JPH05205175A - Body detection device - Google Patents
Body detection deviceInfo
- Publication number
- JPH05205175A JPH05205175A JP4010746A JP1074692A JPH05205175A JP H05205175 A JPH05205175 A JP H05205175A JP 4010746 A JP4010746 A JP 4010746A JP 1074692 A JP1074692 A JP 1074692A JP H05205175 A JPH05205175 A JP H05205175A
- Authority
- JP
- Japan
- Prior art keywords
- image
- difference
- binary
- time
- create
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Processing (AREA)
- Emergency Alarm Devices (AREA)
- Image Analysis (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
Description
【０００１】[0001]
【産業上の利用分野】本発明は、所定時間放置してある
物体を検出する装置に係り、特に、通行人のいる公共場
所に置かれた放置物，トンネルや高速道路の停止車及び
路上の駐停止車の検出に好適な物体検出装置に関する。BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for detecting an object left unattended for a predetermined period of time, and more particularly, to an abandoned object placed in a public place with passersby, a stopped vehicle in a tunnel or a highway, and a road. The present invention relates to an object detection device suitable for detecting a parked vehicle.
【０００２】[0002]
【従来の技術】従来は、特開昭62−136988号公報に記載
されるように、まず、予め登録した正常な参照画像と入
力画像の差分画像を求め一定値で２値化して２値画像を
求め、ラベリングを行って物体抽出画像メモリに格納し
ておく。現時点の物体抽出画像メモリと前時点での物体
抽出画像メモリのラベリングした物体抽出画像メモリに
関して、一致，結合，分離，出現，消失の５通りの状態
に同定して、これを、過去ｎ画面追跡し、画面の外側か
ら警戒領域へ進んだ物体を侵入者として検出するように
構成されていた。2. Description of the Related Art Conventionally, as described in Japanese Patent Laid-Open No. 62-136988, first, a difference image between a normal reference image registered in advance and an input image is first obtained and binarized with a constant value to obtain a binary image. Is obtained, labeled and stored in the object extraction image memory. The labeled object extracted image memory of the object extracted image memory at the present time and the object extracted image memory at the previous time is identified into five states of coincidence, combination, separation, appearance, and disappearance, and the past n screen tracking is performed. However, the object that has advanced from the outside of the screen to the warning area is detected as an intruder.
【０００３】[0003]
【発明が解決しようとする課題】上記の従来技術では、
予め登録した正常な参照画像と入力画像の差分画像を求
め一定値で２値化して２値画像を求め、ラベリングを行
い、このラベリングした結果に関して現時点と前時点で
比較して、一致，結合，分離，出現，消失の５通りの状
態に同定し、これを過去ｎ画面追跡し、画面の外側から
警戒領域へ進んだ物体を侵入者として検出する場合、移
動物体と一定時間静止している物体が混在する状態での
静止物体の抽出、特に移動物体の影に重なって見え隠れ
する静止物体の抽出は困難であり、対象領域に物体が存
在していても静止物体無しと検出し、誤検出してしまう
という問題が有った。SUMMARY OF THE INVENTION In the above prior art,
A difference image between a normal reference image and an input image registered in advance is obtained, binarization is performed by a constant value to obtain a binary image, labeling is performed, and the labeling result is compared with the present time and the previous time, and the matching, combining, When the object is identified in five states of separation, appearance, and disappearance, and the past n screens are traced and an object that has advanced from the outside of the screen to the warning area is detected as an intruder, an object that is stationary for a certain period of time with a moving object. It is difficult to extract a stationary object in a mixed state, especially a stationary object that appears and disappears in the shadow of a moving object.Even if an object exists in the target area, it is detected that there is no stationary object, and false detection is performed. There was a problem that it would end up.
【０００４】本発明の目的は、静止物体と移動物体が混
在する状態、特に、移動物体の影に重なって見え隠れす
る状態から、所定時間放置してある駐停止車両，事故車
両，放置物，危険物等の静止物体を良好に検出し、その
存在を報知する物体検出装置を提供することにある。The object of the present invention is to stop and park a vehicle, an accident vehicle, an abandoned object, a dangerous object which has been left for a predetermined time from a state in which a stationary object and a moving object are mixed, in particular, a state in which a stationary object and a moving object overlap each other in the shadow of the moving object. An object of the present invention is to provide an object detection device that detects a stationary object such as an object satisfactorily and notifies its existence.
【０００５】[0005]
【課題を解決するための手段】上記目的を達成するた
め、本発明では、検出対象シーンを撮影するＩＴＶカメ
ラと該ＩＴＶカメラの画像信号を予め取り込んでディジ
タル信号に変換して背景画像とする参照画像格納手段
と、該参照画像と入力画像の差分画像を求め、該差分画
像から２値画像を求めて、少なくとも前記差分画像及び
前記２値画像のうち何れか一方を累積し、一定時間静止
している物体の領域を抽出する物体抽出手段と、該物体
抽出手段により得られた画像から特徴データを求め、前
記検出対象シーンにおける静止物体の有無を判定する物
体判定手段と、該物体判定手段から得られた結果を報知
する報知手段とから物体検出装置を構成したものであ
る。In order to achieve the above object, according to the present invention, an ITV camera for photographing a scene to be detected and an image signal of the ITV camera are previously captured and converted into a digital signal to be used as a background image. An image storage unit, a difference image between the reference image and the input image is obtained, a binary image is obtained from the difference image, at least one of the difference image and the binary image is accumulated, and the image is stationary for a certain period of time. An object extracting means for extracting a region of an object, an object determining means for determining feature data from an image obtained by the object extracting means, and determining the presence or absence of a stationary object in the detection target scene; and the object determining means. The object detecting device is configured by a notifying unit that notifies the obtained result.
【０００６】[0006]
【作用】本発明によれば、参照画像を入力画像に適合し
た濃度の正常画像に自動更新して、該参照画像と入力画
像から、画素毎の差を求めた差分画像が作成され、当該
差分画像と前回までの累積画像を一定時間静止している
物体を抽出しやすいように加算し、該加算を所定時間繰
り返して累積画像を作成し、該累積画像を２値化して検
出するので、静止物体と移動物体が混在する状態でも、
移動物体が除去され静止物体のみが累積される。特に、
移動物体の影に重なって見え隠れする静止物体が検出さ
れるので、人ごみの中や交通量の多い環境でも精度よく
検出できる。更に、２値画像から面積等の特徴データを
抽出し、これらのデータを判断して報知するので、物体
が検出された場合のみ監視対象シーンに注目すればよ
く、常時モニタ等の報知装置を目視しなくても物体検出
が可能となり、監視における長時間のモニタ凝視による
検出率の低下、単調作業からの脱却を図ることができ
る。According to the present invention, the reference image is automatically updated to a normal image having a density suitable for the input image, and a difference image for which a difference for each pixel is obtained is created from the reference image and the input image. The image and the accumulated images up to the previous time are added so that an object that has been stationary for a certain time can be easily extracted, and the addition is repeated for a predetermined time to create an accumulated image, and the accumulated image is binarized and detected. Even when objects and moving objects are mixed,
Moving objects are removed and only stationary objects are accumulated. In particular,
Since a stationary object that is visible and hidden in the shadow of a moving object is detected, it can be detected accurately even in a crowd of people or in an environment with heavy traffic. Furthermore, since feature data such as area is extracted from the binary image and the data is judged and notified, it is only necessary to pay attention to the monitored scene only when an object is detected. It is possible to detect an object without doing so, and it is possible to reduce the detection rate due to a long-time gaze at the monitor during monitoring, and to avoid monotonous work.
【０００７】[0007]
【実施例】以下、本発明の一実施例を図面を用いて説明
する。図１は、本発明の一実施例を示す物体検出装置の
ブロック図である。本実施例では、まず、検出対象シー
ンを撮影するＩＴＶカメラ１０等とＩＴＶカメラ１０等
から得られた画像を参照画像入力部２０が取り込んでデ
ィジタル信号に変換して、予め基準となる正常な参照画
像を参照画像格納メモリ４０に格納しておく。次に、差
分画像累積部５０は、検出対象シーンを撮影しているＩ
ＴＶカメラ１０等から取り込んだ画像と、参照画像入力
部２０が取り込んで参照画像格納メモリ４０に格納した
正常な参照画像から、画素毎の差（差の絶対値）を求め
て差分画像を作成し、一定時間静止している物体を抽出
しやすいように差分画像を所定時間累積（初期画像は全
画素『０』である）して累積画像を作成する。２値化処
理部６５が、差分画像累積部５０で作成した累積画像を
ユーザの与えた所定の閾値で２値化し、ノイズを除去し
て２値画像を作成する。物体判定部１４０は、２値化処
理部６５で作成した２値画像から面積等の特徴データを
抽出して特徴データが所定値以上の場合、静止物体有り
とし、所定値未満の場合、静止物体無しと判断し、判断
結果等を報知部１５０に報知する。報知装置１６０は、
報知部１５０から報知されたデータ又は／及び結果等を
モニタ又は／及びブザー等の報知装置１６０に表示又は
／及び報知して、監視員に知らせる。DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an object detection device showing an embodiment of the present invention. In this embodiment, first, the reference image input unit 20 captures the ITV camera 10 or the like for capturing the detection target scene and the image obtained from the ITV camera 10 or the like, converts the image into a digital signal, and a normal reference serving as a reference in advance. The image is stored in the reference image storage memory 40. Next, the difference image accumulating unit 50 captures an image of the detection target scene I
From the image captured from the TV camera 10 or the like and the normal reference image captured by the reference image input unit 20 and stored in the reference image storage memory 40, a difference for each pixel (absolute value of the difference) is obtained to create a difference image. A difference image is accumulated for a predetermined period of time (the initial image is all pixels “0”) to create an accumulated image so that an object stationary for a certain period of time can be easily extracted. The binarization processing unit 65 binarizes the cumulative image created by the difference image accumulating unit 50 with a predetermined threshold value given by the user, removes noise, and creates a binary image. The object determination unit 140 extracts feature data such as an area from the binary image created by the binarization processing unit 65, and if the feature data is equal to or greater than a predetermined value, the object is a static object. If the feature data is less than the predetermined value, the object is a static object. It is determined that there is none, and the notification result is notified to the notification unit 150. The notification device 160 is
The data or / and the result or the like notified from the notification unit 150 is displayed or / and notified to the monitor or / and the notification device 160 such as a buzzer to notify the monitor.
【０００８】図２は、本発明の他の実施例を示す物体検
出装置のブロック図である。本実施例では、まず、検出
対象シーンを撮影するＩＴＶカメラ１０等とＩＴＶカメ
ラ１０等から得られた画像を参照画像更新部３０が取り
込んでディジタル信号に変換して監視対象に適した濃度
の正常画像に自動更新して、参照画像格納メモリ４０に
格納しておく。参照画像更新部３０の内部の一実施例
は、特開平3− 118677 号公報と同様である。次に、
差分画像累積部５０，２値化処理部６５，物体判定部１
４０，報知部１５０及び報知装置１６０に関しては、図
１と同様である。FIG. 2 is a block diagram of an object detecting apparatus showing another embodiment of the present invention. In the present embodiment, first, the ITV camera 10 or the like for photographing the detection target scene and the image obtained from the ITV camera 10 or the like are captured by the reference image updating unit 30 and converted into a digital signal to obtain a normal density suitable for the monitoring target. The image is automatically updated and stored in the reference image storage memory 40. An example of the inside of the reference image updating unit 30 is the same as in Japanese Patent Laid-Open No. 118677/1993. next,
Difference image accumulation unit 50, binarization processing unit 65, object determination unit 1
About 40, the alerting | reporting part 150, and the alerting | reporting apparatus 160, it is the same as that of FIG.
