CN115512311A - Vehicle information identification method, intelligent wearable device and readable storage medium - Google Patents

Vehicle information identification method, intelligent wearable device and readable storage medium Download PDF

Info

Publication number
CN115512311A
CN115512311A CN202211098991.2A CN202211098991A CN115512311A CN 115512311 A CN115512311 A CN 115512311A CN 202211098991 A CN202211098991 A CN 202211098991A CN 115512311 A CN115512311 A CN 115512311A
Authority
CN
China
Prior art keywords
vehicle
attention
yolo
model
network model
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
Application number
CN202211098991.2A
Other languages
Chinese (zh)
Inventor
高金龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Goertek Techology Co Ltd
Original Assignee
Goertek Techology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Goertek Techology Co Ltd filed Critical Goertek Techology Co Ltd
Priority to CN202211098991.2A priority Critical patent/CN115512311A/en
Publication of CN115512311A publication Critical patent/CN115512311A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • 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
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
    • 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
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

The application discloses vehicle information identification method, intelligent wearable equipment and readable storage medium, the vehicle information identification method comprises the following steps: the method comprises the steps of obtaining a vehicle image of a vehicle to be identified, inputting the vehicle image into a trained YOLO-V5 improved network model, obtaining vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model, and outputting the vehicle information corresponding to the vehicle image, wherein the YOLO-V5 improved network model is the YOLO-V5 network model embedded with an attention mechanism improvement module. This application can reduce intelligent wearing equipment's use limitation for intelligent wearing equipment can accurately discern the vehicle information of each type vehicle.

Description

Vehicle information identification method, intelligent wearable device and readable storage medium
Technical Field
The application relates to the technical field of image detection, in particular to a vehicle information identification method, intelligent wearable equipment and a readable storage medium.
Background
Due to the rapid development of the automobile industry, the brand and model of the vehicle, the corresponding market price and the factory year of various vehicles and other vehicle information are more and more, so that the identification difficulty of the vehicle information is more and more increased for common users who are not well known in the automobile industry. For example, when people see a vehicle that is relatively interested in the people on the road and cannot know the vehicle information such as the brand, model and market price of the vehicle, the people cannot provide relevant references for the vehicle purchasing demand.
VR (Virtual Reality) devices or AR (Augmented Reality) devices are smart wearable devices that are currently rapidly developing and popularizing. However, the current function of intelligent wearable device mainly focuses on multimedia functions such as music, picture and video playing, and because of the limitation of the use of intelligent wearable device, the vehicle information of various different vehicles can not be accurately identified through the intelligent wearable device, and when people pass through the vehicle which can not be identified by themselves, the vehicle information of the vehicle can not be known in time (for example, the vehicle stops at a parking space on the road, the vehicle owner has left and the like, and the scene of the relevant information of the vehicle can not be known in time to the vehicle owner), thereby the user experience is seriously reduced.
Disclosure of Invention
The application mainly aims to provide a vehicle information identification method, intelligent wearable equipment and a readable storage medium, and aims to reduce the use limitation of the intelligent wearable equipment and enable the intelligent wearable equipment to accurately identify vehicle information of various types of vehicles.
In order to achieve the above object, the present application provides a vehicle information identification method, where the vehicle information identification method is applied to an intelligent wearable device, and the vehicle information identification method includes:
acquiring a vehicle image of a vehicle to be identified;
inputting the vehicle image into a trained YOLO-V5 improved network model, wherein the YOLO-V5 improved network model is a YOLO-V5 network model embedded with an attention mechanism improving module, and the attention mechanism improving module comprises a channel attention module and a space attention module which are arranged in parallel;
and acquiring vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model, and outputting the vehicle information corresponding to the vehicle image.
Optionally, the vehicle information includes a vehicle model and vehicle parameter information corresponding to the vehicle model, and the step of obtaining the vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model and outputting the vehicle information corresponding to the vehicle image includes:
obtaining a vehicle model corresponding to the vehicle image identified by the YOLO-V5 improved network model;
according to the vehicle model, vehicle parameter information corresponding to the vehicle model is searched and obtained from a preset database, wherein the vehicle parameter information comprises at least one of vehicle market quotation, vehicle internal configuration and vehicle delivery year;
and outputting the vehicle model and vehicle parameter information corresponding to the vehicle model.
Optionally, the step of obtaining the vehicle model corresponding to the vehicle image identified by the YOLO-V5 improved network model comprises:
carrying out feature extraction on the vehicle image through a trained YOLO-V5 improved network model, and extracting to obtain vehicle global features corresponding to the vehicle image;
inputting, by the attention mechanism improving module, vehicle global features corresponding to the vehicle image in parallel to the channel attention module and the spatial attention module;
respectively acquiring channel characteristics corresponding to the vehicle images output by the channel attention module and spatial characteristics corresponding to the vehicle images output by the spatial attention module;
performing feature fusion on the channel features and the spatial features to obtain key local features fusing channel attention and spatial attention;
and identifying the vehicle model corresponding to the vehicle image through the key local features.
Optionally, the step of identifying, through the key local features, a vehicle model corresponding to the vehicle image includes:
performing global average pooling on the key local features to obtain a first feature pooling result, and performing global maximum pooling on the key local features to obtain a second feature pooling result;
splicing the first characteristic pooling result and the second characteristic pooling result along a channel direction to obtain a key pooling characteristic;
after convolution processing is carried out on the key pooling features, a key feature attention weight value is obtained through a preset activation function;
multiplying the attention weight of the key feature by the global feature of the vehicle to obtain an attention feature map fusing channel attention and space attention,
and identifying the vehicle model mapped by the attention feature map through a feature map mapping rule obtained by pre-training, and taking the vehicle model mapped by the attention feature map as the vehicle model corresponding to the vehicle image.
Optionally, the step of inputting the vehicle image to the trained YOLO-V5 improved network model is preceded by:
acquiring a plurality of vehicle pictures corresponding to each vehicle model and vehicle type labels of the vehicle models corresponding to the vehicle pictures, and respectively performing label association on the vehicle pictures and the corresponding vehicle type labels to obtain a vehicle type sample picture set;
dividing the vehicle type sample picture set into a vehicle type recognition training sample set and a vehicle type recognition verification sample set according to a preset proportion;
training a preset YOLO-V5 improved network model through the vehicle type recognition training sample set, and verifying the recognition accuracy of the preset YOLO-V5 improved network model through the vehicle type recognition verification sample set;
and if the verified recognition accuracy reaches a preset accuracy threshold, stopping training the improved YOLO-V5 network model to obtain the trained improved YOLO-V5 network model.
Optionally, the step of performing label association on each vehicle picture and the vehicle model label corresponding to the vehicle picture to obtain a vehicle model sample picture set includes:
performing data enhancement processing on each picture in the vehicle type recognition training sample set to obtain a data enhancement sample picture set;
randomly splicing a preset number of pictures in the data enhancement sample picture set into one picture to obtain a Mosaic enhancement sample picture set, wherein the preset number is more than one;
the step of dividing the vehicle type sample picture set into a vehicle type recognition training sample set and a vehicle type recognition verification sample set according to a preset proportion comprises the following steps:
and dividing the Mosaic enhancement sample picture set into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion.
Optionally, the step of randomly splicing a preset number of pictures in the data enhancement sample picture set into one picture to obtain a Mosaic enhancement sample picture set includes:
adding Gaussian noise to each picture in the Mosaic enhancement sample picture set to obtain a Mosaic enhancement sample picture set with the Gaussian noise added;
the step of dividing the Mosaic enhancement sample picture set into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion comprises the following steps:
and dividing the Mosaic enhancement sample picture set added with the Gaussian noise into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion.
Optionally, the step of outputting the vehicle information includes:
displaying the vehicle information through a display screen of the intelligent wearable device; and/or the presence of a gas in the atmosphere,
and playing the vehicle information through a loudspeaker of the intelligent wearable device.
This application still provides an intelligence wearing equipment, intelligence wearing equipment is entity equipment, intelligence wearing equipment includes: a memory, a processor and a program of the vehicle information identification method stored on the memory and executable on the processor, which when executed by the processor, may implement the steps of the vehicle information identification method as described above.
The present application also provides a readable storage medium which is a computer readable storage medium having stored thereon a program for implementing a vehicle information identification method, the program being executed by a processor to implement the steps of the vehicle information identification method as described above.
The present application also provides a computer program product comprising a computer program which, when executed by a processor, carries out the steps of the vehicle information identification method as described above.
The method comprises the steps of acquiring a vehicle image of a vehicle to be identified, inputting the vehicle image into a trained YOLO-V5 improved network model, wherein the YOLO-V5 improved network model is a YOLO-V5 network model embedded with an attention system improving module, the attention system improving module comprises a channel attention module and a space attention module which are arranged in parallel, so that failure or wrong identification of the vehicle information caused by mis-filtering of key features for identifying the vehicle information before the channel attention module and the space attention module are input into the space attention module due to serial connection (CBAM before improvement) of the channel attention module and the space attention module, and the application can accurately enhance the phenomenon that the channel attention module and the space attention module are arranged in parallel (CBAM after improvement) and are embedded into the YOLO-V5 network model, so that the phenomenon of mutual filtering of the channel attention and the space attention of the improved YOLO-V5 network model can not occur, the effect of the key features for identifying the vehicle information can be more accurately extracted, the effect of inhibiting the important features in a picture (image) of the image from being shot by the background information is more accurately, the extracted, the vehicle image, the original vehicle image can be more accurately extracted, and the vehicle image can be more accurately, and the vehicle identification effect of the key features of the vehicle can be more accurately eliminated, and the vehicle identification efficiency of the vehicle identification of the vehicle can be more accurately and the vehicle identification of the vehicle can be more accurately obtained according to the original vehicle identification information can be more accurately, according to the method and the device, the trained YOLO-V5 improved network model is obtained to identify the vehicle information corresponding to the vehicle image, and the vehicle information corresponding to the vehicle image is output, so that the YOLO-V5 improved network model is deployed on a processing platform corresponding to the intelligent wearable device, the vehicle information such as vehicle type parameters and vehicle market prices is identified for the real-time vehicle photos acquired by the intelligent wearable device, the intelligent interaction capacity of the intelligent wearable device is enriched, the use limitation of the intelligent wearable device is reduced, and the light weight, real-time and feature extraction accuracy of the YOLO-V5 improved network model further enables the intelligent wearable device to efficiently and accurately identify the vehicle information of various types of vehicles.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and, together with the description, serve to explain the principles of the application.
