Steel plate blank character recognition method, equipment, storage medium and device

Abstract

The present invention discloses a method, equipment, storage medium and device for recognizing characters of steel plate blanks. The present invention performs image preprocessing on the steel plate blank image by preset camera parameters, extracts pixel feature points from the target frame image, and judges whether the target frame image contains target character features according to the pixel feature points. When the target frame image contains target character features, the target character features are recognized according to a preset deep learning model to obtain a character recognition result. Compared with the prior art that only recognizes the slab moving on a track by an AI algorithm, which results in the inability to completely capture and recognize characters, the present invention performs image preprocessing and judges whether the target character features are contained according to the character features in the target frame image, and performs character recognition on the pixel feature points by a preset deep learning model, thereby achieving complete capture and recognition of slab characters moving on a track under different environments, thereby improving the robustness and stability of the system.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular to a steel plate blank character recognition method, equipment, storage medium and device. Background Art

[0002] With the development of industrial automation, in the steel industry, enterprises have gradually increased their requirements for the level of automation in logistics management. Steel slab information, namely steel slab coding, is important information for the logistics management and production process of steel enterprises. It is usually represented by certain letters or numbers and their combination codes. Automatically obtaining steel slab information is of great significance to steel slab tracking in logistics management and enterprise automation and information production.

[0003] At present, in most steel mills, the recognition of steel slab codes is still based on manual recognition, and there is a lack of measures for automatic recognition of steel slab codes. At the same time, the existing character recognition technology is mainly used in occasions with simple optical imaging backgrounds, but in the actual production environment, the lighting environment is complex, and the slabs are moving on the track. If the algorithm software does not recognize the characters, it is difficult to capture the characters completely, but if the characters are not captured completely, the AI ​​algorithm cannot recognize the characters. This creates a paradox. In addition, the inefficiency of manual recognition seriously affects the automation level of steel slab tracking in logistics management and the improvement of enterprise production systems.

[0004] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of the present invention is to provide a method, equipment, storage medium and device for steel plate slab character recognition, aiming to solve the technical problem in the prior art of using AI algorithm to identify slabs moving on tracks, resulting in the inability to fully capture and recognize characters.

[0006] To achieve the above object, the present invention provides a steel plate slab character recognition method, the steel plate slab character recognition method comprising the following steps:

[0007] Acquire a steel plate slab image, and simultaneously perform image preprocessing on the steel plate slab image according to preset camera parameters to obtain a target frame image;

[0008] Extracting pixel feature points from the target frame image, and judging whether the target frame image contains target character features according to the pixel feature points;

[0009] When the target frame image contains target character features, the target character features are recognized according to a preset deep learning model to obtain a character recognition result.

[0010] Optionally, the step of extracting pixel feature points from the target frame image and determining whether the target frame image contains target character features based on the pixel feature points comprises:

[0011] Determine the threshold range according to the preset sampling rate and the preset OTSU algorithm;

[0012] Determine a threshold value for the operation to be performed according to the threshold range and the preset threshold value;

[0013] Pixel feature points are extracted from the target frame image according to the threshold of the operation to be performed, and at the same time, it is determined whether the target frame image contains target character features according to the pixel feature points.

[0014] Optionally, when the target frame image contains a target character feature, the step of identifying the target character feature according to a preset deep learning model to obtain a character recognition result includes:

[0015] When the target frame image contains target character features, obtaining the number of pixel feature points;

[0016] The target character features are identified according to the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

[0017] Optionally, the step of identifying the target character feature according to the number of pixel feature points and a preset deep learning model to obtain a character recognition result includes:

[0018] Obtain the number of pixel feature points corresponding to each target frame image at a preset acquisition rate, and determine the maximum and minimum values ​​of the number of pixel feature points;

[0019] Obtain the index position values ​​corresponding to the maximum value and the minimum value;

[0020] The target character features are identified according to the index position value, the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

[0021] Optionally, the step of identifying the target character feature according to the index position value, the number of pixel feature points and a preset deep learning model to obtain a character recognition result includes:

[0022] When the index position value of the minimum value appears at the tail of the queue and the number of occurrences is greater than a preset number, a stop copy semaphore is generated;

[0023] Based on the stop copy signal, the image file containing the spray-marked characters that is most recently saved in the hard disk folder is traversed, and the target character features of the spray-marked characters are recognized according to the number of pixel feature points corresponding to the image file and a preset deep learning model to obtain a character recognition result.

