JP2014044525A5

JP2014044525A5 -

Info

Publication number: JP2014044525A5
Application number: JP2012185762A
Authority: JP
Filing date: 2012-08-24
Publication date: 2015-10-08
Anticipated expiration: 2032-08-24

Description

本発明の一側面は、時系列的に供給された画像に含まれる被写体を認識する被写体認識装置に係り、第１の特定の被写体を認識するための特徴データを含む辞書データと、前記第１の特定の被写体以外の第２の特定の被写体を認識するための特徴データを、前記辞書データに移行する前に一時的に含む仮辞書データと、を保持する辞書データベースと、前記画像の一連のフレームの各々において被写体を検出する検出手段と、前記検出手段により検出された被写体ごとに、前記検出手段による検出の結果に基づいて、当該被写体を前記画像の一連のフレームにわたり追跡する追跡手段と、前記検出手段により検出された被写体ごとに、当該被写体の特徴データを抽出し、該抽出された特徴データを前記辞書データベースに保持された特徴データと比較することで当該被写体の認識を行う認識手段と、前記抽出された特徴データを、前記辞書データベースにおける前記辞書データ又は前記仮辞書データに登録する登録手段とを有し、前記登録手段は、前記認識手段により前記辞書データ及び前記仮辞書データを用いて認識が得られなかった被写体については、前記追跡手段の追跡結果に基づいて求められる重要度が所定値以上の場合には、当該被写体の特徴データを前記仮辞書データに登録し、前記認識手段は、前記抽出された特徴データを前記仮辞書データに登録されている特徴データと比較して、前記抽出された特徴データに係る被写体と前記仮辞書データにおける特徴データに係る被写体との類似度を示す認識評価値を算出し、前記認識評価値に基づいて、当該被写体の特徴データを前記辞書データに移行することを特徴とする。 One aspect of the present invention relates to a subject recognition apparatus for recognizing a subject included in an image supplied in time series, dictionary data including feature data for recognizing a first specific subject, and the first A dictionary database that holds temporary dictionary data temporarily including feature data for recognizing a second specific subject other than the specific subject before the transition to the dictionary data, and a series of images Detection means for detecting a subject in each of the frames, and tracking means for tracking the subject over a series of frames of the image, for each subject detected by the detection means, based on the detection result by the detection means; For each subject detected by the detection means, feature data of the subject is extracted, and the extracted feature data is stored in the feature data stored in the dictionary database. Recognition means for recognizing the subject by comparing with, and registration means for registering the extracted feature data in the dictionary data or the temporary dictionary data in the dictionary database, wherein for said dictionary data and the object recognition using the temporary dictionary data is not obtained by the recognition means, when importance degree that is determined based on the tracking result of the tracking means is above a predetermined value, the The feature data of the subject is registered in the temporary dictionary data, and the recognition unit compares the extracted feature data with the feature data registered in the temporary dictionary data, and the subject related to the extracted feature data. And a recognition evaluation value indicating the degree of similarity between the subject and the subject related to the feature data in the temporary dictionary data, and based on the recognition evaluation value, Wherein the migrating data in the dictionary data.

Claims

A subject recognition device for recognizing a subject included in an image supplied in time series,
Before migrating dictionary data including feature data for recognizing a first specific subject and feature data for recognizing a second specific subject other than the first specific subject to the dictionary data A temporary dictionary data temporarily including a dictionary database for holding,
Detecting means for detecting a subject in each of a series of frames of the image;
For each subject detected by the detection means, tracking means for tracking the subject over a series of frames of the image based on the detection result by the detection means;
Recognizing means for recognizing the subject by extracting feature data of the subject for each subject detected by the detecting means and comparing the extracted feature data with feature data held in the dictionary database; ,
Registration means for registering the extracted feature data in the dictionary data or the temporary dictionary data in the dictionary database;
Have
The registration means includes
Wherein for said dictionary data and the object recognition using the temporary dictionary data is not obtained by the recognition means, when importance degree that is determined based on the tracking result of the tracking means is above a predetermined value, the Register feature data of the subject in the temporary dictionary data,
The recognition means is
Recognizing the degree of similarity between the subject related to the extracted feature data and the subject related to the feature data in the temporary dictionary data by comparing the extracted feature data with the feature data registered in the temporary dictionary data Calculate the evaluation value,
Based on the recognition evaluation value, the subject feature data is transferred to the dictionary data.

