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

Object detection device, object detection method, and program

A technology of object detection and detection algorithm, which is applied in biometrics recognition, image data processing, instruments, etc., can solve the problem that the deep learning algorithm is not easy, and achieve the effect of high-speed and high-precision object detection

Active Publication Date: 2020-01-03
OMRON CORP
View PDF20 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Even considering technological innovations in recent years, implementing deep learning algorithms in embedded devices is not easy

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
  • Object detection device, object detection method, and program
  • Object detection device, object detection method, and program
  • Object detection device, object detection method, and program

Examples

Experimental program
Comparison scheme
Effect test

no. 1 approach

[0045] Embodiments described below relate to a face detection device (object detection device) that is mounted on a portable information terminal such as a smartphone and detects a face from an image. However, this is just an example, and the object detected from the image does not necessarily have to be a face, and may be any object. In addition, the face detection device (object detection device) may be mounted on any information processing device (computer) other than the portable information terminal.

[0046] The face detection device of the present embodiment includes a first face detection unit using Haar-like feature values ​​and adaboost, and a second face detection unit using deep learning. However, if deep learning-type face detection is performed on the entire input image, the processing time becomes very long. Therefore, the face detection device of this embodiment performs detection based on the first face detection unit on the entire input image to determine ar...

no. 2 approach

[0095] In the first embodiment, a case where face detection is performed on one image has been described. In this embodiment, face detection is continuously performed on a plurality of frame images constituting a moving image. The face detection processing for each frame image is basically the same as that of the first embodiment, but the integration region creation processing ( figure 2 Step S3) of different. Since the same object exists at the same position in consecutive frame images, in this embodiment, when determining the integration area, information about the current frame and information about past frames are taken into consideration.

[0096] Several methods are conceivable as a method of determining a specific integration area. For example, a method is assumed in which an integration area (hereinafter referred to as a tentative integration area) is determined by the same method as in the first embodiment for the current frame and the past frame (the latest frame)...

no. 3 approach

[0101] In the first embodiment, the second detection unit 114 performs face detection in consideration of only the image information (pixel values) of the integration area. The second detection unit 114 of the present embodiment also receives the detection result of the first detection unit 112 related to the integration area as an input, and performs face detection processing. The “detection result of the first detection unit 112 related to the integrated area” includes face orientation, age, gender, race, expression, and the like.

[0102] In order to enable the second detection unit 114 to perform such recognition, in the learning process (generation process) of the second detection unit 114, the image data including the image data and the detection results (face orientation, age, etc.) of the image data obtained by the first detection unit 112 are used. , gender, race, expression, etc.) learning data.

[0103] In this way, by performing detection processing using the dete...

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

An object detection device detects a predetermined object from an image. The object detection device is provided with: a first detection means which detects from the image a plurality of candidate regions in which the object may exist; a region integration means which determines one or a plurality of integrated regions on the basis of the plurality of candidate regions detected by the first detection means; and a second detection means which, with respect to the integrated regions, detects the object using a detection algorithm different from that of the first detection means. In this way, itbecomes possible to perform faster and more accurate object detection than before.

Description

technical field [0001] The present invention relates to techniques for detecting objects from images. Background technique [0002] Conventionally, a technique for detecting a predetermined object such as a human face from an image is known. For example, methods using Haar-like (Haar-like) feature quantities as image feature quantities and AdaBoost (adaptive boosting algorithm) as a learning algorithm are known (Patent Document 1, Patent Document 2). AdaBoost is a learning / discrimination algorithm for creating a strong classifier by connecting many weak classifiers whose discriminative ability is not so high individually. [0003] In recent years, due to the emergence of deep learning (deep learning), the accuracy of image detection has improved dramatically. However, deep learning uses a multi-layer network called a convolutional neural network (CNN: Convolutional Neural Network), which requires much more computation than previous methods. [0004] As mentioned above, de...

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): G06T7/00G06V10/22
CPCG06T7/00G06V40/166G06V2201/07G06V10/22G06V10/446G06V10/761G06F18/22G06V30/248G06V40/10G06F18/251
Inventor 酒井俊
Owner OMRON CORP
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