JP3682631B2 - Fire detection equipment - Google Patents

Description

【００１３】
このようにＲＧＢ成分であらわすと、炎ＦはＲ成分、Ｇ成分の値（輝度値）が共に高く、テールランプやナトリウム灯などの人工光源は３つの成分のうち、Ｒ成分のみが大きい値をもつことがわかる。つまりＲ成分とＧ成分が共に大きい領域（画素）を抽出することで、監視画像から人工光源を省いて火災の領域のみを抽出することが可能となる。この原理を踏まえて、以下に本発明の動作について説明する。
【００１４】
監視カメラ１によって撮影された監視領域の画像は、カラー画像信号がアナログデジタル変換器２によってデジタル化された後、画像メモリ３に格納される。つまりＲＧＢ信号のそれぞれがＡ／Ｄ変換された後に、Ｒ成分フレームメモリ３Ｒ、Ｇ成分フレームメモリ３Ｇ、Ｂ成分フレームメモリ３Ｂに書き込まれる。画像メモリ３に格納された画像は画素単位で、全ての画像領域にわたって、最小値演算部４によって最小値演算がなされるが、ここでは特に車両Ｃの上記カラー成分で表せられるテールランプＣＴの領域部分が画像処理される場合について説明する。
【００１５】
最小値演算部４が、Ｒ成分フレームメモリ３ＲとＧ成分フレームメモリ３Ｇに格納されたカラー成分信号のＲ成分とＧ成分を同一画素毎に比較して小さい値の成分を出力する。即ち、テールランプＣＴの領域はＲが１６０、Ｇが７５なので小さい値をもつＧ成分が出力される。次にその出力された値を基に火災領域抽出部６により二値化処理が行われる。ここで二値化を行う閾値としての所定値が例えば１８０に設定されていると、最小値演算部４から出力された値は７５であるので、その領域は「０」（黒レベル）とされる。なおナトリウム灯Ｎの領域も同様に最小値演算及び火災領域抽出部６による二値化処理が行われると、その領域は「０」となる。
【００１６】
次に火災時の炎Ｆの場合について説明する。炎ＦもテールランプＣＴやナトリウム灯Ｎと同様にＲ成分とＧ成分では、Ｇ成分の方が小さいので（Ｒ成分が小さい場合もある）、最小値演算部４からはＧ成分の値が出力される。次に火災領域抽出部６により二値化処理が行われるが、炎ＦのＧ成分は２１０であり、所定値１８０よりも大きいので、その領域は「１」になる。なお明るい領域ほど２５５階調で表せられる数値は大きくなるので、車両Ｃの本体部分などの光を発しない領域は、最小値演算部４の結果がいかなる場合でも、火災領域抽出部６の二値化処理の段階で全て「０」になる。図３は二値化メモリ７に格納された画像処理（最小値演算、二値化処理）後の画像で、この図からも明らかなように、画像メモリ３に格納された画像（原画像）から、火災領域のみを抽出して表示させることが可能となる。
【００１７】
なお先に火災検出の原理で説明したように、画像メモリ３からＲ成分とＧ成分が共に大きい領域（画素）を抽出することで、火災領域のみを抽出できる。つまり最小値演算部４の代わりに、Ｒ成分フレームメモリ３Ｒに格納されたカラー成分信号のうちＲ成分が所定値を越える領域を抽出するＲ領域抽出部と、Ｇ成分フレームメモリ３Ｇに格納されたカラー成分信号のうちＧ成分が所定値を越える領域を抽出するＧ領域抽出部を設け、Ｒ領域抽出部によって抽出された領域と、Ｇ領域抽出部によって抽出された領域の重なりあう領域を火災領域として抽出する火災領域抽出部を更に設けても画像メモリ３から火災領域を抽出することができる。この場合、Ｒ領域抽出部、Ｇ領域抽出部及び火災領域抽出部が火災領域抽出手段の一例となる。
【００１８】
この場合、Ｒ成分フレームメモリにおいて画素単位に所定値、例えば１８０を越える画素を探す処理工程、Ｇ成分フレームメモリにおいて画素単位に所定値、例えば１８０を越える画素を探す処理工程、またそれぞれ抽出された画素同士が重なりあう画素を探す処理工程の３つの処理工程が必要となるが、最小値演算部３を使用すればＲ、Ｇの両成分を比較する処理工程と、所定値で二値化する処理工程の２つの処理工程ですむので、いち早く火災領域を検出することが可能となる。しかも単純に２つの値を比較して、その結果だけを出力するものなので、加算、乗算などの演算をする必要もないのでＭＰＵ２３の処理の負担が少なくてすむ。
【００１９】
なお図２に示す車両Ｃの後部に後続車両のヘッドライトの光が強くあたってる場合には、車両Ｃの後ろガラスが鏡面反射を起こし、細い横長の光を放つ領域が発生する。この領域は、最小値演算及び二値化処理を行っても、抽出される恐れがある。そこで現画像のエッジ画像を抽出するエッジ処理部を設け、二値化処理後の二値化画像からこのエッジ画像を差分処理するようにすれば、二値化画像のエッジを削ることができる。つまり二値化画像の抽出領域は、回りが削られ、一回り小さい領域となるので、ある程度の幅（大きさ）を持つ領域のみが残り、幅の小さい領域は、全て削除されてしまう。よってガラスの鏡面反射により発生した細長い形状の領域は、このような処理をすることで領域そのものをなくすことが可能となる。
【００２０】
実施形態２
ところで、トンネルの幅が狭く、双方向の車線があり、監視カメラ方向に向かって走る車両を撮影しなけらばならない場合、車両前部に黄色いフォグランプ（又は黄色のハロゲンランプ）などがあると、これが誤報要因となる。つまり本発明の火災検出原理は、Ｒ、Ｇが共に大きい領域を抽出するもので、これは色にして言い換えると黄色から白色にわたる色の範囲を抽出するものであるから、車両前部に黄色い発光体があると、それを火災領域として抽出してしまう恐れがある。
【００２１】
そこで、この実施形態２では、実施形態１において抽出された領域の特徴を数値化することで、より精度（認識度）の高い火災検出を行う。つまり最小値演算と二値化処理という２つの工程により一次的な火災領域の抽出を行い、外接矩形作成手段１２、火災特徴量演算手段、火災判定手段１４により抽出された火災領域が火災時の炎らしい挙動を示すかどうか、二次的な火災検出を行って、その領域が実火災であるのかどうかを判定する。
【００２２】
図１において、１１は対応判別手段で、監視カメラ１により周期的に撮影された画像に火災領域が連続してある場合、つまり二値化メモリ７に火災領域が連続して格納される場合に、ある時間の前後にわたる火災領域同士の対応関係、即ち同じ炎により抽出された領域なのかどうかを判別するもので、言い換えると監視領域内に所定時間に渡って火災領域があるかどうかを判別するものである。
【００２３】
１２は外接矩形作成手段で、二値化メモリ７に格納された火災領域を外接する最小の矩形で囲み、矩形の対角（例えば左上、右下）の隅の画素の座標（フィレ値）から矩形の高さｄｙと矩形の幅ｄｘ（フィレ径）を演算して、その高さと幅の比｛ｄｘ／ｄｙ｝の値を求めてＲＡＭ２２に格納する（図３参照）。１３は火災特徴量演算手段、１４は火災判定手段で、二値化メモリ７に格納された火災領域がある時間内にわたって存在し、対応判別手段１１により連続して例えば７回程度、火災領域の対応関係がとれた場合、つまりその領域が同じ炎（又は光源）によって抽出されているものと判断された場合、その火災領域が本当の火災領域であるかどうかの判定を行うものである。具体的には火災特徴量演算手段１３は、外接矩形作成手段１２により演算され、ＲＡＭ２２に格納された７個の｛ｄｘ／ｄｙ｝から、最大値と最小値を選び次式に示す火災特徴量を演算する。また火災判定手段は１５は、演算された火災特徴量の値と所定値（１．３〜１．５）の大小関係を調べ、火災特徴量の値の方が大きい場合に火災であると判定し、図示しない表示部や音響部から火災の発生を警報するものである。
【００２４】
火災特徴量＝ｍａｘ｛ｄｘ／ｄｙ｝／ｍｉｎ｛ｄｘ／ｄｙ｝……式１
監視領域内で火災が発生すると、実施形態１に記載のように二値化メモリ７に火災領域が抽出される。この時、時間的に連続して二値化メモリ７に火災領域が格納されると、対応判別手段１１が火災領域同士の対応関係をとる。この対応関係をとる方法は、いくつか有り、例えば二値化メモリ７に格納された、撮影時間（周期）の連続する火災領域を重ねあわせ、その重なり具合が所定値以上の場合に対応有りと判断する方法。また撮影時間（周期）の連続する火災領域の一方の重心座標が他方の火災領域内に入っていれば対応有りと判断する方法などが有り、これらは監視対象物が炎のように全体的な移動量の少ないものの場合に有効で、車両のランプ類が火災領域抽出手段により一次的に火災領域として抽出されてしまっても、車両の移動量は大きいので、領域同士の対応関係がとられることはない。また二値化メモリ７に格納された火災領域は、外接矩形作成手段１２により外接する最小の矩形で囲まれ、矩形の高さと幅の比｛ｄｘ／ｄｙ｝の値が演算され、その値はＲＡＭ２２に一時的に格納される。
【００２５】
監視カメラ１により撮影される画像に所定の明るさをもつ領域が、所定時間に渡ってあり、二値化メモリ７に格納された火災領域が連続して、例えば７回程度対応関係があるものと対応判別手段１１が判断する時、火災特徴量演算手段１３が式１を用いてＲＡＭ２２に格納された｛ｄｘ／ｄｙ｝の７回分のデータから｛ｄｘ／ｄｙ｝の最大値と最小値の比（火災特徴量）を演算する。二値化メモリ７に格納された火災領域が本当の火災である場合、時間の経過に伴いその領域は変化していくはずである。また火災が進展しなくても、炎は激しい高さ方向の変化をもつ。このためある時間内にわたって何回か｛ｄｘ／ｄｙ｝の値を演算すると、それらの値は大きくばらつき、最大値と最小値の比をとると、概ね１．４程度を越える。一方、火災領域がヘッドライトなどの光源である場合、車両が移動して撮影される位置が例え異なっても、形状は変化しないから｛ｄｘ／ｄｙ｝の値は、ある時間内にわたって、一定であり、式１のように最大値と最小値の比をとっても、その値はかぎりなく１に近く、所定値１．３を越すことはない。火災判定手段１４はこの火災特徴量が所定値よりも大きい場合には、火災領域は本当の火災であると判断し、図示しない表示部や音響部から火災の発生を警報する。
【００２６】
上述したように監視領域内を、ある程度の速度で車両が走っていれば、火災領域抽出手段がその車両のランプ類を火災領域として抽出したとしても、対応判別手段１１により火災領域の対応関係がとられることはない。しかし、この車両がトンネル内で停止又はほとんど動かないような場合には、移動量が小さいので火災領域として対応関係がとられることになるが、火災特徴量演算手段１３を設けて、抽出された火災領域が火災らしい挙動を示すかどうかを調べることで、上記のように停止中の車両のランプ類などにより抽出された領域を火災と判定するのを防止できる。
【００２７】
