Title:
IMAGE RECOGNITION APPARATUS, IMAGE RECOGNITION METHOD, AND RECOGNITION DICTIONARY GENERATING METHOD
Document Type and Number:
Japanese Patent JP2022017871
Kind Code:
A
Abstract:
To enhance accuracy in detection of an object in image recognition processing based on a recognition dictionary.SOLUTION: A recognition processing unit 14 calculates a recognition score indicating possibility of including a predetermined object in a partial region of a captured image. The recognition processing unit 14, if an image size of the partial region is less than a threshold value, treats an image having an image size less than a predetermined value as an input image, and calculates a recognition score of the predetermined object in the partial region by using first recognition dictionary data generated by machine learning using, as an output, a recognition score indicating possibility of including the predetermined object in the input image. The recognition processing unit 14, if the image size of the partial region is a threshold value or larger, treats an image having an image size equal to or larger than the predetermined value as an input image, and calculates a recognition score of the predetermined object in the partial region by using second recognition dictionary data generated by machine learning using, as an output, a recognition score indicating possibility of including the predetermined object in the input image.SELECTED DRAWING: Figure 1
Inventors:
OGURA TAKUYA
Application Number:
JP2020120682A
Publication Date:
January 26, 2022
Filing Date:
July 14, 2020
Export Citation:
Assignee:
JVC KENWOOD CORP
International Classes:
G06T7/00
Attorney, Agent or Firm:
Morishita Kenki
Previous Patent: Excavation work support system
Next Patent: Logical program inference device, logical program inference method, and program
Next Patent: Logical program inference device, logical program inference method, and program