CN116883981A - License plate positioning and identifying method, system, computer equipment and storage medium - Google Patents
License plate positioning and identifying method, system, computer equipment and storage medium Download PDFInfo
- Publication number
- CN116883981A CN116883981A CN202310713794.5A CN202310713794A CN116883981A CN 116883981 A CN116883981 A CN 116883981A CN 202310713794 A CN202310713794 A CN 202310713794A CN 116883981 A CN116883981 A CN 116883981A
- Authority
- CN
- China
- Prior art keywords
- license plate
- module
- image
- distance
- camera
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000003860 storage Methods 0.000 title claims abstract description 11
- 238000001514 detection method Methods 0.000 claims abstract description 31
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 24
- 238000012015 optical character recognition Methods 0.000 claims abstract description 10
- 238000000605 extraction Methods 0.000 claims description 40
- 230000006870 function Effects 0.000 claims description 32
- 230000004913 activation Effects 0.000 claims description 20
- 238000004590 computer program Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 10
- 238000012937 correction Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 5
- 238000003709 image segmentation Methods 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 15
- 210000003128 head Anatomy 0.000 description 13
- 238000004364 calculation method Methods 0.000 description 12
- 238000012549 training Methods 0.000 description 10
- 238000003384 imaging method Methods 0.000 description 9
- 230000003287 optical effect Effects 0.000 description 7
- 230000006872 improvement Effects 0.000 description 5
- 230000002776 aggregation Effects 0.000 description 4
- 238000004220 aggregation Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000007726 management method Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000016273 neuron death Effects 0.000 description 1
- 230000003121 nonmonotonic effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
- G06T7/85—Stereo camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/24—Character recognition characterised by the processing or recognition method
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20228—Disparity calculation for image-based rendering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The application provides a license plate positioning and identifying method, a license plate positioning and identifying system, computer equipment and a storage medium, wherein the license plate positioning and identifying method comprises the following steps: acquiring a license plate image to be detected, and performing target detection on the license plate image to be detected by adopting an improved YOLOv7-tiny network model to obtain a license plate positioning area; obtaining license plate position coordinates and an area image to be identified according to the license plate positioning area; obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through a binocular distance measuring algorithm; and judging whether the distance between the license plate and the camera meets the preset distance requirement, and when the distance is met, adopting a preset character recognition model to perform optical character recognition on the image of the area to be recognized to obtain a corresponding license plate number. The application adopts a lighter network model, improves the detection speed and detection precision of the license plate region, and simultaneously realizes simple, efficient and low-cost license plate anti-counterfeiting recognition by combining with binocular range, thereby effectively improving the reliability and application value of license plate recognition.
Description
Technical Field
The application relates to the technical field of artificial intelligence and computer vision, in particular to a license plate positioning and identifying method, a license plate positioning and identifying system, computer equipment and a storage medium.
Background
With the rapid development of economy and the improvement of living standard of people, vehicle travel becomes a very common traffic mode, and meanwhile, the vehicle management difficulty of a parking lot is increased.
The license plate recognition of the existing intelligent parking lot mostly adopts a traditional algorithm of license plate character recognition or a license plate positioning and detecting method based on deep learning, although the intelligence of the vehicle management of the parking lot can be improved to a certain extent, the accuracy and the high efficiency of the license plate recognition are to be improved, and the license plate anti-fake recognition problem is not considered in most cases, a multiplicable machine is provided for people with different interests, for example, after a mobile phone is used for photographing a license plate of a certain vehicle, a camera where a mobile phone is attached to a license plate detection gate is directly used for misleading a camera to directly detect a license plate photo in the mobile phone, so that the license plate recognition is wrong, and the vehicle management is disordered; in addition, although a few license plate recognition systems consider anti-counterfeiting recognition, whether the automobile passes or not is determined by adding a mode of sensing the weight of the automobile through the ground sensing, the ground sensing equipment is high in cost, the cost overhead of license plate recognition is increased, and the application limit is large.
Disclosure of Invention
The application aims to provide a license plate positioning and identifying method, which combines the license plate positioning and identifying method by adopting an improved lighter YOLOv7-tiny network model with the license plate ranging by a binocular ranging algorithm, solves the application defects of the existing license plate positioning and identifying method, improves the license plate area detecting speed and detecting precision, realizes simple, efficient and low-cost license plate anti-counterfeiting identification, and effectively improves the reliability and application value of license plate identification.
In order to achieve the above objective, it is necessary to provide a license plate positioning and identifying method and system for the above technical problems.
In a first aspect, an embodiment of the present application provides a license plate positioning and identifying method, where the method includes the following steps:
acquiring a license plate image to be detected, and performing target detection on the license plate image to be detected by adopting an improved YOLOv7-tiny network model to obtain a license plate positioning area;
obtaining license plate position coordinates and an area image to be identified according to the license plate positioning area;
obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through a binocular distance measuring algorithm;
and judging whether the distance between the license plate and the camera meets the preset distance requirement, and when the distance between the license plate and the camera meets the preset distance requirement, adopting a preset character recognition model to perform optical character recognition on the image of the area to be recognized to obtain a corresponding license plate number.
Further, the improved YOLOv7-tiny network model comprises an Input layer, an improved backhaul layer and an improved Head layer which are connected in sequence; the convolution modules in the improved backhaul layer and the improved Head layer are CB-F modules; the CB-F module comprises a convolution layer, a batch normalization layer and a FReLU activation function which are sequentially connected.