ここで実施形態１、実施形態２の動作説明を図４のフローチャートを用いて説明する。まずステップ１（以下ステップをＳと略す。）で、監視カメラ１が撮影した画像を画像メモリ３に取り込む。画像メモリ３に取り込まれた画像は、最小値演算部４によりＲフレームメモリ３ＲとＧフレームメモリ３Ｇを画素毎に比較して、ＲＧのうち小さい輝度値をもつものが出力される（Ｓ３）。その出力された輝度値を所定値で二値化処理して（Ｓ５）、所定値以上の領域を抽出する。この抽出領域は、光を放つ光源がある領域となる。
【００２８】
Ｓ７で所定値以上の領域、つまり抽出すべき領域があるか否かを判断し、ここで無い場合は新たな画像を取り込むためにＳ１に戻るが、抽出領域があるならば、外接矩形作成手段１２がその抽出した領域の外接矩形を作成し、そしてフィレ径の比｛ｄｘ／ｄｙ｝を演算し、その演算値をＲＡＭ２２に格納しておく。またＳ１１でこの抽出された領域が前回サンプリングした（取り込んだ）画像における抽出領域と対応関係があるかを対応判別手段１１により判別する。対応関係がない場合には新たに発生した領域と認識し、ラベリングを行い（抽出領域に番号をつける）、Ｓ１に戻る。
【００２９】
また対応判別手段１１により対応関係がとれた場合には、Ｓ１３に進んで、所定回に渡って対応関係がとられたかどうかを判断する。例えばまだ２回しか対応関係がとられていなければ、Ｓ１に戻り、所定回、例えば７回対応関係がとれた場合にはＳ１５に進んで、火災特徴量演算手段１３が式１に基づく火災特徴量を演算する。そしてＳ１７で演算した特徴量が所定値よりも大きいと火災判定手段１４が判断したらＳ１９にて火災警報を行う。また所定値以下の場合にはＳ１に戻る。
【００３０】
なお火災特徴量は、数回のデータにおける｛ｄｘ／ｄｙ｝の最大値と最小値の比から求めたが、所定時間内にわたる外接矩形の高さと幅の比の変化量から演算してもよい。つまり｛ｄｘ／ｄｙ｝の値を演算する度に、前回の｛ｄｘ／ｄｙ｝の値との差を絶対値で求めていき、所定回演算した時に、その差を加算した値が所定値よりも大きければ本当の火災領域であると判断するようにしてもよい。また監視領域は球場、アトリウムといった大空間でもよい。
【００３１】
【発明の効果】
Ｒ成分とＧ成分が共に大きい領域を火災領域として画像から抽出することで、Ｇ成分の低いナトリウム灯やテールランプなどの人工光源を省いて炎の領域だけを抽出することができる。また、画像処理部における画素を走査する時間を短くでき、いち早く火災を検出することが可能となる。
【図面の簡単な説明】
【図１】本発明のシステムブロック図である。
【図２】監視カメラにより映される画像の一例である（原画像）。
【図３】二値化メモリに格納された画像処理後の画像の一例である。
【図４】本発明の動作を説明するためのフローチャートである。
【符号の説明】
１ 監視カメラ、 ２ アナログデジタル変換器、 ３ 画像メモリ、
３Ｒ Ｒ成分フレームメモリ、 ３Ｇ Ｇ成分フレームメモリ、
３Ｂ Ｂ成分フレームメモリ、 ４ 最小値演算部、 ６ 火災領域抽出部
７ 二値化メモリ、 ８ 画像処理部、 １１ 対応判別手段、
１２ 外接矩形作成手段、１３ 火災特徴量演算手段、１４ 火災判定手段、
２１ ＲＯＭ、 ２２ ＲＡＭ、 ２３ ＭＰＵ、[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a fire detection apparatus using image processing.
[0002]
[Prior art]
For example, there are the following methods for installing a surveillance camera in the surveillance area and detecting an image captured by the camera using an image processing apparatus. First, the image processing device stores an image obtained by capturing a normal monitoring area, and the image captured by the monitoring camera is subjected to differential processing with the image captured at the normal time, that is, an area different from the normal time, that is, Only capture areas that have changed. Next, if the change area has a predetermined brightness, the area is regarded as a fire area.
[0003]
[Problems to be solved by the invention]
In this conventional method, it is necessary to store an image referred to as a background image for differential processing. However, a normal background image needs to store images in units of time in consideration of changes in brightness. Therefore, the load of the memory capacity becomes large as an image processing apparatus instead of one sheet. Further, if it is determined whether or not the fire is based only on the brightness, when a vehicle enters the monitoring area, lamps such as the headlamps of the vehicle are also recognized as the fire area, and a false alarm is generated. Accordingly, an object of the present invention is to obtain a fire detection device that has a small memory load and does not cause false alarms due to a light source.
[0004]
[Means for Solving the Problems]
The fire detection apparatus according to the present invention is an image capturing means for capturing an image of a monitoring area and outputting a color image signal composed of R, G, B color component signals, and an image for storing an image captured by the image capturing means. And a predetermined value set so that a fire and an artificial light source can be distinguished from an area having a predetermined brightness in a fire detection device that detects a fire by processing an image stored in the image memory. And a minimum value calculation unit for comparing the R component and G component of the color component signal for each pixel from the image stored in the image memory, and outputting a small value component, and the minimum value And a fire area extraction unit that extracts a region where the output signal of the calculation unit exceeds the predetermined value as a fire area.
Fire detection device.
[0006]
DETAILED DESCRIPTION OF THE INVENTION
Embodiment 1
Hereinafter, a first embodiment of the present invention will be described. FIG. 1 is a block diagram showing the present invention. Reference numeral 1 denotes a monitoring camera as an imaging means, which uses a CCD camera, for example, and images a monitoring area at a predetermined sampling period. The surveillance camera 1 outputs a color image signal composed of R (red), G (green), and B (blue) color component signals according to, for example, the NTSC system. For example, the surveillance camera 1 is installed at a position overlooking the entire surveillance area in a tunnel as a surveillance area, monitors a fire that occurs in the tunnel, and determines whether there is a fire area in the captured image. It is detected by the processing unit.
[0007]
FIG. 2 is a drawing showing an image photographed by the surveillance camera 1. As can be seen from this figure, the surveillance camera 1 is installed, for example, on the upper side wall of the tunnel so as to reflect the direction in which the vehicle C runs away. Yes. This is to prevent the headlight of the vehicle C from entering the surveillance camera 1, and the installation of the headlight prevents the headlamp from being captured as a fire area when image processing is performed.
[0008]
An analog-digital converter 2 converts a color image obtained from the monitoring camera 1, that is, an RGB signal into a multi-gradation digital signal in units of pixels. Reference numeral 3 denotes an image memory for storing a digitized video signal. The image memory 3 includes an R component frame memory 3R, a G component frame memory 3G, and a B component frame memory 3B, and stores one screen image taken by the surveillance camera 1. Is. The image memories 3R, 3G, and 3B are configured in plural so that a plurality of images can be stored, and new images are sequentially updated and stored while deleting the oldest image.
[0009]
Reference numeral 4 denotes a minimum value calculation unit (also referred to as a minimum value filter), which compares the R component and G component of the color component signal stored in the R component frame memory 3R and the G component frame memory 3G for each pixel and has a small value. The component (the luminance value) is output. A fire area extraction unit 6 binarizes the output signal of the minimum value calculation unit with a predetermined value and extracts an area exceeding the predetermined value as a fire area. That is, the fire area is represented by “1”, and the other part of the image (the part less than the predetermined value) is represented by “0”. This predetermined value is a value set so that a fire and an artificial light source can be distinguished from an area having a predetermined brightness. Reference numeral 7 denotes a binarization memory for storing an image binarized by the fire area extraction unit 6. The binarization memory 7 includes a plurality of images in the same manner as the image memory 3 and sequentially stores the latest images from the image memory 3.
[0010]
Reference numeral 11 denotes correspondence determination means, reference numeral 12 denotes circumscribed rectangle creation means, reference numeral 13 denotes fire feature amount calculation means, and reference numeral 14 denotes fire determination means, which will be described in detail later. The minimum value calculation unit 4 and the fire area extraction unit 6 are examples of a fire area extraction unit, and extract an area that seems to be a fire from an image based on brightness and the like. The minimum value calculation unit 4, the fire area extraction unit 6, the correspondence determination unit 11, the circumscribed rectangle creation unit 12, the fire feature amount calculation unit 13, and the fire determination unit 14 are collectively referred to as an image processing unit 8 that processes an image. The image processing unit 8 includes a ROM 21 as a storage unit, a RAM 22 as a primary storage unit, and an MPU (microprocessor) 23 as a calculation unit. Various calculation processes and the like in the image processing unit 8 are stored in the ROM 21. This is performed by the MPU 23 based on the programmed program, and the calculated value is stored in the RAM 22 at this time. The ROM 21 stores predetermined values for binarization processing.
[0011]
Next, the principle of fire detection will be briefly described. Now, as shown in FIG. 2, the image taken by the surveillance camera 1 includes an object having three brightness levels, a vehicle C, a lighting sodium lamp N, and a fire flame F as shown in FIG. It is projected. In the figure, CT indicates a tail lamp (including a position lamp) of the vehicle C. Table 1 shows an example of the three color component signals of the tail lamp CT, the sodium lamp N, and the flame F of the vehicle C in 255 gradations.
[0012]
[Table 1]

[0013]
In this way, when expressed in terms of RGB components, the flame F has both high R component and G component values (luminance values), and an artificial light source such as a tail lamp or sodium lamp has only a large R component value among the three components. I understand that. That is, by extracting a region (pixel) where both the R component and the G component are large, it is possible to extract only the fire region from the monitoring image without the artificial light source. Based on this principle, the operation of the present invention will be described below.
[0014]
The image of the monitoring area captured by the monitoring camera 1 is stored in the image memory 3 after the color image signal is digitized by the analog-digital converter 2. That is, each of the RGB signals is A / D converted and then written to the R component frame memory 3R, the G component frame memory 3G, and the B component frame memory 3B. The image stored in the image memory 3 is subjected to the minimum value calculation by the minimum value calculation unit 4 over the entire image area in units of pixels. Here, in particular, the area portion of the tail lamp CT represented by the color component of the vehicle C is here. A case where image processing is performed will be described.
