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

Method and apparatus for obtaining object detectors

A technology of object detection and detector, applied in the field of detection, can solve the problems of low detection accuracy and inability to effectively match objects with changing attitudes and viewing angles

Inactive Publication Date: 2014-07-23
NEC (CHINA) CO LTD
View PDF3 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In the process of matching or detection, the above root classifier and part classifier only use the translation of parts to obtain the best match, and cannot effectively match objects with variable poses and viewing angles. Therefore, the above root classifier and part classifiers are used When the object detector composed of classifiers detects dynamic objects, its detection accuracy is not high

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
  • Method and apparatus for obtaining object detectors
  • Method and apparatus for obtaining object detectors
  • Method and apparatus for obtaining object detectors

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0074] This embodiment provides a method for obtaining an object detector, see figure 1 , the method flow provided by this embodiment is specifically as follows:

[0075] 101: Using SVM to train pre-acquired image training samples to obtain a first root classifier, and obtain a first component classifier corresponding to each object component according to the first root classifier;

[0076] 102: Perform model transformation on the first root classifier and the first part classifier corresponding to each object part, respectively, to obtain the transformed root classifier and the transformed part classifier corresponding to each object part, and the transformation includes at least scale transformation, One of rotation transformation and miscut transformation;

[0077] 103: Use the latent variable SVM to train the transformed root classifier and the transformed part classifiers corresponding to each object part, obtain the second root classifier and the second part classifier ...

Embodiment 2

[0080] Embodiment two

[0081] In combination with the content of the first embodiment above, this embodiment provides a method for acquiring an object detector. Wherein, the object detector includes but is not limited to a human body detector, a vehicle detector, or other types of object detection, etc., and this embodiment does not limit the specific type of the object detector. For the convenience of description, this embodiment takes obtaining a human body detector as an example, and describes the manner of obtaining an object detector in detail. see figure 2 , the method flow provided by this embodiment is specifically as follows:

[0082] 201: Using SVM to train pre-acquired image training samples to obtain a first root classifier, and obtain a first component classifier corresponding to each object component according to the first root classifier;

[0083] Wherein, the pre-acquired image training samples may be images collected by photographing equipment or camera e...

Embodiment 3

[0132]This embodiment provides an apparatus for acquiring an object detector, and the apparatus is used to implement the method for acquiring an object detector provided in the first embodiment or the first or second embodiment above. see Figure 7 , the device consists of:

[0133] The first training module 701 is configured to use SVM to train pre-acquired image training samples to obtain a first root classifier;

[0134] The first obtaining module 702 is used to obtain the first part classifier corresponding to each object part according to the first root classifier trained by the first training module 701;

[0135] The model transformation module 703 is used to perform model transformation on the first root classifier trained by the first training module 701 and the first part classifier corresponding to each object part obtained by the first acquisition module 702, respectively, to obtain the transformed root classifier. The classifier and the transformed part classifie...

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 method and apparatus for obtaining object detectors, and belongs to the technical field of detection. The method comprises: according to an SVM, training an image training sample, obtaining a first root filter, and according to a first root filter, obtaining first part filters corresponding to each object part; performing model transformation on the first root filter and each first part filter, and obtaining a root filer after the transformation and part filters after the transformation; according to a latent SVM, training the root filer after the transformation and the part filters after the transformation, obtaining a second root filter and second part filters, and taking the first root filter, the second root filter, the first part filters and the second part filters as the obtained object detectors. According to the invention, the model transformation is carried out on the first root filter and the first part filters, so that the types of filters can be increased, the search space of the object detectors is expanded, objects with changeable postures and visual angles can be effectively matched, and accordingly the object detection precision is improved.

Description

technical field [0001] The invention relates to the technical field of detection, in particular to a method and device for obtaining an object detector. Background technique [0002] With the advancement of electronic information technology and the popularization of networking, people use various image acquisition devices more and more commonly in their daily life, so that they can quickly acquire a large amount of image and video data. By analyzing this data quickly and intelligently, object detection in images can be performed. At present, object detectors are commonly used in object detection methods to detect objects in images, and object detectors are trained on a large number of object images. Therefore, the performance of object detectors will directly affect the performance of object detection. [0003] There are many ways to obtain an object detector in the prior art. A method for obtaining a human detector is mentioned in a paper titled "Object Detection with Disc...

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): G06K9/62
Inventor 王峰
Owner NEC (CHINA) 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