Further, replacing the ELAN_T module in the improved backhaul layer and the improved Head layer with a PM-ELAN module and an MS-ELAN module, respectively; the PM-ELAN module is obtained by integrating a PConv module and a Mobilene V3 module based on the ELAN_T module; the MS-ELAN module is obtained by integrating a Mobilene V3 module based on the ELAN_T module, and inserting an SE attention module after the Mobilene V3 module.
Further, the PM-ELAN module comprises a first feature extraction module, a feature fusion module and a Mobilene V3 module which are sequentially connected; the first feature extraction module comprises a first feature extraction branch and a second feature extraction branch which are connected in parallel; the first feature extraction branch comprises a CB-F module and two PConv modules which are connected in sequence; the second feature extraction branch comprises a CB-F module;
the MS-ELAN module comprises a second feature extraction module, a feature fusion module, a Mobilene V3 module and an SE attention module which are connected in sequence; the second feature extraction module comprises a third feature extraction branch and a fourth feature extraction branch which are connected in parallel; the third feature extraction branch comprises 3 CB-F modules connected in series; the fourth feature extraction branch includes a CB-F module.
Further, the frame loss function of the improved YOLOv7-tiny network model is expressed as:
Loss＝rR WIOU L IOU
in the method, in the process of the application,
L IOU ＝1-IOU
wherein Loss represents a frame Loss value; r is R WIOU Representing a penalty term; IOU and L IOU Respectively representing the intersection ratio and the loss value of the prediction boundary box and the real boundary box; beta represents an outlier of the prediction bounding box; delta and alpha are hyper-parameters; r represents the radius of the prediction bounding box; (x, y) represents the center coordinates of the prediction bounding box nearest to it; (x) gt ,y gt ) Representing the center coordinates of the minimum prediction bounding box; (w) gt ,h gt ) Representing the width and height of the minimum prediction bounding box.
Further, the step of obtaining license plate position coordinates and an area image to be identified according to the license plate positioning area comprises the following steps:
obtaining a region center coordinate according to the frame coordinates of the license plate positioning region, and taking the region center coordinate as the license plate position coordinate;
and according to the license plate positioning area, image segmentation is carried out on the license plate image to be detected, and the area image to be identified is obtained.
Further, the step of obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through the binocular distance measuring algorithm comprises the following steps:
acquiring internal and external parameters of a binocular camera by adopting a Zhang Zhengyou camera calibration method;
performing binocular correction on the binocular camera in response to completion of acquisition of internal and external parameters of the binocular camera;
and responding to the completion of binocular correction, performing stereo matching through an SGBM algorithm to obtain a parallax depth map, and obtaining the distance between the license plate and a camera according to the parallax depth map and the license plate position coordinates.
In a second aspect, an embodiment of the present application provides a license plate positioning and identifying system, including:
the area positioning module is used for acquiring a license plate image to be detected, and performing target detection on the license plate image to be detected by adopting an improved YOLOv7-tiny network model to obtain a license plate positioning area;
the region processing module is used for obtaining license plate position coordinates and a region image to be identified according to the license plate positioning region;
the license plate distance measuring module is used for obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through a binocular distance measuring algorithm;
and the license plate recognition module is used for judging whether the distance between the license plate and the camera meets the preset distance requirement, and carrying out optical character recognition on the image of the area to be recognized by adopting a preset character recognition model when the distance between the license plate and the camera meets the preset distance requirement, so as to obtain a corresponding license plate number.
In a third aspect, embodiments of the present application further provide a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the above method when executing the computer program.
In a fourth aspect, embodiments of the present application also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the above method.
The application provides a license plate positioning and identifying method, a system, computer equipment and a storage medium, which realize the technical scheme that an improved YOLOv7-tiny network model is adopted to carry out target detection on an acquired license plate image to be detected to obtain a license plate positioning area, license plate position coordinates and an area image to be identified are obtained according to the license plate positioning area, then a license plate and camera distance is obtained according to the license plate position coordinates through a binocular range algorithm, whether the license plate and camera distance meets the preset distance requirement is judged, and when the license plate and camera distance meets the preset distance requirement, a preset character identification model is adopted to carry out optical character identification on the area image to be identified to obtain a corresponding license plate number. Compared with the prior art, the license plate positioning and identifying method reduces network parameters by adopting a lighter YOLOv7-tiny network model, improves the detection speed and detection precision of a license plate region, and simultaneously realizes simple, efficient and low-cost license plate anti-counterfeiting identification by combining with binocular vision, thereby effectively improving the reliability and application value of license plate identification.
Drawings
FIG. 1 is a schematic diagram of an application scenario of a license plate positioning and identifying method in an embodiment of the application;
FIG. 2 is a schematic flow chart of a license plate positioning and identifying method in an embodiment of the application;
FIG. 3 is a schematic diagram of a conventional YOLOv7-tiny network model;
FIG. 4 is a schematic diagram of the structure of a CB-F module of the improved YOLOv7-tiny network model in an embodiment of the application;
FIG. 5 is a schematic diagram of the structure of a PM-ELAN module in the modified backhaul layer of the modified Yolov7-tiny network model in an embodiment of the present application;
FIG. 6 is a schematic diagram of the structure of an MS-ELAN module in the modified Head layer of the modified Yolov7-tiny network model in an embodiment of the present application;
FIG. 7 is a schematic diagram of the structure of a Mobilene V3 module of the PM-ELAN module and the MS-ELAN module in an embodiment of the application;
FIG. 8 is a schematic diagram of the SPPCSPC_tiny layer of the improved Yolov7-tiny network model in accordance with an embodiment of the present application;
FIG. 9 is a schematic diagram of the structure of the improved YOLOv7-tiny network model in accordance with an embodiment of the present application;
FIG. 10 is a schematic diagram of a system imaging model of a binocular camera in an embodiment of the present application;
FIG. 11 is a diagram illustrating the conversion between the physical coordinate system of an image and the pixel coordinate system of the image in a binocular camera according to an embodiment of the present application;
FIG. 12 is a schematic diagram of the principle of binocular distance measurement in an embodiment of the present application;
FIG. 13 is a schematic diagram of a license plate positioning and identifying system according to an embodiment of the present application;
fig. 14 is an internal structural view of a computer device in the embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantageous effects of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples, and it is apparent that the examples described below are part of the examples of the present application, which are provided for illustration only and are not intended to limit the scope of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The license plate positioning and identifying method provided by the application can be applied to a terminal or a server shown in figure 1. The terminal may be, but not limited to, various personal computers, notebook computers, smartphones, tablet computers and portable wearable devices, and the server may be implemented by a separate server or a server cluster formed by a plurality of servers. The server can carry out high-efficiency and accurate license plate positioning recognition on vehicles to be managed entering and exiting the intelligent parking lot by adopting the license plate positioning recognition method according to actual application requirements, so that the anti-counterfeiting recognition capability and the application reliability of the license plate management system of the intelligent parking lot are improved; the following examples will explain the license plate location recognition method of the present application in detail.
In one embodiment, as shown in fig. 2, a license plate positioning and identifying method is provided, which includes the following steps:
s11, acquiring a license plate image to be detected, and performing target detection on the license plate image to be detected by adopting an improved YOLOv7-tiny network model to obtain a license plate positioning area; the license plate positioning area can be understood as an image area in a prediction boundary frame obtained by performing target detection through an improved YOLOv7-tiny network model;
the improved YOLOv7-tiny network model adopted in the embodiment can be understood as a single-stage detection model based on deep learning target detection and obtained based on the improvement of the existing YOLOv7-tiny network architecture, and the target class probability and the corresponding position coordinate value can be directly given out through one stage. As shown in fig. 3, the existing YOLOv7-tiny network structure mainly includes an Input layer, a backhaul layer and a Head layer; the Backbone layer is used as a Backbone network, and is used for carrying out feature extraction after preprocessing an Input picture in the Input layer, and is also called a feature extraction layer; compared with a Yolov7 network model, the Yolov7-tiny adopts an E_ELAN substructure (ELAN_T network structure) in a backbond layer, so that the parameter quantity is less and the detection speed is faster on the basis of ensuring the detection precision; meanwhile, the Head layer is mainly based on the output characteristics of the backhaul layer, and further outputs 3 characteristic diagrams with different sizes, and compared with a YOLOv7 network model, the Head layer is lighter, and mainly adopts an SPPCSPC_tiny layer, a plurality of Conv layers and an MP layer.
In the embodiment, considering the small target recognition application scene requirement of license plate recognition, the back box layer and the Head layer in the existing YOLOv7-tiny network model are correspondingly improved to obtain an improved YOLOv7-tiny network model comprising an Input layer, an improved back box layer and an improved Head layer which are sequentially connected; specifically, the convolution modules in the improved backhaul layer and the improved Head layer are CB-F modules, and as shown in fig. 4, the CB-F modules replace the LeakyReLU activation function with a fralu activation function on the basis of inheriting the original advantages of CBL, and by using an extremely simple implementation manner, the precision is increased while only adding a small amount of parameters, and the sensitivity of space is improved, and the convolution processing module comprises a convolution layer (Conv) responsible for feature extraction, a batch normalization layer (Batch Normalization, BN) for accelerating the training process and reducing problems such as gradient disappearance and gradient explosion, and a fralu activation function capable of solving the possible neuronal death problem of the ReLU activation function, namely the LeakyReLU activation function in the existing CBL (Convolutions with Batch Normalization and Leaky ReLU) module is replaced with the fralu activation function; wherein the FReLU activation function is an activation function that extends the ReLU activation function and the PReLU activation function to 2D activation, which is also an activation function that realizes nonlinear characteristics by adopting a Max function, but adopts a funnel condition T () (a convolution sliding window of 3X 3) in its internal judgment, T (X) is obtained by calculation using the parameters of the convolution window and surrounding pixels, and then, normalization processing is performed on T (X) using batch normalization BN; it should be noted that, the FReLU activation function realizes the pixel level modeling capability by adding negligible space condition overhead, and the funnel condition depends on the space context of surrounding pixels, so as to enlarge the receptive field of the network, capture complex visual layout through conventional convolution, and further effectively improve the robustness of the YOLOv7-tiny recognition task.