[0015]
The minimum value calculation unit 4 compares the R component and G component of the color component signal stored in the R component frame memory 3R and the G component frame memory 3G for each same pixel, and outputs a component having a smaller value. That is, since the tail lamp CT region has R of 160 and G of 75, a G component having a small value is output. Next, the binarization processing is performed by the fire area extraction unit 6 based on the output value. Here, if the predetermined value as the threshold value for binarization is set to 180, for example, the value output from the minimum value calculation unit 4 is 75, so the area is set to “0” (black level). The Similarly, when the area of the sodium lamp N is subjected to the minimum value calculation and binarization processing by the fire area extraction unit 6, the area becomes “0”.
[0016]
Next, the case of the flame F at the time of a fire is demonstrated. Similarly to the tail lamp CT and the sodium lamp N, the flame F is smaller in the G component in the R component and the G component (the R component may be smaller), so the value of the G component is output from the minimum value calculation unit 4. The Next, binarization processing is performed by the fire region extraction unit 6, but the G component of the flame F is 210, which is larger than the predetermined value 180, so that region is “1”. In addition, since the numerical value that can be expressed by 255 gradations becomes larger as the bright region, the region that does not emit light, such as the body portion of the vehicle C, is the binary value of the fire region extraction unit 6 regardless of the result of the minimum value calculation unit 4. All become "0" at the stage of the conversion process. FIG. 3 shows an image after image processing (minimum value calculation, binarization processing) stored in the binarization memory 7. As is clear from this figure, the image (original image) stored in the image memory 3 is shown in FIG. Therefore, only the fire area can be extracted and displayed.
[0017]
As described above in the principle of fire detection, by extracting an area (pixel) having both a large R component and G component from the image memory 3, only the fire area can be extracted. That is, instead of the minimum value calculation unit 4, an R region extraction unit that extracts a region in which the R component exceeds a predetermined value from the color component signal stored in the R component frame memory 3R, and a G component frame memory 3G. A G region extraction unit for extracting a region where the G component exceeds a predetermined value from the color component signal is provided, and a region where the region extracted by the R region extraction unit and the region extracted by the G region extraction unit overlap is a fire region. The fire area can be extracted from the image memory 3 by further providing a fire area extraction unit for extracting as In this case, the R region extraction unit, the G region extraction unit, and the fire region extraction unit are examples of the fire region extraction unit.
[0018]
In this case, the processing step for searching for a pixel exceeding a predetermined value, for example, 180, in the R component frame memory, the processing step for searching for a pixel exceeding a predetermined value, for example, 180, for each pixel in the G component frame memory, and the extracted Three processing steps of searching for pixels where pixels overlap each other are required, but if the minimum value calculation unit 3 is used, the processing step of comparing both R and G components and binarization with a predetermined value are performed. Since the two processing steps of the processing step are sufficient, the fire area can be detected quickly. Moreover, since the two values are simply compared and only the result is output, it is not necessary to perform operations such as addition and multiplication, so that the processing load of the MPU 23 can be reduced.
[0019]
In addition, when the light of the headlight of the following vehicle is shining on the rear part of the vehicle C shown in FIG. 2, the rear glass of the vehicle C causes a specular reflection, and a region where a thin horizontally long light is emitted is generated. This area may be extracted even if the minimum value calculation and binarization processing are performed. Therefore, if an edge processing unit for extracting an edge image of the current image is provided and the edge image is subjected to differential processing from the binarized image after the binarization processing, the edge of the binarized image can be cut. That is, the extraction region of the binarized image is trimmed and becomes a region that is slightly smaller. Therefore, only a region having a certain width (size) remains, and all regions having a small width are deleted. Accordingly, the long and narrow region generated by the specular reflection of the glass can be eliminated by performing such processing.
[0020]
Embodiment 2
By the way, if the tunnel is narrow, there are two-way lanes, and you have to shoot a vehicle that runs in the direction of the surveillance camera, there is a yellow fog lamp (or yellow halogen lamp) in the front of the vehicle, This becomes a false alarm factor. In other words, the fire detection principle of the present invention is to extract a region where both R and G are large. In other words, this is to extract a range of colors ranging from yellow to white. If there is a body, it may be extracted as a fire area.
