IP Library › Granted Patent US 12,633,133
Granted Patent B2
US 12,633,133 · App. 18/693,718 · Granted May 19, 2026

Recognition processing device, recognition processing method, and recognition processing system

Inventor: Satoshi Nakayama (Kanagawa, JP)
Assignee: SONY SEMICONDUCTOR SOLUTIONS CORPORATION
G06V20/58G06V2201/07
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,633,133
App. No.
18/693,718
Granted
May 19, 2026
Kind
B2
Abstract

The present technology relates to a recognition processing device, a recognition processing method, and a recognition processing system capable of realizing more efficient object recognition. The dictionary data storage unit stores a far dictionary for each pattern to be recognized and a nearby dictionary for each pattern, and the recognition processing unit executes recognition processing of recognizing an object included in a wide-angle image captured by a wide-angle camera using the far dictionary and the nearby dictionary. Further, the recognition processing unit recognizes the object by using a plurality of patterns of the far dictionaries and a nearby dictionary of the same pattern as the far dictionary of a high likelihood pattern in which the object is recognized with high likelihood among the far dictionaries. The present technology can be applied to, for example, an image recognition chip that performs pedestrian recognition on a wide-angle image.

Claims (26)

1 . A recognition processing device, comprising:

a dictionary data storage unit that stores a far dictionary for each pattern of a recognition target and a nearby dictionary for each pattern; and

a recognition processing unit that executes recognition processing of recognizing an object included in a wide-angle image captured by a wide-angle camera by using the far dictionary and the nearby dictionary, wherein

the recognition processing unit recognizes the object by using a plurality of patterns of far dictionaries and the nearby dictionary of the same pattern as the far dictionary of a high likelihood pattern in which the object is recognized with high likelihood among the far dictionaries.

2 . The recognition processing device according to claim 1 , wherein the recognition processing unit repeats recognition of the object by using the plurality of patterns of the far dictionaries in a case where a distance to the recognized object is longer than a certain distance, and recognizes the object by using the nearby dictionary of the high likelihood pattern in a case where the distance to the recognized object is shorter than the certain distance.

3 . The recognition processing device according to claim 1 , wherein

the far dictionary and the nearby dictionary respectively include a specialized far dictionary and a specialized nearby dictionary specialized for object recognition of a specific pattern, and

the recognition processing unit recognizes the object by using a plurality of patterns of the far dictionaries and the specialized nearby dictionary of the same pattern as the specialized far dictionary of the high likelihood pattern.

4 . The recognition processing device according to claim 3 , wherein

the far dictionary and the nearby dictionary each further include a general-purpose far dictionary and a general-purpose nearby dictionary for general-purpose object recognition, and

the recognition processing unit recognizes the object by using the far dictionaries of all patterns including the general-purpose far dictionary and the general-purpose nearby dictionary, and then recognizes the object by using the specialized nearby dictionary of the high likelihood pattern instead of the general-purpose nearby dictionary.

5 . The recognition processing device according to claim 4 , wherein, in a case where a distance to the recognized object is longer than a certain distance, the recognition processing unit repeats the recognition of the object by using the far dictionaries of all patterns and the specialized nearby dictionary of the high likelihood pattern.

6 . The recognition processing device according to claim 5 , wherein, in a case where the distance to the recognized object is shorter than a certain distance, the recognition processing unit recognizes the object using the general-purpose far dictionary and the specialized nearby dictionary of the high likelihood pattern.

7 . The recognition processing device according to claim 1 , wherein the recognition processing unit recognizes the object by using only nearby dictionaries of a plurality of patterns at a start of the recognition processing.

8 . The recognition processing device according to claim 1 , wherein the recognition processing unit sets the far dictionary in which the object having a smallest time to collision (TTC) among the objects recognized with the high likelihood using the far dictionary is recognized as the far dictionary of the high likelihood pattern.