【０００９】図３は、本発明における差分画像累積部５
０の内部の一実施例を示すブロック図である。差分画像
累積部５２は、画像入力部５１がＩＴＶカメラ１０等か
ら取り込んで平滑化等のノイズ除去を行った入力画像
と、参照画像更新部３０が入力画像に適した濃度に自動
更新した参照画像から、画素毎の差（差の絶対値）を求
めて差分画像を作成して、一定時間静止している物体を
抽出しやすいように該差分画像を累積（初期画像は全画
素『０』である）し、累積判定部５３によりユーザの与
えた累積時間と比較して累積終了と判定するまで累積を
繰り返して、累積画像を作成する。差分画像累積部５２
における差分画像の累積は、一定時間静止している物体
を容易に抽出できるように指数平滑化手法を用いる。即
ち、次の式で表わされる。FIG. 3 shows the difference image accumulating unit 5 according to the present invention.
3 is a block diagram showing an example of the inside of 0. FIG. The difference image accumulating unit 52 includes an input image captured by the image input unit 51 from the ITV camera 10 or the like and subjected to noise removal such as smoothing, and a reference image automatically updated by the reference image updating unit 30 to a density suitable for the input image. Then, a difference image is created by obtaining the difference (absolute value of the difference) for each pixel, and the difference image is accumulated so that an object stationary for a certain time can be easily extracted (the initial image is all pixels “0”). Then, the accumulation determination unit 53 repeats the accumulation until it is compared with the accumulation time given by the user and determines that the accumulation is completed, and an accumulated image is created. Difference image accumulating unit 52
In the accumulation of the difference images in, the exponential smoothing method is used so that an object that is stationary for a certain time can be easily extracted. That is, it is expressed by the following equation.
【００１０】 ｆ(ｔ)＝αｆ(ｔ−１)＋(１−α)ｇ(ｔ) …（数１） （数１）において、ｆ(ｔ)は累積結果画像を、αは平滑
化指数を、ｆ(ｔ−１)は前回までの累積画像を、ｇ
（ｔ）は現在の差分画像を示す。ここで、０≦α≦１で
あり、例えば、通行人の多い場所で一定時間静止してい
る物体を抽出する場合はαを約０.９０〜０.９８に設定
すればよい。差分画像累積部５２が作成した累積画像
を、２値化処理部６５により、所定の閾値で２値化し２
値画像を作成する。F (t) = α f (t−1) + (1-α) g (t) (Equation 1) In (Equation 1), f (t) is a cumulative result image and α is a smoothing index. F (t-1) is the cumulative image up to the previous time, g
(T) shows the current difference image. Here, 0 ≦ α ≦ 1, and for example, when extracting an object that is stationary for a certain time in a place where many passersby pass, α may be set to about 0.90 to 0.98. The binarization processing unit 65 binarizes the cumulative image created by the difference image accumulating unit 52 with a predetermined threshold value and binarizes it.
Create a value image.
【００１１】図４は、本発明における差分画像累積部５
０及び２値化処理部６５で処理を実行する場合の一実施
例を示す画像メモリの状態変化図である。まず、累積画
像の初期状態５７は値『０』が設定されている。差分画
像累積部５２で求めた１回目の差分画像が格納された濃
淡メモリ５４と初期値『０』が格納された濃淡メモリ５
７を（数１）で示すように加算して濃淡メモリに格納す
ると状態５８の累積画像となる。次に、差分画像累積部
５２で求めた２回目の差分画像が格納された濃淡メモリ
５５と累積画像を格納した濃淡メモリ５８を１回目と同
様に（数１）で示すように加算して濃淡メモリに格納す
ると状態５９の累積画像となる。以下、ユーザが設定し
た累積時間又は回数だけ同様に繰り返す。例えば、ユー
ザがｎ回と設定すると、差分画像累積部５２で求めたｎ
回目の差分画像が格納された濃淡メモリ５６と前回まで
の累積画像を格納した濃淡メモリを上記と同様(数１)で
示すように加算して濃淡メモリに格納すると状態６０の
累積画像となり、状態６０の累積画像を２値化処理部６
５で所定の閾値で２値化すると、状態６１の２値メモリ
となる。２値化処理部６５における閾値は、例えば、参
照画像更新部３０が自動更新した参照画像の平均値又は
予め設定した入力画像の背景領域部（移動物体や静止物
体の存在しない領域）の平均値に基づいてユーザが設定
する。この結果、２値メモリ６１は、一定時間静止して
いる物体が抽出された状態となる。FIG. 4 shows the difference image accumulator 5 of the present invention.
FIG. 9 is a state change diagram of an image memory showing an embodiment when processing is executed by the 0 and binarization processing unit 65. First, the initial state 57 of the cumulative image is set to the value "0". A grayscale memory 54 storing the first differential image obtained by the differential image accumulating unit 52 and a grayscale memory 5 storing an initial value “0”.
When 7 is added as shown in (Equation 1) and stored in the grayscale memory, a cumulative image in state 58 is obtained. Next, the grayscale memory 55 storing the second differential image obtained by the differential image accumulating unit 52 and the grayscale memory 58 storing the cumulative image are added as shown in (Equation 1) as in the first time, and the grayscale is added. When stored in the memory, a cumulative image of state 59 is obtained. Hereinafter, the same time is repeated for the cumulative time or the number of times set by the user. For example, if the user sets n times, n calculated by the difference image accumulation unit 52
When the grayscale memory 56 storing the difference image for the second time and the grayscale memory storing the accumulated images up to the previous time are added as shown in (Equation 1) and stored in the grayscale memory, a cumulative image of state 60 is obtained. The binarization processing unit 6 converts 60 accumulated images.
When binarized with a predetermined threshold value in 5, the binary memory of the state 61 is obtained. The threshold value in the binarization processing unit 65 is, for example, the average value of the reference image automatically updated by the reference image updating unit 30 or the average value of the background area portion (area where no moving object or stationary object exists) of the preset input image. Set by the user. As a result, the binary memory 61 is in a state in which an object stationary for a certain period of time is extracted.
【００１２】図５は、本発明の他の実施例を示す物体検
出装置のブロック図である。本実施例におけるＩＴＶカ
メラ１０等、参照画像入力部２０及び参照画像格納メモ
リ４０に関しては、図１と同様である。差分画像２値化
処理部７０は、検出対象シーンを撮影しているＩＴＶカ
メラ１０等から取り込んだ画像と、参照画像入力部２０
が取り込んで参照画像格納メモリ４０に格納した正常な
参照画像から、画素毎の差（差の絶対値）を求めて差分
画像を作成し、差分画像を所定の閾値で２値化して２値
画像を作成し、２値画像のラベリングを行って小面積部
（ユーザが設定する値）除外等の処理を行ってノイズを
除去した２値画像を作成し、該２値画像を用いて共通領
域の抽出を、ユーザが設定した累積時間又は回数だけ繰
り返して、共通領域部を抽出する。物体判定部１４０，
報知部１５０及び報知装置160に関しても、図１と同様
である。FIG. 5 is a block diagram of an object detecting apparatus showing another embodiment of the present invention. The reference image input unit 20 and the reference image storage memory 40 such as the ITV camera 10 in this embodiment are the same as those in FIG. The differential image binarization processing unit 70 includes an image captured from the ITV camera 10 or the like capturing the detection target scene and the reference image input unit 20.
From the normal reference image stored in the reference image storage memory 40, a difference image is created by obtaining the difference (absolute value of the difference) for each pixel, and the difference image is binarized by a predetermined threshold value to obtain a binary image. , A binary image is labeled, and a small image (value set by the user) is excluded to create a binary image from which noise is removed, and the binary image is used to create a binary image of a common area. The extraction is repeated for the cumulative time or the number of times set by the user to extract the common area part. Object determination unit 140,
The notification unit 150 and the notification device 160 are the same as in FIG.
【００１３】図６は、本発明の他の実施例を示す物体検
出装置のブロック図である。本実施例におけるＩＴＶカ
メラ１０等、参照画像格納メモリ４０，差分画像２値化
処理部７０，物体判定部１４０，報知部１５０及び報知
装置１６０に関しては、図５と同様である。参照画像更
新部３０に関しては、図２と同様である。FIG. 6 is a block diagram of an object detecting apparatus showing another embodiment of the present invention. The reference image storage memory 40, the difference image binarization processing unit 70, the object determination unit 140, the notification unit 150, and the notification device 160, such as the ITV camera 10 in this embodiment, are the same as those in FIG. The reference image update unit 30 is the same as that in FIG.
【００１４】図７は、本発明における差分画像２値化処
理部７０の内部の一実施例を示すブロック図である。差
分画像２値化部７２は、画像入力部７１がＩＴＶカメラ
１０等から取り込んで平滑化等のノイズ除去を行った入
力画像と、参照画像更新部３０が入力画像に適した濃度
に自動更新した参照画像から、画素毎の差（差の絶対
値）を求めて差分画像を作成し、該差分画像を所定の閾
値で２値化して２値画像を作成して、２値画像のラベリ
ングを行って小面積部（ユーザが設定する値）除外等の
処理を行ってノイズを除去した２値画像を作成する。共
通領域抽出部７３は、差分画像２値化部７２がノイズを
除去した２値画像と前回までに抽出した共通領域部の画
像（初期画像は全画素『０』又は『１』である）から共
通領域を抽出する。抽出終了判定部７４は、監視対象シ
ーンに応じてユーザが設定した時間又は回数と比較し共
通領域抽出を繰り返すかどうかの判定を行って、ユーザ
の設定値未満の場合、画像入力部７１へ戻り共通領域抽
出を繰り返し、ユーザの設定値以上の場合、一連の処理
を終了して、物体判定部１４０で静止物体の検出判定を
行う。FIG. 7 is a block diagram showing an embodiment of the inside of the differential image binarization processing unit 70 in the present invention. The difference image binarization unit 72 automatically updates the input image captured by the image input unit 71 from the ITV camera 10 or the like and subjected to noise removal such as smoothing, and the reference image update unit 30 to a density suitable for the input image. From the reference image, a difference (absolute value of the difference) is calculated for each pixel to create a difference image, the difference image is binarized by a predetermined threshold to create a binary image, and the binary image is labeled. A small image (value set by the user) is excluded to create a binary image from which noise has been removed. The common area extraction unit 73 extracts the binary image from which noise has been removed by the difference image binarization unit 72 and the image of the common area extracted up to the previous time (the initial image is all pixels “0” or “1”). Extract common areas. The extraction end determination unit 74 determines whether to repeat the common area extraction by comparing with the time or the number of times set by the user according to the monitoring target scene, and when it is less than the user's set value, returns to the image input unit 71. When the common area extraction is repeated and the value is equal to or larger than the user's set value, the series of processes is ended, and the object determination unit 140 determines whether or not the stationary object is detected.