[0024] Optionally, the step of performing target character feature recognition on the marking characters according to the number of pixel feature points corresponding to the image file and a preset deep learning model to obtain a character recognition result includes:

[0025] Sorting the image files according to the number of pixel feature points corresponding to the image files to obtain target image files in a preset order;

[0026] Based on the target image file, a target image file containing complete marking characters corresponding to the maximum number of pixel feature points is obtained;

[0027] According to the preset deep learning model, character area detection, character direction classification and character recognition operations are performed on the target image file to obtain a character recognition result.

[0028] Optionally, the step of extracting pixel feature points from the target frame image and determining whether the target frame image contains target character features based on the pixel feature points comprises:

[0029] Downsampling the target frame image to obtain first image data corresponding to the target resolution;

[0030] Performing inter-frame difference calculation on the first image data, and judging whether the image change information exceeds a preset threshold according to the difference calculation result;

[0031] When the image change information exceeds a preset threshold, pixel feature points are extracted from the target frame image, and at the same time, it is determined whether the target frame image contains target character features according to the pixel feature points.

[0032] In addition, to achieve the above-mentioned objectives, the present invention also proposes a steel plate slab character recognition device, which includes a memory, a processor, and a steel plate slab character recognition program stored in the memory and runnable on the processor, and the steel plate slab character recognition program is configured to implement the steps of steel plate slab character recognition as described above.

[0033] In addition, to achieve the above-mentioned purpose, the present invention also proposes a storage medium, on which a steel plate slab character recognition program is stored, and when the steel plate slab character recognition program is executed by a processor, the steps of the steel plate slab character recognition method described above are implemented.

[0034] In addition, in order to achieve the above-mentioned purpose, the present invention also proposes a steel plate slab character recognition device, the steel plate slab character recognition device comprising:

[0035] An image acquisition module is used to acquire a steel plate slab image and simultaneously perform image preprocessing on the steel plate slab image according to preset camera parameters to obtain a target frame image;

[0036] A feature extraction module is used to extract pixel feature points from the target frame image, and at the same time determine whether the target frame image contains target character features based on the pixel feature points;

[0037] The feature recognition module is used to identify the target character features according to a preset deep learning model when the target frame image contains target character features to obtain a character recognition result.

[0038] The present invention acquires a target frame image by collecting a steel plate slab image, and performs image preprocessing on the steel plate slab image according to preset camera parameters; simultaneously, pixel feature points are extracted from the target frame image, and simultaneously, it is judged whether the target frame image contains target character features according to the pixel feature points; when the target frame image contains target character features, the target character features are recognized according to a preset deep learning model to obtain a character recognition result. Compared with the prior art in which the slab moving on the track is only recognized by an AI algorithm, which results in the inability to completely capture and recognize characters, the present invention performs image preprocessing on the steel plate slab image by preset camera parameters to obtain a target frame image, and judges whether the target frame image contains target character features according to the character features in the target frame image, and simultaneously, character recognition is performed on the pixel feature points by a preset deep learning model, thereby achieving complete capture and recognition of slab characters moving on the track under different environments, thereby improving the robustness and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of the structure of a steel plate slab character recognition device in a hardware operating environment involved in an embodiment of the present invention;

[0040] Figure 2 It is a schematic flow chart of the first embodiment of the steel plate slab character recognition method of the present invention;

[0041] Figure 3 A schematic diagram of a threshold determination process of a first embodiment of a steel plate slab character recognition method of the present invention;

[0042] Figure 4 A matrix diagram of a first embodiment of a steel plate slab character recognition method according to the present invention;

[0043] Figure 5 A schematic diagram of the system workflow of the first embodiment of the steel plate slab character recognition method of the present invention;

[0044] Figure 6 A character queue diagram of the first embodiment of the steel plate slab character recognition method of the present invention;

[0045] Figure 7 This is a queue analysis flow chart of the first embodiment of the steel plate slab character recognition method of the present invention

[0046] Figure 8 It is a structural block diagram of the first embodiment of the steel plate slab character recognition device of the present invention.

[0047] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0048] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0049] Reference Figure 1 , Figure 1 The present invention is a schematic diagram of the structure of a steel plate slab character recognition device in the hardware operating environment involved in the embodiment of the present invention.

[0050] like Figure 1 As shown, the steel plate slab character recognition device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The wired interface of the user interface 1003 may be a USB interface in the present invention. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (Wireless-Fidelity, Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM), or a stable memory (Non-volatile Memory, NVM), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0051] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the steel plate blank character recognition device, and may include more or less components than those shown in the figure, or a combination of certain components, or a different arrangement of components.

