An OCR-based scrap steel recycling pound note automatic identification system

Abstract

The application discloses a kind of based on OCR's scrap steel recycling pound single automatic identification system, including customer pound single configuration module for defining pound single pound single rule according to the pound single picture uploaded by customer;OCR training module is used to obtain the character in the pound single picture uploaded by customer is corrected and stored;OSS object storage module is used to store the pound single picture uploaded by customer;Pound single OCR image-text recognition module is used to obtain pound single picture and process, text information in pound single picture is identified based on OCR image recognition engine, and text information is converted into structured pound single data according to the pound single rule of customer pound single configuration module;Relational database storage module is used to store the pound single data obtained by pound single OCR image-text recognition module;Manual check module is used to provide pound single picture, pound single data to artificial and recheck.Pound single structured data information is identified automatically according to pound single picture, and rechecked by artificial, improve pound single recognition efficiency and accuracy.

Description

An OCR-based automatic identification system for scrap steel recycling weighbridge slips Technical Field

[0001] This invention relates to the field of recycled resources, and in particular to an OCR-based automatic identification system for scrap steel recycling weighbridge slips. Background Technology

[0002] In the scrap steel recycling industry, due to the large number of downstream customers and their varying levels of IT literacy—some with high IT capabilities and able to provide standardized machine-printed weighbridge slips, while others with lower IT capabilities can only provide handwritten slips—and the format, quality, and clarity of each customer's weighbridge slips differ, resulting in diverse slip styles, quality, and information. Converting information from paper weighbridge slips—including slip number, customer, supplier, weighing date, goods name, goods specifications, license plate number, gross weight, tare weight, deductions, net weight, actual weight, and remarks—into structured data that can be stored in an information system is essential for enterprise informatization. Currently, many recycling companies manually input paper weighbridge slip information into computers to solve this problem. This process is labor-intensive and costly; moreover, the work is monotonous, repetitive, and prone to human error. Summary of the Invention

[0003] This invention provides an OCR-based automatic identification system for scrap steel recycling weighbridge slips to overcome the aforementioned technical problems.

[0004] An OCR-based automatic identification system for scrap steel recycling weighbridge slips includes a customer weighbridge slip configuration module, an OCR training module, an OSS object storage module, a weighbridge slip OCR image and text recognition module, a relational database storage module, and a manual verification module.

[0005] The customer weighbridge configuration module is used to define the weighbridge rules based on the weighbridge image uploaded by the customer. The definition of the weighbridge rules includes determining the fields contained in the weighbridge and the constraints of each field. The constraints are to limit the character length and character types that the fields can contain.

[0006] The OCR training module is used to acquire a set of bill images uploaded by customers, convert the bill images into different formats and generate composite files, and correct the characters in the composite files. The corrected character set is used as the customer's font library. The OCR training module includes an upload sample module, a sample synthesis module, a character correction module, and a sample font library module. The upload sample module is used to acquire a set of bill images uploaded by customers. The sample synthesis module is used to convert the bill images in the set of bill images into TIFF format files and synthesize all the TIFF format files. The character correction module is used to correct the characters in the composite files according to a character set that conforms to the character standard. The sample font library module is used to acquire and save the character set corrected by the character correction module.

[0007] The OSS object storage module is used to store weighbridge images uploaded by customers;

[0008] The bill OCR image and text recognition module is used to acquire bill images stored in the OSS object storage module, process the bill images, and obtain the processed bill images. Based on the character set of the OCR image recognition engine and sample font library module, it recognizes the text information in the bill images and converts the text information into structured bill data according to the bill rules of the customer's bill configuration module. The bill OCR image and text recognition module includes a bill image processing module, a bill image text recognition module, and a bill text conversion module. The bill image processing module is used to acquire and process the bill images stored in the OSS object storage module. The processing includes cropping the bill images, removing noise from the bill images, clustering the pixels of the bill images, and converting the colors of the bill images. The bill image text recognition module is used to acquire the processed bill images and recognize the text information in the bill images based on the OCR image recognition engine. The bill text conversion module is used to convert the text information into structured bill data according to the bill rules of the customer's bill configuration module.

[0009] The relational database storage module is used to store the weighbridge data obtained by the weighbridge OCR image recognition module;

[0010] The manual verification module is used to provide weighbridge images and weighbridge data to humans for review.

[0011] Preferably, the removal of noise from the weighbridge image includes,

[0012] Step 1a: For each pixel in the weighted slip image, sequentially obtain the pixel value of its neighboring pixels, and calculate the sum of the pixel value of the current pixel and the pixel values ​​of its neighboring pixels.

[0013] Step 1b: Calculate the pixel average value based on the sum of pixel values ​​and the number of adjacent pixels.