【００１５】図８は、本発明における差分画像２値化処
理部７０で処理を実行する場合の一実施例を示す画像メ
モリの状態変化図である。まず、共通領域部の画像を格
納する２値メモリの初期状態８１は値『１』が設定され
ている。差分画像２値化処理部７０で求めた１回目の差
分画像が格納された濃淡メモリ７５をユーザが設定した
所定の閾値で２値化してラベリングを行って小面積部
(ユーザが設定する値)除外等の処理を行ってノイズを除
去した画像を２値メモリに格納すると、状態７８の２値
画像となり、状態７８の２値画像と値『１』が設定され
た初期状態８１の２値メモリの共通領域を抽出するため
ＡＮＤをとると状態８２の２値メモリとなる。即ち、状
態７８の２値画像がそのまま抽出される。次に、差分画
像２値化処理部７０で求めた２回目の差分画像が格納さ
れた濃淡メモリ７６を１回目と同様、所定の閾値で２値
化してラベリングを行って小面積部（ユーザが設定する
値）除外等の処理を行ってノイズを除去した画像を２値
メモリに格納すると状態７９の２値画像となり、状態７
９の２値画像と１回目に抽出した共通領域である２値メ
モリの状態８２の共通領域を抽出するためＡＮＤをとる
と状態８３の２値メモリとなる。以下、ユーザが設定し
た時間又は回数だけ同様に繰り返す。この時、静止物体
が移動すると差分画像を２値化して抽出された領域と１
ステップ前までに抽出された領域の共通部分が存在しな
い場合が発生する。そのため、ＡＮＤ処理を行うと共通
領域が消失する場合が発生する。共通領域の消失は面積
等の算出によって判定できる。そこで、静止物体の移動
により共通領域が消失した場合には、共通領域を格納す
る２値メモリを初期状態に戻して『１』を設定する。
今、ユーザが設定した時間又は回数をｎ回と設定する
と、差分画像２値化処理部７０で求めたｎ回目の差分画
像が格納された濃淡メモリ７７を前回と同様、所定の閾
値で２値化してラベリングを行って小面積部（ユーザが
設定する値）除外等の処理を行ってノイズを除去した画
像を２値メモリに格納すると状態８０の２値画像とな
り、状態８０の２値画像とｎ−１回目に抽出した共通領
域である２値メモリの状態の共通領域を抽出するためＡ
ＮＤをとると状態８４の２値メモリとなる。これら、１
〜ｎ回まで連続してＡＮＤをとって共通領域が抽出でき
た場合に静止物体が存在したと判定する。FIG. 8 is a state change diagram of the image memory showing an embodiment in which processing is executed by the differential image binarization processing unit 70 in the present invention. First, the value "1" is set in the initial state 81 of the binary memory that stores the image of the common area portion. The grayscale memory 75 in which the first difference image obtained by the difference image binarization processing unit 70 is stored is binarized by a predetermined threshold set by the user, and labeling is performed to perform a small area portion.
(Value set by user) When the image from which noise has been removed by performing processing such as exclusion is stored in the binary memory, it becomes a binary image in state 78, and the binary image in state 78 and the value “1” are set. When the AND operation is performed to extract the common area of the binary memory in the initial state 81, the binary memory in the state 82 is obtained. That is, the binary image of the state 78 is extracted as it is. Next, similarly to the first time, the grayscale memory 76 in which the second difference image obtained by the difference image binarization processing unit 70 is stored is binarized by a predetermined threshold to perform labeling, and the small area portion (user (Value to be set) When an image from which noise has been removed by performing processing such as exclusion is stored in the binary memory, a binary image in state 79 is obtained, and a state 7
A binary memory of state 83 is obtained by ANDing to extract the binary image of 9 and the common region of state 82 of the binary memory which is the common region extracted at the first time. Hereinafter, the same process is repeated for the time or the number of times set by the user. At this time, when the stationary object moves, the difference image is binarized and the extracted area is 1
There may be a case where there is no common part of the regions extracted before the step. Therefore, when the AND process is performed, the common area may disappear. The disappearance of the common area can be determined by calculating the area or the like. Therefore, when the common area disappears due to the movement of the stationary object, the binary memory storing the common area is returned to the initial state and "1" is set.
Now, when the time or number of times set by the user is set to n times, the grayscale memory 77 storing the nth difference image obtained by the difference image binarization processing unit 70 is binarized with a predetermined threshold value as in the previous time. When the image from which noise has been removed by performing processing such as conversion to labeling and exclusion of small area portions (values set by the user) is stored in the binary memory, a binary image in the state 80 is obtained, and a binary image in the state 80 is obtained. To extract the common area in the state of the binary memory, which is the common area extracted at the (n-1) th time, A
When ND is taken, it becomes a binary memory of state 84. These 1
It is determined that a stationary object exists when the common area can be extracted by continuously ANDing up to n times.
【００１６】図９は、本発明における差分画像２値化処
理部７０で処理を実行する場合の他の実施例を示す画像
メモリの状態変化図である。まず、共通領域部の画像を
格納する２値メモリの初期状態８５は値『０』が設定さ
れている。差分画像２値化処理部７０で求めた１回目の
差分画像が格納された濃淡メモリ７５、状態７８の２値
メモリ、２回目の差分画像が格納された濃淡メモリ７
６、状態７９の２値メモリ、ｎ回目の差分画像が格納さ
れた濃淡メモリ７７、状態８０の２値メモリに関して
は、図８と同様にして求める。状態７８の２値メモリと
値『０』が設定された初期状態８５の濃淡メモリを加算
して共通領域を抽出すると状態８６の濃淡メモリとな
る。次に、状態７９の２値メモリと１回目に抽出した共
通領域である２値メモリの状態８６を加算して共通領域
を抽出すると状態８７の濃淡メモリとなる。以下、ユー
ザが設定した時間又は回数だけ同様に繰り返す。今、ユ
ーザが設定した時間又は回数をｎ回と設定すると、差分
画像２値化処理部７０で求めたｎ回目の状態８０の２値
メモリとｎ−１回目までに抽出した共通領域である濃淡
メモリを加算して共通領域を抽出すると状態８８の濃淡
メモリとなる。この時、静止物体が途中からｎ回目まで
の間に移動して消失していても、ｎ回目に共通領域とし
て抽出された状態８８の濃淡メモリには、静止物体が存
在している結果になる。そこで、少なくとも最後のｎ回
目（共通領域が抽出されてきた時間又は回数なら何回目
でもよい）でマスク処理を併用する。即ち、ｎ回目を例
にとると、ｎ回目の状態８０の２値メモリとｎ回目に共
通領域として抽出された状態８８の濃淡メモリでマスク
処理を行い濃淡画像を作成すると、状態８９の濃淡メモ
リとなる。この状態８９の濃淡メモリを所定の閾値（例
えば、ｎ×２／３付近に設定する）で２値化して２値画
像を作成すると状態９０の２値メモリとなり、状態９０
を用いて静止物体存在の有無を判定する。FIG. 9 is a state change diagram of the image memory showing another embodiment when the processing is executed by the difference image binarization processing unit 70 in the present invention. First, the value “0” is set in the initial state 85 of the binary memory that stores the image of the common area portion. A grayscale memory 75 in which the first differential image obtained by the differential image binarization processing unit 70 is stored, a binary memory in state 78, and a grayscale memory 7 in which the second differential image is stored.
6, the binary memory in the state 79, the grayscale memory 77 in which the nth difference image is stored, and the binary memory in the state 80 are obtained in the same manner as in FIG. When the common memory is extracted by adding the binary memory in the state 78 and the grayscale memory in the initial state 85 in which the value "0" is set, the grayscale memory in the state 86 is obtained. Next, the binary memory in the state 79 and the state 86 of the binary memory which is the common area extracted for the first time are added to extract the common area. Hereinafter, the same process is repeated for the time or the number of times set by the user. Now, when the time or the number of times set by the user is set to n times, the binary memory of the state 80 of the nth time obtained by the differential image binarization processing unit 70 and the light and shade which is the common area extracted up to the (n-1) th time. When the common areas are extracted by adding the memories, the density memory in the state 88 is obtained. At this time, even if the stationary object has moved and disappeared from the middle to the nth time, the stationary object still exists in the density memory of the state 88 extracted as the common area at the nth time. .. Therefore, at least the last n times (any time if the common area is extracted or any number of times may be used) together with the mask processing. That is, taking the nth example as an example, when the grayscale image is created by performing the mask processing with the binary memory of the nth state 80 and the grayscale memory of the state 88 extracted as the common area at the nth time, the grayscale memory of the state 89 is generated. Becomes When the grayscale memory in this state 89 is binarized with a predetermined threshold value (for example, set to around n × 2/3) to create a binary image, the binary memory of the state 90 is obtained.
Is used to determine the presence or absence of a stationary object.
【００１７】図１０は、本発明の他の実施例を示す物体
検出装置のブロック図である。本実施例におけるＩＴＶ
カメラ１０等、参照画像入力部２０及び参照画像格納メ
モリ４０に関しては、図１と同様である。差分画像累積
２値化部１００は、画素毎の差（差の絶対値）を求めて
差分画像を作成して、一定時間静止している物体を抽出
しやすいように該差分画像をユーザが設定した累積時間
又は回数だけ繰り返して累積し累積画像（初期画像は全
画素『０』である）を作成した後、該累積画像をユーザ
の与えた所定の閾値で２値化し、更に、ラベリングを行
い小面積部（ユーザが設定する値）除外等のノイズ除去
を行って２値画像を作成する。物体抽出部１１０は、差
分画像累積２値化部１００で作成した２値画像と前回ま
でに抽出した共通領域部の画像（初期画像は全画素
『０』又は『１』である）の共通領域を抽出する。この
差分画像累積２値化部１００と物体抽出部１１０を、ユ
ーザが設定した累積時間又は回数だけ繰り返して共通領
域部を抽出した２値画像を作成する。物体判定部１４
０，報知部１５０及び報知装置１６０に関しても、図１
と同様である。FIG. 10 is a block diagram of an object detecting apparatus showing another embodiment of the present invention. ITV in this embodiment
The reference image input unit 20 and the reference image storage memory 40 such as the camera 10 are the same as those in FIG. The difference image cumulative binarization unit 100 creates a difference image by obtaining the difference (absolute value of the difference) for each pixel, and the user sets the difference image so that an object stationary for a certain time can be easily extracted. The accumulated image is repeated for the accumulated time or number of times to create an accumulated image (the initial image is all pixels “0”), the accumulated image is binarized by a predetermined threshold value given by the user, and labeling is further performed. A binary image is created by performing noise removal such as exclusion of a small area portion (value set by the user). The object extraction unit 110 is a common area of the binary image created by the differential image cumulative binarization unit 100 and the image of the common area extracted up to the previous time (the initial image is all pixels “0” or “1”). To extract. The difference image cumulative binarization unit 100 and the object extraction unit 110 are repeated for the cumulative time or the number of times set by the user to create a binary image in which the common area portion is extracted. Object determination unit 14
0, the notification unit 150, and the notification device 160 are also shown in FIG.
Is the same as.
