Logistics transfer center vehicle weighing information identification method, apparatus and device, and medium

Abstract

The invention provides a logistics transfer center vehicle weighing information identification method and device, equipment and a medium, and belongs to the technical field of logistics management, and the method comprises the steps: obtaining an initial vehicle weighing image, and carrying out the preprocessing of the obtained initial vehicle weighing image, and obtaining a vehicle weighing image data set; labeling the vehicle weighing image data set to obtain a training set; the training set is used for training an identification model; constructing a recognition model based on the paddleOCRV2, and training the recognition model by using the training set to obtain a trained recognition model; identifying the vehicle weighing information of the logistics transfer center by using the trained identification model; wherein the logistics transfer center vehicle weighing information comprises the license plate number of a weighing vehicle, the cargo weight and cargo name information. According to the invention, the automatic identification and processing of the vehicle weighing information are realized, the requirement of manual intervention is reduced, the working efficiency is improved, the logistics management process is optimized, and the labor cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics management, and in particular to a method, device, equipment and medium for identifying vehicle weighing information in a logistics transfer center. Background Art

[0002] In logistics transfer centers, vehicle weighing is a critical step in the flow of goods, involving the weighing, billing, and recording of logistics information. Traditional weighing methods rely on manual operations, which are not only inefficient but also prone to errors. Especially during peak periods, manual processing speeds cannot meet the rapidly growing logistics needs. In addition, manual recording of weighing information is susceptible to human factors such as fatigue and negligence, resulting in inaccurate data. With the development of artificial intelligence and computer vision technologies, especially advances in optical character recognition (OCR), automated and intelligent vehicle weighing information recognition has become possible. By building an intelligent recognition system, images or video data captured by the camera can be processed in real time to accurately identify relevant information during vehicle weighing, thereby improving the operational efficiency and accuracy of logistics transfer centers. Summary of the Invention

[0003] The main purpose of the present invention is to solve the problems in the prior art that manual weighing and manual recording are inefficient, prone to human errors, and cannot meet the needs of modern logistics management.

[0004] A first aspect of the present invention provides a method for identifying vehicle weighing information in a logistics transfer center, comprising: Acquire an initial vehicle weighing image, and preprocess the acquired initial vehicle weighing image to obtain a vehicle weighing image dataset; Annotate the vehicle weighing image dataset to obtain a training set; the training set is used to train the recognition model; Build a recognition model based on paddleOCR_V2, use the training set to train the recognition model, and obtain the trained recognition model; Use the trained recognition model to identify vehicle weighing information in the logistics transfer center; The vehicle weighing information of the logistics transfer center includes: the license plate number of the weighed vehicle, the weight of the cargo, and the name of the cargo.

[0005] Optionally, in a first implementation of the first aspect of the present invention, obtaining an initial vehicle weighing image and preprocessing the obtained initial vehicle weighing image to obtain a vehicle weighing image dataset includes: Based on the monitoring system, we can obtain the monitoring video of the weighing camera at different time periods within the preset time period; Extracting a frame from the surveillance video at a preset time interval as the initial vehicle weighing image; The initial vehicle weighing images include: vehicle weighing images under different lighting conditions, different vehicle types and different cargo types; The initial vehicle weighing image is generated by adjusting parameters using the open-source tool TextRecognitionDataGenerator to simulate image features in a real environment. The initial vehicle weighing image is preprocessed by the image processing library to obtain a preprocessed initial vehicle weighing image; The synthetic images and the preprocessed initial vehicle weighing images are used to synthesize the vehicle weighing image dataset.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the initial vehicle weighing image is subjected to parameter adjustment using the open source tool TextRecognitionDataGenerator to simulate image features in a real environment to generate a synthetic image, including: Adjust the font type, size, and color of the initial vehicle weighing image to simulate the text logo style in actual logistics scenarios, and obtain the adjusted vehicle weighing image; The tilt angle parameter of the adjusted initial vehicle weighing image is randomly set in the range of -30 degrees to 30 degrees to generate a tilted vehicle weighing image; Setting a blur kernel size and a standard deviation, blurring the tilted vehicle weighing image based on the set blur kernel size and the standard deviation, and obtaining a blurred vehicle weighing image; Using a distortion function to distort the blurred vehicle weighing image according to a preset probability to obtain a composite image; The initial vehicle weighing image is preprocessed by an image processing library to obtain a preprocessed initial vehicle weighing image, including: The initial vehicle weighing image is resized by an image processing library to obtain an adjusted vehicle weighing image; The adjusted vehicle weighing image is cropped through the image processing library to obtain a cropped vehicle weighing image; The cropped vehicle weighing image is color-corrected by an image processing library to obtain a color-corrected vehicle weighing image.

[0007] Optionally, in a third implementation of the first aspect of the present invention, the method further includes: Construct a CarDark model based on a deep learning framework, and train the constructed CarDark model to obtain a trained CarDark model; The CarDark model is used to perform brightness enhancement and contrast enhancement on blurry pictures at night; The trained CarDark model is used to preprocess the acquired nighttime vehicle weighing images, including brightness enhancement and contrast enhancement, to obtain preprocessed nighttime vehicle weighing images.