Considering the high efficiency requirement of license plate target detection application, the embodiment needs to further optimize the improved backhaul layer and the improved Head layer, and the existing ELAN_T module in the improved backhaul layer and the improved Head layer is replaced by a PM-ELAN module and an MS-ELAN module respectively to obtain a structure with smaller parameters, so that the running speed of a network model in practical application is improved; the existing ELAN_T module is divided into two layers, wherein one layer is overlapped with three CBL convolution modules to extract features, the other layer is independently used for extracting features, the two previous layers are fused by using a Concat layer, and the PM-ELAN module and the MS-ELAN module are of lighter structures which are obtained by partially improving the frames based on ELAN-T;
specifically, the PM-ELAN module may be understood as that, on the basis of ELAN-T, two CBL modules are replaced by a lightweight PConv convolution module, and a lightweight network structure mobilet V3 module is added behind the Concat layer, that is, the PM-ELAN module is obtained by integrating the PConv module and the mobilet V3 module based on the elan_t module, as shown in fig. 5, the PM-ELAN module includes a first feature extraction module, a feature integration module and the mobilet V3 module which are sequentially connected, so that parameters of the overall network are reduced, and the overall network is lighter and more suitable for being deployed on embedded devices; the first feature extraction module comprises a first feature extraction branch and a second feature extraction branch which are connected in parallel; the first feature extraction branch comprises a CB-F module and two PConv (pyramid convolution) modules which are connected in series in sequence; the second feature extraction branch comprises a CB-F module; the PConv module can be understood as partial convolution, which only needs to apply the conventional Conv to a part of the input channel to extract the spatial features, keeps the rest channels unchanged, judges for the FLOPs, and for typical r=1/4, only 1/16 of the conventional Conv for the FLOPs of the PConv, so that the access amount of the memory is greatly reduced, and the parameters of the whole network are reduced, namely, the influence of the increased side effect of the memory access is reduced while the FLOPs are reduced, so that the network has lower delay and higher throughput. In the embodiment, after PConv is used for replacing part of CB-F modules, the detection speed is greatly increased, the higher accuracy is maintained, and the parameter quantity is reduced;
meanwhile, the MS-ELAN module may be understood as adding a mobilet V3 lightweight network structure based on ELAN-T, that is, by integrating a mobilet V3 module based on the elan_t module and inserting an SE attention module after the mobilet V3 module, as shown in fig. 6, including a second feature extraction module, a feature fusion module, a mobilet V3 module, and an SE attention module connected in sequence, so as to reduce overall network parameters; the second feature extraction module comprises a third feature extraction branch and a fourth feature extraction branch which are connected in parallel; the third feature extraction branch comprises 3 CB-F modules connected in series; the fourth feature extraction branch comprises a CB-F module; the Mobilene V3 module fused in the PM-ELAN module and the MS-ELAN module comprises a spindle-shaped 3×3 convolution layer, a spindle-shaped 1×1 convolution layer, a Pool pooling layer, an FC full-connection layer and an avgpooling layer, wherein the size of a feature map is reduced from 7×7 to 1×1, and an activation function is an h-swish activation function; the Mobilene V3 module is convenient for light deployment, and can reduce running time and improve the robustness of the whole structure. In the embodiment, the MobilenetV3 module is fused into the backhaul backbone network of YOLOv7-tiny, so that the parameter quantity can be greatly reduced, and the network running speed can be improved;
it should be noted that, the sppcspc_tiny layer in the improved Head layer can be understood as being obtained by replacing all CBL modules in the YOLOv7-tiny network model of the existing YOLOv7-tiny network model shown in fig. 3 with CB-F modules of the frlu activation function given by the embodiment, and the specific structure is shown in fig. 8, which is a spatial pyramid pooling structure, so that the problems of image distortion and the like caused by cutting and scaling operation of an image area can be effectively avoided, the problem of extracting relevant repeated features of the graph by a convolutional neural network is solved, the speed of generating candidate frames is greatly improved, and the calculation cost is saved; in addition, the MP module in the network model can be understood as a maximum pooling layer, which can play roles in reducing the data quantity of the neural network, simplifying the data and accelerating the data processing; the UP module can be understood as an UP-sampling module, which is a technical means for UP-sampling an image to a higher resolution, and sampling a low resolution image to a high resolution image;
thus, the improvement of the network structure related to the improved YOLOv7-tiny network model provided by the application has been described in detail, and the improvement of each structure can be combined to obtain the optimal improved YOLOv7-tiny network model shown in fig. 9, namely, the activation function in the original convolution module is replaced by the FReLU activation function, so that the scope of the convolution receptive field is improved; then the original ELAN-T module is improved, new modules are respectively named as PM-ELAN and MS-ELAN, the PM-ELAN is a PConv convolution module which is replaced by a convolution kernel with the size of 3 multiplied by 3, and a Mobilene V3 module is fused in the PConv convolution module; similarly, the MS-ELAN is added with an SE attention mechanism at the tail part after fusing the mobile net V3 module, so that the feature extraction capability of the network is improved.