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a first embodiment of a vehicle information identification method of the present application;
FIG. 2 is a schematic flow chart illustrating a second embodiment of a vehicle information identification method according to the present application;
FIG. 3 is a vehicle information identification process of a vehicle image by a YOLO-V5 network model;
FIG. 4 is a vehicle information identification process of a vehicle image by a YOLO-V5 improved network model in an embodiment of the present application;
FIG. 5 is a schematic flow chart illustrating the process of extracting and fusing image features in a vehicle image by an attention mechanism improving module according to an embodiment of the present application;
fig. 6 is a schematic device structure diagram of a hardware operating environment related to the intelligent wearable device in the embodiment of the present application.
The objectives, features, and advantages of the present application will be further described with reference to the accompanying drawings.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention more comprehensible, embodiments accompanying figures are described in detail below. It should be apparent that the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In this embodiment, the smart wearable device of the present application may be, for example, a Mixed Reality (Mixed Reality) -MR device (e.g., MR glasses or MR helmet), an Augmented Reality (Augmented Reality) -AR device (e.g., AR glasses or AR helmet), a virtual Reality- (virtual Reality) -VR device (e.g., VR glasses or VR helmet), an Extended Reality (Extended Reality) -XR device, or some combination thereof.
VR (Virtual Reality) devices or AR (Augmented Reality) devices are smart wearable devices that are currently rapidly developing and popularizing. However, the current function of intelligent wearable device mainly focuses on multimedia functions such as music, picture and video playing, and because of the limitation of the use of intelligent wearable device, the vehicle information of various different vehicles can not be accurately identified through the intelligent wearable device, and when people pass through the vehicle which can not be identified by themselves, the vehicle information of the vehicle can not be known in time (for example, the vehicle stops at a parking space on the road, the vehicle owner has left and the like, and the scene of the relevant information of the vehicle can not be known in time to the vehicle owner), thereby the user experience is seriously reduced.
Example one
Based on this, please refer to fig. 1, this embodiment provides a vehicle information identification method, where the vehicle information identification method is applied to an intelligent wearable device, and the steps of the vehicle information identification method include:
step S100, obtaining a vehicle image of a vehicle to be identified;
in an embodiment, the vehicle image of the vehicle to be identified can be acquired through a camera on the intelligent wearable device. In another embodiment, the vehicle image of the vehicle to be identified can be acquired by receiving the vehicle image transmitted by other terminal devices (such as a smart watch, a mobile phone or a smart camera) in communication connection with the smart wearable device. In still another embodiment, the vehicle image of the vehicle to be identified sent by the cloud terminal can be received for obtaining.
In this embodiment, a function of identifying vehicle information of a vehicle image of a vehicle to be identified can be triggered by touching a vehicle identification key on the intelligent wearable device. For example, after the vehicle identification key of the smart wearable device is triggered, the vehicle image of the vehicle to be identified is collected, and the subsequent steps S200 and S300 are performed. The vehicle identification key may be a physical key or a touch screen key. Of course, the triggering mechanism of the vehicle information identification function may also be triggered without a key, and the vehicle information identification function is automatically started after the intelligent wearable device is started.
Step S200, inputting the vehicle image into a trained YOLO-V5 improved network model, wherein the YOLO-V5 improved network model is a YOLO-V5 network model embedded with an attention mechanism improvement module, and the attention mechanism improvement module comprises a channel attention module and a space attention module which are arranged in parallel;
in the present embodiment, it can be understood by those skilled in the art that the YOLO-V5 network model is a neural network model for performing an image detection algorithm in the prior art. The YOLO-V5 network model can be a YOLO-V5s network model, and it can be known that the YOLO-V5s model is the model with the shallowest depth and the least parameters in a YOLO-V5s model series, and is more suitable for intelligent wearable equipment to carry the embedded low-computation model. An improvement of the embodiment lies in that an improved CBAM (conditional Block Attention Module) Module (i.e. Attention improvement Module) is added in the YOLO-V5 network model. The CBAM module is essentially a channel and space mixed type attention mechanism, compared with the single attention mechanism, the CBAM module integrates two mapping processes of a channel and a space, can retain more information, and simultaneously gives heavier weight to more remarkable large features, thereby ensuring that a network can focus correct attention on a main target, and the channel attention mechanism is used for giving larger weight to the content of an important channel and is equivalent to a filter. The spatial attention module mainly highlights the position information of the features and gives greater weight to important positions. That is, the channel attention is to find out which channel has important information, and the spatial attention is to find out at which block the information is focused more based on the direction of the channel, i.e. to find out which position the key information is most in the vehicle image, the spatial attention is to supplement the channel attention,
based on the characteristics of the CBAM, in this embodiment, in order to enhance the objectivity of the recognition result and eliminate the influence of the CBAM model itself, the CBAM module is improved, in the original CBAM module, the original feature map (i.e., the original vehicle image of the vehicle to be recognized) is first input into the channel attention module, then the feature map information (channels considered to be important are screened out and channels considered to be unimportant are filtered) output by the channel attention module is input into the spatial attention module, and the functions of the channel attention module can be understood as: the original feature map is an RGB tricolor map, and areas needing attention are given higher weight on RGB tricolor channels by recognizing gradient or pictures of certain colors and consistency of overall features. The effect of the spatial attention can be understood as paying attention to the spatial features of the picture, and pooling the information of the three dimensions of the CHW so as to obtain the most attention-worthy spatial feature information. However, in view of the entire image, the focus of attention derived from the channel attention is not necessarily a real focus, and if the original CBAM module inputs the feature map processed by the channel attention to the spatial attention module, which is equivalent to inputting the original image into the spatial attention module after "modifying", the obtained result is often deviated from the real target detection area, and the deviation of the area may cause the identification failure or the identification error of the vehicle information.
Therefore, the improvement of the CBAM in this embodiment is mainly realized by changing a single flow direction of an original feature map (a channel attention module is connected in series with a spatial attention module, that is, a channel attention module flows from the channel attention module to the spatial attention module) into a dual parallel flow direction, inputting the original feature map into the spatial attention module and the channel attention module which are arranged in parallel, processing the feature map in parallel by the two modules, paying more attention to the weight of each feature plane for channel attention, paying more attention to the weight of each local region for spatial attention, separating the two modules to give better consideration to the dual features of the feature plane and the local space, and obtaining a feature map which is an attention feature map fusing the channel attention and the spatial attention, so that when the intelligent wearable device recognizes a vehicle image through a trained YOLO-V5 improved network model, the method avoids that key features for identifying vehicle information are mistakenly filtered by a channel attention module and a space attention module (CBAM before improvement) because the channel attention module and the space attention module are connected in series, and further causes failure or identification errors of the vehicle information, but the embodiment improves the CBAM, and embeds the channel attention module and the space attention module which are arranged in parallel (the improved CBAM) into a YOLO-V5 network model, so that the weight of important features in a vehicle image can be accurately enhanced, the weight of non-important features is inhibited, more comprehensive and rich information after the integration of a shallow layer and a deep layer is extracted, more discriminative features of the vehicle are obtained, and further key feature information of the vehicle image can be efficiently and accurately identified and extracted, and identifying the vehicle information of the vehicle to be identified according to the key characteristic information, thereby improving the accuracy of vehicle information identification.
And step S300, acquiring the vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model, and outputting the vehicle information corresponding to the vehicle image.
In this embodiment, the vehicle information may include one or more of a vehicle brand, a vehicle model, a market quote, a vehicle interior configuration, and a vehicle shipping year.
As one example, the step of outputting the vehicle information includes:
displaying the vehicle information through a display screen of the intelligent wearable device; and/or playing the vehicle information through a loudspeaker of the intelligent wearable device.
