional two-dimensional code, robust two-dimensional code has fewer pixels, so it is more robust to environmental changes.

[0076] See also Figure 4 In step 1, the two-dimensional code is a robust two-dimensional code; the code area of ​​the robust two-dimensional code is divided into four parts, each part is in the shape of a square, each square is evenly divided into 4 blocks, a hollow circle is used in each block to represent "1" in the binary code, and a blank space represents "0" in the binary code, and the converted binary code of each block is arranged in a "Z" shape sequence to identify the 4-digit binary number of different molten iron torpedo tanks.

[0077] In practical application, the present invention first uses the SQR coding rule to generate a corresponding robust two-dimensional code for each molten iron torpedo tank, and the two-dimensional code is composed of 4*4 pixel blocks. The robust two-dimensional code is printed on a heat-resistant material and fixed at a specific position of the molten iron torpedo tank.

[0078] Since the numbers of all molten iron torpedo tanks are composed of 4 digits, and the usage scenario is a steel plant. The present invention draws on the traditional QR code and makes some simplifications when designing the code, and designs a set of special coding rules (SQR code). The SQR code divides the code area into four parts, each of which displays the 4 digits of the code. The display method of each digit draws on the basic method of QR coding, and converts the coded number into a 4-bit binary code. The converted binary code is displayed in the area in a "Z" shape. The SQR code uses a hollow circle to represent the "1" in the binary code, and does not perform any operation on the code "0". The two-dimensional code generated by this coding rule is simple, easy to generate, and robust.

[0079] Step 2: Fix the QR code at a preset position of the corresponding molten iron torpedo tank;

[0080] Step 3: Use a camera to capture the QR code to form a QR code image;

[0081] Step 4: preprocessing the two-dimensional code image to obtain a preprocessed two-dimensional code image;

[0082] Furthermore, step 4 includes:

[0083] Step 4.1: Convert the two-dimensional code image into RGB color space to form a first image;

[0084] Step 4.2: performing feature enhancement processing on the first image based on the values ​​of each color channel in the first image to form a feature-enhanced first image;

[0085] Among them, step 4.2 includes:

[0086] Calculate the mean of each color channel in the first image;

[0087] Calculating a pixel mean of the first image according to the mean of each color channel;

[0088] Among them, the pixel mean calculation formula of the first image is:

[0089]

[0090] Among them, R ave represents the mean value of the red color channel, G ave represents the mean value of the green color channel, B ave represents the mean value of the blue color channel, M represents the length of the first image, N represents the width of the first image, and f R (i, j) represents the value of the red channel at point (i, j), f G (i, j) represents the value of the green channel at point (i, j), f B(i, j) represents the value of the blue color channel at point (i, j), and K represents the pixel mean of the first image;

[0091] Calculating an enhancement coefficient according to a pixel mean of the first image;

[0092] Performing feature enhancement on the first image using the enhancement coefficient to form a feature-enhanced first image;

[0093] In the implementation of the present invention, the formula can be used:

[0094]

[0095] The first image is feature enhanced to form a first image after feature enhancement; wherein the first enhancement coefficient The second enhancement factor The third enhancement factor represents the first image, Represents the first image after feature enhancement.

[0096] The present invention can effectively suppress the color deviation caused by light, rain or the illumination of the camera itself by balancing the color tone of the image, thereby ensuring the color stability of the two-dimensional code image.

[0097] Gaussian noise is added to the feature-enhanced first image and the original two-dimensional code image to expand the data samples to form a preprocessed two-dimensional code image.

[0098] Step 5: Set corresponding labels for the preprocessed QR code images to form a data set;

[0099] Step 6: Input the data set into the deep recognition model for training to obtain a molten iron torpedo tank number visual recognition model;

[0100] The present invention uses a data set expansion method that adds Gaussian noise. Gaussian noise is a common type of image noise that has randomness and continuity. The deep recognition model can learn features that are robust to noise, thereby improving the adaptability of the model in the actual environment and improving the robustness of the model. In addition, adding Gaussian noise to the image is also a method of data enhancement. Through data enhancement, the scale of the training data set can be expanded, more changes and diversity can be introduced, thereby improving the generalization ability of the model and preventing overfitting. After image preprocessing, labelme software is used to annotate the pictures in the data set, and a label file is generated for each picture containing a QR code. For the YOLOv5 model, the content of the label file is the class number class of the QR code and the relative position x, y, w, h of the QR code area.