[0014] Step 1c: Use the average pixel value as the pixel value of the pixel.

[0015] Preferably, the color conversion of the bill image includes converting the color bill image to a black and white image according to formula (1).

[0016] gray=(max-mid)*ratio_max+(mid-min)*ratio_max_mid+min(1)

[0017] Where gray is the pixel grayscale value, max, mid and min are the maximum, median and minimum pixel values ​​of the R, G and B components of the color image, respectively, ratio_max is the ratio of the component color represented by the maximum pixel value, and ratio_max_mid is the ratio of the composite color formed by the maximum and median pixel values.

[0018] Preferably, the character set recognition of text information in the billboard image based on the OCR image recognition engine and sample font library module includes recognizing text information in the billboard image through a convolutional recurrent neural network, comparing the text information with the character set of the sample font library module, and obtaining the compared text information. The convolutional recurrent neural network includes three modules: convolutional neural network, recurrent neural network, and connectivity temporal classification.

[0019] Preferably, the clustering process of the pixels in the weighing slip image includes,

[0020] Step 2a: Randomly select k pixels from the weighbridge image, ensuring that the k pixels are evenly distributed across the image.

[0021] Step 2b: Determine the neighboring regions of each pixel, calculate the gradient values ​​of all pixels within the neighboring regions of the current pixel, and replace the pixel value of the current pixel with the pixel value of the pixel with the smallest gradient value within the neighboring regions.

[0022] Step 2c: Determine the clustering region for each pixel, perform clustering processing on the image based on the SLIC algorithm, and obtain the cluster center for each clustering region.

[0023] This invention provides an OCR-based automatic recognition system for scrap steel recycling weighbridge slips. The system obtains a font library for customer weighbridge slips through an OCR training module, significantly improving the efficiency and accuracy of slip recognition. By uploading photos of paper weighbridge slips to the OCR-based automatic recognition system for scrap steel recycling weighbridge slips, the system automatically recognizes the structured data information of the weighbridge slips from the images. A manual verification module further verifies and checks the data, ensuring 100% accuracy. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 is a system structure diagram of the present invention;

[0026] Figure 2 is a diagram of the convolutional recurrent neural network structure of the present invention;

[0027] Figure 3 is a flowchart of the application of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Figure 1 is a system structure diagram of the present invention. As shown in Figure 1, the method of this embodiment may include:

[0030] An OCR-based automatic identification system for scrap steel recycling weighbridge slips includes a customer weighbridge slip configuration module, an OCR training module, an OSS object storage module, a weighbridge slip OCR image and text recognition module, a relational database storage module, and a manual verification module.

[0031] The customer weighbridge configuration module is used to define weighbridge rules based on the weighbridge image uploaded by the customer. Defining the weighbridge rules includes determining the fields included in the weighbridge and the constraints on each field. The constraints limit the character length and character types that can be included in the fields. The fields included in the weighbridge include weighbridge number, customer, supplier, weighing date, goods name, goods specifications, license plate number, gross weight, tare weight, deductions, net weight, actual weight, and remarks. For example, the constraint on the weighbridge number is that it can only consist of numbers and cannot contain characters or special symbols. The constraint on the license plate number is that it must include seven characters, with the first character limited to Chinese characters, the second character limited to characters, and the third to seventh characters allowed to be numbers or letters. The constraint on the net weight of the goods is that it is calculated based on the gross weight and tare weight of the goods. The constraints on each field are limited according to actual usage.

[0032] The OCR training module is used to acquire a set of weighbridge slip images uploaded by the customer, convert the formats of each image, generate a composite file, correct the characters in the composite file, and use the corrected character set as the customer's font library.

[0033] The OCR training module includes a sample uploading module, a sample synthesis module, a character correction module, and a sample font library module.

[0034] The sample upload module is used to obtain a collection of weighbridge slip images uploaded by customers. Customers package these weighbridge slip image files into a zip archive and upload it to the system's backend server. The weighbridge slip image file must contain no fewer than 1000 images.

[0035] The sample synthesis module is used to convert the baggage slip images in the baggage slip image set into TIFF format files, and then combine all the TIFF format files into a single file.

[0036] The character correction module is used to correct characters in the synthesized file according to a set of characters that conform to a character standard. The character standard includes standard formats for commonly used Chinese characters, all numbers, all letters, and commonly used mathematical symbols.

[0037] The sample font library module is used to acquire and save the character set corrected by the character correction module;

[0038] The OSS object storage module is used to store weighbridge images uploaded by customers;

[0039] The billing slip OCR image recognition module is used to acquire billing slip images stored in the OSS object storage module. It processes these images, including cropping, removing noise, clustering pixels, and color conversion. The resulting processed image is then used to identify text information within the billing slip image based on the OCR image recognition engine and the character set of the sample font library module. Finally, according to the billing rules configured in the customer's billing slip configuration module, the text information is converted into structured billing slip data.