In addition, considering that the target detection performance depends greatly on the used loss function, the boundary box loss function is taken as an important component of the target detection loss function, the defined precision degree is critical to the improvement of the model performance, and a scene that an actual target is not completely matched with a prediction boundary box exists in the actual target detection application, in order to further improve the performance of the target detection model, the embodiment preferably replaces the CIOU loss function adopted in the existing YOLOv7-tiny network model with a dynamic non-monotonic focusing mechanism, and trains the improved YOLOv7-tiny network model by the loss function of describing the quality of the prediction boundary box in an outlier degree so as to obtain a stable model with optimal parameters for license plate target detection in the actual application; specifically, the frame loss function of the improved YOLOv7-tiny network model is expressed as:
Loss＝rR WIOU L IOU
in the method, in the process of the application,
L IOU ＝1-IOU
wherein, lossRepresenting a frame loss value; r is R WIOU Representing a penalty term; IOU and L IOU Respectively representing the intersection ratio and the loss value of the prediction boundary box and the real boundary box; beta represents the outlier of the prediction boundary frames, and the outlier of one prediction boundary frame is smaller, which means that the quality of the prediction boundary frame is higher, so that the prediction boundary frame can be better focused on the prediction boundary frame with higher quality to carry out boundary frame regression, and similarly, if the outlier of the prediction boundary frame is larger, the smaller gradient gain is required to be allocated, and the harmful gradient of a low-quality example to a model can be effectively prevented; delta and alpha are hyper-parameters, delta being such that r=1 when beta=delta, the prediction bounding box will obtain the highest gradient gain when the outlier of the prediction bounding box satisfies beta=c (C is a constant value); r represents the radius of the prediction bounding box; (x, y) represents the center coordinates of the prediction bounding box nearest to it; (x) gt ,y gt ) Representing the center coordinates of the real bounding box; (w) gt ,h gt ) Representing the width and height of the real bounding box; b and B gt Representing a prediction bounding box region and a real bounding box region, respectively;represents L IOU The momentum of m is a moving average value of m, and the value is dynamic, so that the quality division standard of the prediction boundary box is also dynamic, the frame loss function can make a gradient gain distribution strategy which is most in line with the current situation at each moment, and the problem of low convergence speed in the later stage of training can be effectively solved;
in order to prevent the low-quality prediction bounding box from falling behind in the early stage of training, the low-quality prediction bounding box is initialized during trainingSo that L IOU The prediction bounding box of =1 has the highest gradient gain; meanwhile, in order to maintain the same strategy in the early training phase, a small power m needs to be set to delay +.>Near true value +.>Is a time of (a) to be used. If the total data volume of the data for training is N and the set batch size is B, the batch number for training is n=n/B; if the rate of elevation of the mAP is significantly slowed in the first t rounds of training, it is recommended that the momentum be set in this case to:after t-rounds of training in this setting there is +.>In the training process of target detection, the loss function distributes small gradient gain to a low-quality prediction boundary box in the middle and later stages so as to reduce the influence of harmful gradients, and meanwhile, the loss function focuses on the prediction boundary box with common quality more, so that the positioning performance of the model is improved.
S12, obtaining license plate position coordinates and an area image to be identified according to the license plate positioning area; the license plate position coordinates can be understood as center point coordinates of a license plate positioning area, and the area image to be recognized can be understood as a local area image which can be used for carrying out subsequent optical character recognition in the license plate positioning area in the license plate image to be detected;
specifically, the step of obtaining license plate position coordinates and an area image to be identified according to the license plate positioning area includes:
obtaining a region center coordinate according to the frame coordinates of the license plate positioning region, and taking the region center coordinate as the license plate position coordinate;
according to the license plate positioning area, image segmentation is carried out on the license plate image to be detected, and the area image to be identified is obtained; the specific manner of the image segmentation process is implemented by adopting the prior art, and will not be described in detail herein.
S13, obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through a binocular distance measuring algorithm; the binocular range algorithm can be realized by adopting any binocular range algorithm in principle, but in consideration of the requirements of high efficiency and accuracy of range finding, the embodiment preferably adopts an SGBM algorithm to realize the binocular range function;
the system imaging model of the binocular camera, the conversion mode of the physical coordinate system of the image and the pixel coordinate system, and the binocular distance principle used in the present embodiment are described in detail as follows:
1) System imaging model
Since the following four coordinate systems are involved in the imaging process of the camera: a camera coordinate system, a world coordinate system, an image pixel coordinate system, and an image physical coordinate system; the camera system imaging model is shown in FIG. 10, wherein O c The coordinate system is a camera coordinate system, O p The coordinate system is an image pixel coordinate system which reflects the specification of the detector, and as can be seen from the figure, O l Physical coordinate system of image and O l The coordinates are at the midpoint of the image pixel coordinate system, and as the camera coordinate system is projected into the image coordinate system in perspective, the image coordinate system is transformed from 3D to 2D space, which in the middle loses one-dimensional depth information, namely the focal length f.
Specifically, an image pixel coordinate system O P Uv: is a two-dimensional coordinate system describing the pixel arrangement in the camera chip, which represents a discretization of the physical coordinate system of the image, the u, v axes under this coordinate system are generally parallel to both sides of the imaging target surface, and the coordinate units are pixels (pixels);
image physical coordinate system O I -xy: also a two-dimensional coordinate system, is used for describing the actual physical position of the projection of the space object point on the imaging target surface, O l The coordinates of which lie in the image pixel coordinate system O P And the x, y axes are parallel to the u, v axes of the image pixel coordinate system, respectively, in millimeters (mm);
camera coordinate system O C -X C Y C Z C : the coordinate system is a three-dimensional coordinate system, and is used for connecting the bridge between the world coordinate system and the image coordinate system and the optical center O of the camera lens C Recorded as coordinatesIs the origin and X C 、Y C Axes being parallel to the x-axis, the y-axis, Z in the image coordinate system, respectively C The axis is coincident with the optical axis of the camera, the unit of the coordinate system is millimeters (mm);
world coordinate system O W -X W Y W Z W : is a three-dimensional coordinate system that, if referenced, may be used to describe the spatial position between the camera and the object under test, and is in millimeters (mm).
