Procedure and device for detecting the movement of a target and their applications

• US 5,883,969 A
• Filed: 05/24/1996
• Issued: 03/16/1999
• Est. Priority Date: 06/01/1995
• Status: Expired due to Fees

• Current Assignee: Aerospatiale Societe Nationale Industrielle
• Original Assignee: Aerospatiale Societe Nationale Industrielle
• Inventors: Schlossers, Christophe, Le Gouzouguec, Anne
• Primary Examiner(s): Mancuso, Joseph
• Assistant Examiner(s): PATEL, JAYANTI K
• Application Number:
  US08/653,332
• Time in Patent Office: 1,026 Days
• Field of Search: 382/103, 382/236, 348/169, 348/699, 348/700, 348/701
• US Class Current: 382/103
• CPC Class Codes: G01S 3/7865 using correlation of the live video image with a stored image
• CPC Class Codes: G06T 7/254 involving subtraction of images
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes

Abstract

A device and a process for detecting the movement of a target observed in the field of view of an optical detector in which a differential image is determined from two successive pictures of the field of view, a binary image is formed from the differential image and from a binary conversion threshold, moving zones in the binary image are individualized, individual envelopes surrounding the zones are determined, an overall envelope surrounding the individual envelopes is determined, and it is determined whether there is movement of the target from the individual envelopes and the overall envelope.

Claims

1. A process for detecting the movement of a target observed in the field of view of an optical detector from two successive pictures of said field of view, each of said pictures being formed by a matrix of pixels, said process comprising the steps of:

a) determining a differential image from two identical size analysis images corresponding respectively to at least part of said successive pictures, said differential image corresponding to a matrix of pixels, and the value of each of the pixels of said matrix corresponding to the absolute value of the difference in the gray levels of the pixels located at the same position respectively in said two analysis images;

b) forming a binary image from said differential image and from a defined binary conversion threshold, the pixels of said binary image taking a first value if luminance values of the corresponding pixels of the differential image are greater than or equal to said binary conversion threshold, and a second value if not;

c) individualizing a plurality of zones on said binary image, the pixels of said zones having said first value;

d) forming an overall envelope of defined shape surrounding said individual envelopes;

e) detecting whether or not movement of said target has occurred based on the whether said overall envelope has an area which is less than a predefined area; and

f) if movement of said target has been detected, determining a characteristic outline corresponding to said movement of said target, said characteristic outline relating to the position of said target in a successive one of said pictures.

2. The process as claimed in claim 1, wherein the direction of movement of said target in said pictures is known and wherein said step f) comprises, for each of the lines of said overall envelope, the steps of:

detecting the presence of a configuration formed by two groups of the same number of adjacent pixels having said first value and an intermediate zone of pixels having said second value, said intermediate zone being located between said groups; and

if said configuration exists, determining a characteristic segment formed by said intermediate zone and by the group located downstream in the direction of movement of said target; and

grouping together a plurality of said characteristic segments to obtain said characteristic outline.

3. The process as claimed in claim 1, wherein said overall envelope has a rectangular shape.

4. The process as claimed in claim 1, additionally comprising the step of determining said binary conversion threshold, wherein said step of determining said binary conversion threshold comprises the following steps:

determining a luminance value of all the pixels of said differential image;

forming a histogram representing the number of pixels per increasing luminance value;

determining an upper luminance value, corresponding to a luminance value in said histogram, from a first percentage of the total number of said pixels;

determining a lower luminance value, corresponding to a luminance value in said histogram, from a second percentage of the total number of said pixels, said second percentage being less than said first percentage; and

computing said binary conversion threshold as the average of said upper luminance value and said lower luminance value.

5. The process as claimed in claim 1, wherein the detection is made from at least three successive pictures and wherein at least two successive binary images are formed from said pictures, additionally comprising the step of temporally filtering said binary image by employing a logic OR between said binary image and the preceding binary image.

