METHOD OF ESTIMATING OPTICAL FLOW ON THE BASIS OF AN ASYNCHRONOUS LIGHT SENSOR

• US 20140363049A1
• Filed: 12/21/2012
• Published: 12/11/2014
• Est. Priority Date: 12/21/2011
• Status: Active Grant

• Current Assignee: Centre National De La Recherche Scientifique, Sorbonne Universite
• Original Assignee: Centre National De La Recherche Scientifique, University Pierre et Marie Curie
• Inventors: Benosman, Ryad, Ieng, Siohoi
• Granted Patent: US 9,213,902 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06T 7/269 using gradient-based methods
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69
• CPC Class Codes: H04N 25/00 Circuitry of solid-state image sensors [SSIS]; Control thereof

Abstract

A computer receives asynchronous information originating from a light sensor (10) having a pixel matrix disposed opposite a scene. The asynchronous information comprises, for each pixel of the matrix, successive events originating from this pixel and depending on variations in light in the scene. For a place of estimation (p) in the matrix of pixels and an estimation time (t), the computer selects a set (Sp t) of events originating from pixels included in a spatial neighbourhood (πρ) of the place of estimation and which have occurred in a time interval (Θ) defined with respect to the estimation time, such that this set has at most one event per pixel of the spatial neighbourhood. The computer quantifies the variations in the times of occurrence of the events of the set selected as a function of the positions, in the matrix, of the pixels from which these events originate.

Claims

1. A method for estimating optical flow, comprising:

receiving asynchronous information originating from a light sensor having a pixel matrix arranged facing a scene, the asynchronous information comprising, for each pixel of the matrix, successive events originating from said pixel and depending on light variations in the scene;

for a place of estimation in the pixel matrix and an estimation time, selecting a set of events originating from pixels lying in a spatial neighborhood of the place of estimation and having arisen in a time interval defined with respect to the estimation time, the set having at most one event per pixel of the spatial neighborhood; and

quantizing variations in the times of occurrence of the events of the selected set as a function of positions, in the matrix, of the pixels from which said events originate.

2. The method as claimed in claim 1, wherein the quantization of the variations comprises estimating a slope that the events of the selected set exhibit in a space-time representation around the place of estimation.

3. The method as claimed in claim 2, comprising, before estimating the slope, a smoothing operation, in space-time representation, on the selected set of events.

4. The method as claimed in claim 2, wherein the pixel matrix is two-dimensional and the quantization of the variations comprises determining a plane exhibiting a minimum distance with respect to the events of the selected set in the space-time representation.

5. The method as claimed in claim 2, wherein the pixel matrix is one-dimensional and the quantization of the variations comprises determining a straight line exhibiting a minimum distance with respect to the events of the selected set in space-time representation.

6. The method as claimed in claim 2, wherein the quantization of the variations further comprises estimating second-order derivatives that the events of the selected set exhibit in the space-time representation around the place of estimation.

7. The method as claimed in claim 1, wherein the events originating from pixels of the matrix each have a polarity from among several possible polarities, and wherein the set of events selected for the place of estimation and the estimation time is composed of events of any polarity.

8. The method as claimed in claim 1, wherein, for each pixel of the matrix, the time of occurrence of the most recent event originating from said pixel is memorized, and wherein the selection of the set of events for the place of estimation and the estimation time comprises:

including in said set each event having a time of occurrence in said time interval defined with respect to the estimation time and memorized for a pixel of said spatial neighborhood of the place of estimation.

9. The method as claimed in claim 1, wherein the events originating from the pixels of the matrix each have a polarity from among two possible polarities, and wherein the set of events selected for the place of estimation and the estimation time is composed of events of the same polarity.

10. The method as claimed in claim 9, wherein, for each pixel of the matrix and each possible polarity, the time of occurrence of the most recent event originating from said pixel with said polarity is memorized, and wherein the selection of the set of events for the place of estimation and the estimation time comprises, following receipt of an event having a given polarity at the place of estimation:

including in said set each event having the given polarity and a time of occurrence in the time interval defined with respect to the estimation time and memorized for a pixel of the spatial neighborhood of the place of estimation.

11. The method as claimed in claim 1, wherein said time interval is a period of predefined duration ending at the estimation time.

12. The method as claimed in claim 1, wherein the steps of selecting a set of events and of quantizing the variations in the times of occurrence are executed by taking as place of estimation the position of a pixel where an event is detected and as estimation time an instant of detection of said event.

13. The method as claimed in claim 12, wherein the selection of the set of events is carried out in response to detection of an event originating from a pixel of the matrix at an instant of detection, and wherein the quantization of the variations in the times of occurrence is carried out to update motion information on condition that the set of events contains a number of events above a threshold.

14. A device for estimating optical flow, comprising:

a light sensor having a pixel matrix to be arranged facing a scene and adapted to output asynchronous information comprising, for each pixel of the matrix, successive events originating from said pixel and depending on light variations in the scene; and

a computer to execute the following steps for a place of estimation in the pixel matrix and an estimation time;

selecting a set of events originating from pixels lying in a spatial neighborhood of the place of estimation and having arisen in a time interval defined with respect to the estimation time, the set having at most one event per pixel of the spatial neighborhood; and

quantizing variations in the times of occurrence of the events of the selected set as a function of positions, in the matrix, of the pixels from which said events originate.