The recognizing unit compares the extracted feature data with the feature data registered in the dictionary data, and determines the similarity between the subject related to the extracted feature data and the subject related to the feature data in the dictionary data. The subject recognition apparatus according to claim 1, wherein when the similarity is greater than or equal to a threshold value, the subject related to the extracted feature data is recognized as a subject registered in the dictionary data.

The recognizing means, when the subject related to the extracted feature data is not recognized as a subject registered in the dictionary data, the extracted feature data is registered in the temporary dictionary data; The subject recognition apparatus according to claim 2, wherein the recognition evaluation value is calculated by comparison.

Before Kikasane iodide is object recognition apparatus according to any one of claims 1 to 3, wherein the object has a value representing a duration which is tracked by the tracking means.

The detecting means selects a main subject from a plurality of detected subjects,
Before Kikasane iodide is characterized in that the duration is tracked by the tracking means as main object, or a value indicating a selected number of times as the main subject while being tracked by the tracking means The subject recognition apparatus according to any one of claims 1 to 3.

In the recognition means, at least one of the number of times the recognition evaluation value exceeds the threshold, the frequency at which the recognition evaluation value exceeds the threshold within a predetermined time, and the average value of the recognition evaluation values is equal to or greater than a predetermined value. 6. The subject recognition apparatus according to claim 1, wherein the feature data of the subject is transferred to the dictionary data.

The registration unit does not register the subject in the temporary dictionary data when the number of subjects detected in the series of frames of the image by the detection unit is a predetermined number or more. The subject recognition apparatus according to any one of claims 1 to 6 .

It said registration means, when said period of recognition evaluation value is not updated is equal to or greater than a predetermined value, according to any one of claims 1 to 7, characterized in that to remove the temporary dictionary data relating to the subject Subject recognition device.

Imaging means for supplying the captured images in time series;
The subject recognition apparatus according to any one of claims 1 to 8 , wherein a subject included in an image supplied in time series by the imaging unit is recognized.
Control means for controlling an imaging condition when the imaging means captures an image according to subject information by the subject recognition device;
An imaging device comprising:

Display means for displaying images supplied in time series;
A subject recognition device according to any one of claims 1 to 8 ,
Control means for controlling display conditions when the display means displays an image in accordance with subject information by the subject recognition device;
A display device comprising:

Dictionary data including feature data for recognizing a first specific subject and feature data for recognizing a second specific subject other than the first specific subject are transferred to the dictionary data. A temporary dictionary data temporarily included in a dictionary database that holds a subject, and a method for controlling a subject recognition device that recognizes a subject included in an image supplied in time series,
A detecting step for detecting a subject in each of the series of frames of the image;
For each subject detected in the detection step, the tracking means tracks the subject over a series of frames of the image based on the detection result in the detection step;
Recognizing means extracts feature data of the subject for each subject detected in the detection step, and compares the extracted feature data with the feature data held in the dictionary database to recognize the subject. Recognition steps to perform,
A registration step in which the registration means registers the extracted feature data in the dictionary data or the temporary dictionary data in the dictionary database;
Have
In the registration step,
Wherein said dictionary data and said recognition step for object recognition using the temporary dictionary data is not obtained, if importance degree that is determined based on the tracking result of the tracking means is above a predetermined value, the Register feature data of the subject in the temporary dictionary data,
In the recognition step,
Recognizing the degree of similarity between the subject related to the extracted feature data and the subject related to the feature data in the temporary dictionary data by comparing the extracted feature data with the feature data registered in the temporary dictionary data Calculate the evaluation value,
Based on the recognition evaluation value, the feature data of the subject is transferred to the dictionary data.

Program for executing the steps of the control method of an object recognition apparatus according to a computer to claim 1 1.