[0021]
Therefore, in the second embodiment, fire detection with higher accuracy (recognition degree) is performed by digitizing the features of the region extracted in the first embodiment. In other words, the primary fire area is extracted by two processes of the minimum value calculation and the binarization process, and the fire area extracted by the circumscribed rectangle creating means 12, the fire feature amount calculating means, and the fire determining means 14 is the fire area at the time of the fire. A secondary fire detection is performed to determine whether the area is a real fire or not.
[0022]
In FIG. 1, reference numeral 11 denotes correspondence determination means, in the case where fire areas are consecutive in images periodically taken by the monitoring camera 1, that is, when fire areas are continuously stored in the binarization memory 7. , To determine the correspondence between fire areas before and after a certain time, that is, whether the areas are extracted by the same flame, in other words, whether there is a fire area in the monitoring area for a predetermined time Is.
[0023]
A circumscribed rectangle creating means 12 encloses the fire area stored in the binarization memory 7 with the smallest circumscribed rectangle, and is based on the pixel coordinates (fillet value) of the corner of the rectangle (for example, upper left, lower right). The height dy of the rectangle and the width dx (fillet diameter) of the rectangle are calculated, and the value of the ratio of height to width {dx / dy} is obtained and stored in the RAM 22 (see FIG. 3). Reference numeral 13 denotes a fire feature amount calculation means, and reference numeral 14 denotes a fire determination means. The fire area stored in the binarized memory 7 exists for a certain period of time. When the correspondence is established, that is, when it is determined that the area is extracted by the same flame (or light source), it is determined whether or not the fire area is a real fire area. Specifically, the fire feature quantity calculation means 13 is calculated by the circumscribed rectangle creation means 12 and selects the maximum value and the minimum value from the seven {dx / dy} stored in the RAM 22. Is calculated. Further, the fire determination means 15 examines the magnitude relationship between the calculated fire feature value and a predetermined value (1.3 to 1.5), and determines that the fire is found when the fire feature value is larger. Then, the occurrence of a fire is warned from a display unit and an acoustic unit (not shown).
[0024]
Fire feature amount = max {dx / dy} / min {dx / dy} Equation 1
When a fire occurs in the monitoring area, the fire area is extracted to the binarization memory 7 as described in the first embodiment. At this time, when fire areas are stored in the binarized memory 7 continuously in time, the correspondence determining means 11 takes a correspondence relationship between the fire areas. There are several methods for taking this correspondence, for example, there is a correspondence when the fire areas stored in the binarization memory 7 with continuous shooting time (cycles) are overlapped and the degree of overlap is a predetermined value or more. How to judge. In addition, there is a method of determining that there is a response if the center of gravity coordinates of one of the fire areas with consecutive shooting times (cycles) are within the other fire area. Effective when the amount of movement is small. Even if the lamps of the vehicle are temporarily extracted as a fire area by the fire area extraction means, the movement amount of the vehicle is large, so the correspondence between the areas should be taken. There is no. The fire area stored in the binarized memory 7 is surrounded by the smallest rectangle circumscribed by the circumscribed rectangle creating means 12, and the value of the ratio of the height and width {dx / dy} of the rectangle is calculated. It is temporarily stored in the RAM 22.
[0025]
An area having a predetermined brightness in an image captured by the monitoring camera 1 over a predetermined period of time, and fire areas stored in the binarization memory 7 have a correspondence relationship, for example, about 7 times. When the correspondence determination unit 11 determines, the fire feature amount calculation unit 13 calculates the maximum value and the minimum value of {dx / dy} from the data of seven times {dx / dy} stored in the RAM 22 using Equation 1. Calculate the ratio (fire feature). If the fire area stored in the binarization memory 7 is a real fire, the area should change with time. Even if the fire does not progress, the flame has a severe change in height. For this reason, if the value of {dx / dy} is calculated several times within a certain period of time, these values greatly vary, and if the ratio between the maximum value and the minimum value is taken, it generally exceeds about 1.4. On the other hand, when the fire area is a light source such as a headlight, the shape does not change even if the position where the vehicle is moved and shot is different, so the value of {dx / dy} is constant over a period of time. Yes, even if the ratio between the maximum value and the minimum value is taken as shown in Equation 1, the value is infinitely close to 1 and does not exceed the predetermined value 1.3. When the fire feature amount is larger than a predetermined value, the fire determination unit 14 determines that the fire area is a real fire, and warns the occurrence of a fire from a display unit and an acoustic unit (not shown).
[0026]
As described above, if the vehicle is traveling in the monitoring area at a certain speed, even if the fire area extracting means extracts the lamps of the vehicle as the fire area, the correspondence determining means 11 causes the correspondence relationship of the fire areas. Never taken. However, when this vehicle stops or hardly moves in the tunnel, the movement amount is small, so that the correspondence relationship is taken as the fire area, but the fire feature amount calculation means 13 is provided and extracted. By examining whether the fire area behaves like a fire, it is possible to prevent the area extracted by the lamps of the stopped vehicle from being determined as a fire as described above.
[0027]
Here, the operation of the first and second embodiments will be described with reference to the flowchart of FIG. First, in step 1 (hereinafter, step is abbreviated as S), an image taken by the monitoring camera 1 is taken into the image memory 3. As for the image taken into the image memory 3, the minimum value calculation unit 4 compares the R frame memory 3R and the G frame memory 3G for each pixel, and an image having a small luminance value among the RGs is output (S3). The output luminance value is binarized with a predetermined value (S5), and an area having a predetermined value or more is extracted. This extraction area is an area where there is a light source that emits light.
[0028]
In S7, it is determined whether or not there is an area equal to or greater than a predetermined value, that is, an area to be extracted. If not, the process returns to S1 to capture a new image. 12 creates a circumscribed rectangle of the extracted area, calculates a fillet diameter ratio {dx / dy}, and stores the calculated value in the RAM 22. In step S11, the correspondence determination unit 11 determines whether the extracted region has a corresponding relationship with the extracted region in the previously sampled (captured) image. If there is no correspondence relationship, it is recognized as a newly generated area, labeling is performed (numbers are extracted areas), and the process returns to S1.
[0029]
If the correspondence is determined by the correspondence determination unit 11, the process proceeds to S13 to determine whether or not the correspondence has been taken a predetermined time. For example, if the correspondence relationship has only been taken twice, the process returns to S1, and if the correspondence relationship has been taken a predetermined number of times, for example, seven times, the procedure proceeds to S15, where the fire feature quantity calculating means 13 is based on the formula 1 Calculate the quantity. And if the fire determination means 14 judges that the feature-value calculated by S17 is larger than a predetermined value, a fire alarm will be performed in S19. If it is equal to or less than the predetermined value, the process returns to S1.
[0030]
The fire feature amount is obtained from the ratio between the maximum value and the minimum value of {dx / dy} in several times of data, but may be calculated from the amount of change in the ratio between the height and width of the circumscribed rectangle within a predetermined time. . That is, every time the value of {dx / dy} is calculated, the difference from the previous value of {dx / dy} is obtained as an absolute value, and when the calculation is performed a predetermined number of times, the value obtained by adding the difference is greater than the predetermined value. May be determined to be a real fire area. The monitoring area may be a large space such as a stadium or an atrium.
[0031]
【The invention's effect】
By extracting a region where both the R component and the G component are large from the image as a fire region, it is possible to extract only the flame region while omitting an artificial light source such as a sodium lamp or a tail lamp having a low G component. In addition, the time for scanning the pixels in the image processing unit can be shortened, and a fire can be detected quickly.
[Brief description of the drawings]
FIG. 1 is a system block diagram of the present invention.
FIG. 2 is an example of an image projected by a surveillance camera (original image).
FIG. 3 is an example of an image after image processing stored in a binarized memory.
FIG. 4 is a flowchart for explaining the operation of the present invention.
[Explanation of symbols]
1 surveillance camera, 2 analog-digital converter, 3 image memory,
3R R component frame memory, 3G G component frame memory,
3B B component frame memory, 4 minimum value calculation unit, 6 fire area extraction unit 7 binarization memory, 8 image processing unit, 11 correspondence determination means,
12 circumscribed rectangle creating means, 13 fire feature quantity calculating means, 14 fire determining means,
21 ROM, 22 RAM, 23 MPU,

Claims (1)

An image capturing unit that captures an image of a monitoring area and outputs a color image signal composed of R, G, and B color component signals, and an image memory for storing an image captured by the image capturing unit are provided in the image memory. In a fire detection device that detects fire by processing stored images,
Storage means for storing a predetermined value set so that a fire and an artificial light source can be distinguished from an area having a predetermined brightness;
A minimum value calculation unit that outputs a component having a small value by comparing the R component and the G component of the color component signal for each pixel from the image stored in the image memory ;
A fire detection device comprising: a fire region extraction unit that extracts a region where an output signal of the minimum value calculation unit exceeds the predetermined value as a fire region .

Images