2) Conversion between physical coordinate system of image and pixel coordinate system of image
Since the projection point of the optical center of the camera on the imaging plane is the origin of the image physical coordinate system, in other words, the origin of the image physical coordinate system is located at the midpoint of the image pixel coordinate system, and the unit of the image physical coordinate system is mm, the conversion relationship between the two coordinate systems is that the actual size represented by the rows and columns of the dx and dy pixel points is assumed to be how many mm per grid, then the image physical coordinate is converted into the image pixel coordinate as follows:
(1) Coordinate representation:
wherein, (u) 0 ,v 0 ) As shown in fig. 11, the origin of the image physical coordinate system is generally located at the midpoint of the image pixel coordinate system; (x, y) is the coordinate value of the point P to be measured on the physical coordinate system of the image, so (u, v) is the coordinate value of converting the physical coordinate of the image into the pixel coordinate of the image;
(2) Matrix representation:
3) Principle of binocular distance measurement
As shown in fig. 12, P is a point to be measured; p (P) l And P r Is PCoordinates of points in an image physical coordinate system of the left and right eye cameras; o (O) L And O R Left and right eyes of the binocular camera respectively; z is the distance of the object to be measured; b is a base line, namely the distance between the optical centers of the binocular cameras; f is the focal length of the camera; since parallax d, i.e., X, occurs when the same point P is observed simultaneously using a binocular camera l -X r (II), (III), (V), (; according to the principle of similar triangles: Δpp l P r ～ΔPO L O R The equation Z may be listed. Let P be l To P r The distance dis between them is: dis=b-d, then it is known from a similar triangle:
therefore, the distance z can be found as:
the binocular range finding system established based on the binocular range finding related basic theory can be used for calculating the actual distance between the license plate and the camera according to the license plate position coordinates obtained by carrying out target detection through the improved YOLOv7-tiny network model; specifically, the step of obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through the binocular distance measuring algorithm comprises the following steps:
acquiring internal and external parameters of a binocular camera by adopting a Zhang Zhengyou camera calibration method; the double-sided camera can be any type of binocular camera in principle, but in order to ensure the accuracy of ranging, the embodiment preferably adopts an HBV-1780-2S2.0 binocular camera; meanwhile, the Zhang Zhengyou camera calibration method selected in the embodiment has the advantages of strong imaging constraint, simple calibration process, high algorithm robustness and the like, is already packaged in OpenCV and MATLAB, and can be used for carrying out binocular camera calibration in practical application, so that the internal and external parameters of the binocular camera are obtained, and the specific calibration process is not described in detail here;
performing binocular correction on the binocular camera in response to completion of acquisition of internal and external parameters of the binocular camera; the binocular correction is understood to be that the images of the two cameras are corrected to be coplanar and aligned in rows, and the left and right optical centers of the binocular cameras are parallel and correspond to each other through the binocular correction, so that preparation is made for subsequent stereo matching, only one-dimensional search along polar lines is needed when the subsequent stereo matching is carried out, and the distance from the camera to an object to be detected is also conveniently and accurately calculated.
In response to the completion of binocular correction, performing stereo matching through an SGBM algorithm to obtain a parallax depth map, and obtaining the distance between the license plate and a camera according to the parallax depth map and the license plate position coordinates;
in binocular vision, stereo matching is also called disparity estimation, which means that a close correspondence between pixels is established in a polar corrected stereo image pair, thereby obtaining disparity. The stereo matching generally comprises the steps of cost calculation, cost aggregation, parallax calculation, parallax optimization and the like; the cost calculation is usually Census calculation, and when the hamming distance obtained by using Census cost calculation is the required cost, the specific process of Census cost calculation is as follows:
the following two matrices are assumed:
when calculating the hamming distance, the formula needed to be used is:wherein (1)>Representing exclusive OR;
wherein center is the value of the center of the matrix; t is each value except the center point value in the matrix; for example, in w1, center is 45, t is 33, 54, 36, 23, 125, 123, 44, 79, respectively;
therefore, the following steps are obtained:
then the first time period of the first time period,
it is possible to obtain a solution,
that is, the hamming distance is 3;
cost aggregation can be understood as aggregation of the cost obtained through the steps, and in an SGBM algorithm, a method mainly used for cost aggregation is to firstly conduct one-dimensional dynamic planning and then conduct multi-directional superposition on the one-dimensional dynamic planning so as to obtain a better optimized depth map;
the parallax calculation can be understood as calculation of a parallax depth map, after the binocular correction and the stereo matching are completed, a corresponding SGBM object can be constructed based on an SGBM algorithm with parameters set in OpenCV-python, the corresponding parallax depth map is obtained through calculation, and a corresponding depth value, namely the distance between a license plate and a camera, can be obtained based on a calculation formula in the binocular vision distance principle according to the acquired license plate position coordinates.
S14, judging whether the distance between the license plate and the camera meets the preset distance requirement, and when the distance between the license plate and the camera meets the preset distance requirement, adopting a preset character recognition model to perform optical character recognition on the image of the area to be recognized to obtain a corresponding license plate number; the preset distance requirement can be selected according to actual application requirements, and is not particularly limited herein; for example, when the preset distance is 1 meter, when the measured distance between the license plate and the camera is smaller than or equal to 1 meter, the subsequent step of license plate positioning and recognition is stopped, the subsequent step of obtaining the corresponding license plate number based on the image recognition of the region to be recognized is not performed, the license plate which is currently detected is directly judged not to pass through the gate, and the gate is not released; when the measured distance between the license plate and the camera exceeds one meter, judging that the current license plate meets the preset distance requirement, and enabling the gate to further enter a specific license plate content recognition step so as to obtain a precise license plate number; the preset character recognition model in the present embodiment may be understood as an optical character recognition model capable of recognizing a license plate number, and may be a recognition model based on template matching, a recognition model based on statistical feature analysis, a recognition model based on deep learning, or the like, which is not particularly limited herein.
According to the embodiment of the application, the improved YOLOv7-tiny network model is adopted to carry out target detection on the acquired license plate image to be detected to obtain the license plate positioning area, the license plate position coordinates and the image of the area to be identified are obtained according to the license plate positioning area, then the license plate and the camera distance are obtained according to the license plate position coordinates through the binocular range algorithm, whether the license plate and the camera distance meet the preset distance requirement is judged, and when the license plate and the camera distance meet the preset distance requirement, the preset character recognition model is adopted to carry out optical character recognition on the image of the area to be identified to obtain the corresponding license plate number, so that the application defect of the existing license plate positioning recognition method is effectively overcome, the network parameters are reduced through adopting the lighter YOLOv7-tiny network model, the license plate area detection speed and the detection precision are improved, and the reliability and the application value of the license plate recognition are effectively improved through the combination with the binocular range.