【００１８】図１１は、本発明の他の実施例を示す物体
検出装置のブロック図である。本実施例におけるＩＴＶ
カメラ１０等、参照画像格納メモリ４０，差分画像累積
２値化部１００，物体抽出部１１０，物体判定部１４
０，報知部１５０及び報知装置１６０に関しては、図１
０と同様である。参照画像更新部３０に関しては、図２
と同様である。FIG. 11 is a block diagram of an object detecting apparatus showing another embodiment of the present invention. ITV in this embodiment
The camera 10, etc., the reference image storage memory 40, the difference image cumulative binarization unit 100, the object extraction unit 110, the object determination unit 14
0, the notification unit 150, and the notification device 160 are shown in FIG.
The same as 0. Regarding the reference image updating unit 30, FIG.
Is the same as.
【００１９】図１２は、本発明における差分画像累積２
値化部１００の内部の一実施例を示すブロック図であ
る。画像入力部１０１がＩＴＶカメラ１０等から取り込
んで平滑化等のノイズ除去を行った入力画像と、参照画
像更新部３０が入力画像に適した濃度に自動更新した参
照画像から、差分画像累積部１０２は、画素毎の差（差
の絶対値）を求めて差分画像を作成し、一定時間静止し
ている物体を抽出しやすいように差分画像を累積（初期
画像は全画素『０』である）し、累積終了判定部１０３
がユーザの与えた累積時間と比較して累積終了と判定す
るまで累積を繰り返して累積画像を作成する。差分画像
累積部１０２における差分画像の累積は、一定時間静止
している物体を容易に抽出できるように（数１）で示し
た指数平滑化手法を用いる。２値化処理部１０４は、差
分画像累積部１０２が作成した累積画像を所定の閾値で
２値化し２値画像を作成して、ノイズ除去部１０５が、
２値化処理部１０４で作成した２値画像をラベリングを
行い小面積部（ユーザが設定する値）除外等のノイズ除
去を行って２値画像を作成する。ノイズ除去部１０５で
作成した２値画像を用いて物体抽出部１１０が静止物体
抽出処理を行う。FIG. 12 shows the difference image accumulation 2 according to the present invention.
3 is a block diagram showing an example of the inside of the digitization unit 100. FIG. The difference image accumulating unit 102 uses the input image captured by the image input unit 101 from the ITV camera 10 or the like and subjected to noise removal such as smoothing, and the reference image automatically updated by the reference image updating unit 30 to a density suitable for the input image. Calculates the difference (absolute value of the difference) for each pixel, creates a difference image, and accumulates the difference images so that an object stationary for a certain time can be easily extracted (the initial image is all pixels “0”). Then, the accumulation end determination unit 103
Compares the accumulated time given by the user and repeats the accumulation until it is determined that the accumulation is completed, to create an accumulated image. For the accumulation of the difference images in the difference image accumulating unit 102, the exponential smoothing method shown in (Equation 1) is used so that an object stationary for a certain time can be easily extracted. The binarization processing unit 104 binarizes the cumulative image created by the difference image cumulative unit 102 with a predetermined threshold to create a binary image, and the noise removal unit 105
The binary image created by the binarization processing unit 104 is labeled to remove noise such as exclusion of small area parts (values set by the user), and a binary image is created. The object extracting unit 110 uses the binary image created by the noise removing unit 105 to perform the still object extracting process.
【００２０】図１３は、本発明における物体抽出部１１
０の内部の一実施例を示すブロック図である。差分画像
累積２値化部１００でノイズを除去した２値画像から、
２値画像補正部１１１は膨張及び収縮等の処理を行って
２値画像を補正する。共通領域抽出部１１２は、２値画
像補正部１１１の２値画像と前回までに抽出した共通領
域を記憶した画像（初期画像は全画素『１』である）か
ら、更に、共通領域を抽出して、抽出終了判定部１１３
がユーザの与えた時間又は回数と比較して抽出終了と判
定するまで共通領域の抽出を繰り返す。２値化処理部１
１４は、共通領域抽出部１１２が抽出した共通領域の画
像を所定の閾値で２値化し２値画像を作成して、ノズ除
去部１１５が、共通領域抽出部１１２で作成した２値画
像のラベリングを行い、比較的小面積部（検出対象の大
きさに応じてユーザが適宜設定する値）のノイズを除去
して２値画像を作成する。ノイズ除去部１１５で作成し
た２値画像を用いて物体判定部１４０が検出物体の判定
処理を行う。FIG. 13 shows the object extracting unit 11 according to the present invention.
3 is a block diagram showing an example of the inside of 0. FIG. From the binary image from which noise has been removed by the differential image cumulative binarization unit 100,
The binary image correction unit 111 corrects the binary image by performing processing such as expansion and contraction. The common area extraction unit 112 further extracts a common area from the binary image of the binary image correction unit 111 and the image that stores the common area extracted up to the previous time (the initial image is all pixels “1”). The extraction end determination unit 113
Repeats extraction of the common area until it is determined that the extraction is completed by comparing with the time or the number of times given by the user. Binarization processing unit 1
Reference numeral 14 binarizes the image of the common area extracted by the common area extraction unit 112 with a predetermined threshold value to create a binary image, and the nose removal unit 115 labels the binary image created by the common area extraction unit 112. Is performed to remove noise in a relatively small area (a value appropriately set by the user according to the size of the detection target) to create a binary image. The object determining unit 140 performs the determination process of the detected object using the binary image created by the noise removing unit 115.
【００２１】図１４は、本発明における差分画像累積２
値化部１００及び物体抽出部１１０で処理を実行する場
合の一実施例を示す画像メモリの状態変化図である。ま
ず、１回目に関し、差分画像累積部１０２で（数１）に
示した指数平滑化手法で求めた累積画像を格納した濃淡
メモリ１１６を２値化処理部１０４で所定の閾値で２値
化すると状態１１９の２値メモリとなる。２回目に関し
ても、１回目と同様にして求めた累積画像を格納した濃
淡メモリ１１７を所定の閾値で２値化すると状態１２０
の２値メモリとなる。以下、ユーザが設定した累積時間
又は回数（例えば、ｎ回）だけ同様に繰り返し、ｎ回目
の累積画像を格納した濃淡メモリ１１８を所定の閾値で
２値化すると状態１２１の２値メモリとなる。差分画像
累積２値化部１００で求めた１回目の２値メモリ１１９
と『０』が設定された初期状態１２２の濃淡メモリを物
体抽出部１１０で加算して共通領域を抽出すると状態１
２３の濃淡メモリとなる。次に、状態１２０の２値メモ
リと１回目に抽出した共通領域である濃淡メモリの状態
１２３を加算して共通領域を抽出すると状態１２４の濃
淡メモリとなる。以下、ユーザが設定した時間又は回数
だけ同様に繰り返す。例えば、ユーザがｋ回と設定する
と、差分画像累積２値化部１００で求めたｋ回目の状態
１２１の２値メモリと物体抽出部１１０で求めたｋ−１
回目までに抽出した共通領域である濃淡メモリを加算し
て共通領域を抽出すると状態１２５の濃淡メモリとな
る。この時、静止物体が途中からｋ回目までの間に移動
して消失していても、ｋ回目に共通領域として抽出され
た状態１２５の濃淡メモリには、静止物体が存在してい
る結果になる。そこで、少なくてもｋ回目（共通領域が
抽出されてきた時間又は回数なら何回目でもよい）でマ
スク処理を併用する。即ち、ｋ回目を例にとると、ｋ回
目の状態１２１の２値メモリとｋ回目に共通領域として
抽出された状態１２５の濃淡メモリでマスク処理を行い
濃淡画像を作成すると、状態１２６の濃淡メモリとな
る。この状態１２６の濃淡メモリを所定の閾値（例え
ば、ｋ×２／３付近に設定する）で２値化して２値画像
を作成すると状態１２７の２値メモリとなり、状態１２
７を用いて静止物体存在の有無を判定する。FIG. 14 shows the difference image accumulation 2 according to the present invention.
FIG. 6 is a state change diagram of an image memory showing an embodiment when processing is executed by the binarization unit 100 and the object extraction unit 110. First, regarding the first time, when the grayscale memory 116 storing the accumulated image obtained by the exponential smoothing method shown in (Equation 1) in the difference image accumulating unit 102 is binarized by a predetermined threshold value in the binarizing processing unit 104. It becomes the binary memory of the state 119. Also for the second time, if the grayscale memory 117 storing the accumulated image obtained in the same manner as the first time is binarized with a predetermined threshold value, the state 120
It becomes a binary memory. Hereinafter, the same operation is repeated for the cumulative time or the number of times (for example, n times) set by the user, and the grayscale memory 118 storing the n-th cumulative image is binarized with a predetermined threshold value to become a binary memory of the state 121. The first binary memory 119 obtained by the differential image cumulative binarization unit 100.
When the common area is extracted by adding the grayscale memory of the initial state 122 in which “0” and “0” are set in the object extraction unit 110, the state 1
It becomes 23 shade memory. Next, the binary memory in the state 120 and the state 123 of the grayscale memory, which is the common area extracted for the first time, are added to extract the common area, and the grayscale memory in the state 124 is obtained. Hereinafter, the same process is repeated for the time or the number of times set by the user. For example, when the user sets k times, the binary memory of the k-th state 121 obtained by the difference image cumulative binarization unit 100 and k−1 obtained by the object extraction unit 110
When the grayscale memory which is the common area extracted up to the first time is added to extract the common area, the grayscale memory in the state 125 is obtained. At this time, even if the stationary object has moved and disappeared from the middle to the kth time, the stationary object still exists in the grayscale memory of the state 125 extracted as the common area at the kth time. .. Therefore, the masking process is used together at least at the k-th time (any time if the common area has been extracted or any number of times). That is, taking the kth time as an example, when the grayscale image is created by performing mask processing with the binary memory of the state 121 of the kth time and the grayscale memory of the state 125 extracted as the common area at the kth time, the grayscale memory of the state 126 is generated. Becomes When the grayscale memory in this state 126 is binarized with a predetermined threshold value (for example, set to around k × 2/3) to create a binary image, it becomes the binary memory in state 127, and the state 12
The presence / absence of a stationary object is determined using 7.
【００２２】図１４では、共通領域を抽出する物体抽出
部１１０の方法として加算する例を示したが、他の例と
して、図８と同様にＡＮＤをとる方法でもよく、共通領
域を抽出できるならばどんな方法でもよい。In FIG. 14, an example of addition is shown as a method of the object extracting unit 110 for extracting a common area, but as another example, an AND method may be used as in FIG. 8 as long as the common area can be extracted. Any method will do.
【００２３】図１５は、本発明における物体判定部１４
０の内部の一実施例を示すブロック図である。物体抽出
部１１０で抽出した共通領域を示す２値画像又は、濃淡
画像ならば一定の閾値で２値化した２値画像から、面積
抽出部１４１で面積を、重心抽出部１４２で重心を、投
影分布抽出部１４３で投影分布データ等を算出する。即
ち、特徴データを示すデータなら何でも良い。これらの
面積及び／又は重心及び／又は投影分布データ等から、
判定処理部１４４が、検出対象物体の有無を、例えば、
面積の場合、検出対象の大きさに応じてユーザが適宜設
定した値未満の場合『無し』と、以上の場合『有り』と
判定して、報知部１５０に報知する。報知部１５０は、
報知装置１６０に監視員に判るように報知する。FIG. 15 is a block diagram of the object determining unit 14 according to the present invention.
3 is a block diagram showing an example of the inside of 0. FIG. From the binary image showing the common area extracted by the object extracting unit 110 or the binary image binarized with a certain threshold in the case of a grayscale image, the area extracting unit 141 projects the area, and the centroid extracting unit 142 projects the centroid. The distribution extraction unit 143 calculates projection distribution data and the like. That is, any data indicating the characteristic data may be used. From these area and / or center of gravity and / or projection distribution data,
The determination processing unit 144 determines whether or not there is a detection target object, for example,
In the case of the area, if there is less than a value appropriately set by the user according to the size of the detection target, it is determined to be “absent”, and if it is greater than or equal to it is determined to be present and the notification unit 150 is notified. The notification unit 150
The notification device 160 is informed so that the observer can understand.
【００２４】図１６は、本発明における物体判定部１４
０の内部の他の実施例を示す状態変化図である。まず、
１回目の濃淡メモリの状態２０１に関しては、濃淡メモ
リ７５又は濃淡メモリ１１６と同様に、２値メモリの状
態２０４に関しては、２値メモリ７８又は２値メモリ１
１９と同様に、２回目の濃淡メモリの状態２０２に関し
ては、濃淡メモリ７６又は濃淡メモリ１１７と同様に、
２値メモリの状態２０５に関しては、２値メモリ７９又
は２値メモリ１２０と同様に、ｎ回目の濃淡メモリの状
態２０３に関しては、濃淡メモリ７７又は濃淡メモリ１
１８と同様に、２値メモリの状態２０６に関しては、２
値メモリ８０又は２値メモリ１２１と同様に求める。次
に、面積又は／及び位置等の類似度の初期状態２１０
は、各々『１』が設定されている。１回目に求めた２値
メモリの状態２０４から面積又は／及び位置２０７等を
算出するが、１回目の類似度は自分自身と比較するた
め、面積の類似度２１０を１，位置の類似度２１０を１
とする。１回目に求めた面積又は／及び位置２０７等と
２回目の２値メモリの状態２０５から面積又は／及び位
置２０８等を算出して、これらを比較して、誤差範囲
（例えば、約±２０％以内）ならば、面積の類似度２１
１を１，位置の類似度２１１を１とする。以下、ユーザ
が設定した時間又は回数（例えば、ｎ回）だけ同様に繰
り返し、ｎ−１回目に求めた面積又は／及び位置等とｎ
回目の２値メモリの状態２０６から面積又は／及び位置
２０９等を算出して、これらを比較して、誤差範囲（例
えば、約±２０％以内）ならば、面積の類似度２１２を
１，位置の類似度２１２を１とする。これら、１〜ｎ回
まで連続して面積又は／及び位置等の類似度が１になっ
た場合に静止物体が存在すると判定する。FIG. 16 is a block diagram of the object determining unit 14 according to the present invention.
It is a state change figure which shows the other Example inside 0. First,
Regarding the state 201 of the first grayscale memory, similarly to the grayscale memory 75 or the grayscale memory 116, regarding the state 204 of the binary memory, the binary memory 78 or the binary memory 1
Similarly to 19, regarding the state 202 of the second grayscale memory, similarly to the grayscale memory 76 or the grayscale memory 117,
Regarding the state 205 of the binary memory, similarly to the binary memory 79 or the binary memory 120, regarding the state 203 of the nth density memory, the density memory 77 or the density memory 1
Similar to 18, regarding the state 206 of the binary memory, 2
It is calculated in the same manner as the value memory 80 or the binary memory 121. Next, the initial state 210 of the degree of similarity such as area or / and position
Is set to "1". The area or / and the position 207 is calculated from the state 204 of the binary memory obtained at the first time. However, since the similarity at the first time is compared with itself, the area similarity 210 is 1, and the position similarity 210 is 1
And The area or / and position 207 or the like obtained at the first time and the area or / and position 208 or the like are calculated from the state 205 of the binary memory at the second time, and these are compared, and an error range (for example, about ± 20% Within 21), the area similarity is 21
1 is 1 and the position similarity 211 is 1. Hereinafter, the time or the number of times (for example, n times) set by the user is similarly repeated, and the area or / and the position etc. obtained at the (n-1) th time and n
The area or / and the position 209 is calculated from the state 206 of the second binary memory, and these are compared, and if the error range (for example, within about ± 20%), the area similarity 212 is 1, the position. The similarity 212 of 1 is set to 1. It is determined that a stationary object exists when the degree of similarity of the area or / and the position or the like becomes 1 continuously from 1 to n times.
【００２５】図１７は、本発明における物体判定部１４
０の内部の他の実施例を示す状態変化図である。１回目
〜ｎ回目までの濃淡メモリの状態２２１，２２２，２２
３，２２６，２２７，２２８，２２９及び２値メモリの
状態２３０に関しては、図４と同様にして求める。又、
２値メモリの状態２３０を算出する方法として、図８，
図９及び図１４と同様でも良い。ここで、まず、２値メ
モリの状態２３０から面積等を算出して所定値以上の場
合、『静止物体有り』と判定する。次に、該静止物体が
放置物又は危険物等であるか否かを判定するため、静止
物体有りと検出した場合、２値メモリ２３０の白と黒を
反転して状態２３１の２値画像にし、２値画像２３０と
ｎ回目の濃淡メモリの画像２２３のマスク処理を行って
静止物体を除外した状態２２４の濃淡画像を求め、状態
２２４の濃淡画像を所定の閾値で２値化して状態２２５
の２値画像を作成する。２値画像２２５をラベリングし
て小面積のノイズを除去してから、２値画像の面積等を
算出し、該面積が所定値以上ならば静止物体の付近に人
物が存在しているものとして、放置物又は危険物無し
と、該面積が所定値未満ならば静止物体の付近に人物が
存在していないものとして、放置物又は危険物有りと判
定する。FIG. 17 is a block diagram of the object determining section 14 according to the present invention.
It is a state change figure which shows the other Example inside 0. First to nth density memory states 221, 222, 22
3, 226, 227, 228, 229 and the state 230 of the binary memory are obtained in the same manner as in FIG. or,
As a method of calculating the state 230 of the binary memory, as shown in FIG.
It may be similar to FIGS. 9 and 14. Here, first, when the area or the like is calculated from the state 230 of the binary memory and the area is equal to or more than a predetermined value, it is determined that "there is a stationary object". Next, in order to determine whether or not the stationary object is an abandoned object, a dangerous object, or the like, when it is detected that there is a stationary object, white and black in the binary memory 230 are inverted to form a binary image in the state 231. The binary image 230 and the nth grayscale image 223 are masked to obtain a grayscale image in the state 224 excluding static objects, and the grayscale image in the state 224 is binarized with a predetermined threshold to obtain the state 225.
Create a binary image of. After the small area noise is removed by labeling the binary image 225, the area of the binary image or the like is calculated. If the area is equal to or larger than a predetermined value, it is assumed that a person exists near the stationary object. If there is no abandoned material or dangerous material, and if the area is less than a predetermined value, it is determined that there is no person near the stationary object and that there is an abandoned material or dangerous material.
【００２６】図１８は、本発明における物体判定部１４
０の内部の他の実施例を示す状態変化図である。いま、
時刻ｔの入力画像をｆ（ｔ）、時刻ｔ＋Δｔの入力画像
をｆ（ｔ＋Δｔ）と記述し、時刻ｔからΔｔの間隔で入
力される画像を用いて、次の手順で静止物体を検出す
る。まず、入力画像ｆ（ｔ）２４１と入力画像ｆ（ｔ＋
Δｔ）２４２で画素毎の差（差の絶対値）を求め、差分
画像ｇ（ｔ＋Δｔ）２４４を作成し、同様に、入力画像
ｆ(ｔ＋Δｔ)と入力画像ｆ（ｔ＋２・Δｔ）で画素毎の
差（差の絶対値）を求め、差分画像ｇ（ｔ＋２・Δｔ）
２４５を作成する。以下、時間軸に対して同様な操作を
繰り返し、差分画像ｇ(ｔ＋３・Δｔ)２４６を作成す
る。次に、得られた差分画像を用い、ｇ（ｔ＋Δｔ）２
４４とｇ（ｔ＋２・Δｔ）２４５で同一物体（移動して
いる物体）の抽出処理（ＡＮＤ処理）を行ない、ｋ（ｔ
＋Δｔ）２４７の画像を作成する。同様に、ｇ（ｔ＋２
・Δｔ）とｇ(ｔ＋３・Δｔ)からｋ(ｔ＋２・Δｔ)２４
８の画像を作成する。これらは、濃淡画像であるので、
同一物体の抽出処理は、差分画像ｇ（ｔ＋Δｔ）２４４
とｇ(ｔ＋２・Δｔ)２４５の最小値をとる画像間の最小
値演算を行ったり、または、２枚の差分画像をある閾値
でそれぞれ２値化し、２値画像間のＡＮＤ処理を行な
い、もとの差分画像のうちいずれかの差分とＡＮＤ処理
してもよい。要するに、差分画像で同じような画像とな
っている画素を残すことができればどのような処理でも
かまわない。更に、入力画像からＡＮＤ処理の画像を減
算する画像間の差分処理を実行すると、差分画像２４９
となる。同様に画像間の差分処理を実行して、差分画像
２５０を作成する。このとき、ＡＮＤ処理画像ｋ（ｔ＋
Δｔ）２４７と入力画像ｆ（ｔ＋Δｔ）２４２のように
時間が一致する画像で差分処理を行う。この処理によ
り、一定時間停止している物体のみが検出される。以上
のような処理を時間軸に対して、常に繰り返して実行す
ることにより、静止している物体をリアルタイムに検出
することが可能になる。なお、入力画像が複雑で、背景
が種々存在する場合は、前述した参照画像更新部３０で
入力画像に適した濃度に更新した参照画像との差分画像
を用いることが望ましい。また、各々の時刻において、
静止物体の有無を判定可能であるが、上記処理を繰り返
し実行し、ある時間連続して、静止物体があるか否かを
逐次判定することにより検出精度の向上を図ることがで
きる。FIG. 18 is a block diagram of the object determining unit 14 according to the present invention.
It is a state change figure which shows the other Example inside 0. Now
The input image at time t is described as f (t), the input image at time t + Δt is described as f (t + Δt), and a stationary object is detected by the following procedure using the images input at intervals of Δt from time t. First, the input image f (t) 241 and the input image f (t +
Δt) 242 is used to obtain the difference (absolute value of the difference) for each pixel to create a difference image g (t + Δt) 244. Similarly, for the input image f (t + Δt) and the input image f (t + 2 · Δt), The difference (absolute value of the difference) is calculated, and the difference image g (t + 2 · Δt)
245 is created. Hereinafter, the same operation is repeated on the time axis to create the difference image g (t + 3 · Δt) 246. Next, using the obtained difference image, g (t + Δt) 2
44 and g (t + 2 · Δt) 245 perform the same object (moving object) extraction processing (AND processing), and k (t
+ Δt) 247 image is created. Similarly, g (t + 2
・ Δt) and g (t + 3 ・ Δt) to k (t + 2 ・ Δt) 24
Create 8 images. Since these are grayscale images,
The same object is extracted by the difference image g (t + Δt) 244.