[0008] Optionally, in a fourth implementation of the first aspect of the present invention, constructing a CarDark model based on a convolutional neural network and training the constructed CarDark model to obtain a trained CarDark model includes: Constructing a CarDark model based on a deep learning framework, wherein the CarDark model includes: an input layer, a feature extraction layer, and an output layer; The feature extraction layer includes a convolution layer, a pooling layer, and an activation function. The feature extraction layer includes a convolution layer, a pooling layer, and an activation function, and optimizes the features of low-light and blurred images that meet preset requirements based on the convolution layer, the pooling layer, and the activation function in the feature extraction layer; The CarDark model is optimized through the back-propagation algorithm so that the CarDark model can learn to convert blurry night images into clear and recognizable images.

[0009] Optionally, in the fifth implementation method of the first aspect of the present invention, the labeling processing of the vehicle weighing image dataset includes: using the labelimg labeling tool to label the vehicle weighing image dataset, recording the location information and category information including the vehicle, cargo, weighing weight number and license plate number.

[0010] Optionally, in a sixth implementation of the first aspect of the present invention, building a recognition model based on paddleOCR_V2 and training the recognition model using a training set include: Get two different training sets and mix them using Mixup data enhancement technology to get a new training set. in, and Represents two different training sets; Represents image data from two different training sets; Indicates the labels corresponding to the images in two different training sets; The recognition model is trained using the new training set to obtain a trained recognition model.

[0011] A second aspect of the present invention provides a vehicle weighing information identification device for a logistics transfer center, comprising: An image acquisition module is used to acquire an initial vehicle weighing image and pre-process the acquired initial vehicle weighing image to obtain a vehicle weighing image dataset; A training set construction module is used to label the vehicle weighing image dataset to obtain a training set; the training set is used to train the recognition model; The recognition model building module is used to build a recognition model based on paddleOCR_V2, train the recognition model using the training set, and obtain the trained recognition model; The vehicle weighing information recognition module is used to identify the vehicle weighing information of the logistics transfer center using the trained recognition model; The vehicle weighing information of the logistics transfer center includes: the license plate number of the weighed vehicle, the weight of the cargo, and the name of the cargo.

[0012] Optionally, in a first implementation of the second aspect of the present invention, the image acquisition module includes: Based on the monitoring system, we can obtain the monitoring video of the weighing camera at different time periods within the preset time period; Extracting a frame from the surveillance video at a preset time interval as the initial vehicle weighing image; The initial vehicle weighing images include: vehicle weighing images under different lighting conditions, different vehicle types and different cargo types; The initial vehicle weighing image is generated by adjusting parameters using the open-source tool TextRecognitionDataGenerator to simulate image features in a real environment. The initial vehicle weighing image is preprocessed by the image processing library to obtain a preprocessed initial vehicle weighing image; The synthetic images and the preprocessed initial vehicle weighing images are used to synthesize the vehicle weighing image dataset.

[0013] Optionally, in a second implementation of the second aspect of the present invention, the initial vehicle weighing image is subjected to parameter adjustment using the open source tool TextRecognitionDataGenerator to simulate image features in a real environment to generate a synthetic image, including: Adjust the font type, size, and color of the initial vehicle weighing image to simulate the text logo style in actual logistics scenarios, and obtain the adjusted vehicle weighing image; The tilt angle parameter of the adjusted initial vehicle weighing image is randomly set in the range of -30 degrees to 30 degrees to generate a tilted vehicle weighing image; Setting a blur kernel size and a standard deviation, blurring the tilted vehicle weighing image based on the set blur kernel size and the standard deviation, and obtaining a blurred vehicle weighing image; Using a distortion function to distort the blurred vehicle weighing image according to a preset probability to obtain a composite image; The initial vehicle weighing image is preprocessed by an image processing library to obtain a preprocessed initial vehicle weighing image, including: The initial vehicle weighing image is resized by an image processing library to obtain an adjusted vehicle weighing image; The adjusted vehicle weighing image is cropped through the image processing library to obtain a cropped vehicle weighing image; The cropped vehicle weighing image is color-corrected by an image processing library to obtain a color-corrected vehicle weighing image.

[0014] Optionally, in a third implementation of the second aspect of the present invention, the device further includes: a night image preprocessing module, configured to construct a CarDark model based on a deep learning framework, and train the constructed CarDark model to obtain a trained CarDark model; and perform preprocessing including brightness enhancement and contrast enhancement on the acquired night vehicle weighing image using the trained CarDark model to obtain a preprocessed night vehicle weighing image; The CarDark model is used to perform brightness enhancement and contrast enhancement on blurry pictures at night.

[0015] Optionally, in a fourth implementation of the second aspect of the present invention, constructing a CarDark model based on a convolutional neural network and training the constructed CarDark model to obtain a trained CarDark model includes: Constructing a CarDark model based on a deep learning framework, wherein the CarDark model includes: an input layer, a feature extraction layer, and an output layer; The feature extraction layer includes a convolution layer, a pooling layer, and an activation function. The feature extraction layer includes a convolution layer, a pooling layer, and an activation function, and optimizes the features of low-light and blurred images that meet preset requirements based on the convolution layer, the pooling layer, and the activation function in the feature extraction layer; The CarDark model is optimized through the back-propagation algorithm so that the CarDark model can learn to convert blurry night images into clear and recognizable images.

[0016] Optionally, in the fifth implementation of the second aspect of the present invention, the training set construction module includes: using the labelimg labeling tool to label the vehicle weighing image dataset, recording the location information and category information including the vehicle, cargo, weighing weight number and license plate number.

[0017] Optionally, in a sixth implementation of the second aspect of the present invention, the recognition model building module includes: Get two different training sets and mix them using Mixup data enhancement technology to get a new training set. in, and Represents two different training sets; Represents image data from two different training sets; Indicates the labels corresponding to the images in two different training sets; The recognition model is trained using the new training set to obtain a trained recognition model.

[0018] A third aspect of the present invention provides an electronic device, comprising a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the electronic device to execute the various steps of the logistics transfer center vehicle weighing information identification method as described above.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed by a processor, implement the various steps of the above-mentioned method for identifying vehicle weighing information in a logistics transfer center.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention can effectively process images under blurry and low-light conditions at night, significantly improving the recognition accuracy of vehicle weighing information; 2. This paper uses TextRecognitionDataGenerator to generate synthetic images and adjusts parameters to simulate various image features in real environments, making the training dataset richer and more diverse. This diversity helps the model generalize across different scenarios and improves its performance in practical applications. 3. The present invention adopts Mixup data enhancement technology, which can effectively increase the diversity of training samples, improve the robustness of the model, and reduce the risk of overfitting; 4. The present invention realizes the automatic identification and processing of vehicle weighing information, reduces the need for manual intervention, and improves work efficiency. Through the automated system, the transfer center can obtain weighing information in real time, optimize the logistics management process, and reduce labor costs. 5. The present invention constructs a CarDark model that can effectively cope with nighttime operations and complex environmental conditions. The flexibility and adaptability of the model enable it to process images with different lighting and blur levels, meeting the actual needs of logistics transfer centers. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings: Figure 1 This is a first flow chart of a method for identifying vehicle weighing information in a logistics transfer center provided by an embodiment of the present invention.

[0022] Figure 2 This is a first flow chart of a method for identifying vehicle weighing information in a logistics transfer center provided by an embodiment of the present invention.

[0023] Figure 3 This is a first flow chart of a method for identifying vehicle weighing information in a logistics transfer center provided by an embodiment of the present invention.

[0024] Figure 4 A schematic structural diagram of a vehicle weighing information identification device for a logistics transfer center provided by an embodiment of the present invention.

[0025] Figure 5 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The embodiments of the present invention provide a method, device, equipment, and medium for identifying vehicle weighing information in a logistics transfer center. The method comprises obtaining an initial vehicle weighing image, pre-processing the obtained initial vehicle weighing image, and obtaining a vehicle weighing image dataset. The vehicle weighing image dataset is annotated to obtain a training set. The training set is used to train a recognition model. A recognition model is constructed based on paddleOCR_V2, and the recognition model is trained using the training set to obtain a trained recognition model. The trained recognition model is used to identify vehicle weighing information in a logistics transfer center. The vehicle weighing information in the logistics transfer center includes the license plate number of the weighed vehicle, the weight of the cargo, and the name of the cargo. The present invention solves the problem in the prior art that manual weighing and manual recording are inefficient, prone to human errors, and cannot meet the needs of modern logistics management.

[0027] The terms "first," "second," "third," "fourth," and so on (if any) in the description and claims of the present invention and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that shown or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.

[0028] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 The first embodiment of the method for identifying vehicle weighing information in a logistics transfer center according to the present invention includes: 101. Acquire an initial vehicle weighing image, and preprocess the acquired initial vehicle weighing image to obtain a vehicle weighing image dataset; In this embodiment, surveillance video footage of the vehicle weighing camera at different time periods within a certain period of time is obtained based on the monitoring system; a frame of the video is extracted from the surveillance video every 5 minutes as the initial vehicle weighing image; these samples cover different lighting conditions, vehicle types and cargo types to ensure the comprehensiveness and diversity of the data set.

[0029] To account for varying lighting conditions, we specifically selected images during the early morning, midday, evening, and nighttime hours, paying particular attention to weighing scenes under cloudy, sunny, and rainy weather conditions to ensure diverse lighting conditions. To encompass a wide range of vehicle types, we recorded weighing moments for common trucks, including vans, flatbed trucks, dump trucks, trailers, and small cargo vehicles, taking into account vehicle color, size, and load status. Regarding cargo types, we collected images of weighing operations for a variety of cargo, including building materials, fresh food, and daily necessities, ensuring a rich variety of cargo appearance and packaging.