Although the steps in the flowcharts described above are shown in order as indicated by arrows, these steps are not necessarily executed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders.
In one embodiment, as shown in fig. 13, there is provided a license plate location recognition system, the system comprising:
the area positioning module 1 is used for acquiring a license plate image to be detected, and performing target detection on the license plate image to be detected by adopting an improved YOLOv7-tiny network model to obtain a license plate positioning area;
the area processing module 2 is used for obtaining license plate position coordinates and an area image to be identified according to the license plate positioning area;
the license plate distance measuring module 3 is used for obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through a binocular distance measuring algorithm;
and the license plate recognition module 4 is used for judging whether the distance between the license plate and the camera meets the preset distance requirement, and carrying out optical character recognition on the image of the area to be recognized by adopting a preset character recognition model when the distance between the license plate and the camera meets the preset distance requirement, so as to obtain a corresponding license plate number.
The specific limitation of the license plate positioning recognition system can be referred to the limitation of the license plate positioning recognition method, and the corresponding technical effects can be obtained equally, and are not repeated here. All or part of each module in the license plate positioning and identifying system can be realized by software, hardware and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
Fig. 14 shows an internal structural diagram of a computer device, which may be a terminal or a server in particular, in one embodiment. As shown in fig. 14, the computer device includes a processor, a memory, a network interface, a display, a camera, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by the processor, implements a license plate location recognition method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those of ordinary skill in the art that the architecture shown in fig. 14 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer devices to which the present inventive arrangements may be applied, and that a particular computing device may include more or fewer components than shown, or may combine some of the components, or have the same arrangement of components.
In one embodiment, a computer device is provided comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the above method when the computer program is executed.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, implements the steps of the above method.
In summary, the license plate positioning and identifying method and system provided by the embodiment of the application realize the technical scheme that the improved YOLOv7-tiny network model is adopted to carry out target detection on the acquired license plate image to be detected to obtain a license plate positioning region, after license plate position coordinates and the region image to be identified are obtained according to the license plate positioning region, the distance between the license plate and a camera is obtained according to the license plate position coordinates through a binocular range algorithm, and whether the distance between the license plate and the camera meets the preset distance requirement is judged, and when the distance between the license plate and the camera meets the preset distance requirement, the preset character identification model is adopted to carry out optical character identification on the region image to be identified to obtain the corresponding license plate number.
In this specification, each embodiment is described in a progressive manner, and all the embodiments are directly the same or similar parts referring to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments. It should be noted that, any combination of the technical features of the foregoing embodiments may be used, and for brevity, all of the possible combinations of the technical features of the foregoing embodiments are not described, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples represent only a few preferred embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the application. It should be noted that modifications and substitutions can be made by those skilled in the art without departing from the technical principles of the present application, and such modifications and substitutions should also be considered to be within the scope of the present application. Therefore, the protection scope of the patent of the application is subject to the protection scope of the claims.
Claims (10)
1. The license plate positioning and identifying method is characterized by comprising the following steps of:
acquiring a license plate image to be detected, and performing target detection on the license plate image to be detected by adopting an improved YOLOv7-tiny network model to obtain a license plate positioning area;
obtaining license plate position coordinates and an area image to be identified according to the license plate positioning area;
obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through a binocular distance measuring algorithm;
and judging whether the distance between the license plate and the camera meets the preset distance requirement, and when the distance between the license plate and the camera meets the preset distance requirement, adopting a preset character recognition model to perform optical character recognition on the image of the area to be recognized to obtain a corresponding license plate number.
2. The license plate positioning and identifying method according to claim 1, wherein the improved YOLOv7-tiny network model comprises an Input layer, an improved backhaul layer and an improved Head layer which are connected in sequence; the convolution modules in the improved backhaul layer and the improved Head layer are CB-F modules; the CB-F module comprises a convolution layer, a batch normalization layer and a FReLU activation function which are sequentially connected.
3. The license plate location recognition method of claim 2, wherein elan_t modules in the modified backhaul layer and the modified Head layer are replaced with PM-ELAN modules and MS-ELAN modules, respectively; the PM-ELAN module is obtained by integrating a PConv module and a Mobilene V3 module based on the ELAN_T module; the MS-ELAN module is obtained by integrating a Mobilene V3 module based on the ELAN_T module, and inserting an SE attention module after the Mobilene V3 module.
4. The license plate positioning and identifying method according to claim 3, wherein the PM-ELAN module comprises a first feature extraction module, a feature fusion module and a Mobilene V3 module which are connected in sequence; the first feature extraction module comprises a first feature extraction branch and a second feature extraction branch which are connected in parallel; the first feature extraction branch comprises a CB-F module and two PConv modules which are connected in sequence; the second feature extraction branch comprises a CB-F module;
the MS-ELAN module comprises a second feature extraction module, a feature fusion module, a Mobilene V3 module and an SE attention module which are connected in sequence; the second feature extraction module comprises a third feature extraction branch and a fourth feature extraction branch which are connected in parallel; the third feature extraction branch comprises 3 CB-F modules connected in series; the fourth feature extraction branch includes a CB-F module.
