A video small target detection method based on a depth convolution neural network

A small target detection, neural network technology, applied in the field of intelligent video surveillance, can solve the problems of noise interference, low efficiency, false detection, etc., to achieve the effect of rapid detection, improved effect and efficiency

Inactive Publication Date: 2019-02-22
ZHEJIANG GONGSHANG UNIVERSITY
View PDF4 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the small targets in the video are very susceptible to noise interference due to their small scale, which leads to false detection and missed detection. The small target detection in the existing technology is not effective and efficient, which affects subsequent target tracking, Accuracy of target re-identification

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 video small target detection method based on a depth convolution neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0020] Embodiments of the present invention will be described in further detail below in conjunction with the accompanying drawings.

[0021] Such as figure 1 As shown, a video small target detection method based on a deep convolutional neural network includes the following steps:

[0022] Step (1) Model pre-training: start the iterative training of the network based on the pre-trained VGG model, and take 9 candidate windows at each position of the 256-channel image with a size of 51×39, that is, three kinds of areas {128 2 ,256 2 ,512 2}×three ratios {1:1,1:2,2:1}, the candidate window is used as anchors, that is, the anchor point;

[0023] Step (2) Feature extraction based on deep convolutional neural network: the convolutional layer with residual structure is used to extract feature maps, and the loss function used by the network is ;

[0024] Among them, i represents the anchor index value, pi represents the softmax prediction probability of the foreground, Indicat...

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 video small target detection method based on a depth convolution neural network,. Based on the characteristics of multi-layer nonlinear structure of deep convolution neural network, the video object features are modeled as a whole. Combined with the fast and accurate detection of common video objects by Faster RCNN and the effective path added by ResNet in the feature extraction layer of depth network, With the simplification of the network structure, a depth network structure ERF-Net (Efficient Residual Faster rcnn) based video target detection method is proposed, which can detect video objects quickly and small objects at the same time.. The invention has the advantages that the targets with different distances and different scales in the video are accurately andquickly detected, the effect and efficiency of small target detection are improved, and a good foundation is provided for subsequent target tracking, target recognition, and the like.

Description

technical field [0001] The invention relates to the technical field of intelligent video monitoring in computer vision and big data processing, in particular to a detection method based on a deep convolutional neural network that can quickly and accurately detect small-scale targets of interest from massive video data. Background technique [0002] How to efficiently analyze the massive video data acquired by the video surveillance system using the method of artificial intelligence is a frontier topic that has attracted much attention in the fields of computer vision and big data in recent years. [0003] With the rapid development of computer science and technology and video surveillance hardware, the industry has higher and higher requirements for intelligent video surveillance technology. The so-called intelligent video surveillance processing technology mainly refers to the use of computer vision video analysis methods to parse the video into a video sequence and automat...

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): G06K9/32G06N3/04G06N3/08
CPCG06N3/08G06V10/255G06N3/045
Inventor 王慧燕
Owner ZHEJIANG GONGSHANG UNIVERSITY
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