The embodiment obtains the vehicle image of the vehicle to be identified, and inputs the vehicle image into a trained YOLO-V5 improved network model, wherein the YOLO-V5 improved network model is a YOLO-V5 network model embedded with an attention system improving module, the attention system improving module comprises a channel attention module and a spatial attention module which are arranged in parallel, so as to avoid that the identification failure or the identification error of the vehicle information is caused by mis-filtering the key features of the identified vehicle information before the channel attention module and the spatial attention module are input into the spatial attention module due to the serial connection of the channel attention module and the spatial attention module (CBAM before improvement), while the embodiment can accurately enhance the phenomenon that the channel attention module and the spatial attention module are arranged in parallel (CBAM after improvement) to be embedded into the YOLO-V5 network model by improving the CBAM, so that the channel attention and the spatial attention of the improved YOLO-V5 network model do not conflict with each other, thereby accurately enhancing the effect of extracting the important features (characteristic information in a picture) of the vehicle image, and inputting the vehicle image into the YOLO-V5 improved network model, and further enhancing the overall identification effect of the extracted vehicle according to the extracted features of the key features of the YOLO-V5 network model, thereby more accurately enhancing the image and the overall and the extracted vehicle identification efficiency of the vehicle, in the embodiment, vehicle information corresponding to the vehicle images and recognized by the trained YOLO-V5 improved network model are obtained, and the vehicle information corresponding to the vehicle images is output, so that the YOLO-V5 improved network model is deployed on a processing platform corresponding to the intelligent wearable device, vehicle information such as vehicle type parameters and vehicle market prices is recognized for real-time vehicle photos collected by the intelligent wearable device, the intelligent interaction capacity of the intelligent wearable device is enriched, the use limitation of the intelligent wearable device is reduced, and the lightweight, real-time performance and feature extraction accuracy of the YOLO-V5 improved network model further enable the intelligent wearable device to efficiently and accurately recognize the vehicle information of various types of vehicles.
In an implementable manner, the vehicle information includes a vehicle model and vehicle parameter information corresponding to the vehicle model, and the step of obtaining vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model and outputting the vehicle information corresponding to the vehicle image includes:
step A10, obtaining a vehicle model corresponding to the vehicle image identified by the YOLO-V5 improved network model;
in the present embodiment, it is understood that the vehicle model refers to more accurate vehicle information than the vehicle brand, for example, the vehicle brand is a red-flag vehicle, and the vehicle model may include a red flag L5, a red flag H7, a red flag H5, a red flag HS7, and the like. Also for example, the vehicle brand is a biddi vehicle, and the vehicle model thereof may include biddi F3, biddi F6, biddi FO, biddi F3R, biddi G3, and biddi S8, and the like.
Step A20, according to the vehicle model, searching a preset database to obtain vehicle parameter information corresponding to the vehicle model, wherein the vehicle parameter information comprises at least one of vehicle market quotation, vehicle internal configuration and vehicle delivery year;
those skilled in the art will appreciate that different vehicle models will often correspond to different vehicle parameter information. In this embodiment, the predetermined database may be a self-established database or an internet database. When the database is a self-built database, the vehicle parameter information corresponding to the vehicle model can be directly inquired from the self-built database. When the database is an internet database, vehicle parameter information corresponding to the vehicle model can be obtained from the internet database in a crawling mode. For example, vehicle information related to each vehicle model may be obtained from each vehicle website (i.e., internet database) based on a crawler technology, wherein the crawler technology has been extensively studied by those skilled in the art and is not described herein, and the vehicle website includes, but is not limited to, a website platform such as a brand official website, a home of an automobile, and an automobile quote. Vehicle interior configurations that the user needs to know include, but are not limited to, displacement, engine/transmission type, body size, chassis, brakes, safety devices and functions, etc. A vehicle website may not include all vehicle internal configuration information, so that complete and accurate information may be obtained from multiple vehicle websites.
And A30, outputting the vehicle model and vehicle parameter information corresponding to the vehicle model.
According to the embodiment, the vehicle model corresponding to the vehicle image identified by the YOLO-V5 improved network model is obtained firstly, then the vehicle parameter information corresponding to the vehicle model is searched from the preset database according to the vehicle model, and finally the vehicle model and the vehicle parameter information corresponding to the vehicle model are output, so that the vehicle information of each type of the vehicle to be identified is accurately obtained and displayed to a user, the use limitation of the intelligent wearable device is reduced, the vehicle information of each type of the vehicle can be efficiently and accurately identified by the intelligent wearable device, and the vehicle purchasing requirement of the user can be effectively provided with relevant references.
In one possible embodiment, the step of obtaining the vehicle model corresponding to the vehicle image identified by the YOLO-V5 improved network model comprises:
b10, extracting the features of the vehicle image through the trained YOLO-V5 improved network model, and extracting the vehicle global features corresponding to the vehicle image;
specifically, the trained YOLO-V5 improved network model is used for extracting the features of the vehicle image to obtain a plurality of common local features of the vehicle image, and then feature fusion is carried out on each common local feature to obtain the vehicle global feature corresponding to the vehicle image.
In the present embodiment, it should be noted that the common local feature in the vehicle image refers to a local feature in which an important feature (a feature that facilitates identification of vehicle information) and an unimportant feature (a feature that does not facilitate identification of vehicle information) are not distinguished.
Step B20, the attention mechanism improving module parallelly inputs the vehicle global features corresponding to the vehicle images into the channel attention module and the space attention module;
step B30, respectively acquiring a channel feature corresponding to the vehicle image output by the channel attention module and a spatial feature corresponding to the vehicle image output by the spatial attention module;
b40, performing feature fusion on the channel features and the spatial features to obtain key local features fusing channel attention and spatial attention;
in the present embodiment, the key local features in the vehicle image refer to important local features that facilitate recognition of the vehicle information.
And B50, identifying the vehicle model corresponding to the vehicle image through the key local features.
In this embodiment, it is known in the art that the corresponding channel feature and spatial feature of the vehicle image may be feature fused by the kronecker product operation. For many ways of performing feature fusion, those skilled in the art have conducted some intensive studies and will not be described herein. It should be noted that any conventional way of performing feature fusion is intended to be within the scope of the present application.
In the embodiment, a trained YOLO-V5 improved network model is used to perform feature extraction on the vehicle image, a vehicle global feature corresponding to the vehicle image is extracted, then a vehicle global feature corresponding to the vehicle image is input to the channel attention module and the spatial attention module in parallel through an attention mechanism improvement module, then a channel feature corresponding to the vehicle image output by the channel attention module and a spatial feature corresponding to the vehicle image output by the spatial attention module are obtained respectively, finally the channel feature and the spatial feature are subjected to feature fusion to obtain a key local feature fusing channel attention and spatial attention, so that the separately extracted channel attention feature and spatial attention feature are fused to be associated with each other, different features are enhanced with each other to form a complete key local feature, and the problem that the key features of the vehicle information identified by the channel attention module and the spatial attention module are mistakenly filtered to cause vehicle information identification failure or identification error is avoided because the channel attention module and the spatial attention module are connected in series (CBAM before improvement), so that the key features of the channel attention module and the spatial attention module are set in series (CBAM before improvement), and the CBAM is set to be more accurately embedded in a cbo-V5 improved network model, and a vehicle information extraction layer with a shallow attention filtering and a low-V5 parallel connection is set to obtain a vehicle information extraction important feature extraction layer, so that the important feature extraction and a vehicle information extraction layer is not more accurately judged that the vehicle information is more accurately judged by the improved by the channel attention module and the improved by the improved yol-V system, and the improved network model, therefore, the key characteristic information of the vehicle image can be efficiently and accurately identified and extracted, the vehicle information of the vehicle to be identified is identified according to the key characteristic information, the robustness of the original YOLO-V5 network model for identifying the vehicle information is further enhanced, the influence of light, background environment and shooting angle on the vehicle information identification is eliminated, and the technical effect of improving the accuracy of the vehicle information identification is achieved.
Further, in a possible implementation, the step of identifying, through the key local features, a vehicle model corresponding to the vehicle image includes:
step C10, performing global average pooling on the key local features to obtain a first feature pooling result, and performing global maximum pooling on the key local features to obtain a second feature pooling result;
step C20, splicing the first characteristic pooling result and the second characteristic pooling result along the channel direction to obtain a key pooling characteristic;
step C30, after the key pooling features are subjected to convolution processing, obtaining a key feature attention weight value through a preset activation function;
step C40, multiplying the attention weight of the key feature by the global feature of the vehicle to obtain an attention feature map fusing the attention of the channel and the attention of the space,
and step C50, identifying the vehicle model mapped by the attention feature map through a feature map mapping rule obtained through pre-training, and taking the vehicle model mapped by the attention feature map as the vehicle model corresponding to the vehicle image.
