Driver-vehicle classification method and device

A classification method and inter-class difference technology, which is applied in the direction of neural learning methods, instruments, biological neural network models, etc., can solve problems that affect accuracy, expand the number of layers, affect speed, etc., and achieve a balance between speed and accuracy, Good balance between speed and precision

Inactive Publication Date: 2015-03-25
深圳市华尊科技股份有限公司
View PDF0 Cites 23 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The traditional method directly expands the number of convolutional layers, which may expand the inappropriate number of layers, resulting in an inappropriate emphasis on the speed and accuracy of the implementation of the method, that is, the traditional method may expand the layer under the condition of high precision requirements. If the number is too small, the accuracy will be affected, or if the accuracy requirement is not high, the number of extended layers will be too large, which will affect the speed

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
  • Driver-vehicle classification method and device
  • Driver-vehicle classification method and device
  • Driver-vehicle classification method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] A very important aspect of the human-vehicle classification method based on the deep convolutional neural network is the expansion of the number of neural network layers. If the number of convolutional layers is expanded too much, the speed of running the human-vehicle classification method will slow down and affect the work. Efficiency; if the number of convolutional layers is expanded too little, the calculation accuracy will be reduced, affecting the quality of work. In practical applications, the number of convolutional layers is usually extended to 3, 4 or 5 layers. The increase in the number of convolutional layers is generally beneficial to the improvement of accuracy, but it will slow down the operation speed. How to balance the speed and accuracy has a great impact on the realization of the method. In actual work, the expansion of the number of layers in the traditional way may not be the most suitable choice. For example, when the number of layers is expanded ...

Embodiment 2

[0100] The device for classifying people and vehicles in this embodiment includes two modules: a training module and a classification module. The following describes in detail how the two modules work during a process of classifying people and vehicles.

[0101] The training module is used to read the training sample set offline, and use the training sample set to expand the number of convolutional layers of the deep convolutional neural network, as shown in equations (4) and (5), to calculate the intra-class difference of the training sample set s1 and inter-class difference s2, calculate the ratio of s1 and s2, the ratio of s1 and s2 is the difference ratio s1 / s2. As shown in formula (6), the difference ratio is compared with the threshold, and the number of convolutional layers N of the deep convolutional neural network is adaptively expanded according to the comparison result. The threshold includes the first threshold d1 and the second threshold d2, which is Obtained thro...

Embodiment 3

[0110] The method for classifying people and vehicles in this embodiment is divided into two processes: a training process and a classification process. The detailed steps of each process are as follows.

[0111] Such as Figure 6 Shown is the flow chart of the training process, and the following content describes the specific steps of the training process in detail. Receive image samples from the training sample set, and set the learning rate and number of iterations of the BP algorithm. Use the training sample set to expand the number of convolutional layers of the deep convolutional neural network, as shown in equations (4) and (5), calculate the intra-class difference s1 and the inter-class difference s2 of the training sample set, and calculate s1 and The ratio of s2, the ratio of s1 and s2 is the difference ratio s1 / s2. As shown in formula (6), the difference ratio is compared with the threshold, and the number of convolutional layers N of the deep convolutional neural...

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 discloses a driver-vehicle classification method and device. The method includes the steps that a training sample set is read; difference within a category and difference among categories of the training sample set are calculated, a difference specific value is calculated, and the difference specific value is the specific value between the difference within the category and difference among categories; the difference specific value is compared with a preset threshold value, and the number of convolution layers of a deep convolution neural network is determined according to the comparison result; the deep convolution neural network is initialized; the initialized deep convolution neural network is trained in combination with the number of the convolution layers, and a deep convolution neural network model is obtained; input images to be tested are classified through the deep convolution neural network model. By means of the method, the problem of improper emphasis on the aspects of the speed and accuracy in the method implementing process generated due to the fact that through a traditional mode, an error number of the convolution layers is set is avoided, and the requirement for balance between the speed and accuracy can be better met.

Description

technical field [0001] The present application relates to a method and device for classifying people and vehicles. Background technique [0002] With the development of my country's economy, the level of urbanization has been further improved. Smart city has become an important direction of urban modernization development. As an important part of smart city, video surveillance system is also developing in the direction of intelligence and networking. In intelligent video surveillance, the classification of pedestrians and vehicles is an important issue in the pre-processing process of public security image detection and traffic status analysis. Therefore, the classification method of people and vehicles based on image processing and machine learning is the key technology of intelligent video surveillance, and has become a research hotspot in related fields at home and abroad. [0003] The traditional process of a person-vehicle classification system includes three steps: ...

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): G06K9/62
CPCG06N3/088G06F18/2413
Inventor 刘凯吴伟华
Owner 深圳市华尊科技股份有限公司
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