[0030] The initial vehicle weighing images were processed using the open-source tool TextRecognitionDataGenerator, with parameters adjusted to simulate image features in a real environment. This generated synthetic images to enhance the complexity and robustness of the dataset. The initial vehicle weighing image is preprocessed through the image processing library to obtain the preprocessed initial vehicle weighing image to meet the input requirements of the recognition model; The synthetic images and the preprocessed initial vehicle weighing images are used to synthesize the vehicle weighing image dataset.

[0031] The initial vehicle weighing image is parameterized using the open-source tool TextRecognitionDataGenerator to simulate image features in a real environment, generating a synthetic image, including: Adjust the font type, size, and color of the initial vehicle weighing image to simulate the text logo style in actual logistics scenarios, and obtain the adjusted vehicle weighing image; The tilt angle parameter of the adjusted initial vehicle weighing image is randomly set within the range of -30 degrees to 30 degrees to generate a tilted vehicle weighing image; Set the blur kernel size and standard deviation, and blur the tilted vehicle weighing image based on the set blur kernel size and standard deviation to obtain a blurred vehicle weighing image to simulate the blur effect caused by inaccurate camera focus or fast-moving vehicles. The blurred vehicle weighing image is distorted using a distortion function according to a preset probability to obtain a synthetic image, which is similar to the impact of vibration on the image when a vehicle passes through an uneven road.

[0032] The initial vehicle weighing image is preprocessed by an image processing library to obtain a preprocessed initial vehicle weighing image, including: The initial vehicle weighing image is resized by the image processing library to obtain an adjusted composite image, thereby avoiding waste of computing resources and slow model training caused by excessive image size; The adjusted composite image is cropped by an image processing library to remove background areas irrelevant to the weighing key information in the image, thereby obtaining a cropped composite image; The cropped composite image is subjected to color correction processing through an image processing library, and the color cast of the image is automatically corrected to ensure the color consistency of the image under different lighting conditions, thereby obtaining a pre-processed initial vehicle weighing image.

[0033] 102. Label the vehicle weighing image dataset to obtain a training set; the training set is used to train the recognition model; In this embodiment, the labelimg annotation tool is used to annotate the vehicle weighing image dataset, record the location information and category information including the vehicle, cargo, weighing weight and license plate number, and convert it into a format that can be recognized by the recognition model.

[0034] 103. Build a recognition model based on paddleOCR_V2, use the training set to train the recognition model, and obtain the trained recognition model; In this embodiment, two different training sets are obtained and mixed by using the Mixup data enhancement technology to obtain a new training set. in, and Represents two different training sets; Represents image data from two different training sets; Indicates the labels corresponding to the images in two different training sets; The recognition model is trained using the new training set to obtain a trained recognition model.

[0035] 104. Use the trained recognition model to identify the vehicle weighing information of the logistics transfer center; In this example, a pre-processed test image is fed into a trained recognition model, the paddleOCR_V2 model, which incorporates Mixup data augmentation technology. The model predicts the text in the image based on learned image features and text patterns, outputting a probability distribution for each character's category. Post-processing algorithms, such as taking the character with the highest probability as the prediction result and concatenating consecutive characters according to specific language rules, form a complete text recognition result. For example, key information such as the vehicle's license plate number, cargo weight, and cargo name can be accurately identified.

[0036] See also Figure 2 The second embodiment of the method for identifying vehicle weighing information in a logistics transfer center according to the present invention includes: 201. Acquire an initial vehicle weighing image, and pre-process the acquired initial vehicle weighing image to obtain a vehicle weighing image dataset; In this embodiment, the monitoring system obtains the vehicle weighing camera monitoring video at different time periods within a preset time period; extracts a frame from the monitoring video at each preset time interval as the initial vehicle weighing image; The initial vehicle weighing images include: vehicle weighing images under different lighting conditions, different vehicle types and different cargo types; The initial vehicle weighing image is processed using the open-source tool TextRecognitionDataGenerator, with parameters adjusted to simulate image features in a real environment, such as tilt, blur, and distortion, to generate a synthetic image. The initial vehicle weighing image is pre-processed by the image processing library including resizing, cropping and color correction to obtain the pre-processed initial vehicle weighing image; The synthetic images and the preprocessed initial vehicle weighing images are used to synthesize the vehicle weighing image dataset.

[0037] 202. Constructing a CarDark model based on a deep learning framework and training the constructed CarDark model to obtain a trained CarDark model; using the trained CarDark model to preprocess the acquired nighttime vehicle weighing image, including brightness enhancement and contrast enhancement, to obtain a preprocessed nighttime vehicle weighing image; In this embodiment, the CarDark model is used to perform brightness enhancement and contrast enhancement on blurry pictures at night; Constructing a CarDark model based on a deep learning framework, wherein the CarDark model includes: an input layer, a feature extraction layer, and an output layer; The feature extraction layer includes a convolution layer, a pooling layer, and an activation function, and optimizes the features of low-light and blurred images that meet preset requirements based on the convolution layer, the pooling layer, and the activation function in the feature extraction layer; The CarDark model is optimized through the back-propagation algorithm so that the CarDark model can learn to convert blurry night images into clear and recognizable images.