In the embodiment, a first feature pooling result is obtained by performing global average pooling on key local features, a second feature pooling result is obtained by performing global maximum pooling on the key local features, the first feature pooling result and the second feature pooling result are spliced along a channel direction to obtain key pooling features, so that the key local features are subjected to average pooling and maximum pooling in channel dimensions, the key pooling features are subjected to convolution processing, a key feature attention weight is obtained through a preset activation function, the key feature attention weight is multiplied by the vehicle global features to obtain an attention feature map with channel attention and space attention fused, so that the weight coefficient of the key local features is extracted by fusing the space domain attention and the channel domain attention, the weight coefficient is multiplied by the original features to obtain a weighted result of each feature information in a vehicle image, the weight of an important feature in the vehicle image is enhanced, the weight of an unimportant feature is suppressed to obtain a more characteristic of the vehicle, a feature map obtained by mapping the obtained in advance is trained to identify the weighted result of each feature information in the vehicle image, and the vehicle identification efficiency is further improved by mapping rules, and the vehicle identification efficiency is further improved.
Example two
Based on the first embodiment of the present application, in another embodiment of the present application, the same or similar contents as those in the first embodiment may refer to the above description, and are not repeated herein. On the basis, please refer to fig. 2, the step of inputting the vehicle image into the trained YOLO-V5 improved network model is preceded by:
step S400, obtaining a plurality of vehicle pictures corresponding to each vehicle model and vehicle model labels of the vehicle models corresponding to the vehicle pictures, and respectively performing label association on the vehicle pictures and the corresponding vehicle model labels to obtain a vehicle model sample picture set;
in this embodiment, the plurality of vehicle pictures corresponding to each vehicle model may be obtained through crawling on the internet based on a crawler manner, for example, the plurality of vehicle pictures corresponding to each vehicle model may be obtained from each vehicle website based on a crawler technology, where a skilled person in the crawler technology has a certain intensive research and is not repeated herein, and the vehicle website includes, but is not limited to, a website platform such as a brand official website, a car owner, a car quote, and the like. In addition, the vehicle type label of the vehicle type corresponding to each vehicle picture can be obtained through manual active input, that is, the vehicle type label of the vehicle type corresponding to each vehicle picture can be obtained through a mode of manually marking each vehicle picture.
Step S500, dividing the vehicle type sample picture set into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion;
in this embodiment, the preset ratio is not specifically limited in this embodiment, so as to better improve the training convergence efficiency of the YOLO-V5 improved network model. For example, the preset ratio is 4:1.
step S600, training a preset YOLO-V5 improved network model through the vehicle type recognition training sample set, and verifying the recognition accuracy of the preset YOLO-V5 improved network model through the vehicle type recognition verification sample set;
step S700, if the verified recognition accuracy reaches a preset accuracy threshold, stopping training the improved YOLO-V5 network model to obtain the trained improved YOLO-V5 network model.
In the embodiment, a plurality of vehicle pictures corresponding to each vehicle model and vehicle type labels corresponding to the vehicle models of each vehicle picture are obtained, each vehicle picture is respectively subjected to label association with the corresponding vehicle type label to obtain a vehicle type sample picture set, then the vehicle type sample picture set is divided into a vehicle type recognition training sample set and a vehicle type recognition verification sample set according to a preset proportion, a preset YoLO-V5 improved network model is trained through the vehicle type recognition training sample set, the recognition accuracy of the preset YoLO-V5 improved network model is verified through the vehicle type recognition verification sample set, and when the verified recognition accuracy reaches a preset accuracy threshold value, the training of the YOLO-V5 improved network model is stopped, the trained YOLO-V5 improved network model is obtained, so that the accuracy of vehicle information of a vehicle to be recognized is effectively ensured through the YOLO-V5 improved network model, and the efficiency and robustness of convergence of the YOLO-V5 improved network model are further improved.
In a possible implementation manner, the step of respectively tag-associating each vehicle picture with its corresponding vehicle model to obtain a vehicle model sample picture set includes:
step D10, performing data enhancement processing on each picture in the vehicle type identification training sample set to obtain a data enhancement sample picture set;
in this embodiment, the data enhancement processing may include at least one of image scale normalization, image random cropping, image value normalization, image flipping, image scaling, brightness adjustment, image rotation, and image tilting of each picture in the vehicle type recognition training sample set.
In the embodiment, after data enhancement processing is performed on each picture in the vehicle type recognition training sample set, the picture is input to the YOLO-V5 improved network model for training, so that the influence of brightness, an environmental background, an image acquisition angle and the like on the training of the YOLO-V5 improved network model is eliminated as much as possible, and the efficiency and the robustness of training convergence of the YOLO-V5 improved network model are further improved.
Step D20, respectively randomly splicing a preset number of pictures in the data enhancement sample picture set into one picture to obtain a Mosaic enhancement sample picture set, wherein the preset number is more than one;
the preset number is not specifically limited in this embodiment, for example, the preset number is 9, that is, the data enhancement sample pictures obtained after the preliminary data enhancement can be spliced in a nine-in-one manner to obtain the Mosaic enhancement sample pictures.
The step of dividing the vehicle type sample picture set into a vehicle type recognition training sample set and a vehicle type recognition verification sample set according to a preset proportion comprises the following steps:
and D30, dividing the Mosaic enhancement sample picture set into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion.
In the embodiment, a preset number of pictures in the data enhancement sample picture set are respectively randomly spliced into one picture to obtain a Mosaic enhancement sample picture set, and the YOLO-V5 improved network model is trained and verified according to the Mosaic enhancement sample picture set, so that the training effect of the multi-in-one Mosaic enhancement sample picture is better than that of a single picture (a single picture without picture splicing) and the convergence speed of the model is higher, and the efficiency and robustness of training convergence of the YOLO-V5 improved network model are further improved.
In an implementable manner, the step of randomly splicing a preset number of pictures in the data enhancement sample picture set into one picture to obtain a Mosaic enhancement sample picture set includes:
step E10, adding Gaussian noise to each picture in the Mosaic enhancement sample picture set to obtain a Mosaic enhancement sample picture set added with Gaussian noise;
in the present embodiment, it is understood by those skilled in the art that gaussian noise refers to a type of noise whose probability density function follows a gaussian distribution (i.e., a normal distribution).
The step of dividing the Mosaic enhancement sample picture set into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion comprises the following steps:
and E20, dividing the Mosaic enhancement sample picture set added with the Gaussian noise into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion.
In the embodiment, after Gaussian noise is added to each picture in the Mosaic enhancement sample picture set, the Mosaic enhancement sample picture set after the Gaussian noise is added is obtained, so that the pictures are overall and average, a wrong attention direction is avoided in a subsequent attention mechanism, meanwhile, the picture to be used for training is enabled to restore the actual image shooting effect of a user wearing the intelligent wearable device to a vehicle to be recognized, and the robustness for training the YOLO-V5 improved network model is improved.
To aid in understanding the technical concepts of the present application, a specific embodiment is set forth:
in this specific embodiment, due to the limitation of an application scenario, an algorithm used in the present application performs network training in a self-established data set manner during training, 10000 vehicle type pictures after 2000 years are acquired from a network in a crawler manner, the obtained pictures are screened, the obtained pictures are evaluated by a sampling method, the obtained pictures are blurred in view angle, the vehicle picture is not the picture positioned in the center of the picture, and then the obtained screened pictures are subjected to an improved data enhancement operation, wherein the general data enhancement is to rotate, translate, partially cut and the like a single picture, and since the target is a vehicle and has a large volume, the amount of information contained in the single picture is not enough, and multiple rounds of training are often required to make the network model better converge, the present application improves a data enhancement method, namely a mosaic-9 data enhancement method: and splicing the data pictures obtained after the preliminary data enhancement according to a nine-in-one mode, and adding Gaussian noise to the whole situation after the splicing is finished, so that the pictures are overall average, and the problem that a subsequent attention mechanism obtains wrong attention direction is avoided. And after the data set is manufactured, data marking is needed, and the correct information of the vehicle type is correctly identified for each picture. The labeled data set is divided into a training set and a verification set according to the proportion of 4.
The method for enhancing the mosaics-9 data has the advantages that: the target detection and identification object is a vehicle, so that the condition of small target detection does not exist, and the condition of missed detection cannot occur after nine-in-one pictures; the nine-in-one picture has better training effect than a single picture (a single picture without picture splicing), so that the model convergence speed is higher.
The YOLO-V5s model adopted by the application is improved, and the purpose is to better fit a low-cost platform such as AR glasses and to be more suitable for recognition of non-small targets such as vehicle type recognition.
The specific network model training process may be:
referring to fig. 3, in an original training process of the YOLO-V5s model, an original image is input and then sequentially passes through five parts, namely an input end, a backbone network, a neck, a head and an output end, the input end inputs the original image, the specification of the original image input by the input end in the model is 640 × 3, and then the original image enters a backbone to be subjected to feature extraction, a Focus structure unique to the YOLO-V5s network model performs a slicing operation on the image, and the operation changes the length and the width of the image and the number of channels to obtain a feature map; then entering a CBL module, wherein the module comprises a convolution layer, a batch normalization layer and an activation layer, and further processing the characteristics of the characteristic diagram; and finally, entering an SPP module, wherein the module mainly performs pooling operation on the characteristic diagram, so that the input and output matrixes are kept consistent, and the next module is convenient to process. And entering a neck part, mainly performing feature fusion, and performing feature fusion on the feature graph obtained by the backbone, wherein the stage is to perform object detection object highlighting globally and facilitate the identification of objects by a subsequent head module.