[0040] The character set recognition of text information in a billboard image based on an OCR image recognition engine and a sample font library module, as described in section 0, includes recognizing text information in the billboard image through a convolutional recurrent neural network, comparing the text information with the character set of the sample font library module, and obtaining the compared text information. The convolutional recurrent neural network includes convolutional neural networks, recurrent neural networks, and connectivity-based temporal classification.

[0041] The system consists of three modules, as shown in Figure 2. The convolutional neural network extracts features from the input bill image to obtain a feature map. The recurrent neural network is used to predict the feature sequence in the feature map, learn each feature vector in the sequence, and output the predicted label (true value) distribution. The connectivity temporal classification is used to transform a series of label distributions obtained from the recurrent layers of the recurrent neural network into the final label sequence.

[0042] The weighbridge slip OCR image recognition module includes a weighbridge slip image processing module, a weighbridge slip image text recognition module, and a weighbridge slip text conversion module.

[0043] The aforementioned bill of lading image processing module is used to obtain the bill of lading images stored in the OSS object storage module and process them. The processing includes cropping the bill of lading images, removing noise from the images, clustering the pixels of the images, and converting the colors of the images.

[0044] The removal of noise from the weighbridge image includes...

[0045] Step 1a: For each pixel in the weighted slip image, sequentially obtain the pixel value of its neighboring pixel 5, and calculate the sum of the pixel value of the current pixel and the pixel values ​​of its neighboring pixels.

[0046] Step 1b: Calculate the pixel average value based on the sum of pixel values ​​and the number of adjacent pixels.

[0047] Step 1c: Use the average pixel value as the pixel value of the pixel.

[0048] The process of converting the bill image color includes converting the color bill image to a black and white image according to formula (1).

[0049] gray=(max-mid)*ratio_max+(mid-min)*ratio_max_mid+min(1)

[0050] 0 Where gray is the pixel gray value, max, mid and min are the maximum, median and minimum pixel values ​​of the R, G and B components of the color image, respectively, ratio_max is the ratio of the component color represented by the maximum pixel value, and ratio_max_mid is the ratio of the compound color formed by the maximum pixel value and the median pixel value.

[0051] The clustering process for the pixels of the weighing slip image includes...

[0052] Step 2a: Randomly select k pixels from the weighbridge image, ensuring that the k pixels are evenly distributed across the image.

[0053] Step 2b: Determine the adjacent area of each pixel. The adjacent area is an area composed of n*n pixels including the current pixel, generally 3*3. Calculate the gradient values of all pixels in the adjacent area of the current pixel, and replace the pixel value of the current pixel with the pixel value of the pixel with the minimum gradient value in the adjacent area.

[0054] Step 2c: Determine the clustering area of each pixel. The clustering area is an area composed of 2S*2S pixels including the current pixel. Perform clustering processing on the image based on the SLIC algorithm to obtain the clustering center of each clustering area, where S = sqrt(N / k), and N is the number of image pixels.

[0055] The weigh bill picture text recognition module is used to obtain the processed weigh bill picture and recognize the text information in the weigh bill picture based on the character set of the OCR picture recognition engine and the sample font library module.

[0056] The weigh bill text conversion module is used to convert the text information into structured weigh bill data according to the weigh bill rules of the customer weigh bill configuration module. For example, if the text information is "2323332 Jialiao A1234567 000kg", according to the weigh bill rules, first create a blank record, and then extract the text information in sequence. For the order number with 7 characters, obtain the first seven data in the text information and store them in the order number corresponding to the blank record. Obtain the license plate number according to the regular expression and store it in the license plate number. Obtain the gross weight of the goods according to the commonly used unit of weighing and the regular expression and store it in the gross weight.

[0057] The relational database storage module is used to store the weigh bill data obtained by the weigh bill OCR graphic and text recognition module.

[0058] The manual verification module is used to provide the weigh bill picture and weigh bill data to the manual for review.

[0059] The application method of this embodiment is shown in Figure 3. The customer uploads the weigh bill picture to the OSS object storage module, analyzes the picture file into a byte stream, crops the picture, removes the picture noise, sets the picture background to black and white, recognizes the text information, extracts the customer weigh bill configuration rules according to the customer, and analyzes the structured data of the customer weigh bill according to the customer weigh bill configuration rules. Save the structured data of the customer weigh bill to the relational database, and manually verify and review the weigh bill.