[0101] In the present invention, the robust two-dimensional code visual detection and recognition module uses a robust two-dimensional code detection and recognition model based on YOLOv5. As a commonly used target detection model, the YOLOv5 model has the advantages of fast speed, high accuracy and lightweight model. The classic YOLOv5 model consists of three parts: the DarkNet53 backbone network, the neck structure and the prediction head. The backbone network is responsible for extracting the features of the input image, the neck structure is used to better fuse the extracted features, and the prediction head completes the prediction of the target confidence. Since the backbone network of YOLOv5 can better complete the detection and recognition tasks and is relatively lightweight compared to the two-stage detection networks such as R-CNN, the present invention uses the corresponding parts of the YOLOv5 network in the selection of the backbone network and the selection of the prediction head. The problem of poor real-time detection caused by the use of the two-stage detection network is avoided, and the detection speed of the network is improved.

[0102] In the present invention, in order to further enhance the real-time and robustness of the YOLOv5 network, the original DarkNet53 network is replaced by the MobileNet network as the backbone network, and a lightweight YOLOv5lite network is proposed. MobileNet is a lightweight convolutional neural network that aims to provide efficient computing performance and good image classification and feature extraction capabilities. Replacing the backbone in the YOLOv5 network with MobileNet has many advantages, which can effectively improve its real-time and robustness in the molten iron torpedo tank number recognition task.

[0103] MobileNet is a lightweight convolutional neural network with fewer parameters and computational complexity than the default backbone network (DarkNet53 network) of YOLOv5. Due to this streamlined design, using MobileNet as the backbone of YOLOv5 can significantly improve the inference speed and real-time performance of the model. This is very important for applications that require target detection and tracking in real-time scenarios. In the task of identifying the tank number of a molten iron torpedo tank, the real-time requirement is high. Therefore, the present invention uses the MobileNet network as the backbone network to greatly improve the detection real-time performance.

[0104] MobileNet is designed with not only lightness but also robustness of the model in mind. It uses a depthwise separable convolution structure to reduce the amount of computation while maintaining good feature representation capabilities. This structure can better capture details and local information in the image, allowing YOLOv5 to better adapt to target detection tasks in different scenarios and have better robustness. In the present invention, for the problem of molten iron torpedo tank number recognition in different scenarios, using the MobileNet network as the backbone network of YOLOv5 can also improve the robustness of molten iron torpedo tank number recognition.

[0105] Step 7: Use the molten iron torpedo tank number visual recognition model to complete the recognition of the target molten iron torpedo tank number.

[0106] The identification method of the present invention is further described below in conjunction with specific embodiments:

[0107] The first step is based on Figure 4 The SQR encoding rule shown generates a robust two-dimensional code, which can represent all numbers from 0 to 9999. According to the numbering requirements of the molten iron torpedo tank, the present invention generates a total of 86 robust two-dimensional codes from 2601 to 2686, and fixes the two-dimensional code at a fixed position in the molten iron torpedo tank to prepare for the data set production.

[0108] The second step is to use Hikvision cameras to shoot QR code images and build a robust QR code image dataset. The images in the dataset are then preprocessed. The specific implementation steps are to add Gaussian noise to the image and expand the data samples.

[0109] The third step is to generate a corresponding label file for each image. The label file indicates the QR code position information x, y, w, h, and target category information class corresponding to each image in the dataset. In order to distinguish the data used for training and testing, the dataset is divided into training set and test set in a ratio of 3:1.

[0110] Step 4: Modify the original YOLOv5 network structure and use the MobileNet network to replace the original DarkNet53 network as the backbone network of YOLOv5. The specific operation is to convert the original convolution (Conv) into a deep separable convolution (DConv) in the original CBL network structure.

[0111] Step 5: Initialize the YOLOv5lite network, specify the training set and test set, and perform iterative training according to the preset loss function. In the present invention, the original YOLOv5 default training parameters are used to implement training, and a complete training model is obtained after the training is completed.

[0112] Step 6: Pass the QR code images in the test set into the complete training model for forward reasoning, and finally output the detection results.

[0113] The implementation results are as follows:

[0114] This example selects 1 / 4 of the images in the QR code dataset as test objects, and the image resolution is 1920*1680. The detection results are as follows: Figure 5-10 As shown in the figure, the left column is the original QR code image, the right column is the network detection result, and the number in the upper left corner represents the recognition result, indicating that the detection and recognition model detects the area where the QR code is located and realizes the QR code classification according to the encoding features. It can be seen from the figure that the proposed improved network can accurately detect the location of the QR code under different numbers and identify the molten iron torpedo tank number corresponding to the QR code (accurate classification). Figure 6 It also shows that the network in this embodiment is more robust to environmental changes and still has good results under high brightness conditions. In terms of detection speed, the inference speed of the YOLOv5lite visual detection model proposed in the present invention on a server equipped with a 3080ti graphics card can reach more than 30FPS, meeting the real-time standard.

[0115] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the method disclosed in the embodiment, since it corresponds to the device disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the device part description.