[0038] 203. Use the labelimg annotation tool to annotate the vehicle weighing image dataset and the pre-processed nighttime vehicle weighing images to obtain a training set; the annotated records include the location and category information of the vehicle, cargo, weighing weight, and license plate number; In this example, after opening an image, carefully select the location of objects such as vehicles, cargo name labels, weighing numbers, and license plates, and annotate each selected area with the corresponding category information. The annotation results are converted to a format commonly used by models such as PASCAL VOC or COCO, recording each object's category ID and bounding box coordinates (top-left and bottom-right corner coordinates) so that the model can accurately read the annotation information for training.

[0039] 204. Build a recognition model based on paddleOCR_V2, use the training set to train the recognition model, and obtain the trained recognition model; 205. Use the trained recognition model to identify the vehicle weighing information of the logistics transfer center; The vehicle weighing information of the logistics transfer center includes: the license plate number of the weighed vehicle, the weight of the cargo, and the name of the cargo.

[0040] See also Figure 3 The second embodiment of the method for identifying vehicle weighing information in a logistics transfer center according to the present invention includes: 301. Obtain vehicle weighing image samples monitored by the weighing camera in the logistics transfer center, and pre-process and annotate the images; In this example, to construct an efficient dataset for camera-based weighing monitoring at a logistics transfer center, we first need to collect image samples of various weighing situations from the actual monitoring system. We then obtain camera-based weighing monitoring footage from different time periods over a period of time, such as the past month. We then extract a frame from the footage at regular intervals, for example, every 5 minutes, to serve as the initial image sample.

[0041] The open-source tool TextRecognitionDataGenerator was used to generate synthetic images. Parameters were adjusted to simulate real-world image features, such as tilt, blur, and distortion, to enhance the dataset's complexity and robustness. Font type, size, and color were adjusted to simulate common text labeling styles found in actual logistics scenarios. For example, a variety of handwritten fonts were used to simulate driver-written weighing slips, while printed fonts were used to simulate formal cargo labels. To simulate image features, the tilt angle parameter was randomly set between -30 and 30 degrees to generate tilted text images. The blur kernel size and standard deviation were set to produce varying degrees of blur in the image, simulating blurring caused by camera inaccuracy or rapid vehicle movement. A distortion function was then applied to slightly distort the image with a certain probability (e.g., 20%), similar to the vibration caused by a vehicle traversing an uneven road.

[0042] The image is then resized, cropped, and color-corrected using image processing libraries such as PIL or OpenCV to meet the input requirements of the recognition model. Using the Python Imaging Library (PIL), oversized images are resized according to the model input requirements (e.g., scaling the long side to 640 pixels and the short side proportionally), avoiding wasted computing resources and slow model training caused by oversized images. Cropping removes background areas irrelevant to the key weighing information, such as distant parking lots and warehouse corners, highlighting the core objects such as the vehicle, cargo, and weighing display. A color correction algorithm automatically corrects color casts based on the image's histogram distribution, ensuring color consistency under varying lighting conditions and facilitating subsequent text recognition.

[0043] Use annotation tools such as labelimg to accurately label images; The synthetic images are combined with actual samples to form a complete training dataset, which is used to train and test the recognition model to achieve automatic recognition and processing of weighing information.

[0044] 302. Train the paddleOCR_V2 model using Mixup data enhancement technology; In this embodiment, Mixup data enhancement technology is used during model training. Mixup is a data enhancement technology that mixes two different training samples by linear interpolation to generate new training samples. This method can not only increase the diversity of training data, but also improve the generalization ability and robustness of the model. Mixup technology is used for two training samples ( , )and( , ),in is the image data, For the corresponding label, Mixup generates new samples = +(1- ) and ŷ = +(1-λ) , It is a parameter randomly sampled from the Beta distribution, with a value range between 0 and 1. Locate the configuration file corresponding to the paddleOCR_V2 model, which is usually a file ending in .yml or .json. The file contains key information such as the structural parameters of the model (such as the settings of the convolutional layer, pooling layer, and fully connected layer), training parameters (such as learning rate, number of iterations, and optimizer type). Modify the parameters in the configuration file according to the characteristics of the dataset and actual needs. For example, if the dataset is large, you can appropriately reduce the learning rate (such as from the initial 0.001 to 0.0005) to prevent the model from skipping the optimal solution during training; increase the number of iterations (such as from 100 to 150) to ensure that the model fully learns the data features; adjust the momentum parameter of the optimizer to improve the convergence speed of the model.