[0052] like Figure 1 As shown, the memory 1005 identified as a computer storage medium may include an operating system, a network communication module, a user interface module, and a steel plate blank character recognition program.

[0053] exist Figure 1 In the steel plate and slab character recognition device shown, the network interface 1004 is mainly used to connect to the background server and communicate data with the background server; the user interface 1003 is mainly used to connect to the user device; the steel plate and slab character recognition device calls the steel plate and slab character recognition program stored in the memory 1005 through the processor 1001, and executes the steel plate and slab character recognition method provided by the embodiment of the present invention.

[0054] Based on the above hardware structure, an embodiment of the steel plate slab character recognition method of the present invention is proposed.

[0055] Reference Figure 2 , Figure 2 It is a flow chart of the first embodiment of the steel plate slab character recognition method of the present invention, and proposes the first embodiment of the steel plate slab character recognition method of the present invention.

[0056] In this embodiment, the steel plate blank character recognition method comprises the following steps:

[0057] Step S10: collecting a steel plate slab image, and performing image preprocessing on the steel plate slab image according to preset camera parameters to obtain a target frame image.

[0058] It should be noted that the execution subject of this embodiment can be a device with steel plate and slab character recognition function, such as a computer, notebook, mobile phone, etc. This embodiment does not limit this. The device with steel plate and slab character recognition function in this embodiment is suitable for identifying slab character images in different environments. This embodiment uses a computer as the execution subject to illustrate this embodiment.

[0059] It is understandable that the industrial camera hardware system is connected through the IP setting connection module that comes with the industrial camera, and data is exchanged with the industrial camera through the gigabit network port. After a series of command interactions, the camera is successfully opened without any abnormalities. After the camera is successfully opened, the first main thread Processfunc() can be started to obtain the industrial camera data and transmit it to the gigabit network card of the industrial control host. If there is data in the gigabit network card of the industrial control host, it will be determined whether the data is normal. If it is abnormal, an error will be reported. If it is normal, the second thread MstreamProc() can be started to collect the image data of the steel plate slab.

[0060] It should be understood that the preset camera parameters may refer to the preset camera parameters for capturing the steel plate slab image, and the camera parameters include pixels, shooting speed, aperture, and other parameters. The slab image is captured according to the preset camera parameters, and the captured image is saved to the computer for image preprocessing to obtain the target frame image. Image preprocessing may refer to optical preprocessing of the captured steel plate slab image. For example: the captured image is optically processed to separate the image foreground and background, the image background color is converted to a dark color, and the area color corresponding to the character features in the image foreground part is converted to gray or white, so as to facilitate character recognition later.

[0061] In the specific implementation, in order to avoid the influence of ambient light, the collected image is optically processed, that is, the steel plate slab spray-marked character image has been optically processed to a dark brown background, and the target character can be white. The optically processed image is saved as the target frame image for later recognition.

[0062] Furthermore, in order to be applicable to slab character recognition under various ambient lighting conditions, the step S10 includes: determining a threshold range according to a preset sampling rate and a preset OTSU algorithm; determining a threshold for an operation to be performed according to the threshold range and the preset threshold; extracting pixel feature points from the target frame image according to the threshold for the operation to be performed, and at the same time judging whether the target frame image contains target character features based on the pixel feature points.

[0063] It should be noted that the preset sampling rate refers to the preset image acquisition rate and downsampling rate, and the preset OTSU algorithm is a preset algorithm for obtaining the adaptive image processing threshold corresponding to the target frame image. The preset threshold refers to a fixed empirical threshold (pixel grayscale value) preset for image grayscale processing. The fixed threshold is not limited to one value, but can be a set of thresholds with different numerical spans, such as 100, 120, 150, etc. The above values ​​are only used as examples and are not specifically limited. The fixed value can be set according to the empirical threshold, or it can be an image processing threshold under different environments obtained through model training. The empirical threshold can be obtained by manually selecting a preset number of images containing spray-marked characters, downsampling and graying the target frame image, and obtaining an empirical threshold of 120 (pixel grayscale value); write this value into the software algorithm configuration document, and automatically read it each time the software algorithm runs.

