JP3827128B2 - Fire detection equipment - Google Patents

Description

図４は、図１の火災判定部９による図２の炎３と壁面反射炎５ａを区別するための判別処理の説明図である。即ち、図４は図２を平面的に見ており、図３の三角測量の原理に基づく火源位置Ｐ（Ｘ，Ｙ，Ｚ）の検出結果から、左側カメラ１及び右側カメラ２の光軸方向での炎３及び壁面反射炎５ａの距離はＺとなっている。
【００２６】
ここで監視空間の壁４ａの位置が例えば左側カメラ１のカメラ光軸１ａを基準としたＸ軸方向の距離として設定されていたとすると、カメラ光軸１ａから壁面４ａまでの設定境界距離はＸｔｈとなっている。そこで、炎３の三角測量に基づく火源位置座標に基づき、カメラ光軸１ａからのＸ軸方向の距離Ｘ０を求め、また壁面反射炎５ａについて三角測量の原理に基づいて検出している火源位置から同じくカメラ光軸１ａに対するＸ軸方向の距離Ｘｒを求める。
【００２７】
このようにして求めた炎３の距離Ｘ０及び壁面反射炎５ａの距離Ｘｒを、予め設定した壁４ａの設定境界距離Ｘｔｈと比較する。炎３の距離Ｘ０は壁４ａの設定境界距離Ｘｔｈより小さいことから、炎３は監視空間にあることが分かる。これに対し壁面反射炎５ａの距離Ｘｒは壁４ａの設定境界距離Ｘｔｈを超えており、監視範囲の外に存在していることが分かる。
【００２８】
このように監視範囲の外に炎が存在することがないため、距離Ｘｒとなった炎５ａは反射による炎として処理対象から除外し、炎３に対してのみ火災による炎と判断して処理を行う。
【００２９】
図５は、図２の監視空間における炎３と床面反射炎５ｂを区別するための図１の火災判定部９による判定処理の説明図である。図５においても、図３の三角測量の原理に基づいて、火源３に加え床４ｂの床面反射炎５ｂについて火源位置が検出されている。
【００３０】
図１の火災判定部９にあっては、例えば側面から見た左側カメラ１及び右側カメラ２の光軸１ａ，２ａを基準に、床４ｂまでの距離を設定境界距離Ｙｔｈとして予め登録している。この状態で炎３及び床面反射炎５ｂの火源位置が検出できたならば、カメラ光軸１ａ，２ａを基準に炎３までの距離Ｙ０と床面反射炎５ｂまでの距離Ｙｒを検出し、それぞれ予め設定した床４ｂまでの設定境界距離Ｙｔｈと比較する。
【００３１】
この場合、炎３までの距離Ｙ０は設定境界距離Ｙｔｈより小さいことから、炎３は監視範囲にある火災による炎であると判定できる。これに対し炎５ｂの距離Ｙｒは設定境界距離Ｙｔｈを超えており、床４ｂの下に存在することになる。このような設定境界距離Ｙｔｈを超える位置の炎５ｂは、火災による炎ではなく床４ｂによる床面反射炎５ｂであることから、これを処理対象から除外する。
【００３２】
次に図６，図７のフローチャートを参照して図１の第１実施形態の火災検出処理を説明する。ここで図６のステップＳ１から図７のステップＳ１０までが図１の画像処理部８による火源位置の検出処理であり、図７のステップＳ１１〜Ｓ１４が図１の火災判定部９による火災判定処理である。
【００３３】
まず図６のステップＳ１で、例えば右側カメラ２で撮影されて右画像メモリ７に記憶された右画像を画像処理部８に取り込み、ステップＳ２で火源の可能性のある画素を１（白画素）とし、火源の可能性のない画素を０（黒画素）とする。続いてステップＳ３で火源の可能性ありとして１にセットされた画素の領域を画素のオブジェクトとし、オブジェクトごとに固有の番号をつけるラベリングを行う。
【００３４】
図８は図６のステップＳ１〜Ｓ３の処理画像である。まず図８（Ａ）が例えば右画像の読取りで得られた元画像であり、炎画像１２ａ、壁面反射炎画像１２ｂ及び床面反射炎画像１２ｃの３つが映っている。この図８（Ａ）の元画像に対し、各画素単位に火災による火源の可能性ありで１（白画素）をセットし、火源の可能性なしのときには０（黒画素）をセットして、図８（Ｂ）の二値化画像とする。
【００３５】
この二値化画像を生成する際の火源の可能性の判定は、例えばカメラとして赤外線カメラを使用している場合には、予め定めておいた温度以上の画素について火災の可能性ありとすればよい。また通常のカメラの場合には、炎の放射光により輝度が高いことから、所定輝度以上の画素を火災の可能性ありと判定すればよい。
【００３６】
この場合、炎特有の特徴であるゆらぎ、移動の有無、重心移動等を捉えて火災の可能性ありの判定を行うことで判定精度を高めることができる。更にカラーカメラの場合には、輝度に加えて火災時の炎の色を判定すればよい。即ち、火災時の炎の色は通常赤色（拡散炎）であることから、例えばカラー画像がＲＧＢデータであった場合には、赤となるＲＧＢ成分の画素を火災の可能性ありと判定して１（白画素）をセットすればよい。
【００３７】