The CBAM module is essentially a channel and space hybrid attention mechanism, compared with the single attention mechanism, the CBAM integrates two mapping processes of a channel and a space, can retain more information, and simultaneously gives heavier weight to more obvious large features, thereby ensuring that a network can focus correct attention on a main target. The spatial attention module mainly highlights the position information of the features and gives greater weight to important positions.
The original CBAM module inputs the original feature map into the channel attention module first and then into the spatial attention module, and the function of the channel attention module can be understood as follows: the original feature map is an RGB tricolor map, and areas needing attention are given higher weight on RGB tricolor channels by recognizing gradient or pictures of certain colors and consistency of overall features. The role of spatial attention can be understood as focusing on the spatial features of the picture, and performing pooling operation on the information of three dimensions of the CHW so as to obtain the spatial feature information which is most worth focusing.
Based on the characteristics of the CBAM, in order to enhance the objectivity of the recognition result and eliminate the influence of the model, the CBAM module is improved, and the improvement is different from the traditional YOLO-V5s model, and the improvement of the CBAM module is realized by adding an improved CBAM (attention mechanism improvement module) module into the YOLO-V5s network model, as shown in FIG. 4. Since the focus of attention obtained by the channel attention is not necessarily the real focus in the overall view of the picture, and then the original CBAM module directly inputs the feature map processed by the channel attention into the spatial attention module, which is equivalent to inputting the spatial attention module after "modifying" the original image, the obtained result is often deviated from the real target detection area, and the deviation of the area may cause failure of vehicle type identification. The improvements of CBAM of the present application are therefore mainly reflected inThe single flow direction of the feature map (from the channel attention module to the spatial attention module) is changed into the dual flow direction, the original feature maps enter the spatial attention module and the channel attention module respectively, and the two modules perform the processing of the feature maps respectively, as shown in fig. 5. The channel attention pays more attention to the weight of each feature plane, the space attention pays more attention to the weight of each local feature, the two modules are separated, the dual features of the feature planes and the local space can be considered better, the original feature map is processed by the two attention modules respectively to generate two feature maps, and the two feature maps are subjected to Crohn's product operation
Figure BDA0003839176440000181
The two processed feature maps can be understood as two matrixes with the same size (because the input feature map matrixes are the same), the kronecker product operation on the two matrixes with the same size is to multiply the matrixes to obtain a fused feature matrix, then the weight emphasis of the coincidence feature is carried out through the global pooling and activating function, the obtained feature map is the attention feature map fusing the channel attention and the space attention, and the obtained attention feature map is subsequently input into a head part of the deep learning model to judge the result.
Experiments prove that although the size of the model is slightly increased on the basis of the original YOLO-V5s, the algorithm provided by the application can meet the requirements of real-time performance and light weight of AR glasses and can achieve the target of expected 85% accuracy.
It should be noted that the details of the specific embodiment are only for understanding the technical idea of the present application, and do not constitute a limitation to the present application, and the technical idea of the present application should be protected within the scope of the present application by making more simple changes.
EXAMPLE III
The embodiment of the invention provides an intelligent wearing device, which comprises: at least one processor; and a memory communicatively coupled to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the vehicle information identification method in the first embodiment.
Referring now to fig. 6, a schematic diagram of a smart wearable device suitable for use in implementing embodiments of the present disclosure is shown. The smart wearable device in the embodiments of the present disclosure may include, but is not limited to, a Mixed Reality (Mixed Reality) -MR device (e.g., MR glasses or MR helmet), an Augmented Reality (Augmented Reality) -AR device (e.g., AR glasses or AR helmet), a virtual Reality- (virtual Reality) -VR device (e.g., VR glasses or VR helmet), an Extended Reality (Extended Reality) -XR device, or some combination thereof, and the like. The smart wearable device shown in fig. 6 is only an example, and should not bring any limitation to the functions and the use range of the embodiment of the present disclosure.
As shown in fig. 6, the smart wearable device may include a processing means 1001 (e.g., a central processing unit, a graphic processor, etc.) which may perform various appropriate actions and processes according to a program stored in a read only memory (ROM 1002) or a program loaded from a storage means into a random access memory (RAM 1004). In the RAM1004, various programs and data necessary for the operation of the smart wearable device are also stored. The processing device 1001, the ROM1002, and the RAM1004 are connected to each other via a bus 1005. An input/output (I/O) interface is also connected to bus 1005.
Generally, the following systems may be connected to the I/O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, or the like; output devices 1008 including, for example, liquid Crystal Displays (LCDs), speakers, vibrators, and the like; a storage device 1003 including, for example, a magnetic tape, a hard disk, or the like; and a communication device 1009. Communication means 1009 may allow smart wearable device to communicate wirelessly or wiredly with other devices to exchange data. While the figures illustrate a smart wearable device with various systems, it is to be understood that not all illustrated systems are required to be implemented or provided. More or fewer systems may alternatively be implemented or provided.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication means, or installed from the storage means 1003, or installed from the ROM 1002. The computer program, when executed by the processing device 1001, performs the above-described functions defined in the methods of the embodiments of the present disclosure.
By adopting the vehicle information identification method in the first embodiment or the second embodiment, the intelligent wearable device provided by the invention can reduce the use limitation of the intelligent wearable device, so that the intelligent wearable device can accurately identify the vehicle information of various types of vehicles. Compared with the prior art, the beneficial effects of the intelligent wearable device provided by the embodiment of the invention are the same as those of the vehicle information identification method provided by the first embodiment, and other technical features of the intelligent wearable device are the same as those disclosed by the method of the first embodiment, which are not repeated herein.
It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof. In the foregoing description of embodiments, the particular features, structures, materials, or characteristics may be combined in any suitable manner in any one or more embodiments or examples.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Example four
The embodiment of the invention provides a readable storage medium which is a computer readable storage medium and has computer readable program instructions stored thereon, wherein the computer readable program instructions are used for executing the vehicle information identification method in the first embodiment.
The computer readable storage medium provided by the embodiments of the present invention may be, for example, a USB flash disk, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, or device, or any combination thereof. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present embodiment, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, or device. Program code embodied on a computer readable storage medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.
The computer readable storage medium may be contained in a smart wearable device; or the device can be independently arranged and not assembled into the intelligent wearable device.
The computer readable storage medium carries one or more programs which, when executed by the smart wearable device, cause the smart wearable device to: acquiring a vehicle image of a vehicle to be identified; inputting the vehicle image into a trained YOLO-V5 improved network model, wherein the YOLO-V5 improved network model is a YOLO-V5 network model embedded with an attention mechanism improving module, and the attention mechanism improving module comprises a channel attention module and a space attention module which are arranged in parallel; and acquiring vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model, and outputting the vehicle information corresponding to the vehicle image.
Computer program code for carrying out operations for aspects of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The modules described in the embodiments of the present disclosure may be implemented by software or hardware. Wherein the names of the modules do not in some cases constitute a limitation of the unit itself.
The computer-readable storage medium provided by the invention stores computer-readable program instructions for executing the vehicle information identification method, so that the use limitation of the intelligent wearable device can be reduced, and the intelligent wearable device can accurately identify the vehicle information of various types of vehicles. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the embodiment of the invention are the same as those of the vehicle information identification method provided by the first embodiment or the second embodiment, and are not repeated herein.
EXAMPLE five
Embodiments of the present invention further provide a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the vehicle information identification method as described above are implemented.
The application provides a computer program product can reduce intelligent wearing equipment's use limitation for intelligent wearing equipment can accurately discern the vehicle information of each type of vehicle. Compared with the prior art, the beneficial effects of the computer program product provided by the embodiment of the present invention are the same as those of the vehicle information identification method provided by the first embodiment or the second embodiment, and are not described herein again.
The above description is only a preferred embodiment of the present application, and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings, or which are directly or indirectly applied to other related technical fields, are included in the scope of the present application.