[0064] It is understandable that if the color of the area corresponding to the character feature is converted to white, in order to extract these white pixel feature points, the image binarization method is used to separate the target white pixels from the background pixels. For the binarization method, the setting of the threshold is the key, and a reasonable threshold can correctly separate the target white pixels from the background pixels. However, the selection of the reasonable threshold is a difficult problem, because different lighting environments, image changes, etc. will affect the selection of the threshold. The traditional method is to use an adaptive algorithm to determine the threshold, but the inventors have found that the traditional method is not advisable; the classic adaptive threshold method is the OTSU maximum inter-class variance method. It is found that for the collected 1280*1024 resolution image, the OTSU algorithm runs for 14ms, and the image acquisition rate is 60 frames per second. It takes 14ms*60=840ms to process 60 frames; it can be seen that the processing speed cannot meet the real-time requirements, so this method is not advisable. Therefore, in this embodiment, the threshold range is determined by the preset sampling rate and the preset OTSU algorithm, and the threshold for the operation to be executed is determined according to the threshold range and the preset threshold, so as to obtain a more reasonable threshold.

[0065] In the specific implementation, images are acquired at a preset acquisition rate, and the acquired original images are optically processed, and the target frame images after optical processing are downsampled and grayscaled, that is, the acquired target frame images are downsampled by 16x and grayscaled, and the adaptive threshold range is obtained by the preset OTSU algorithm, and the threshold value (pixel grayscale value) of the operation to be performed is obtained from the adaptive threshold range according to the preset empirical threshold. The binarization operation is performed according to the threshold value of the operation to be performed, and the operation efficiency is extremely high, so as to extract pixel feature points from the target frame image later, and at the same time determine whether there are target character features in the target frame image according to the pixel feature points. In order to ensure the recognition rate of characters, refer to Figure 3Schematic diagram of the threshold determination process. After the target frame image is binarized, the resolution remains unchanged, still 1280*1024; then it is divided into two branches for simultaneous operation, and the input data of the two branches are in a copy relationship; one branch enters the "inter-frame analysis" module; one branch calculates the number of pixels, and the number value is temporarily stored, waiting for the signal of the "inter-frame analysis" module. If the inter-frame analysis finds a suspected target, the pixel number value enters the "analysis queue" module as the input data of the "analysis queue" module; if the inter-frame analysis signal indicates that no suspected target is found, the temporarily stored value is refreshed by the next frame. The fixed threshold value obtained in advance through the offline process of manual intervention is reliable after field experiments. Unless there is a major change in the lighting in the factory, such as the installation of lights in the factory, or the removal and maintenance of the factory ceiling, causing a large amount of light to enter the field of view, in this case, it is only necessary to manually select 10 images containing spray-labeled characters, re-run the threshold acquisition step, obtain the threshold under the new environment, and write it into the configuration document. The above situation rarely occurs, so this method will not cause the trouble of frequent modifications while ensuring the efficiency of the algorithm.

[0066] Step S20: extracting pixel feature points from the target frame image, and determining whether the target frame image contains target character features based on the pixel feature points.

[0067] It should be noted that the pixel feature point is a feature point obtained by binarizing the target frame image according to the threshold of the operation to be performed. The color corresponding to the pixel feature point is not specifically limited, and can be white or other colors. The specific color can be set according to actual conditions.

[0068] In a specific implementation, the target frame image is binarized according to the threshold of the operation to be performed to obtain white pixel feature points, and at the same time, it is determined whether the target frame image contains target character features according to the white pixel feature points (or other color pixel feature points).

[0069] Furthermore, the step S20 includes: downsampling the target frame image to obtain first image data corresponding to the target resolution; performing inter-frame difference calculation on the first image data, and judging whether the image change information exceeds a preset threshold value based on the difference calculation result; when the image change information exceeds the preset threshold value, extracting pixel feature points from the target frame image, and judging whether the target frame image contains target character features based on the pixel feature points.

[0070] It should be noted that, through the collected image data stream, the analysis between frames is used to determine whether there is a suspected target in the field of view; if not, continue to collect images and perform feature extraction; if so, obtain the collected image and copy it to the analysis queue. In order to enable the algorithm to process in real time, the actual storage location of the data is the memory. The image change information includes information on the changes between background image frames and character image frames during the movement of the steel plate slab. The preset threshold is a pre-set threshold used to determine whether the background and characters change during the movement of the steel plate slab.

[0071] It is understandable that the first image data is data obtained by downsampling the target frame image, for example, the target frame image is downsampled by 8x, that is, the resolution is reduced from 1280*1024 to a frame image of 160*128.