[0045] 303. Build our own CarDark model for night-time blurred images and pre-process them. In this example, considering that logistics transfer centers primarily operate at night, a proprietary CarDark model for nighttime blurred images was constructed. Brightness and contrast enhancement were performed on a number of blurred nighttime images, and the model was trained and subsequently applied to the preprocessing stage of the predicted images. Representative blurred images were selected from the logistics transfer center's nighttime surveillance footage. These images should contain varying degrees of dim lighting, glare, and noise caused by low illumination. Key information in these nighttime images, such as vehicle outlines, license plates, and weighing data display areas, was manually annotated to provide accurate supervision for subsequent model training. A custom convolutional neural network model was built based on a deep learning framework. The model architecture referenced existing image enhancement and deblurring model architectures and was optimized based on the characteristics of nighttime logistics images. For example, convolutional layers sensitive to low-light features were added, and pooling strategies were employed to preserve detailed information. During model training, a dataset of blurred nighttime images was constructed, with a clear image after brightness and contrast enhancement as the target output. A loss function was defined, and the model parameters were continuously adjusted through a backpropagation algorithm, enabling the model to learn the mapping relationship that converts blurred nighttime images into clear, recognizable images. Prediction Image Preprocessing Application: During the actual prediction phase, the input nighttime prediction image is first fed into the trained CarDark model. The model outputs an image with enhanced brightness and contrast, which is visually clearer and more prominent in text. This enhanced image is then normalized to ensure that its pixel values ​​meet the input requirements of the subsequent OCR model. For example, pixel values ​​are normalized to between 0 and 1 to ensure stable text recognition.

[0046] 304, Model Prediction and Deployment; In this embodiment, a preprocessed test image is input into a trained OCR model, the paddleOCR_V2 model trained with Mixup. Based on learned image features and text patterns, the model predicts the textual information in the image and outputs a probability distribution for each character's category. A post-processing algorithm, such as selecting the character with the highest probability as the prediction result, joins consecutive characters according to specific language rules to form a complete text recognition result. This can accurately identify key information such as the vehicle's license plate number, cargo weight, and cargo name. Deployment is determined based on the logistics transfer center's actual business processes and IT infrastructure. If the transfer center already has a mature server cluster, the model can be encapsulated as a RESTful API service and deployed on the server. Other business systems (such as the logistics management system and the weighing and billing system) can then invoke the OCR service by sending HTTP requests to obtain real-time weighing information recognition results. Considering edge computing requirements, for weighing stations with extremely high real-time requirements and unstable network conditions, the model can be lightweighted and deployed on local edge devices (such as industrial-grade gateways and smart cameras). Image acquisition and text recognition can be performed directly on the device, reducing data transmission latency and ensuring efficient business operations. During the deployment process, thorough performance testing is also required, including indicators such as the model's inference speed, concurrent processing capabilities, and resource utilization. Based on the test results, the deployment configuration can be further optimized to ensure the model's stable and reliable operation in actual logistics scenarios.

[0047] The above describes the method for identifying vehicle weighing information in a logistics transfer center according to an embodiment of the present invention. The following describes the device for identifying vehicle weighing information in a logistics transfer center according to an embodiment of the present invention. Figure 4 In one embodiment of the present invention, a device for identifying vehicle weighing information in a logistics transfer center includes: The image acquisition module 401 is used to acquire an initial vehicle weighing image and pre-process the acquired initial vehicle weighing image to obtain a vehicle weighing image dataset; In this embodiment, the image acquisition module 401 includes: Based on the monitoring system, we obtain surveillance videos of vehicle weighing cameras at different time periods over a period of time. We extract a frame from the surveillance video every 5 minutes as the initial vehicle weighing image. These samples cover different lighting conditions, vehicle types, and cargo types to ensure the comprehensiveness and diversity of the dataset.

[0048] To account for varying lighting conditions, we specifically selected images during the early morning, midday, evening, and nighttime hours, paying particular attention to weighing scenes under cloudy, sunny, and rainy weather conditions to ensure diverse lighting conditions. To encompass a wide range of vehicle types, we recorded weighing moments for common trucks, including vans, flatbed trucks, dump trucks, trailers, and small cargo vehicles, taking into account vehicle color, size, and load status. Regarding cargo types, we collected images of weighing operations for a variety of cargo, including building materials, fresh food, and daily necessities, ensuring a rich variety of cargo appearance and packaging.

[0049] The initial vehicle weighing images were processed using the open-source tool TextRecognitionDataGenerator, with parameters adjusted to simulate image features in a real environment. This generated synthetic images to enhance the complexity and robustness of the dataset. The initial vehicle weighing image is preprocessed through the image processing library to obtain the preprocessed initial vehicle weighing image to meet the input requirements of the recognition model; The synthetic images and the preprocessed initial vehicle weighing images are used to synthesize the vehicle weighing image dataset.

[0050] The initial vehicle weighing image is parameterized using the open-source tool TextRecognitionDataGenerator to simulate image features in a real environment, generating a synthetic image, including: Adjust the font type, size, and color of the initial vehicle weighing image to simulate the text logo style in actual logistics scenarios, and obtain the adjusted vehicle weighing image; The tilt angle parameter of the adjusted initial vehicle weighing image is randomly set within the range of -30 degrees to 30 degrees to generate a tilted vehicle weighing image; Set the blur kernel size and standard deviation, and blur the tilted vehicle weighing image based on the set blur kernel size and standard deviation to obtain a blurred vehicle weighing image to simulate the blur effect caused by inaccurate camera focus or fast-moving vehicles. The blurred vehicle weighing image is distorted using a distortion function according to a preset probability to obtain a synthetic image, which is similar to the impact of vibration on the image when a vehicle passes through an uneven road.

