Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Target detection method, device and system

A target detection and target technology, applied in the field of big data, can solve the problems of excessive interference information, low division precision, and low accuracy rate, and achieve the effects of high division accuracy, high target recognition efficiency, and improved accuracy rate

Active Publication Date: 2019-08-06
HUAWEI TECH CO LTD
View PDF7 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In the prior art, the division of target candidate areas is based on rectangular image blocks, the division accuracy is low and each of the divided rectangular image blocks contains more interference information, and the accuracy of reflecting the attitude change or occlusion of the target is low. poor applicability

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
  • Target detection method, device and system
  • Target detection method, device and system
  • Target detection method, device and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0068] see image 3 , is a schematic flowchart of the target detection method provided in the embodiment of the present application. The method for target detection provided by the embodiment of the present application may include steps:

[0069] S301. Acquire target candidate regions in an image to be detected.

[0070] In some implementation manners, after the image to be detected is input into the target detection system 10 through the input device 14 , the feature extraction unit 111 extracts image features with pedestrians as detection objects from the image to be detected. In the deep learning pedestrian detection method based on the deep neural network model, the convolutional neural network (CNN) can be used first to extract the depth features of the image to be detected, and then the region proposal network (RPN) can be used according to the convolutional neural network. The extracted deep features extract local area candidate boxes from the image to be detected, su...

Embodiment 2

[0108] In the pedestrian detection method described in the first embodiment above, on the one hand, the component candidate regions corresponding to each pedestrian component are obtained by constructing an image segmentation network, as well as the local image features corresponding to each component candidate region, and then learned through a bidirectional LSTM model Part relationship features between pedestrian parts. On the other hand, by fusing the component relationship features learned by the bidirectional LSTM model with the overall image features of the pedestrian candidate area to achieve pedestrian detection in the image to be detected, it further enhances the target detection system's ability to detect pedestrian pose changes in complex application scenarios. The image processing capability of the mask and occlusion state realizes the optimal detection of pedestrians in the actual video surveillance scene.

[0109]In addition, the embodiment of the present applica...

Embodiment 3

[0128] see Figure 12 , Figure 12 is another schematic structural diagram of the target detection device provided in the embodiment of the present application. In the embodiment of the present application, the object detection device 21 may include: an object candidate area extraction unit 211 , an image segmentation unit 213 , a component relationship learning unit 214 and an object prediction unit 215 .

[0129] The target candidate area extracting unit 211 is configured to acquire a target candidate area in the image to be detected with the target as the detection object.

[0130] The image segmentation unit 213 is configured to determine at least two component candidate areas from the target candidate areas extracted by the target candidate area extraction unit 211 through the image segmentation network, each component candidate area corresponds to a component of the target to be detected, and from the target candidate area to be detected The local image features corres...

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 target detection method and device. The method comprises the steps of obtaining a target candidate area in a to-be-detected image; determining at least two component candidateregions from the target candidate regions through an image segmentation network, each component candidate region corresponding to one component of the to-be-detected target, and extracting the localimage features corresponding to each component candidate region from the to-be-detected image; learning each local image feature of each part candidate region through a bidirectional long short time memory network (LSTM) to obtain a part relation feature for describing a mutual relation between the part candidate regions; and detecting the to-be-detected target in the to-be-detected image according to the component relationship characteristics. According to the invention, the image data processing precision of target detection can be improved, the diversity of application scenarios of target detection is enhanced, and the accuracy of the target detection is improved.

Description

technical field [0001] The embodiments of the present application relate to the field of big data, and in particular to a method and device for object detection. Background technique [0002] In the context of the era of building a safe city, image search has become one of the important technical means to assist public security organizations (such as the people's police) to quickly locate the crime location and action trajectory of the target (such as a criminal suspect). Image search is to use a query image containing a certain target to find the image data containing the target from a large amount of surveillance video data, and determine the appearance of the target in the surveillance video according to the image data containing the target. information such as time and place to determine the target's action trajectory. Searching images by image includes two processes of establishing the target database and target query. In the process of establishing the target databas...

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): G06K9/00G06K9/46G06K9/62G06V40/10G06V10/50G06V10/764G06V10/776
CPCG06T7/215G06T7/246G06T2207/10016G06T2207/20021G06T2207/20081G06T2207/20084G06V40/10G06V10/44G06F18/243G06F18/253G06V10/50G06V10/40G06V10/764G06V10/776Y02D10/00G06T7/11G06T2207/30196G06F18/217
Inventor 杨怡蒋宇豪陈茂林杨双
Owner HUAWEI TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com