6. The process as claimed in claim 1, wherein a reference pattern of said target is formed by the steps of:

identifying the position of said characteristic outline in said analysis image of the last of said pictures; and

extracting the pixels corresponding to said position, the pixels thus extracted forming said reference pattern of the target.

7. A process of determining the location of a target observed in the field of view of an optical detector, said method being based on the steps set forth in claim 1 and said method comprising the steps of:

taking a picture of the field of view of said optical detector;

determining a model of the background, including at least one partial image of the background;

determining a model of said target, including at least one reference pattern representing at least one partial image of the target; and

repeating the following steps;

taking a current picture of the field of view;

isolating at least one background search window from said current picture;

seeking the partial image of the background in said isolated search window;

computing the movement of the optical detector between the preceding picture and said current picture from the position of this partial image in said search window;

isolating a target search window from the current picture, taking into account the computed movement of the optical detector and of the position of the target in the preceding picture;

seeking the reference pattern of the target in said isolated search window;

computing the movement of the target with respect to the background from the position of this reference pattern in said search window;

comparing said computed movement of the target with two, minimum and maximum, movement values determined from the kinematic characteristics of the target so that, if the computed movement lies between said two values, the location of the target obtained by this computed movement is validated, and if not an estimated position of the target is determined from the kinematic characteristics of the latter; and

conditionally updating the background and target models, wherein the process specified in claim 1 is performed for each picture and the target model is updated, at least partially, when a movement of the target is detected.

8. A process of determining the location of a target observed in the field of view of an optical detector, said method being based on the steps set forth in claim 6 and said method comprising the steps of:

taking a picture of the field of view of said optical detector;

determining a model of the background, including at least one partial image of the background;

determining a model of said target, including at least one reference pattern representing at least one partial image of the target; and

repeating the following steps;

taking a current picture of the field of view;

isolating at least one background search window from said current picture;

seeking the partial image of the background in said isolated search window;

computing the movement of the optical detector between the preceding picture and said current picture from the position of this partial image in said search window;

isolating a target search window from the current picture, taking into account the computed movement of the optical detector and of the position of the target in the preceding picture;

seeking the reference pattern of the target in said isolated search window;

computing the movement of the target with respect to the background from the position of this reference pattern in said search window;

comparing said computed movement of the target with two, minimum and maximum, movement values determined from the kinematic characteristics of the target so that, if the computed movement lies between said two values, the location of the target obtained by this computed movement is validated, and if not an estimated position of the target is determined from the kinematic characteristics of the latter; and

conditionally updating the background and target models, wherein the process specified in claim 6 is performed for each picture and the target model is updated, at least partially, when a movement of the target is detected.

9. A device for detecting the movement of a target observed in the field of view of an optical detector from two successive pictures of said field of view, each of said pictures being formed by a matrix of pixels, said device comprising:

means for determining a differential image from two identical size analysis images corresponding respectively to at least part of said successive pictures, said differential image corresponding to a matrix of pixels, and the value of each of the pixels of said matrix corresponding to the absolute value of the difference in the gray levels of the pixels located at the same position respectively in said two analysis images;

means for forming a binary image from said differential image and from a defined binary conversion threshold, the pixels of said binary image taking a first value if luminance values of the corresponding pixels of the differential image are greater than or equal to said binary conversion threshold, and a second value if not;

means for individualizing a plurality of zones on said binary image, the pixels of said zones having said first value;

means for determining an individual envelope of defined shape surrounding each of said individualized zones;

means for detecting whether or not movement of said target has occurred based on the number of said individual envelopes; and

means for determining a characteristic outline corresponding to said movement of said target if movement of said target has been detected, said characteristic outline relating to the position of said target in a successive one of said pictures.

10. The device as claimed in claim 9, additionally comprising means for selecting said individualized zones having an area which is greater than a predefined area and using only said selected individualized zones.

11. The device as claimed in claim 9, additionally comprising means for determining said binary conversion threshold.

12. The device as claimed in claim 9, additionally comprising means for determining a reference pattern of said target.