[0051] The initial vehicle weighing image is preprocessed by an image processing library to obtain a preprocessed initial vehicle weighing image, including: The initial vehicle weighing image is resized by the image processing library to obtain an adjusted composite image, thereby avoiding waste of computing resources and slow model training caused by excessive image size; The adjusted composite image is cropped by an image processing library to remove background areas irrelevant to the weighing key information in the image, thereby obtaining a cropped composite image; The cropped composite image is subjected to color correction processing through an image processing library, and the color cast of the image is automatically corrected to ensure the color consistency of the image under different lighting conditions, thereby obtaining a pre-processed initial vehicle weighing image.

[0052] The night image preprocessing module 402 is configured to construct a CarDark model based on a deep learning framework, train the constructed CarDark model, and obtain a trained CarDark model; and perform preprocessing including brightness enhancement and contrast enhancement on the acquired night vehicle weighing images using the trained CarDark model to obtain preprocessed night vehicle weighing images. In this embodiment, the nighttime image preprocessing module 402 includes: The CarDark model is used to perform brightness enhancement and contrast enhancement on blurry pictures at night; Constructing a CarDark model based on a deep learning framework, wherein the CarDark model includes: an input layer, a feature extraction layer, and an output layer; The feature extraction layer includes a convolution layer, a pooling layer, and an activation function, and optimizes the features of low-light and blurred images that meet preset requirements based on the convolution layer, the pooling layer, and the activation function in the feature extraction layer; The CarDark model is optimized through the back-propagation algorithm so that the CarDark model can learn to convert blurry night images into clear and recognizable images.

[0053] The training set construction module 403 is used to label the vehicle weighing image dataset and the pre-processed nighttime vehicle weighing images to obtain a training set; the training set is used to train the recognition model; In this embodiment, the labelimg annotation tool is used to annotate the vehicle weighing image dataset and the pre-processed nighttime vehicle weighing images to obtain a training set; wherein the annotation records include the location and category information of the vehicle, cargo, weighing weight, and license plate number; More specifically, after opening the image, carefully select the locations of objects such as vehicles, cargo name labels, weighing numbers, and license plates, and annotate each selected area with the corresponding category information. The annotation results are converted to a format commonly used by models such as PASCAL VOC or COCO, recording each object's category ID and bounding box coordinates (top-left and bottom-right corner coordinates) so that the model can accurately read the annotation information for training.

[0054] The recognition model building module 404 is used to build a recognition model based on paddleOCR_V2, train the recognition model using the training set, and obtain a trained recognition model; In this embodiment, the recognition model building module 404 includes: Get two different training sets and mix them using Mixup data enhancement technology to get a new training set. in, and Represents two different training sets; Represents image data from two different training sets; Indicates the labels corresponding to the images in two different training sets; The recognition model is trained using the new training set to obtain a trained recognition model.

[0055] The vehicle weighing information recognition module 405 is used to recognize the vehicle weighing information of the logistics transfer center using the trained recognition model; In this example, a pre-processed test image is fed into a trained recognition model, the paddleOCR_V2 model, which incorporates Mixup data augmentation technology. The model predicts the text in the image based on learned image features and text patterns, outputting a probability distribution for each character's category. Post-processing algorithms, such as taking the character with the highest probability as the prediction result and concatenating consecutive characters according to specific language rules, form a complete text recognition result. For example, key information such as the vehicle's license plate number, cargo weight, and cargo name can be accurately identified.

[0056] above Figure 4 The vehicle weighing information identification device of the logistics transfer center in the embodiment of the present invention is described in detail from the perspective of modular functional entities, and the electronic device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0057] Figure 5is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device 700 may vary significantly due to different configurations or performance, and may include one or more processors (central processing units, CPUs) 710 (for example, one or more processors), a memory 720, and one or more storage media 730 (for example, one or more mass storage devices) storing application programs 733 or data 732. The memory 720 and storage medium 730 may be either transient or persistent storage. The program stored in the storage medium 730 may include one or more modules (not shown), each of which may include a series of instruction operations on the electronic device 700. Furthermore, the processor 710 may be configured to communicate with the storage medium 730 to execute the series of instruction operations in the storage medium 730 on the electronic device 700.

[0058] The electronic device 700 may further include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input and output interfaces 750, and / or one or more operating systems 731, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. It will be understood by those skilled in the art that Figure 5 The illustrated electronic device structure does not constitute a limitation on the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0059] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of a method for identifying vehicle weighing information in a logistics transfer center.

[0060] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0061] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0062] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying vehicle weighing information in a logistics transfer center, characterized in that: include: Acquire an initial vehicle weighing image, and preprocess the acquired initial vehicle weighing image to obtain a vehicle weighing image dataset; Annotate the vehicle weighing image dataset to obtain a training set; the training set is used to train the recognition model; Build a recognition model based on paddleOCR_V2, use the training set to train the recognition model, and obtain the trained recognition model; Use the trained recognition model to identify vehicle weighing information in the logistics transfer center; The vehicle weighing information of the logistics transfer center includes: the license plate number of the weighed vehicle, the weight of the cargo, and the name of the cargo.