[0072] In the specific implementation, the target frame image is first downsampled by 8x, that is, the resolution is reduced from 1280*1024 to 160*128. The relationship between the image frames is observed. In a certain time interval, the movement of the steel plate blank is directional, that is, either from left to right or from right to left. After the collected images are inter-frame differentiated, if the preset threshold is set to 0, if the background does not change, the subtraction of the two frames is 0, but the target object spray-marked characters are in motion, so the subtraction of the two frames is not 0, and the change pattern is striped distribution;

[0073] From the matrix point of view, the matrix Md obtained by subtracting the two frames (the matrix size is the same as the size of the image after 8x downsampling, 160*128), the part of the matrix corresponding to the background is 0, and the part corresponding to the target object spray mark characters is not equal to 0, and is either greater than 0 or less than 0. Next, the frame difference matrix Md is transformed into the frame difference mark matrix Kd using the following formula:

[0074]

[0075] Through the above formula, a new frame difference marker matrix Kd is generated, such as Figure 4 As shown in the matrix diagram (a), by summing the matrix Kd by columns, a one-dimensional vector is obtained, such as Figure 4 As shown in (b), let the vector be D(Kd); search for the maximum value in D(Kd) and obtain the index position corresponding to the maximum value, which is the Y-axis coordinate: Y D(Kd) Note that the 1 in row 1 and 4 columns, and the 0 in row 4 and 4 columns in Figure (a) are simulated noise conditions. The noise effect can be eliminated by summing them into a vector. Continuously collect 6 frames and perform inter-frame difference in order to obtain 3 sets of difference matrices Md0, Md1, and Md2; follow the above steps to obtain 3 sets of Y D(Kd0) , Y D(Kd1) , Y D(Kd2); When calculating the difference between the three of them, if the difference is small, less than 1 / 10 of the image height, that is, less than 128*1 / 10≈13, it is determined that the image change pattern is a striped distribution; from this, it can be seen that through the collected image data stream, the analysis between frames determines whether there is a suspected target appearing in the field of view. If no suspected target is recognized to enter, continue to collect images and extract pixel feature points from the target frame image and then perform inter-frame analysis. If a suspected target is recognized to enter the field of view, it is determined whether the target frame image contains target character features based on the pixel feature points.

[0076] Step S30: When the target frame image contains target character features, the target character features are identified according to a preset deep learning model to obtain a character recognition result.

[0077] It should be noted that the preset deep learning model is a pre-set model for recognizing target character features.

[0078] It can be understood that when the target frame image is identified to contain the target character features, the third thread capture character thread ProcessfuncMat() is started, the image files containing the spray-marked characters saved in the hard disk folder are traversed, and the integrity of the character image is judged. If the complete character image is not recognized, the above-mentioned continuous image acquisition is returned to and pixel feature points are extracted from the target frame. When there is a complete character image, one branch transmits the data to the software interface for display, and the other enters the fourth thread PaddleOCR() that has been started for character recognition. The fourth thread captures data containing character images from the software interface in real time. This operation is to avoid thread logic confusion.

[0079] In the specific implementation, in order to improve the recognition efficiency, the above four threads are operated simultaneously. Figure 5 From the system workflow diagram, we can see that after turning on the industrial camera, the first main thread Processfunc() is started to obtain the industrial camera data, the second thread MstreamProc() collects the steel plate slab image data, and the third thread captures the character thread ProcessfuncMat(). The above three threads are started and operated at the same time, that is, the image is collected and copied while capturing the characters. After successfully obtaining the image data containing the characters in the fourth thread PaddleOCR(), the recognition process is carried out, and the target character features are recognized through the pre-trained deep learning model, and the recognized text data is obtained and output to the software interface. A monitoring timer is always running inside the program to calculate the time of each step, sum it up, and finally output the entire processing time to the software interface.

[0080] Furthermore, the step S30 includes: when the target frame image contains target character features, obtaining the number of pixel feature points; identifying the target character features according to the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

[0081] It should be noted that the number of pixel feature points is obtained in the step of extracting pixel feature points from the target frame image, and the number value is temporarily stored. When a suspected target is identified as entering the field of view in the target frame image, the number value of the pixel feature points is extracted, and the target character features are identified based on the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

[0082] Furthermore, the step of identifying the target character features according to the number of pixel feature points and a preset deep learning model to obtain a character recognition result includes: obtaining the number of pixel feature points corresponding to each target frame image in a preset acquisition rate, and determining the maximum and minimum values ​​of the number of pixel feature points; obtaining the index position values ​​corresponding to the maximum and minimum values; identifying the target character features according to the index position values, the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

[0083] It should be noted that the preset acquisition rate refers to the rate at which the industrial camera captures images in advance. Obtaining the number of pixel feature points corresponding to each target frame image in the preset acquisition rate refers to obtaining the pixel feature points corresponding to each image in the acquired image. For example: if the acquisition rate of the industrial camera is 60 frames per second, a FIFO first-in-first-out queue with a queue length of 60 is constructed to analyze whether the internal data is increasing or decreasing; if it is increasing, it indicates that a spray-marked character target has entered the field of view, and the "spray-marked character image copy module" starts to save the collected original image data to the hard disk; if it is decreasing, it indicates that the spray-marked character target has left the field of view, and the "spray-marked character image copy module" ends the save operation.

[0084] It is understandable that in order to accurately determine whether the 60 values ​​in the queue are increasing or decreasing, this embodiment determines the index position value corresponding to the maximum and minimum values ​​of the number of pixel feature points to determine whether the marked characters are in the field of view, and performs character feature recognition based on the judgment result.

[0085] Furthermore, the step of identifying the target character features according to the index position value, the number of pixel feature points and a preset deep learning model to obtain a character recognition result includes: generating a stop copy signal when the index position value of the minimum value appears at the tail of the queue and the number of occurrences is greater than a preset number of times; based on the stop copy signal, traversing the image file containing the spray-marked characters that is most recently saved to the hard disk folder, and performing target character feature identification on the spray-marked characters according to the number of pixel feature points corresponding to the image file and a preset deep learning model to obtain a character recognition result.

[0086] It is understandable that if a spray-marked character target enters the field of view, the maximum value of the elements in the queue will repeatedly appear at the tail of the queue, that is, the position just entering the queue; if a spray-marked character target leaves the field of view, the minimum value of the elements in the queue will also repeatedly appear at the tail of the queue.

[0087] In the specific implementation, the analysis queue is designed to find the maximum and minimum values ​​in a queue of length 60; if the maximum value appears at the tail more than 10 times, it is determined that the target of the spray-on character has entered; if the minimum value appears at the tail more than 10 times, it is determined that the target of the spray-on character has left. Figure 6 Character queue diagram and reference Figure 7 Flowchart of cohort analysis.

[0088] Furthermore, the step of performing target character feature recognition on the marked characters according to the number of pixel feature points corresponding to the image file and a preset deep learning model to obtain a character recognition result includes: sorting the image files according to the number of pixel feature points corresponding to the image files to obtain target image files in a preset order; acquiring a target image file containing complete marked characters that satisfies the maximum number of pixel feature points based on the target image files; performing character area detection, character direction classification and character recognition operations on the target image file according to the preset deep learning model to obtain a character recognition result.

[0089] It should be noted that the image files are sorted according to the number of pixel feature points corresponding to each image, and the image file containing complete spray-marked characters in the image with the largest number of pixel feature points is output, and character area detection, character direction classification and character recognition operations are performed on the image file according to the preset deep learning model.

[0090] It is understandable that the 4th thread PaddleOCR() is started. This thread can use the OCR optical character recognition function module in the Baidu PaddlePaddle deep learning framework structure. This module contains 3 sub-processes, namely: 1. Character region detection, which locates the regions with characters in the image, and the result is that single characters are wrapped by positioning frames; 2. Character direction classification, which classifies the directions of the characters in the image, because the angle of the characters will affect recognition. For example, if the character "人" is upside down, it will be recognized as the letter "Y"; 3. Character recognition, which recognizes the character image data into corresponding text data. The above 3 sub-process steps are all obtained by reasoning through a pre-trained deep learning model reference.

[0091] In specific implementation, the text data obtained by character recognition is output to the software interface. A monitoring timer is always running inside the program to calculate the time of each step and sum them up, and finally output the entire processing time to the software interface.

[0092] In this embodiment, by collecting the steel plate slab image and performing image preprocessing on the steel plate slab image according to the preset camera parameters, a target frame image is obtained; at the same time, pixel feature points are extracted from the target frame image, and whether the target frame image contains target character features is judged according to the pixel feature points; when the target frame image contains target character features, the target character features are recognized according to the preset deep learning model to obtain a character recognition result. Compared with the prior art in which only the slab moving on the track is recognized by the AI algorithm, resulting in the inability to completely capture and recognize characters, in this embodiment, a target frame image is obtained by performing image preprocessing on the steel plate slab image according to the preset camera parameters, and whether the target frame image contains target character features is judged according to the character features in the target frame image. At the same time, character recognition is performed on the pixel feature points through the preset deep learning model, realizing the recognition of the characters of the slab moving on the track under different environments, improving the robustness and stability of the system.

[0093] In addition, to achieve the above object, the present invention also proposes a steel plate slab character recognition device, the steel plate slab character recognition device includes a memory, a processor, and a steel plate slab character recognition program stored on the memory and executable on the processor, and the steel plate slab character recognition program is configured to implement the steps of the steel plate slab character recognition as described above.

[0094] In addition, to achieve the above object, the present invention also proposes a storage medium, on which a steel plate slab character recognition program is stored, and when the steel plate slab character recognition program is executed by a processor, it implements the steps of the steel plate slab character recognition method as described above.

[0095] Refer to Figure 8 , Figure 8 It is a structural block diagram of the first embodiment of the steel plate slab character recognition device of the present invention.

[0096] like Figure 8 As shown, the steel plate slab character recognition device proposed in the embodiment of the present invention includes:

[0097] The image acquisition module 10 is used to acquire the steel plate slab image and simultaneously perform image preprocessing on the steel plate slab image according to preset camera parameters to obtain a target frame image;

[0098] A feature extraction module 20 is used to extract pixel feature points from the target frame image, and at the same time determine whether the target frame image contains target character features according to the pixel feature points;

[0099] The feature recognition module 30 is used to recognize the target character features according to a preset deep learning model when the target frame image contains target character features to obtain a character recognition result.

[0100] This embodiment acquires a target frame image by acquiring a steel plate slab image, and performs image preprocessing on the steel plate slab image according to preset camera parameters; simultaneously, pixel feature points are extracted from the target frame image, and simultaneously, it is determined whether the target frame image contains target character features according to the pixel feature points; when the target frame image contains target character features, the target character features are recognized according to a preset deep learning model to obtain a character recognition result. Compared with the prior art in which the slab moving on the track is only recognized by an AI algorithm, which results in the inability to completely capture and recognize the characters, this embodiment performs image preprocessing on the steel plate slab image by preset camera parameters to obtain a target frame image, and determines whether the target frame image contains target character features according to the character features in the target frame image, and simultaneously, character recognition is performed on the pixel feature points by a preset deep learning model, thereby achieving complete capture and recognition of slab characters moving on the track under different environments, thereby improving the robustness and stability of the system.

[0101] Furthermore, the feature extraction module 20 is also used to determine a threshold range based on a preset sampling rate and a preset OTSU algorithm; determine a threshold for the operation to be performed based on the threshold range and the preset threshold; extract pixel feature points from the target frame image based on the threshold for the operation to be performed, and at the same time determine whether the target frame image contains target character features based on the pixel feature points.

[0102] Furthermore, the feature recognition module 30 is also used to obtain the number of pixel feature points when the target frame image contains target character features; identify the target character features according to the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

[0103] Furthermore, the feature recognition module 30 is also used to obtain the number of pixel feature points corresponding to each target frame image in a preset acquisition rate, and determine the maximum and minimum values ​​of the number of pixel feature points; obtain the index position values ​​corresponding to the maximum and minimum values; identify the target character features according to the index position values, the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

[0104] Furthermore, the feature recognition module 30 is also used to generate a stop copying signal when the index position value of the minimum value appears at the tail of the queue and the number of occurrences is greater than a preset number of times; based on the stop copying signal, traverse the image file containing the spray-marked characters that is most recently saved to the hard disk folder, and perform target character feature recognition on the spray-marked characters according to the number of pixel feature points corresponding to the image file and a preset deep learning model to obtain a character recognition result.

[0105] Furthermore, the feature recognition module 30 is also used to sort the image files according to the number of pixel feature points corresponding to the image files to obtain target image files in a preset order; obtain a target image file containing complete spray-marked characters corresponding to the maximum number of pixel feature points based on the target image files; perform character area detection, character direction classification and character recognition operations on the target image files according to a preset deep learning model to obtain character recognition results.

[0106] Furthermore, the feature extraction module 20 is also used to downsample the target frame image to obtain first image data corresponding to the target resolution; perform inter-frame difference calculation on the first image data, and judge whether the image change information exceeds a preset threshold based on the difference calculation result; when the image change information exceeds the preset threshold, extract pixel feature points from the target frame image, and at the same time judge whether the target frame image contains target character features based on the pixel feature points.

[0107] It should be understood that the above is only an example and does not constitute any limitation on the technical solution of the present invention. In specific applications, technicians in this field can make settings as needed, and the present invention does not limit this.

[0108] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of them according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0109] In addition, for technical details that are not described in detail in this embodiment, reference can be made to the steel plate slab character recognition method provided in any embodiment of the present invention, and will not be repeated here.

[0110] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0111] The serial numbers of the embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In a unit claim that lists several means, several of these means may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order and these words may be interpreted as names.

[0112] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory image (ROM) / random access memory (RAM), a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0113] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for character recognition of steel plate blanks, characterized in that: The steel plate slab character recognition method comprises the following steps: Acquire a steel plate slab image, and simultaneously perform image preprocessing on the steel plate slab image according to preset camera parameters to obtain a target frame image; Extracting pixel feature points from the target frame image, and judging whether the target frame image contains target character features according to the pixel feature points; When the target frame image contains a target character feature, the target character feature is recognized according to a preset deep learning model to obtain a character recognition result; Wherein, when the target frame image contains a target character feature, the step of identifying the target character feature according to a preset deep learning model to obtain a character recognition result also includes: When the target frame image contains the target character feature, the number of pixel feature points corresponding to each target frame image in a preset acquisition rate is obtained, and the maximum and minimum values ​​of the number of pixel feature points are determined; Obtain the index position values ​​corresponding to the maximum value and the minimum value; The target character features are identified according to the index position value, the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

2. The steel plate slab character recognition method according to claim 1, characterized in that: The step of extracting pixel feature points from the target frame image and determining whether the target frame image contains target character features based on the pixel feature points comprises: Determine the threshold range according to the preset sampling rate and the preset OTSU algorithm; Determine a threshold value for the operation to be performed according to the threshold range and the preset threshold value; Pixel feature points are extracted from the target frame image according to the threshold of the operation to be performed, and at the same time, it is determined whether the target frame image contains target character features according to the pixel feature points.

3. The steel plate slab character recognition method according to claim 1, characterized in that: The step of identifying the target character feature according to the index position value, the number of pixel feature points and a preset deep learning model to obtain a character recognition result includes: When the index position value of the minimum value appears at the tail of the queue and the number of occurrences is greater than a preset number, a stop copy semaphore is generated; Based on the stop copy signal, the image file containing the spray-marked characters that is most recently saved in the hard disk folder is traversed, and the target character features of the spray-marked characters are recognized according to the number of pixel feature points corresponding to the image file and a preset deep learning model to obtain a character recognition result.

4. The steel plate slab character recognition method according to claim 3, characterized in that: The step of performing target character feature recognition on the marking characters according to the number of pixel feature points corresponding to the image file and a preset deep learning model to obtain a character recognition result comprises: Sorting the image files according to the number of pixel feature points corresponding to the image files to obtain target image files in a preset order; Based on the target image file, a target image file containing complete marking characters corresponding to the maximum number of pixel feature points is obtained; According to the preset deep learning model, character area detection, character direction classification and character recognition operations are performed on the target image file to obtain a character recognition result.

5. The steel plate slab character recognition method according to claim 1, characterized in that: The step of extracting pixel feature points from the target frame image and determining whether the target frame image contains target character features based on the pixel feature points comprises: Downsampling the target frame image to obtain first image data corresponding to the target resolution; Performing inter-frame difference calculation on the first image data, and judging whether the image change information exceeds a preset threshold according to the difference calculation result; When the image change information exceeds a preset threshold, pixel feature points are extracted from the target frame image, and at the same time, it is determined whether the target frame image contains target character features according to the pixel feature points.

6. A steel plate slab character recognition device, characterized in that: The steel plate slab character recognition device includes: a memory, a processor, and a steel plate slab character recognition program stored in the memory and executable on the processor. When the steel plate slab character recognition program is executed by the processor, the steps of the steel plate slab character recognition method as described in any one of claims 1 to 5 are implemented.

7. A storage medium, characterized in that: The storage medium stores a steel plate slab character recognition program, and when the steel plate slab character recognition program is executed by the processor, the steps of the steel plate slab character recognition method as described in any one of claims 1 to 5 are implemented.

8. A steel plate slab character recognition device, characterized in that: The steel plate slab character recognition device comprises: An image acquisition module is used to acquire a steel plate slab image and simultaneously perform image preprocessing on the steel plate slab image according to preset camera parameters to obtain a target frame image; A feature extraction module is used to extract pixel feature points from the target frame image, and at the same time determine whether the target frame image contains target character features based on the pixel feature points; A feature recognition module is used to recognize the target character features according to a preset deep learning model when the target frame image contains target character features, so as to obtain a character recognition result; The feature recognition module is also used to obtain the number of pixel feature points corresponding to each target frame image in a preset acquisition rate and determine the maximum and minimum values ​​of the number of pixel feature points when the target frame image contains target character features; obtain the index position values ​​corresponding to the maximum and minimum values; and identify the target character features according to the index position values, the number of pixel feature points and a preset deep learning model to obtain a character recognition result.

Images