IP Library › Granted Patent US 10,728,450
Granted Patent B2
US 10,728,450 · App. 14/859,146 · Granted Jul 28, 2020

Event based computer vision computation

Inventors: Alok Govil (Fremont, CA); Soo Youn Kim (San Diego, CA); Evgeni Petrovich Gousev (Saratoga, CA)
Assignee: QUALCOMM Incorporated
H04N5/23241G06F3/0425G06K9/00986H01L27/14643H04N5/378H04N5/3745G06K2009/4666
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 10,728,450
App. No.
14/859,146
Granted
Jul 28, 2020
Kind
B2
Abstract

Certain techniques are provided for triggering events based on detecting changes in CV features, such as edges, corners etc., by generating computed results based on sensor readings. Yet other techniques also provided for detecting CV features once an event is detecting based on a change in a sensor reading for a sensor element. In certain aspects, the event detection logic and the feature calculation CV vision operations may be performed, individually, or in combination in circuitry on a sensor or in software/firmware on a computing device.

Claims (42)

1. A method, comprising:

comparing a previous localized computer vision (CV) feature associated with a subject sensor element of a sensor element array computed using CV feature computation hardware and a current localized CV feature associated with the subject sensor element computed using the CV feature computation hardware, wherein, for both the previous localized CV feature and the current localized CV feature, the CV feature computation hardware computes the respective CV feature using a respective reading from the subject sensor element and a respective reading from a neighboring sensor element to the subject sensor element, and wherein:

the sensor element array comprises a plurality of sensor elements, the plurality of sensor elements arranged along at least a first dimension and a second dimension of the sensor element array,

a pixel comprises the subject sensor element and in-pixel circuitry, and

the CV feature computation hardware is implemented as part of the in-pixel circuitry;

generating, based on the comparing, an event associated with the subject sensor element, the event comprising a location of the subject sensor element and a feature payload comprising a plurality of bits representing the computed current localized CV feature associated with the subject sensor element; and

sending the event to a processor.

2. The method of claim 1 , wherein the localized CV feature comprises a local binary pattern (LBP) label or a variation of the LBP label.

3. The method of claim 1 , wherein the localized CV feature comprises a histogram of signed gradient (HSG) label.

4. The method of claim 1 , wherein the sensor element array is a vision sensor and each of the plurality of sensor elements of the sensor element array comprises at least one photodiode.

5. The method of claim 1 , wherein the event is generated using dynamic vision sensor (DVS) circuitry.

6. A vision sensor comprising:

a sensor element array comprising a plurality of sensor elements, the plurality of sensor elements arranged along at least a first dimension and a second dimension of the sensor element array, each of the plurality of sensor elements capable of generating sensor readings based on light incident upon the sensor elements;

a computer vision (CV) feature computation hardware configured to compute a previous localized CV feature associated with a subject sensor element of the sensor element array and a current localized CV feature associated with the subject sensor element, wherein, for both the previous localized CV feature and the current localized CV feature, the CV feature computation hardware computes the respective CV feature using a respective reading from the subject sensor element and a respective reading from a neighboring sensor element to the subject sensor element, and wherein:

a pixel comprises the subject sensor element and in-pixel circuitry, and

the CV feature computation hardware is implemented as part of the in-pixel circuitry;

a comparator configured to compare the previous localized CV feature associated with the subject sensor element and the current localized CV feature associated with the subject sensor element; and

the CV feature computation hardware further configured to generate, based on the comparison by the comparator, an event associated with the subject sensor element, the event comprising a location of the subject sensor element and a feature payload comprising a plurality of bits representing the computed current localized CV feature associated with the subject sensor element, and send the event to a processor.

7. The vision sensor of claim 6 , wherein the localized CV feature comprises a local binary pattern (LBP) label or a variation of the LBP label.

8. The vision sensor of claim 6 , wherein the localized CV feature comprises a histogram of signed gradient (HSG) label.

9. The vision sensor of claim 6 , wherein each of the plurality of sensor elements of the sensor element array comprises at least one photodiode.

10. The vision sensor of claim 6 , wherein the event is generated using dynamic vision sensor (DVS) circuitry.

11. An apparatus, comprising:

means for comparing a previous localized computer vision (CV) feature associated with a subject sensor element of a sensor element array computed using CV feature computation hardware and a current localized CV feature associated with the subject sensor element computed using the CV feature computation hardware, wherein, for both the previous localized CV feature and the current localized CV feature, the CV feature computation hardware computes the respective CV feature using a respective reading from the subject sensor element and a respective reading from a neighboring sensor element to the subject sensor element, and wherein:

the sensor element array comprises a plurality of sensor elements, the plurality of sensor elements arranged along at least a first dimension and a second dimension of the sensor element array,

a pixel comprises the subject sensor element and in-pixel circuitry, and

the CV feature computation hardware is implemented as part of the in-pixel circuitry;

means for generating, based on the comparing, an event associated with the subject sensor element, the event comprising a location of the subject sensor element and a feature payload comprising a plurality of bits representing the computed current localized CV feature associated with the subject sensor element; and

means for sending the event to a processor.

12. The apparatus of claim 11 , wherein the localized CV feature comprises a local binary pattern (LBP) label or a variation of the LBP label.

13. The apparatus of claim 11 , wherein the localized CV feature comprises a histogram of signed gradient (HSG) label.

14. The apparatus of claim 11 , wherein the sensor element array is a vision sensor and each of the plurality of sensor elements of the sensor element array comprises at least one photodiode.

15. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium comprises instructions executable by a dedicated computer vision (CV) processing module for:

comparing a previous localized CV feature associated with a subject sensor element of a sensor element array computed using CV feature computation hardware and a current localized CV feature associated with the subject sensor element computed using the CV feature computation hardware, wherein, for both the previous localized CV feature and the current localized CV feature, the CV feature computation hardware computes the respective CV feature using a respective reading from the subject sensor element and a respective reading from a neighboring sensor element to the subject sensor element, and wherein:

the sensor element array comprises a plurality of sensor elements, the plurality of sensor elements arranged along at least a first dimension and a second dimension of the sensor element array,

a pixel comprises the subject sensor element and in-pixel circuitry, and

the CV feature computation hardware is implemented as part of the in-pixel circuitry;

generating, based on the comparing, an event associated with the subject sensor element, the event comprising a location of the subject sensor element and a feature payload comprising a plurality of bits representing the computed current localized CV feature associated with the subject sensor element; and

sending the event to a processor.

16. The non-transitory computer-readable storage medium of claim 15 , wherein the localized CV feature comprises a local binary pattern (LBP) label or a variation of the LBP label.

17. The non-transitory computer-readable storage medium of claim 15 , wherein the localized CV feature comprises a histogram of signed gradient (HSG) label.

18. The non-transitory computer-readable storage medium of claim 15 , wherein the sensor element array is a vision sensor and each of the plurality of sensor elements of the sensor element array comprises at least one photodiode.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 17, 2015
From: GOVIL, ALOK; KIM, SOO YOUN; GOUSEV, EVGENI PETROVICH
To: QUALCOMM INCORPORATED
Reel/Frame 037319/0086 →
Continuity (2)
Provisional Application 62057927 · Sep 30, 2014
Related Publication 20160094787A1 · Mar 31, 2016
Cited By (1)
US 12,445,744