[0116] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for visually identifying the tank number of a molten iron torpedo tank, characterized in that: include: Step 1: Generate a QR code according to the serial number of the molten iron torpedo tank; Step 2: Fix the QR code at a preset position of the corresponding molten iron torpedo tank; Step 3: Use a camera to capture the QR code to form a QR code image; Step 4: preprocessing the two-dimensional code image to obtain a preprocessed two-dimensional code image; Step 5: Set corresponding labels for the preprocessed QR code images to form a data set; Step 6: Input the data set into the deep recognition model for training to obtain a molten iron torpedo tank number visual recognition model; Step 7: Use the molten iron torpedo tank number visual recognition model to complete the recognition of the target molten iron torpedo tank number.

2. A method for visually identifying the tank number of a molten iron torpedo tank as claimed in claim 1, characterized in that: In step 1, the two-dimensional code is a robust two-dimensional code; the code area of ​​the robust two-dimensional code is divided into four parts, each part is in the shape of a square, each square is evenly divided into 4 blocks, a hollow circle in each block represents "1" in the binary code, and a blank represents "0" in the binary code, and the converted binary code of each block is arranged in a "Z" shape sequence to identify the 4-digit binary number of different molten iron torpedo tanks.

3. A method for visually identifying the tank number of a molten iron torpedo tank as claimed in claim 2, characterized in that: In step 6, the deep recognition model is a YOLOv5lite network; the YOLOv5lite network is constructed by replacing the DarkNet53 network in the YOLOv5 network with the MobileNet network.

4. A method for visually identifying the tank number of a molten iron torpedo tank as claimed in claim 1, characterized in that: The step 4: preprocessing the two-dimensional code image to obtain a preprocessed two-dimensional code image, comprises: Convert the two-dimensional code image into RGB color space to form a first image; Performing feature enhancement processing on the first image based on the values ​​of each color channel in the first image to form a feature-enhanced first image; Gaussian noise is added to the feature-enhanced first image and the original two-dimensional code image to expand the data samples to form a preprocessed two-dimensional code image.

5. A method for visually identifying the tank number of a molten iron torpedo tank as claimed in claim 4, characterized in that: Performing feature enhancement processing on the first image based on the values ​​of each color channel in the first image to form a feature-enhanced first image includes: Calculate the mean of each color channel in the first image; Calculating a pixel mean of the first image according to the mean of each color channel; Among them, the pixel mean calculation formula of the first image is: Among them, R ave represents the mean value of the red color channel, G ave represents the mean value of the green color channel, B ave represents the mean of the blue color channel, M represents the length of the first image, N represents the width of the first image, and f R (i, j) represents the value of the red channel at point (i, j), f G (i, j) represents the value of the green color channel at point (i, j), f B (i, j) represents the value of the blue color channel at point (i, j), and K represents the pixel mean of the first image; Calculating an enhancement coefficient according to a pixel mean of the first image; The enhancement coefficient is used to perform feature enhancement on the first image to form a feature-enhanced first image.

6. A method for visually identifying the tank number of a molten iron torpedo tank as claimed in claim 5, characterized in that: Performing feature enhancement on the first image by using the enhancement coefficient to form a feature-enhanced first image includes: Using the formula: Performing feature enhancement on the first image to form a first image after feature enhancement; wherein the first enhancement coefficient The second enhancement factor The third enhancement factor represents the first image, Represents the first image after feature enhancement.

7. A device for visually identifying the tank number of a molten iron torpedo tank, characterized in that: include: A molten iron torpedo tank, on which a QR code is fixed; A camera is used to take a picture of the QR code and upload the QR code image to a host computer; The host computer is used to identify the serial number of the molten iron torpedo tank according to the QR code image.

8. A molten iron torpedo tank number visual recognition device as claimed in claim 7, characterized in that: Also includes: The Bluetooth module is connected to the host computer for communication and is used for receiving the two-dimensional code image sent by the camera and uploading the two-dimensional code image to the host computer.

9. A molten iron torpedo tank number visual recognition device as claimed in claim 8, characterized in that: Also includes: A target molten iron torpedo tank reminder module, wherein the target molten iron torpedo tank reminder module comprises: A current limiting resistor (R1), one end of which is connected to a host computer; A transistor (Q), wherein the base of the transistor (Q) is connected to the other end of the current limiting resistor (R1), and the emitter of the transistor (Q) is grounded; A shunt resistor (R2), one end of which is connected to the base of the transistor, and the other end of which is grounded; A voltage regulator tube (D), wherein the anode of the voltage regulator tube (D) is connected to the collector of the triode (Q), the cathode of the voltage regulator tube (D) is connected to the positive electrode of the buzzer, and the negative electrode of the buzzer is connected to the collector of the triode (Q).

Images