And g (t + 2 · Δt) 245, the minimum value operation between the images having the minimum value is performed, or the two difference images are binarized by a certain threshold value, and the AND process between the binary images is performed. The AND process may be performed with any one of the difference images between and. In short, any processing may be used as long as pixels having the same image in the difference image can be left. Furthermore, when difference processing between images is performed by subtracting the AND processing image from the input image, a difference image 249
Becomes Similarly, difference processing between images is executed to create a difference image 250. At this time, the ANDed image k (t +
The difference processing is performed on images having the same time, such as Δt) 247 and the input image f (t + Δt) 242. Through this process, only objects that have been stationary for a certain period of time are detected. It is possible to detect a stationary object in real time by constantly repeating the above-described processing with respect to the time axis. When the input image is complicated and there are various backgrounds, it is desirable to use the difference image with the reference image updated to the density suitable for the input image by the reference image update unit 30 described above. Also, at each time
Although it is possible to determine the presence / absence of a stationary object, it is possible to improve the detection accuracy by repeatedly executing the above processing and continuously determining whether or not there is a stationary object for a certain period of time.
【００２７】上記処理は当然ながら次のような処理に置
き換えることも可能である。即ち、 （１）入力画像同士での差分画像を上記のように作成す
る。Of course, the above processing can be replaced with the following processing. That is, (1) The difference image between the input images is created as described above.
【００２８】（２）差分画像をある所定の２値化しきい
値で２値化する。(2) The difference image is binarized with a certain predetermined binarization threshold value.
【００２９】（３）得られた２値画像間で画像間のＡＮ
Ｄ処理を実行する（移動している領域は‘１’，移動い
ていない領域は‘０’となる）。(3) AN between images between the obtained binary images
The D processing is executed (the moving area becomes "1", and the non-moving area becomes "0").
【００３０】（４）入力画像と上記ＡＮＤ処理画像で論
理演算する。論理演算はＡＮＤ処理画像の‘０’の画素
（移動していない領域の画素）に対応する入力画像のみ
を取り出す処理を行う。(4) A logical operation is performed on the input image and the ANDed image. The logical operation performs a process of extracting only the input image corresponding to the pixel of “0” (pixel of the area that has not moved) of the AND processed image.
【００３１】物体報知部１５０は、報知装置１６０に監
視員の五感に感知するように報知又は／及び対策のガイ
ダンス等を行う。監視員の五感に感知させる例として、
報知物体が検出された場合、監視シーンに１−１に対応
したランプ等を中央の監視装置に設けて、シーンに対応
したランプの点燈及び点滅又は／及びブザー鳴動又は／
及び放送等の方法を用いる。又、モニタにその映像及び
ズームした映像を写す方法でも良い。更には、検出時の
みＶＴＲ及びプリンタ等に出力して過去の履歴を報知装
置に記録しても良い。The object notifying unit 150 gives notification to the notifying device 160 so that the five senses of the observer can sense it, and / or provides guidance on countermeasures. As an example to let the five senses of the observer sense it,
When a notification object is detected, a lamp corresponding to 1-1 in the monitoring scene is provided in the central monitoring device, and the lamp corresponding to the scene is turned on and blinks or / and the buzzer sounds or /
And methods such as broadcasting are used. Alternatively, a method of displaying the image and the zoomed image on the monitor may be used. Furthermore, the past history may be recorded in the notification device by outputting to the VTR, the printer, etc., only at the time of detection.
【００３２】以上のとおり、上述の各実施例によれば、
参照画像を入力画像から処理対象に適合した濃度の正常
画像に自動更新するため、屋内外の天候及び明るさが変
動しても、処理対象となる画像の濃度の正規化を行わな
いでも、精度良く検出できるという効果がある。又、背
景が更新された参照画像と処理対象の入力画像から、移
動物体の影になって見え隠れする一定時間静止している
物体を容易に抽出できるように差を累積して検出するの
で、移動物体が除去され静止物体のみが検出されるの
で、検出精度が向上するという効果がある。更に、検出
物体を判定して報知するので、監視現場に出向かないで
中央の監視所で物体が検出された場合のみ、監視対象シ
ーンに注目すればよく、常時モニタ等の報知装置を目視
しなくても物体検出が可能となり、監視における長時間
のモニタ凝視による検出率の低下、単調作業からの脱却
を図ることができるという効果がある。As described above, according to the above embodiments,
Since the reference image is automatically updated from the input image to a normal image with a density suitable for the processing target, the accuracy can be improved even if the indoor and outdoor weather and brightness change and the density of the processing target image is not normalized. It has the effect of being able to detect well. In addition, the difference is accumulated and detected from the reference image with the updated background and the input image to be processed so that a stationary object that appears and disappears in the shadow of a moving object can be easily extracted for a certain period of time. Since the object is removed and only the stationary object is detected, there is an effect that the detection accuracy is improved. Further, since the detected object is determined and notified, only when the object is detected at the central monitoring station without going to the monitoring site, it suffices to pay attention to the monitoring target scene, without constantly watching the notification device such as a monitor. Even if it is possible to detect an object, it is possible to reduce the detection rate due to a long-time gaze at the monitor during monitoring, and to avoid monotonous work.
【００３３】[0033]
【発明の効果】本発明によれば、静止物体と移動物体が
混在する状態、特に、移動物体の影に重なって見え隠れ
する状態から、所定時間放置してある駐停止車両，事故
車両，放置物，危険物等の静止物体を良好に検出するこ
とができる。また、その検出された静止物体の存在を報
知するので監視における長時間のモニタ凝視による検出
率の低下、単調作業からの脱却を図ることができる。According to the present invention, a parked vehicle, an accident vehicle, or an abandoned object that has been left for a predetermined time from a state in which a stationary object and a moving object are mixed, in particular, a state in which the object appears and disappears in a shadow of the moving object. , It can detect stationary objects such as dangerous goods. Further, since the presence of the detected stationary object is notified, it is possible to reduce the detection rate due to a long-time monitor gaze in monitoring and to avoid monotonous work.
【図１】本発明の一実施例を示す物体検出装置のブロッ
ク図である。FIG. 1 is a block diagram of an object detection device showing an embodiment of the present invention.
【図２】本発明の他の実施例を示す物体検出装置のブロ
ック図である。FIG. 2 is a block diagram of an object detection device showing another embodiment of the present invention.
【図３】本発明の差分画像累積部５０の内部の一実施例
を示すブロック図である。FIG. 3 is a block diagram showing an embodiment inside the difference image accumulating unit 50 of the present invention.
【図４】本発明の差分画像累積部５０及び２値化か処理
部６５を行う場合の一実施例を示す画像メモリの状態変
化図である。FIG. 4 is a state change diagram of an image memory showing an embodiment when the difference image accumulating unit 50 and the binarizing or processing unit 65 of the present invention are performed.
【図５】本発明の他の実施例を示す物体検出装置のブロ
ック図である。FIG. 5 is a block diagram of an object detection device showing another embodiment of the present invention.
【図６】本発明の他の実施例を示す物体検出装置のブロ
ック図である。FIG. 6 is a block diagram of an object detection apparatus showing another embodiment of the present invention.
【図７】本発明の差分画像２値化処理部７０の内部の一
実施例を示すブロック図である。FIG. 7 is a block diagram showing an example of the inside of a differential image binarization processing unit 70 of the present invention.
【図８】本発明の差分画像２値化処理部７０を行う場合
の一実施例を示す画像メモリの状態変化図である。FIG. 8 is a state change diagram of an image memory showing an embodiment when the difference image binarization processing unit 70 of the present invention is performed.
【図９】本発明の差分画像２値化処理部７０を行う場合
の他の実施例を示す画像メモリの状態変化図である。FIG. 9 is a state change diagram of an image memory showing another embodiment when the difference image binarization processing unit 70 of the present invention is performed.
【図１０】本発明の他の実施例を示す物体検出装置のブ
ロック図である。FIG. 10 is a block diagram of an object detection apparatus showing another embodiment of the present invention.
【図１１】本発明の他の実施例を示す物体検出装置のブ
ロック図である。FIG. 11 is a block diagram of an object detection apparatus showing another embodiment of the present invention.
【図１２】本発明における差分画像累積２値化部１００
の内部の一実施例を示すブロック図である。FIG. 12 is a differential image cumulative binarization unit 100 according to the present invention.
3 is a block diagram showing an example of the inside of FIG.
【図１３】本発明における物体抽出部１１０の内部の一
実施例を示すブロック図である。FIG. 13 is a block diagram showing an example of the inside of the object extraction unit 110 in the present invention.
【図１４】本発明における差分画像累積２値化部１００
及び物体抽出部１１０を行う場合の一実施例を示す画像
メモリの状態変化図である。FIG. 14 is a differential image cumulative binarization unit 100 according to the present invention.
FIG. 7 is a state change diagram of an image memory showing an embodiment in which the object extracting unit 110 is performed.
【図１５】本発明における物体判定部１４０の内部の一
実施例を示すブロック図である。FIG. 15 is a block diagram showing an example of the inside of the object determination unit 140 according to the present invention.
【図１６】本発明における物体判定部１４０の内部の他
の実施例を示す状態変化図である。16 is a state change diagram showing another embodiment of the inside of the object determination unit 140 in the present invention. FIG.
【図１７】本発明における物体判定部１４０の内部の他
の実施例を示す状態変化図である。FIG. 17 is a state change diagram showing another embodiment of the inside of the object determination unit 140 according to the present invention.
【図１８】本発明における物体判定部１４０の内部の他
の実施例を示す状態変化図である。FIG. 18 is a state change diagram showing another embodiment inside the object determining unit 140 according to the present invention.
１０…ＩＴＶカメラ、２０…参照画像入力部、３０…参
照画像更新部、５０…差分画像累積部、６５…２値化処
理部、７０…差分画像２値化処理部、１００…差分画像
累積２値化部、１１０…物体抽出部、１４０…物体判定
部、１５０…報知部。10 ... ITV camera, 20 ... Reference image input unit, 30 ... Reference image updating unit, 50 ... Difference image accumulation unit, 65 ... Binarization processing unit, 70 ... Difference image binarization processing unit, 100 ... Difference image accumulation 2 Quantization unit, 110 ... Object extraction unit, 140 ... Object determination unit, 150 ... Notification unit.
───────────────────────────────────────────────────── フロントページの続き (72)発明者 北村 忠明 茨城県日立市久慈町4026番地 株式会社日 立製作所日立研究所内 (72)発明者 田代 勝男 茨城県日立市大みか町五丁目２番１号 株 式会社日立製作所大みか工場内 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Tadaaki Kitamura 4026 Kujimachi, Hitachi City, Ibaraki Prefecture Hitachi Research Laboratory, Hitachi Ltd. (72) Inventor Katsuo Tashiro 5-2-1 Omikacho, Hitachi City, Ibaraki Prefecture Ceremony company Hitachi Ltd. Omika factory
Claims (17)
該ＩＴＶカメラの画像信号を予め取り込んでディジタル
信号に変換して背景画像とする参照画像格納手段と、該
参照画像と入力画像の差分画像を求め、該差分画像から
２値画像を求めて、少なくとも前記差分画像及び前記２
値画像のうち何れか一方を累積し、一定時間静止してい
る物体の領域を抽出する物体抽出手段と、該物体抽出手
段により得られた画像から特徴データを求め、前記検出
対象シーンにおける静止物体の有無を判定する物体判定
手段と、該物体判定手段から得られた結果を報知する報
知手段とを具備することを特徴とする物体検出装置。1. An ITV camera for photographing a scene to be detected, a reference image storage means for previously capturing an image signal of the ITV camera and converting the image signal into a digital signal to be a background image, and a difference image between the reference image and the input image. Then, a binary image is obtained from the difference image, and at least the difference image and the binary image are obtained.