[0060] Overall beneficial effects:

[0061] This invention provides an OCR-based automatic recognition system for scrap steel recycling weighbridge slips. The system obtains a font library for customer weighbridge slips through an OCR training module, significantly improving the efficiency and accuracy of slip recognition. By uploading photos of paper weighbridge slips to the OCR-based automatic recognition system for scrap steel recycling weighbridge slips, the system automatically recognizes the structured data information of the weighbridge slips from the images. A manual verification module further verifies and checks the data, ensuring 100% accuracy.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An OCR-based automatic identification system for scrap steel recycling weighbridge slips, characterized in that, The system includes a customer bill configuration module, an OCR training module, an OSS object storage module, a bill OCR image and text recognition module, a relational database storage module, and a manual verification module. The customer bill configuration module defines the bill rules based on the bill images uploaded by the customer. Defining the bill rules includes determining the fields included in the bill and the constraints on each field. The constraints limit the character length and character types contained in the fields. The OCR training module acquires the set of bill images uploaded by the customer, converts the formats of each bill image, generates a composite file, and verifies the characters in the composite file. The OCR training module performs corrections and uses the corrected character set as the client's font library. It includes a sample upload module, a sample synthesis module, a character correction module, and a sample font library module. The sample upload module acquires the set of bill of lading images uploaded by the client. The sample synthesis module converts each bill of lading image in the set into a TIFF file and then synthesizes all the TIFF files. The character correction module corrects the characters in the synthesized file based on a character set that conforms to character standards. The sample font library module acquires and saves the characters corrected by the character correction module. A character set; the OSS object storage module is used to store the bill images uploaded by customers; the bill OCR image recognition module is used to obtain the bill images stored in the OSS object storage module, process the bill images, obtain the processed bill images, recognize the text information in the bill images based on the character set of the OCR image recognition engine and sample font library module, and convert the text information into structured bill data according to the bill rules of the customer's bill configuration module. The bill OCR image recognition module includes a bill image processing module, a bill image text recognition module, and a bill text conversion module. The bill image processing module uses... The system acquires and processes the bill of lading images stored in the OSS object storage module. This processing includes cropping the bill of lading images, removing noise, clustering pixels, and converting the colors. A bill of lading image text recognition module acquires the processed images and uses an OCR image recognition engine to identify the text information within them. A bill of lading text conversion module converts the text information into structured bill of lading data according to the bill of lading rules configured in the customer's bill of lading configuration module. A relational database storage module stores the bill of lading data acquired by the bill of lading OCR image recognition module. The manual verification module is used to provide weighbridge images and weighbridge data to humans for review.

2. The OCR-based automatic identification system for scrap steel recycling weighbridge slips according to claim 1, characterized in that, The process of removing noise from the bill of quantities image includes the following steps: Step 1a: For each pixel in the bill of quantities image, sequentially obtain the pixel value of its neighboring pixels and calculate the sum of the pixel value of the current pixel and the pixel values ​​of its neighboring pixels; Step 1b: Calculate the average pixel value based on the number of neighboring pixels according to the sum of the pixel values; Step 1c: Use the average pixel value as the pixel value of the current pixel.

3. The automatic identification system for scrap steel recycling weighbridge slips based on OCR according to claim 1, characterized in that, The color conversion of the bill image includes converting the color bill image to a black and white image according to formula (1), where gray = (max - mid) ratio_max + (mid - min) ratio_max_mid + min(1) where gray is the pixel gray value, max, mid and min are the maximum, median and minimum pixel values ​​of the R, G and B components of the color image, respectively, ratio_max is the ratio of the component color represented by the maximum pixel value, and ratio_max_mid is the ratio of the two component colors formed by the maximum pixel value and the median pixel value.

4. The OCR-based automatic identification system for scrap steel recycling weighbridge slips according to claim 1, characterized in that, The character set recognition of text information in the billboard image based on the OCR image recognition engine and sample font library module includes recognizing text information in the billboard image through a convolutional recurrent neural network, comparing the text information with the character set of the sample font library module, and obtaining the compared text information. The convolutional recurrent neural network includes three modules: convolutional neural network, recurrent neural network, and connectivity temporal classification.

5. The OCR-based automatic identification system for scrap steel recycling weighbridge slips according to claim 1, characterized in that, The clustering process for the pixels of the bill of lading image includes: step 2a, randomly selecting k pixels in the bill of lading image so that the k pixels are evenly distributed on the image; step 2b, determining the neighboring regions of each pixel, calculating the gradient values ​​of all pixels in the neighboring regions of the current pixel, and replacing the pixel value of the current pixel with the pixel value of the pixel with the smallest gradient value in the neighboring regions; step 2c, determining the clustering region of each pixel, performing clustering processing on the image based on the SLIC algorithm, and obtaining the cluster center of each clustering region.

Images