2. The method for identifying vehicle weighing information in a logistics transfer center according to claim 1, characterized in that: The step of obtaining an initial vehicle weighing image and preprocessing the obtained initial vehicle weighing image to obtain a vehicle weighing image dataset includes: Based on the monitoring system, we can obtain the monitoring video of the weighing camera at different time periods within the preset time period; Extracting a frame from the surveillance video at a preset time interval as the initial vehicle weighing image; The initial vehicle weighing images include: vehicle weighing images under different lighting conditions, different vehicle types and different cargo types; The initial vehicle weighing image is generated by adjusting parameters using the open-source tool TextRecognitionDataGenerator to simulate image features in a real environment. The initial vehicle weighing image is preprocessed by the image processing library to obtain a preprocessed initial vehicle weighing image; The synthetic images and the preprocessed initial vehicle weighing images are used to synthesize the vehicle weighing image dataset.

3. The method for identifying vehicle weighing information in a logistics transfer center according to claim 2, characterized in that: The initial vehicle weighing image is parameterized using the open-source tool TextRecognitionDataGenerator to simulate image features in a real environment, generating a synthetic image, including: Adjust the font type, size, and color of the initial vehicle weighing image to simulate the text logo style in actual logistics scenarios, and obtain the adjusted vehicle weighing image; The tilt angle parameter of the adjusted initial vehicle weighing image is randomly set within the range of -30 degrees to 30 degrees to generate a tilted vehicle weighing image; Setting a blur kernel size and a standard deviation, blurring the tilted vehicle weighing image based on the set blur kernel size and the standard deviation, and obtaining a blurred vehicle weighing image; Using a distortion function to distort the blurred vehicle weighing image according to a preset probability to obtain a composite image; The initial vehicle weighing image is preprocessed by an image processing library to obtain a preprocessed initial vehicle weighing image, including: The initial vehicle weighing image is resized by an image processing library to obtain an adjusted vehicle weighing image; The adjusted vehicle weighing image is cropped through the image processing library to obtain a cropped vehicle weighing image; The cropped vehicle weighing image is color-corrected by an image processing library to obtain a color-corrected vehicle weighing image.

4. The method for identifying vehicle weighing information in a logistics transfer center according to claim 3, characterized in that: The method further comprises: Construct a CarDark model based on a deep learning framework, and train the constructed CarDark model to obtain a trained CarDark model; The CarDark model is used to perform brightness enhancement and contrast enhancement on blurry pictures at night; The trained CarDark model is used to preprocess the acquired nighttime vehicle weighing images, including brightness enhancement and contrast enhancement, to obtain preprocessed nighttime vehicle weighing images.

5. The method for identifying vehicle weighing information in a logistics transfer center according to claim 4, characterized in that: The CarDark model is constructed based on a convolutional neural network, and the constructed CarDark model is trained to obtain a trained CarDark model, including: Constructing a CarDark model based on a deep learning framework, wherein the CarDark model includes: an input layer, a feature extraction layer, and an output layer; The feature extraction layer includes a convolution layer, a pooling layer, and an activation function, and optimizes the features of low-light and blurred images that meet preset requirements based on the convolution layer, the pooling layer, and the activation function in the feature extraction layer; The CarDark model is optimized through the back-propagation algorithm so that the CarDark model can learn to convert blurry night images into clear and recognizable images.

6. The method for identifying vehicle weighing information in a logistics transfer center according to claim 1, characterized in that: The labeling process of the vehicle weighing image dataset includes: labeling the vehicle weighing image dataset using the labelimg labeling tool to record the location information and category information including the vehicle, cargo, weighing weight and license plate number.

7. The method for identifying vehicle weighing information in a logistics transfer center according to claim 1, characterized in that: The recognition model is constructed based on paddleOCR_V2, and the recognition model is trained using the training set, including: Get two different training sets and mix them using Mixup data enhancement technology to get a new training set. in, and Represents two different training sets; Represents image data from two different training sets; Indicates the labels corresponding to the images in two different training sets; The recognition model is trained using the new training set to obtain a trained recognition model.

8. A vehicle weighing information identification device for a logistics transfer center, characterized in that: include: An image acquisition module is used to acquire an initial vehicle weighing image and pre-process the acquired initial vehicle weighing image to obtain a vehicle weighing image dataset; A training set construction module is used to label the vehicle weighing image dataset to obtain a training set; the training set is used to train the recognition model; The recognition model building module is used to build a recognition model based on paddleOCR_V2, train the recognition model using the training set, and obtain the trained recognition model; The vehicle weighing information recognition module is used to identify the vehicle weighing information of the logistics transfer center using the trained recognition model; The vehicle weighing information of the logistics transfer center includes: the license plate number of the weighed vehicle, the weight of the cargo, and the name of the cargo.

9. An electronic device comprising a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory so that the electronic device executes each step of the method for identifying vehicle weighing information in a logistics transfer center as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the method for identifying vehicle weighing information in a logistics transfer center as described in any one of claims 1-7 are implemented.

Images