Object extracting means for accumulating one of the value images to extract a region of an object stationary for a certain period of time, and feature data obtained from the image obtained by the object extracting means to obtain a stationary object in the detection target scene. An object detection device, comprising: an object determination means for determining the presence or absence of the object; and an informing means for informing a result obtained from the object determination means.
該ＩＴＶカメラの画像信号を取り込んでディジタル信号
に変換して監視シーンに適した背景に自動更新して背景
画像とする参照画像格納手段と、該参照画像と入力画像
の差分画像を求め、該差分画像から２値画像を求めて、
少なくとも前記該差分画像及び前記２値画像のうち何れ
か一方を累積し、一定時間静止している物体の領域を抽
出する物体抽出手段と、該物体抽出手段により得られた
画像から特徴データを求めて、前記検出対象シーンにお
ける静止物体の有無を判定する物体判定手段と、該物体
判定手段から得られた結果を報知する報知手段とを具備
することを特徴とする物体検出装置。2. An ITV camera for photographing a scene to be detected, and a reference image storage means for taking in an image signal of the ITV camera, converting it into a digital signal, and automatically updating it to a background suitable for a surveillance scene to obtain a background image. A difference image between the reference image and the input image is obtained, a binary image is obtained from the difference image,
At least one of the difference image and the binary image is accumulated, and an object extraction unit that extracts a region of an object that has been stationary for a certain period of time, and feature data is obtained from the image obtained by the object extraction unit. An object detection device comprising: an object determination means for determining the presence or absence of a stationary object in the detection target scene; and an informing means for informing a result obtained from the object determination means.
該ＩＴＶカメラの画像信号を取り込んでディジタル信号
に変換してΔｔの時間間隔で画像を入力し、該入力画像
に対し所定の画像処理を実行して処理対象画像を作成
し、時刻ｔの処理対象画像ｆ(ｔ)と時刻ｔ＋Δｔの処理
対象画像ｆ(ｔ＋Δｔ)で画素毎の差を求め、差分画像ｇ
(ｔ＋Δｔ)を作成し、同様に、時刻ｔ＋Δｔの処理対象
画像ｆ(ｔ＋Δｔ)と時刻ｔ＋２・Δｔの処理対象画像ｆ
（ｔ＋２・Δｔ）で画素毎の差を求め、差分画像ｇ(ｔ
＋２・Δｔ)を作成し、該差分画像ｇ(ｔ＋Δｔ)とｇ
（ｔ＋２・Δｔ）で同一物体を抽出する処理を実行し、
同一物体抽出画像を作成し、更に、入力画像ｆ（ｔ＋Δ
ｔ）から同一物体抽出画像を減算する画素毎の差を求め
る処理を実行し、上記処理を繰返し実行することによ
り、前記検出対象シーンにおける静止物体の有無を判定
することを特徴とする物体検出装置。3. An ITV camera for photographing a scene to be detected and an image signal of the ITV camera are captured and converted into a digital signal, the images are input at time intervals of Δt, and predetermined image processing is executed on the input image. Then, a processing target image is created, a difference for each pixel is calculated between the processing target image f (t) at time t and the processing target image f (t + Δt) at time t + Δt, and the difference image g
(t + Δt) is created, and similarly, the processing target image f (t + Δt) at the time t + Δt and the processing target image f at the time t + 2 · Δt are generated.
The difference for each pixel is calculated by (t + 2 · Δt), and the difference image g (t
+ 2 · Δt) is created, and the difference image g (t + Δt) and g
The process of extracting the same object is executed at (t + 2 · Δt),
The same object extraction image is created, and the input image f (t + Δ
The object detection apparatus characterized in that the presence or absence of a stationary object in the detection target scene is determined by executing a process for obtaining a difference for each pixel by subtracting the same object extracted image from t) and repeating the above process. ..
該ＩＴＶカメラの画像信号を取り込んでディジタル信号
に変換してΔｔの時間間隔で画像を入力し、該入力画像
に対し所定の画像処理を実行して処理対象画像を作成
し、時刻ｔの処理対象画像ｆ(ｔ)と時刻ｔ＋Δｔの処理
対象画像ｆ(ｔ＋Δｔ)で画素毎の差を求め、差分画像ｇ
(ｔ＋Δｔ)を作成し、同様に、時刻ｔ＋Δｔの処理対象
画像ｆ(ｔ＋Δｔ)と時刻ｔ＋２・Δｔの処理対象画像ｆ
(ｔ＋２・Δｔ)で画素毎の差を求め、差分画像ｇ(ｔ＋
２・Δｔ)を作成し、該差分画像ｇ（ｔ＋Δｔ）とｇ(ｔ
＋２・Δｔ)をある所定の２値化しきい値で２値化して
２値画像ｂ（ｔ＋Δｔ）とｂ（ｔ＋２・Δｔ）を作成
し、該２値画像ｂ（ｔ＋Δｔ）とｂ（ｔ＋２・Δｔ）の
ＡＮＤ処理を実行して同一物体を抽出する処理を実行
し、同一物体抽出画像を作成し、更に、入力画像ｆ（ｔ
＋Δｔ）と同一物体抽出画像の論理演算を実行し、上記
処理を繰返し実行することにより、前記検出対象シーン
における静止物体の有無を判定することを特徴とする物
体検出装置。4. An ITV camera for photographing a scene to be detected and an image signal of the ITV camera are taken and converted into a digital signal, the image is input at a time interval of Δt, and predetermined image processing is executed on the input image. Then, a processing target image is created, a difference for each pixel is calculated between the processing target image f (t) at time t and the processing target image f (t + Δt) at time t + Δt, and the difference image g
(t + Δt) is created, and similarly, the processing target image f (t + Δt) at the time t + Δt and the processing target image f at the time t + 2 · Δt are generated.
The difference for each pixel is calculated by (t + 2 · Δt), and the difference image g (t +
2 · Δt), and the difference images g (t + Δt) and g (t
+ 2 · Δt) is binarized by a certain predetermined binarization threshold value to create binary images b (t + Δt) and b (t + 2 · Δt), and the binary images b (t + Δt) and b (t + 2 · Δt). ) AND processing to extract the same object to create the same object extracted image, and further to input image f (t
+ Δt), the same object extracted image is logically operated, and the above processing is repeatedly executed to determine the presence or absence of a stationary object in the detection target scene.
入力画像と正常な参照画像から、画素毎の差を求めて差
分画像を作成し、一定時間静止している物体を抽出しや
すいように該差分画像を所定時間累積して累積画像を作
成し、該累積画像を所定の閾値で２値化して、ノイズを
除去した２値画像から、特徴データを抽出して検出する
ことを特徴とする物体検出装置。5. The object extracting means according to claim 1,
From the input image and the normal reference image, a difference image is created by obtaining the difference for each pixel, and the difference image is accumulated for a predetermined time so as to easily extract an object that is stationary for a certain period of time to create a cumulative image, An object detection device characterized in that the accumulated image is binarized by a predetermined threshold value, and feature data is extracted and detected from a binary image from which noise has been removed.
入力画像と正常な参照画像から、画素毎の差を求めて差
分画像を作成し、該差分画像を所定の閾値で２値化して
２値画像を作成し、該２値画像からノイズを除去した２
値画像を作成し、ノイズを除去した該２値画像を指定し
た累積時間又は回数だけＡＮＤ処理又は加算処理を繰り
返して検出することを特徴とする物体検出装置。6. The object extracting means according to claim 1,
From the input image and the normal reference image, a difference for each pixel is obtained to create a difference image, the difference image is binarized with a predetermined threshold to create a binary image, and noise is removed from the binary image. Two
An object detecting apparatus, wherein a value image is created and the binary image from which noise has been removed is repeatedly detected by AND processing or addition processing for a designated cumulative time or number of times.
入力画像と正常な参照画像から、画素毎の差を求めて差
分画像を作成し、該差分画像を所定の閾値で２値化して
２値画像を作成し、該２値画像からノイズを除去した２
値画像を作成し、ノイズを除去した該２値画像を指定し
た累積時間又は回数だけ加算処理及び少なくとも最後に
マスク処理を繰り返して検出することを特徴とする物体
検出装置。7. The object extracting means according to claim 1,
From the input image and the normal reference image, a difference for each pixel is obtained to create a difference image, the difference image is binarized with a predetermined threshold to create a binary image, and noise is removed from the binary image. Two
An object detecting apparatus, wherein a value image is created and the noise-removed binary image is detected by repeating addition processing and at least lastly mask processing for a designated cumulative time or number of times.
入力画像と正常な参照画像から、画素毎の差を求めて差
分画像を作成し、該差分画像を指定した累積時間又は回
数だけ繰り返して累積し累積画像を作成した後、該累積
画像を所定の閾値で２値化して２値画像を作成し、該２
値画像からノイズを除去した２値画像を作成し、ノイズ
を除去した該２値画像を指定した累積時間又は回数だけ
ＡＮＤ処理又は加算処理を繰り返して検出することを特
徴とする物体検出装置。8. The object extracting means according to claim 1,
From the input image and the normal reference image, the difference for each pixel is obtained to create a difference image, and the difference image is repeatedly accumulated for a designated accumulation time or number of times to create an accumulated image, and then the accumulated image is set to a predetermined value. Binarize with a threshold value to create a binary image,
An object detection apparatus, wherein a binary image from which noise is removed is created from a value image, and the binary image from which noise is removed is repeatedly detected by AND processing or addition processing for a designated cumulative time or number of times.
入力画像と正常な参照画像から、画素毎の差を求めて差
分画像を作成し、該差分画像を指定した累積時間又は回
数だけ繰り返して累積し累積画像を作成した後、該累積
画像を所定の閾値で２値化して２値画像を作成し、該２
値画像からノイズを除去した２値画像を作成し、ノイズ
を除去した該２値画像を指定した累積時間又は回数だけ
加算処理及び少なくとも最後にマスク処理を繰り返して
検出することを特徴とする物体検出装置。9. The object extracting means according to claim 1,
From the input image and the normal reference image, the difference for each pixel is obtained to create a difference image, and the difference image is repeatedly accumulated for a designated accumulation time or number of times to create an accumulated image, and then the accumulated image is set to a predetermined value. Binarize with a threshold value to create a binary image,
Object detection characterized in that a binary image from which noise is removed from a value image is created, and the binary image from which noise is removed is detected by repeating addition processing and at least lastly masking processing for a designated cumulative time or number of times. apparatus.