9 . The recognition processing device according to claim 1 , wherein the recognition processing unit executes a series of recognition processing using the far dictionary and the nearby dictionary for one frame of the wide-angle image acquired at a predetermined timing.

10 . The recognition processing device according to claim 1 , wherein the recognition processing unit recognizes the object included in the cylindrically corrected wide-angle image.

11 . The recognition processing device according to claim 1 , wherein the wide-angle camera includes an in-vehicle camera mounted on a vehicle.

12 . The recognition processing device according to claim 11 , wherein the wide-angle camera captures the wide-angle image with a rear of the vehicle as an imaging range.

13 . The recognition processing device according to claim 12 , wherein the recognition processing unit recognizes a pedestrian existing around the vehicle as the object included in the wide-angle image.

14 . A recognition processing method, comprising:

recognizing, by a recognition processing device that executes recognition processing of recognizing an object included in a wide-angle image captured by a wide- angle camera using a far dictionary for each pattern to be recognized and a nearby dictionary for each pattern, the object by using far dictionaries of a plurality of patterns and the nearby dictionary of the same pattern as the far dictionary of a high likelihood pattern in which the object is recognized with high likelihood among the far dictionaries.

15 . A recognition processing system, comprising:

a wide-angle camera; and

a recognition processing device that stores a far dictionary for each pattern to be recognized and a nearby dictionary for each pattern, and executes recognition processing of recognizing an object included in a wide-angle image captured by the wide-angle camera using the far dictionary and the nearby dictionary, wherein

the recognition processing device recognizes the object by using a plurality of patterns of far dictionaries and the nearby dictionary of the same pattern as the far dictionary of a high likelihood pattern in which the object is recognized with high likelihood among the far dictionaries.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Mar 20, 2024
From: NAKAYAMA, SATOSHI
To: SONY SEMICONDUCTOR SOLUTIONS CORPORATION
Reel/Frame 066842/0993 →
Priority Claims (1)
JP 2021-162501 · Oct 1, 2021 · national
Continuity (1)
Related Publication 20240386724A1 · Nov 21, 2024
References Cited (30)
US 20150269448A1 · Kosaki · 2015 [cited by examiner]
US 20150269449A1 · Kosaki · 2015 [cited by examiner]
US 20160086333A1 · Scholl · 2016 [cited by examiner]
US 20210136292A1 · Hoch · 2021 [cited by examiner]
US 20210192680A1 · Masuda · 2021 [cited by examiner]
US 20230077082A1 · Liu · 2023 [cited by examiner]
US 20240386724A1 · Nakayama · 2024 [cited by examiner]
CA 2975139A1 · 2018 [cited by examiner]
CA 2988303C · 2018 [cited by examiner]
CA 3035710A1 · 2019 [cited by examiner]
CA 3149005A1 · 2022 [cited by examiner]
CN 104954664A · 2015 [cited by examiner]
CN 108734697A · 2018 [cited by examiner]
CN 110443142A · 2019 [cited by examiner]
CN 110458137A · 2019 [cited by examiner]
CN 106952308B · 2020 [cited by examiner]
CN 111814720A · 2020 [cited by examiner]
CN 112069980A · 2020 [cited by examiner]
JP 2009009209A · 2009 [cited by applicant]
JP 2014092975A · 2014 [cited by applicant]
JP 2015184908A · 2015 [cited by examiner]
JP 2017044530A · 2017 [cited by applicant]
JP 2020027380A · 2020 [cited by applicant]
KR 20230174730A · 2003 [cited by examiner]
KR 101481648B1 · 2015 [cited by examiner]
WO WO2018203512A1 · 2018 [cited by examiner]
WO WO2019102938A1 · 2019 [cited by examiner]
WO WO2022215409A1 · 2022 [cited by examiner]
WO WO2022230413A1 · 2022 [cited by examiner]
International Search Report and Written Opinion of PCT Application No. PCT/JP2022/035093, issued on Oct. 25, 2022, 09 pages of ISRWO. [cited by applicant]