5. The license plate location recognition method of claim 1, wherein the modified YOLOv7-tiny network model's frame loss function is expressed as:
Loss＝rR WIOU L IOU
in the method, in the process of the application,
L IOU ＝1-IOU
wherein Loss represents a frame Loss value; r is R WIOU Representing a penalty term; IOU and L IOU Respectively representing the intersection ratio and the loss value of the prediction boundary box and the real boundary box; beta represents an outlier of the prediction bounding box; delta and alpha are hyper-parameters; r represents the radius of the prediction bounding box; (x, y) represents the center coordinates of the prediction bounding box nearest to it; (x) gt ,y gt ) Representing the center coordinates of the minimum prediction bounding box; (w) gt ,h gt ) Representing the width and height of the minimum prediction bounding box.
6. The license plate positioning and identifying method according to claim 1, wherein the step of obtaining license plate position coordinates and an image of an area to be identified according to the license plate positioning area comprises:
obtaining a region center coordinate according to the frame coordinates of the license plate positioning region, and taking the region center coordinate as the license plate position coordinate;
and according to the license plate positioning area, image segmentation is carried out on the license plate image to be detected, and the area image to be identified is obtained.
7. The license plate positioning and identifying method according to claim 1, wherein the step of obtaining the distance between the license plate and the camera according to the license plate position coordinates through a binocular ranging algorithm comprises:
acquiring internal and external parameters of a binocular camera by adopting a Zhang Zhengyou camera calibration method;
performing binocular correction on the binocular camera in response to completion of acquisition of internal and external parameters of the binocular camera;
and responding to the completion of binocular correction, performing stereo matching through an SGBM algorithm to obtain a parallax depth map, and obtaining the distance between the license plate and a camera according to the parallax depth map and the license plate position coordinates.
8. A license plate location identification system, the system comprising:
the area positioning module is used for acquiring a license plate image to be detected, and performing target detection on the license plate image to be detected by adopting an improved YOLOv7-tiny network model to obtain a license plate positioning area;
the region processing module is used for obtaining license plate position coordinates and a region image to be identified according to the license plate positioning region;
the license plate distance measuring module is used for obtaining the distance between the license plate and the camera according to the position coordinates of the license plate through a binocular distance measuring algorithm;
and the license plate recognition module is used for judging whether the distance between the license plate and the camera meets the preset distance requirement, and carrying out optical character recognition on the image of the area to be recognized by adopting a preset character recognition model when the distance between the license plate and the camera meets the preset distance requirement, so as to obtain a corresponding license plate number.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310713794.5A CN116883981A (en) | 2023-06-15 | 2023-06-15 | License plate positioning and identifying method, system, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310713794.5A CN116883981A (en) | 2023-06-15 | 2023-06-15 | License plate positioning and identifying method, system, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116883981A true CN116883981A (en) | 2023-10-13 |
Family
ID=88268863
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310713794.5A Pending CN116883981A (en) | 2023-06-15 | 2023-06-15 | License plate positioning and identifying method, system, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116883981A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117292370A (en) * | 2023-11-23 | 2023-12-26 | 合肥天帷信息安全技术有限公司 | Icon character recognition method and device |
-
2023
- 2023-06-15 CN CN202310713794.5A patent/CN116883981A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117292370A (en) * | 2023-11-23 | 2023-12-26 | 合肥天帷信息安全技术有限公司 | Icon character recognition method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021004312A1 (en) | Intelligent vehicle trajectory measurement method based on binocular stereo vision system | |
| CN109997170B (en) | Apparatus and method for obtaining distance information from view | |
| US9646212B2 (en) | Methods, devices and systems for detecting objects in a video | |
| WO2018049998A1 (en) | Traffic sign information acquisition method and device | |
| CN112444242B (en) | Pose optimization method and device | |
| CN111476827B (en) | Target tracking method, system, electronic device and storage medium | |
| Won et al. | Omnimvs: End-to-end learning for omnidirectional stereo matching | |
| CN106529538A (en) | Method and device for positioning aircraft | |
| CN111091023B (en) | Vehicle detection method and device and electronic equipment | |
| CN108537214B (en) | Automatic construction method of indoor semantic map | |
| WO2020207172A1 (en) | Method and system for optical monitoring of unmanned aerial vehicles based on three-dimensional light field technology | |
| US9323989B2 (en) | Tracking device | |
| CN111598065A (en) | Depth image acquisition method, living body identification method, apparatus, circuit, and medium | |
| Shi et al. | An improved lightweight deep neural network with knowledge distillation for local feature extraction and visual localization using images and LiDAR point clouds | |
| WO2023093086A1 (en) | Target tracking method and apparatus, training method and apparatus for model related thereto, and device, medium and computer program product | |
| CN111325828B (en) | Three-dimensional face acquisition method and device based on three-dimensional camera | |
| CN116883981A (en) | License plate positioning and identifying method, system, computer equipment and storage medium | |
| CN109829401A (en) | Traffic sign recognition method and device based on double capture apparatus | |
| CN116012609A (en) | Multi-target tracking method, device, electronic equipment and medium for looking around fish eyes | |
| CN116051736A (en) | Three-dimensional reconstruction method, device, edge equipment and storage medium | |
| CN112069997B (en) | Unmanned aerial vehicle autonomous landing target extraction method and device based on DenseHR-Net | |
| CN112215036B (en) | Cross-mirror tracking method, device, equipment and storage medium | |
| CN114140659A (en) | Social distance monitoring method based on human body detection under view angle of unmanned aerial vehicle | |
| CN114608522A (en) | Vision-based obstacle identification and distance measurement method | |
| CN103236053A (en) | MOF (motion of focus) method for detecting moving objects below mobile platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |