A method and system for searching for a vehicle by a graph

A vehicle and license plate technology, applied in the field of vehicle monitoring, can solve the problems of long running time, large memory usage, poor generalization, etc., and achieve the effect of reducing difficulty, ensuring accuracy, and running speed balance

Inactive Publication Date: 2018-12-18
GOSUNCN TECH GRP
View PDF7 Cites 12 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] CN106156750A "A Method for Searching Cars Based on Convolutional Neural Networks" only uses a single model to extract features for comparison and sorting, and there is no subdivision of the usage scene
Therefore, in practical applications, generalization will be poor, and it will be easily affected by interference factors such as lighting, posture, and occlusion.
At the same time, although the classic feature comparison models used in the existing technology, such as Resnet (residual network), Inception, VGG, etc.) have high accuracy, they have disadvantages such as large memory usage and long running time in practical applications, which affect Product real-time and economical

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A method and system for searching for a vehicle by a graph
  • A method and system for searching for a vehicle by a graph
  • A method and system for searching for a vehicle by a graph

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0048] figure 1 It is a schematic flow chart of this embodiment, a method of searching for a car with a picture, including,

[0049] Step 101, pre-training a model for vehicle filtering, the model includes a vehicle type model, a color model, a sub-brand model, a license plate information model, and a feature area classification model;

[0050] The method for searching cars by image disclosed in this embodiment adopts the object detection model of deep learning, because deep learning has powerful feature learning ability, it can improve the accuracy and efficiency of object detection. Specifically, this embodiment uses a network structure improved by classical deep learning algorithms such as vgg and resnet, and simultaneously meets the requirements of accuracy and speed.

[0051] In this embodiment, model training is performed on vehicle images collected from cameras at public security traffic checkpoints to obtain a variety of filtering models, where the filtering models in...

Embodiment 2

[0068] In this embodiment, the target vehicle of the present invention is defined as a licensed vehicle, figure 2 It is a schematic flow chart of this embodiment, a method of searching for a car with a picture, including:

[0069] Step 201, pre-training a model for vehicle filtering, where the model includes a vehicle type model, a color model, a sub-brand model, a license plate information model, and a vehicle face area classification model;

[0070] The training method and process of pre-training for vehicle detection in this embodiment are the same as step 101 in Embodiment 1, but it should be noted that the feature area classification model in this embodiment uses the vehicle face area for training. This is because it is easier to learn the character features of the license plate by using the vehicle face area model compared to the classification model of the whole vehicle area training vehicle, and for a licensed vehicle, the license plate information is used as the uniq...

Embodiment 3

[0086] In this embodiment, the target vehicle of the present invention is defined as an unlicensed vehicle, image 3 It is a schematic flow chart of this embodiment, a method of searching for a car with a picture, including:

[0087] Step 301, pre-training a model for vehicle filtering, where the model includes a vehicle type model, a color model, a sub-brand model, a license plate information model, and a whole vehicle area classification model;

[0088] In this embodiment, the training method and process for pre-training for vehicle detection are the same as step 101 in Embodiment 1, but it should be noted that the feature area classification model uses the entire vehicle area for training. This is because for unlicensed vehicles, the vehicle information in the vehicle face area is limited. In order to ensure the accuracy of target vehicle detection, it is necessary to perform feature extraction on the entire vehicle area. The entire vehicle area classification model is used...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention provides a method and a system for searching for a vehicle by a graph. The method comprises the following steps: a model for vehicle filtering is trained in advance, wherein the model comprises a vehicle type model, a color model, a sub-brand model, a license plate information model and a feature region classification model. A front overhead picture of a target vehicle to be detectedis obtained, and license plate information of the target vehicle is detected. The retrieval result is: a licensed vehicle or an unlicensed vehicle; according to the different conditions of the targetvehicle license plate information, the retrieval of the licensed vehicle or the retrieval of the unlicensed vehicle is started. A variety of models perform filtering first to reduce the impact of interference factors, and the feature region classification model is used for feature alignment ranking. Moreover, different feature comparison models are used according to the usage scenario. At the same time, according to the characteristics of vehicle samples, a feature comparison model is specially designed to ensure the balance among accuracy, memory occupation and running speed in the application process.

Description

[0001] The technical solution relates to the field of vehicle monitoring, in particular to a method and system for searching vehicles by classifying pictures. Background technique [0002] With the research and application of computer vision technology, the intelligence and reliability of tracking and detecting targets can be used in real-life scenarios such as government control, illegal locking, and suspect tracking. It has been used in transportation, public security, military, energy, etc. And many other fields, there is a trend of further in-depth development. [0003] Searching cars with pictures is an image-based vehicle retrieval technology, which is widely used in traffic vehicle management, criminal suspect tracking and other fields. Its purpose is to replace manual retrieval of target vehicles from massive data. The technical route of searching for a car by image can be summarized as follows: extract the features of the target vehicle image and all the images in the...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06F17/30G06K9/62
CPCG06V2201/08G06F18/24
Inventor 黎邹邹毛亮李超熊伟郑康元文莉薛坤南黄仝宇汪刚宋一兵侯玉清刘双广
Owner GOSUNCN TECH GRP
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap