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

A target detection method based on discriminative area mining

A target detection and discriminative technology, applied in the field of computer vision, which can solve the problems of detector foreground target and background area interference, etc.

Active Publication Date: 2019-06-28
SUN YAT SEN UNIV
View PDF7 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the technical defect that the detector used in the above-mentioned existing image target detection algorithm is easily disturbed by the apparently similar foreground target and background area, the present invention provides a An Object Detection Method Based on Discriminative Region Mining

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 target detection method based on discriminative area mining
  • A target detection method based on discriminative area mining
  • A target detection method based on discriminative area mining

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0062] Such as figure 1 As shown, an object detection method based on discriminative region mining includes the following steps:

[0063] S1: Perform feature extraction through the feature extraction network to obtain the feature stream s 1 , s 2 ;

[0064] S2: Construct a local discriminative region mining LDRM module, the feature stream s 1 , s 2 Perform local discriminative feature learning;

[0065] S3: Construct a context-based discriminative region mining CDRM module, and learn the feature stream s after local discriminative features 1 , s 2 Carry out context discriminant feature learning;

[0066] S4: Construct feature stream s 3 , using the detector to pair the feature stream s 3 The feature map is detected and the final detection result is obtained.

[0067] More specifically, such as figure 2 As shown, the step S1 includes the following steps:

[0068] S11: Input an image with an input size of 300×300 and its ×2 upsampled image into the same feature extr...

Embodiment 2

[0109] More specifically, on the basis of Example 1, in order to reflect the technical effect of the present invention, experiments were carried out on two data sets of PASCAL VOC 2007 [9] and MS COCO [10], and compared with other advanced target detection methods .

[0110] In the specific implementation process, the input image of the experiment has two resolutions, which are 300×300 and 512×512. In the experiment, ResNet-18 and ResNet-34 pre-trained by the ImageNet dataset were selected as the feature extraction network.

[0111]In the specific implementation process, this experiment is trained on the trainval datasets of PASCAL VOC 2007 and VOC 2012, and tested on the PASCAL VOC 2007test dataset, and the feature extraction network is ResNet-18. During training, the batch size is set to 16 and the number of iterations is 120k. The initial learning rate is set to 1×10 -3 , when the number of iv iterations is 80k and 100k, the learning rate is adjusted to 1×10 -4 and 1×10...

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 based on discriminative area mining, and the method comprises the steps: carrying out the feature extraction through a feature extraction network, andobtaining a feature flow; constructing an LDRM module, and performing local discriminative feature learning on the feature flow; constructing a CDRM module, and carrying out context discriminant feature learning on the feature flow subjected to the local discriminant feature learning; constructing a characteristic flow s3, and detecting the characteristic pattern of the characteristic flow s3 by using a detector to obtain a final detection result. The invention provides a target detection method based on discriminative area mining. According to the method, the characteristic expression of thediscriminative area is generated based on the characteristic graph of the receptive field, so that a large amount of calculation amount introduced by extracting the discriminative area from the original image and then performing characteristic extraction is avoided, the method is ensured to perform target detection with high efficiency, and interference of a foreground target and a background areawhich are similar in appearance is prevented; and by fusing the generated discriminative region features with the features of the candidate target, and by combining the features of different receptive fields, the feature expression is optimized.

Description

technical field [0001] The present invention relates to the technical field of computer vision, and more specifically, to a target detection method based on discriminative region mining. Background technique [0002] As an important means of security monitoring, video surveillance is of great significance to safeguarding the safety of people's lives and properties, and combating crimes. At present, video surveillance is spread all over the corners of the city, and the existing video surveillance systems are usually only used to record, store and retrieve video images, and cannot predict and alarm abnormal situations. In order to realize real-time monitoring, staff are required to analyze the video data. With the expansion of data scale, it is difficult for manual analysis to maintain high accuracy and processing efficiency. Therefore, people hope that the computer can automatically analyze the video and complete the preset visual tasks, such as target recognition, target d...

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
IPC IPC(8): G06K9/46G06K9/62G06N3/04
Inventor 郑慧诚陈绿然严志伟李烨
Owner SUN YAT SEN UNIV
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