State estimation apparatus, state estimation method, and integrated circuit

Patent metadata

Field	Value
Publication number	US-9805285-B2
Application number	US-201615355532-A
Country	US
Kind code	B2
Filing date	Nov 18, 2016
Priority date	May 22, 2014
Publication date	Oct 31, 2017
Grant date	Oct 31, 2017

Abstract

Provided is a state estimation apparatus that enables more accurate and robust detection and tracking of an object by obtaining a plurality of sets of observation data for a tracking target object and estimating the internal state of the object using a plurality of likelihoods calculated from the obtained sets of observation data. The state estimation apparatus obtains first observation data and second observation data, each of which is composed of a plurality of pieces of observation data, and obtains possibility measurement data and necessity measurement data from the obtained plurality of pieces of observation data. In the state estimation apparatus, a likelihood obtaining unit obtains a first likelihood wp and a second likelihood wn from the possibility measurement data and the necessity measurement data. Using the obtained first likelihood wp and second likelihood wn enables the internal state of the object to be estimated.

First claim

The invention claimed is:

1. A state estimation apparatus for estimating an internal state of an observation target, the apparatus comprising:
first observation obtaining circuitry configured to obtain first observation data from an observable event at predetermined time intervals;
second observation obtaining circuitry configured to obtain second observation data from the observable event at predetermined time intervals;
possibility measurement obtaining circuitry configured to obtain possibility measurement data by performing addition processing or logical OR processing with the first observation data and the second observation data;
necessity measurement obtaining circuitry configured to obtain necessity measurement data by performing multiplication processing or logical AND processing with the first observation data and the second observation data;
prediction circuitry configured to perform prediction using posterior probability distribution data, which is probability distribution, obtained at a preceding time t−1, of an internal state of the observation target to obtain predictive probability distribution data, which is probability distribution of an internal state of the observation target at a current time t;
likelihood obtaining circuitry configured to obtain a first likelihood based on the predictive probability distribution data and the possibility measurement data and to obtain a second likelihood based on the predictive probability distribution data and the necessity measurement data;
posterior probability distribution estimation circuitry configured to estimate, from the first likelihood and the predictive probability distribution data, posterior probability distribution data, which is probability distribution of the internal state of the observation target at the current time t; and
prior probability distribution output circuitry configured to output prior probability distribution data based on the posterior probability distribution data estimated by the posterior probability distribution estimation circuitry as prior probability distribution data at a next time t+1 in order to detect and track the observation target in a moving image.

2. The state estimation apparatus according to claim 1 , wherein
the posterior probability distribution estimation circuitry sets an internal variable of a particle sampled in accordance with the posterior probability distribution at the current time t such that the internal variable includes a value set based on the second likelihood obtained by the likelihood obtaining circuitry.

3. The state estimation apparatus according to claim 2 , wherein
the posterior probability distribution estimation circuitry sets an internal variable of a particle sampled in accordance with the posterior probability distribution at the current time t such that the internal variable includes a value defined by w t|t (i) =wn, where S t|t is the posterior probability distribution data, wn t|t (i) is an internal variable of an i-th particle obtained in accordance with the posterior probability distribution data S t|t , wn is the second likelihood, corresponding to the particle, that is obtained by the likelihood obtaining circuitry.

4. The state estimation apparatus according to claim 2 , wherein
the posterior probability distribution estimation circuitry sets an internal variable of a particle sampled in accordance with the posterior probability distribution at the current time t such that the internal variable includes a value defined by

wn

t
|
t

(
i
)

=

wp

t
|
t

(
i
)

×
wn
⁢

/

⁢
wp

=

(

1
⁢

/

⁢
M
⁢

⁢
1

)

×
wn
⁢

/

⁢
wp

where S t|t is the posterior probability distribution data, wn t|t (i) is an internal variable of an i-th particle obtained in accordance with the posterior probability distribution data S t|t , wp is the first likelihood, corresponding to the particle, that is obtained by the likelihood obtaining circuitry, wn is the second likelihood, corresponding to the particle, that is obtained by the likelihood obtaining circuitry, M1 (M1 is a natural number) is the number of particles sampled in accordance with the posterior probability distribution at the current time t, and wp t|t (i) is an internal variable based on the first likelihood wp of the particle sampled in accordance with the posterior probability distribution at the current time t.

5. The state estimation apparatus according to claim 2 , wherein
the posterior probability distribution estimation circuitry sets an internal variable of a particle sampled in accordance with the posterior probability distribution at the current time t such that the internal variable includes a value defined by

wn

t
|
t

(
i
)

=

wp

t
|
t

(
i
)

-

(

wp
-
wn

)

Assignees

• Megachips Corp
• Kyushu Inst Technology

Inventors

• Ikoma Norikazu
• Hasegawa Hiromu

Classifications

• G06T7/277 Primary

  involving stochastic approaches, e.g. using Kalman filters · CPC title

• G06K9/4671 Primary

  Physics · mapped topic

• G06T2207/10016

  Video; Image sequence · CPC title

• G06T2207/10004

  Still image; Photographic image · CPC title

• G06T2207/20076

  Probabilistic image processing · CPC title