Claims (10)

1. A vehicle information identification method is applied to intelligent wearable equipment, and is characterized by comprising the following steps:
acquiring a vehicle image of a vehicle to be identified;
inputting the vehicle image into a trained YOLO-V5 improved network model, wherein the YOLO-V5 improved network model is a YOLO-V5 network model embedded with an attention mechanism improving module, and the attention mechanism improving module comprises a channel attention module and a space attention module which are arranged in parallel;
and acquiring vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model, and outputting the vehicle information corresponding to the vehicle image.
2. The vehicle information identification method according to claim 1, wherein the vehicle information includes a vehicle model and vehicle parameter information corresponding to the vehicle model, and the step of obtaining the vehicle information corresponding to the vehicle image identified by the trained YOLO-V5 improved network model and outputting the vehicle information corresponding to the vehicle image includes:
obtaining a vehicle model corresponding to the vehicle image identified by the YOLO-V5 improved network model;
according to the vehicle model, vehicle parameter information corresponding to the vehicle model is searched and obtained from a preset database, wherein the vehicle parameter information comprises at least one of vehicle market quotation, vehicle internal configuration and vehicle delivery year;
and outputting the vehicle model and vehicle parameter information corresponding to the vehicle model.
3. The vehicle information identification method according to claim 2, wherein the step of obtaining the vehicle model corresponding to the vehicle image identified by the YOLO-V5 improved network model comprises, before:
carrying out feature extraction on the vehicle image through a trained YOLO-V5 improved network model, and extracting to obtain a vehicle global feature corresponding to the vehicle image;
inputting, by the attention mechanism improvement module, vehicle global features corresponding to the vehicle image in parallel to the channel attention module and the spatial attention module;
respectively acquiring channel characteristics corresponding to the vehicle images output by the channel attention module and spatial characteristics corresponding to the vehicle images output by the spatial attention module;
performing feature fusion on the channel features and the spatial features to obtain key local features fusing channel attention and spatial attention;
and identifying the vehicle model corresponding to the vehicle image through the key local features.
4. The vehicle information identification method according to claim 3, wherein the step of identifying the vehicle model corresponding to the vehicle image by the key local feature comprises:
performing global average pooling on the key local features to obtain a first feature pooling result, and performing global maximum pooling on the key local features to obtain a second feature pooling result;
splicing the first characteristic pooling result and the second characteristic pooling result along the channel direction to obtain a key pooling characteristic;
after carrying out convolution processing on the key pooling characteristics, obtaining a key characteristic attention weight through a preset activation function;
multiplying the attention weight of the key feature by the global feature of the vehicle to obtain an attention feature map fusing the channel attention and the space attention,
and identifying the vehicle model mapped by the attention feature map through a feature map mapping rule obtained by pre-training, and taking the vehicle model mapped by the attention feature map as the vehicle model corresponding to the vehicle image.
5. The vehicle information recognition method of claim 1, wherein the step of inputting the vehicle image to a trained YOLO-V5 improved network model is preceded by:
acquiring a plurality of vehicle pictures corresponding to each vehicle model and vehicle type labels of the vehicle models corresponding to the vehicle pictures, and respectively performing label association on the vehicle pictures and the corresponding vehicle type labels to obtain a vehicle type sample picture set;
dividing the vehicle type sample picture set into a vehicle type recognition training sample set and a vehicle type recognition verification sample set according to a preset proportion;
training a preset YOLO-V5 improved network model through the vehicle type recognition training sample set, and verifying the recognition accuracy of the preset YOLO-V5 improved network model through the vehicle type recognition verification sample set;
and if the verified recognition accuracy reaches a preset accuracy threshold, stopping training the improved YOLO-V5 network model to obtain the trained improved YOLO-V5 network model.
6. The vehicle information identification method according to claim 5, wherein the step of performing label association between each vehicle picture and its corresponding vehicle type label to obtain a vehicle type sample picture set comprises:
performing data enhancement processing on each picture in the vehicle type recognition training sample set to obtain a data enhancement sample picture set;
respectively randomly splicing a preset number of pictures in the data enhancement sample picture set into one picture to obtain a Mosaic enhancement sample picture set, wherein the preset number is more than one;
the step of dividing the vehicle type sample picture set into a vehicle type recognition training sample set and a vehicle type recognition verification sample set according to a preset proportion comprises the following steps:
and dividing the Mosaic enhancement sample picture set into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion.
7. The vehicle information identification method according to claim 6, wherein the step of randomly splicing a preset number of pictures in the data enhancement sample picture set into one picture to obtain a Mosaic enhancement sample picture set, respectively, comprises:
adding Gaussian noise to each picture in the Mosaic enhancement sample picture set to obtain a Mosaic enhancement sample picture set with the Gaussian noise added;
the step of dividing the Mosaic enhancement sample picture set into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion comprises the following steps:
and dividing the Mosaic enhancement sample picture set added with the Gaussian noise into a vehicle type identification training sample set and a vehicle type identification verification sample set according to a preset proportion.
8. The vehicle information identification method according to claim 1, characterized in that the step of outputting the vehicle information includes:
displaying the vehicle information through a display screen of the intelligent wearable device; and/or the presence of a gas in the gas,
and playing the vehicle information through a loudspeaker of the intelligent wearable device.
9. The utility model provides an intelligence wearing equipment which characterized in that, intelligence wearing equipment includes:
at least one processor; and (c) a second step of,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the vehicle information identification method of any one of claims 1 to 8.
10. A readable storage medium, characterized in that the readable storage medium is a computer-readable storage medium having stored thereon a program for implementing a vehicle information identification method, the program being executed by a processor to implement the steps of the vehicle information identification method according to any one of claims 1 to 8.
CN202211098991.2A 2022-09-08 2022-09-08 Vehicle information identification method, intelligent wearable device and readable storage medium Pending CN115512311A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211098991.2A CN115512311A (en) 2022-09-08 2022-09-08 Vehicle information identification method, intelligent wearable device and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211098991.2A CN115512311A (en) 2022-09-08 2022-09-08 Vehicle information identification method, intelligent wearable device and readable storage medium

Publications (1)

Publication Number Publication Date
CN115512311A true CN115512311A (en) 2022-12-23

Family

ID=84504210

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211098991.2A Pending CN115512311A (en) 2022-09-08 2022-09-08 Vehicle information identification method, intelligent wearable device and readable storage medium

Country Status (1)

Country Link
CN (1) CN115512311A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116872961A (en) * 2023-09-07 2023-10-13 北京捷升通达信息技术有限公司 Control system for intelligent driving vehicle

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116872961A (en) * 2023-09-07 2023-10-13 北京捷升通达信息技术有限公司 Control system for intelligent driving vehicle
CN116872961B (en) * 2023-09-07 2023-11-21 北京捷升通达信息技术有限公司 Control system for intelligent driving vehicle

Similar Documents

Publication Publication Date Title
US11367281B2 (en) Systems and methods for augmented reality navigation
CN109003297B (en) Monocular depth estimation method, device, terminal and storage medium
CN113179368A (en) Data processing method and device for vehicle damage assessment, processing equipment and client
CN111310770B (en) Target detection method and device
CN107084740B (en) Navigation method and device
CN111078940B (en) Image processing method, device, computer storage medium and electronic equipment
CN115031758A (en) Live-action navigation method, device, equipment, storage medium and program product
CN108170751B (en) Method and apparatus for handling image
CN112712036A (en) Traffic sign recognition method and device, electronic equipment and computer storage medium
CN115512311A (en) Vehicle information identification method, intelligent wearable device and readable storage medium
CN112232311A (en) Face tracking method and device and electronic equipment
CN115577768A (en) Semi-supervised model training method and device
CN114529890A (en) State detection method and device, electronic equipment and storage medium
CN112396060B (en) Identification card recognition method based on identification card segmentation model and related equipment thereof
CN116186354B (en) Method, apparatus, electronic device, and computer-readable medium for displaying regional image
CN116258756B (en) Self-supervision monocular depth estimation method and system
CN116844129A (en) Road side target detection method, system and device for multi-mode feature alignment fusion
CN115731370A (en) Large-scene element universe space superposition method and device
CN116188587A (en) Positioning method and device and vehicle
CN115393423A (en) Target detection method and device
CN112052863B (en) Image detection method and device, computer storage medium and electronic equipment
CN111353441B (en) Road extraction method and system based on position data fusion
CN114332798A (en) Processing method and related device for network car booking environment information
CN114399696A (en) Target detection method and device, storage medium and electronic equipment
CN114386481A (en) Vehicle perception information fusion method, device, equipment and storage medium

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