は、入力画像と正常な参照画像から、画素毎の差を求め
て差分画像を作成し、該差分画像を指定した累積時間又
は回数だけ繰り返して累積し累積画像を作成した後、該
累積画像を所定の閾値で２値化して２値画像を作成し、
該２値画像からノイズを除去した２値画像を用いて、特
徴データを算出して、該特徴データを比較して類似度を
求め、指定した時間又は回数を繰り返して類似度を求め
て検出することを特徴とする物体検出装置。10. The object extracting means according to claim 1, wherein a difference image is created by obtaining a difference for each pixel from an input image and a normal reference image, and the difference image is repeated for a designated cumulative time or number of times. After accumulating and creating a cumulative image, the cumulative image is binarized with a predetermined threshold value to create a binary image,
Using the binary image from which noise has been removed from the binary image, the characteristic data is calculated, the characteristic data is compared to obtain the similarity, and the similarity is obtained by detecting the similarity by repeating the designated time or number of times. An object detection device characterized by the above.
は、入力画像と正常な参照画像から、画素毎の差を求め
て差分画像を作成し、該差分画像を所定の閾値で２値化
して２値画像を作成し、該２値画像からノイズを除去し
た２値画像を用いて、特徴データを算出して、該特徴デ
ータを比較して類似度を求め、指定した時間又は回数を
繰り返して類似度を求めて検出することを特徴とする物
体検出装置。11. The object extracting means according to claim 1, wherein a difference image is created by calculating a difference for each pixel from an input image and a normal reference image, and the difference image is binarized by a predetermined threshold value. A binary image is created, feature data is calculated using the binary image from which noise has been removed from the binary image, the feature data is compared to obtain a similarity, and a specified time or number of times is repeated. An object detection device characterized by obtaining and detecting a degree of similarity.
は、入力画像と正常な参照画像から、画素毎の差を求め
て差分画像を作成し、該差分画像を指定した累積時間又
は回数だけ繰り返して累積し累積画像を作成した後、該
差分画像又は該累積画像を所定の閾値で２値化して２値
画像を作成し、該２値画像からノイズを除去した２値画
像を作成し、ノイズを除去した該２値画像を指定した累
積時間又は回数だけＡＮＤ処理又は加算処理を繰り返し
て静止物体を検出し、更に、静止物体が検出された場合
に、人物の存在の有無を判定する手段を備え、人物が存
在しないで静止物体が存在する場合に警告を報知するこ
とを特徴とする物体検出装置。12. The object extracting means according to claim 1, wherein a difference image is created by obtaining a difference for each pixel from an input image and a normal reference image, and the difference image is repeated for a designated cumulative time or number of times. Then, the difference image or the accumulated image is binarized with a predetermined threshold value to create a binary image, and a binary image in which noise is removed from the binary image is created. Means for determining the presence / absence of a person when a stationary object is detected by repeating AND processing or addition processing for the specified accumulated time or number of times of the binary image from which the stationary object has been removed. An object detection device, comprising: a warning when a person does not exist and a stationary object exists.
理は、２枚の差分画像の各画素毎の明るさを比較し、最
小値をとる方の差分画像の画素データを取り出す画像間
最小値演算を実行することを特徴とする物体検出装置。13. The minimum value between images for extracting the pixel data of the difference image having the minimum value by comparing the brightness of each pixel of the two difference images and extracting the pixel data of the difference image in the same object extraction process according to claim 3. An object detection device characterized by executing an operation.
理は、２枚の差分画像をある閾値でそれぞれ２値化処理
し、２値化した２枚の画像のＡＮＤ処理を行ない、該Ａ
ＮＤ処理画像と２枚の差分画像のうちいずれかの差分画
像とで、更に、ＡＮＤ処理することを特徴とする物体検
出装置。14. The same object extracting process according to claim 3, wherein two difference images are binarized by a certain threshold value, and an AND process of the binarized two images is performed.
An object detection device characterized by further performing an AND process on an ND processed image and one of the two difference images.
像を作成するための所定の画像処理は、予め格納してい
る背景画像と入力画像との差分処理を実行することを特
徴とする物体検出装置。15. The object detection according to claim 3 or 4, wherein the predetermined image processing for creating the processing target image is a difference processing between a background image stored in advance and an input image. apparatus.
の何れかにおいて、該差分画像を指定した累積時間又は
回数だけ繰り返して累積し累積画像を作成する方法は指
数平滑化手法を用い、該指数平滑化を行う場合のαを
０.９ 以上に設定することを特徴とする物体検出装置。16. The exponential smoothing method according to any one of claims 5, 8, 9, 10 and 12, wherein the difference image is repeatedly accumulated for a designated accumulation time or number of times to create an accumulated image. An object detecting apparatus characterized in that α is set to 0.9 or more when the exponential smoothing is performed.
れた場合に、人物の存在の有無を判定する処理は、最後
に求めた２値画像と濃淡画像のマスク処理を行って静止
物体を除外した濃淡画像を求め、該濃淡画像を所定の閾
値で２値化して２値画像を作成して、該２値画像のノイ
ズを除去し特徴データを算出して検出することを特徴と
する物体検出装置。17. The method according to claim 12, wherein when a stationary object is detected, the process of determining the presence or absence of a person excludes the stationary object by performing masking processing of the finally obtained binary image and grayscale image. An object detection characterized by obtaining the grayscale image, binarizing the grayscale image with a predetermined threshold to create a binary image, removing noise from the binary image, and calculating and detecting characteristic data. apparatus.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP4010746A JPH05205175A (en) | 1992-01-24 | 1992-01-24 | Body detection device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP4010746A JPH05205175A (en) | 1992-01-24 | 1992-01-24 | Body detection device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| JPH05205175A true JPH05205175A (en) | 1993-08-13 |
Family
ID=11758876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP4010746A Pending JPH05205175A (en) | 1992-01-24 | 1992-01-24 | Body detection device |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JPH05205175A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001325672A (en) * | 2000-05-16 | 2001-11-22 | Mega Chips Corp | Invader monitoring system |
| JP2005310068A (en) * | 2004-04-26 | 2005-11-04 | Noritsu Koki Co Ltd | Method for correcting white of eye, and device for executing the method |
| JP2007172504A (en) * | 2005-12-26 | 2007-07-05 | Daihatsu Motor Co Ltd | Adhering matter detection device and adhering matter detection method |
| JP2008005399A (en) * | 2006-06-26 | 2008-01-10 | Matsushita Electric Ind Co Ltd | Device and method for detecting a neglected object |
| JP2008243009A (en) * | 2007-03-28 | 2008-10-09 | Secom Co Ltd | Unidentified object monitoring device |
| JP2009152748A (en) * | 2007-12-19 | 2009-07-09 | Mitsubishi Electric Corp | Image processing device for monitoring neglected object |
| JP2010266308A (en) * | 2009-05-14 | 2010-11-25 | Kobe Steel Ltd | Welding workpiece shape measuring device and program for the same |
| JP2011258214A (en) * | 2011-07-07 | 2011-12-22 | Mega Chips Corp | Object detector |
| KR101350777B1 (en) * | 2006-10-23 | 2014-01-15 | 가부시키가이샤 히타치세이사쿠쇼 | Image monitoring apparatus and system |
| WO2018056355A1 (en) * | 2016-09-23 | 2018-03-29 | 株式会社日立国際電気 | Monitoring device |
| US10026019B2 (en) | 2014-03-11 | 2018-07-17 | Mitsubishi Electric Corporation | Person detecting device and person detecting method |
| JP2020095554A (en) * | 2018-12-14 | 2020-06-18 | 三菱電機インフォメーションシステムズ株式会社 | Left-behind object detection device, left-behind object detection method, and left-behind object detection program |
-
1992
- 1992-01-24 JP JP4010746A patent/JPH05205175A/en active Pending
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001325672A (en) * | 2000-05-16 | 2001-11-22 | Mega Chips Corp | Invader monitoring system |
| JP2005310068A (en) * | 2004-04-26 | 2005-11-04 | Noritsu Koki Co Ltd | Method for correcting white of eye, and device for executing the method |
| JP2007172504A (en) * | 2005-12-26 | 2007-07-05 | Daihatsu Motor Co Ltd | Adhering matter detection device and adhering matter detection method |
| JP2008005399A (en) * | 2006-06-26 | 2008-01-10 | Matsushita Electric Ind Co Ltd | Device and method for detecting a neglected object |
| KR101350777B1 (en) * | 2006-10-23 | 2014-01-15 | 가부시키가이샤 히타치세이사쿠쇼 | Image monitoring apparatus and system |
| JP2008243009A (en) * | 2007-03-28 | 2008-10-09 | Secom Co Ltd | Unidentified object monitoring device |
| JP2009152748A (en) * | 2007-12-19 | 2009-07-09 | Mitsubishi Electric Corp | Image processing device for monitoring neglected object |
| JP2010266308A (en) * | 2009-05-14 | 2010-11-25 | Kobe Steel Ltd | Welding workpiece shape measuring device and program for the same |
| JP2011258214A (en) * | 2011-07-07 | 2011-12-22 | Mega Chips Corp | Object detector |
| US10026019B2 (en) | 2014-03-11 | 2018-07-17 | Mitsubishi Electric Corporation | Person detecting device and person detecting method |
| WO2018056355A1 (en) * | 2016-09-23 | 2018-03-29 | 株式会社日立国際電気 | Monitoring device |
| JPWO2018056355A1 (en) * | 2016-09-23 | 2019-07-25 | 株式会社日立国際電気 | Monitoring device |
| US10691947B2 (en) | 2016-09-23 | 2020-06-23 | Hitachi Kokusai Electric Inc. | Monitoring device |
| JP2020095554A (en) * | 2018-12-14 | 2020-06-18 | 三菱電機インフォメーションシステムズ株式会社 | Left-behind object detection device, left-behind object detection method, and left-behind object detection program |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10051246B2 (en) | Stopped object detection | |
| KR101237970B1 (en) | Image survailance system and method for detecting left-behind/taken-away of the system | |
| Albiol et al. | Detection of parked vehicles using spatiotemporal maps | |
| US6985172B1 (en) | Model-based incident detection system with motion classification | |
| KR100259737B1 (en) | Method and apparatus for sensing object located within visual field of imaging device | |
| US7982774B2 (en) | Image processing apparatus and image processing method | |
| US20100201820A1 (en) | Intrusion alarm video-processing device | |
| JP3423861B2 (en) | Method and apparatus for monitoring a moving object | |
| CN111444854B (en) | Abnormal event detection method, related device and readable storage medium | |
| JP2000090277A (en) | Reference background image updating method, method and device for detecting intruding object | |
| JP4611776B2 (en) | Image signal processing device | |
| JPH05205175A (en) | Body detection device | |
| JP3569163B2 (en) | Moving object monitoring device | |
| WO2013132711A1 (en) | Object detection device, object detection method and program | |
| JP3995671B2 (en) | Image processing device | |
| JP2001160146A (en) | Method and device for recognizing image | |
| Li et al. | Robust real-time detection of abandoned and removed objects | |
| JP3294468B2 (en) | Object detection method in video monitoring device | |
| JP2001043383A (en) | Image monitoring system | |
| JP2008152586A (en) | Automatic identification monitor system for area surveillance | |
| JPH0514891A (en) | Image monitor device | |
| JP3736836B2 (en) | Object detection method, object detection apparatus, and program | |
| JPH05183901A (en) | Moving body monitor device | |
| JP3505924B2 (en) | Vehicle monitoring device | |
| Dinh et al. | Detection and localization of road area in traffic video sequences using motion information and fuzzy-shadowed sets |