図８（Ｂ）のような二値化により、１で埋められた二値化画像１３ａ，１３ｂが火源画像と認識できる。続いて図８（Ｃ）のように、図８（Ｂ）の二値化画像１３ａ，１３ｂについて火源の可能性ありとして１（白画素）にセットされた二値化画像１３ａ，１３ｂについて、オブジェクト１，２，３としてラベリングしたラベリング画像１４ａ，１４ｂ，１４ｃを生成する。
【００３８】
このような取込画像に対する二値化とラベリングは、図６のステップＳ４〜Ｓ６において、左画像についても同様にして行う。続いてラベリングが済んだ右画面及び左画面について１つのオブジェクトに着目し、ステップＳ８で右画像について着目したオブジェクトの右重心の計測を行い、またステップＳ９で左画像について注目したオブジェクトの左重心の計測を行う。
【００３９】
このステップＳ８，Ｓ９によって、注目したオブジェクトの火源位置が特定される。即ち、図３の左画像１１Ｌ，右画像１１Ｒにおける画像上の火源位置Ｐ１，Ｐ２の座標位置が求められる。続いてステップＳ３で、図３に示す三角測量の原理に基づき、監視範囲の火源位置Ｐの三次元位置を算出する。
【００４０】
次にステップＳ１１に進み、例えば図４，図５に示したように、予め設定した壁４ａや床４ｂに対する設定境界距離Ｘｔｈ，Ｙｔｈとの比較で監視範囲内か否か判別する。監視範囲内であればステップＳ１２に進み、現在処理対象としているオブジェクトは火災による火源と判定する。
【００４１】
ステップＳ１１で監視範囲になければ、ステップＳ１３に進み、全てのオブジェクトの判定が終了したか否かチェックし、終了していなければステップＳ７に戻り、次のオブジェクトについて重心計測を行って三次元位置を求め、ステップＳ１１で監視範囲か否か判定する。
【００４２】
このため、処理対象となるオブジェクトが監視範囲内にあれば直ちに火災判定が行われる。また火源の可能性のあるオブジェクトが実際には火災によるものでなかった場合には、ステップＳ１３でオブジェクトが監視範囲内になく且つ全てのオブジェクトの判定が終了した段階で、ステップＳ４で非火災の判定が行われる。
【００４３】
図９は本発明による火災検出装置の第２実施形態のブロック図である。この第２実施形態にあっては、監視画像から判定した火源位置にスポット光を照射し、スポット光を照射した画像の状態から、火災による炎か炎の反射画像かを判定するようにしたことを特徴とする。
【００４４】
図９において、左側カメラ１、右側カメラ２、左画像メモリ６、右画像メモリ７は図１の実施形態と同じであるが、第２実施形態にあっては更に、スポット光投影部１５、投影制御部１６、左画像処理部１７、右画像処理部１８及び定部１９を設けている。左画像処理部１７と右画像処理部１８は、図１の画像処理部８と同様、三角測量の原理に基づいて火源位置を検出する。即ち、図６及び図７のステップＳ１〜ＳＳ１０と同じ処理により、火源位置が検出される。
【００４５】
スポット光投影部１５は、図１０の平面図のように、例えばカメラ光軸が平行配置された左側カメラ１と右側カメラ２のうちの左側カメラ１の上部に旋回自在に配置されている。このスポット光投影部１５は例えばレーザ発射装置であり、その指向方向を制御することで例えば監視範囲の火源３の位置にスポット光２０を照射できるようにしている。
【００４６】
このスポット光投影部１５を用いた図１０の炎３と、例えば壁４ａに映った壁面反射炎５ａとの判別処理は次のようになる。まず図９の左画像処理部１７は、左側カメラ１で撮影した監視範囲１０Ｌの画像を左画像メモリ６から取り込み、左画面上での炎３及び壁面反射炎５ａの位置を三角測量の原理に基づき検出する。
【００４７】
左側カメラ１の撮影画像から炎３及び壁面反射炎５ａの位置が検出できたならば、炎３に向けてスポット光投影部１５の光軸を光軸１５ａに指向させて、炎３にスポット光２０を照射する。この炎３に対するスポット光２０の照射状態で左側カメラ１及び右側カメラ２により撮影した画像を取り込むと、図１１（Ａ）（Ｂ）のようになる。
【００４８】
図１１（Ａ）の左画像２６Ｌには炎画像２７Ｌと壁面反射炎画像２８Ｌが映っており、炎画像２７Ｌにはスポット光投影部１５で照射したスポット光３０Ｌの輝点が映っている。図１１（Ｂ）の右画像２６Ｒにも炎画像２７Ｒと壁面反射炎画像２８Ｒが映っており、炎画像２７Ｒにはスポット光投影部１５から照射したスポット光３０Ｒの輝点が映っている。
【００４９】
次に図１０でスポット光投影部１５の光軸を光軸１５ｂのように壁面反射炎５ａに指向制御してビームスポットを照射する。しかしながら、光軸１５ｂによるスポット光の照射は、壁４ａに当たってスポット光２３となる。この光軸１５ｂによるスポット光の照射状態で左側カメラ１及び右側カメラ２で撮影した画像を取り込むと、図１２（Ａ）（Ｂ）のようになる。
【００５０】
まず図１２（Ａ）の左画像２６Ｌにあっては、スポット光投影部１５からの光軸１５ｂの一直線上に壁面反射炎５ａと壁４ａに当たったスポット光２３が位置することから、壁面反射炎画像２８Ｌの部分にスポット光３１Ｌの輝点が映っている。
【００５１】
これに対し図１２（Ｂ）の右画像２６Ｒにあっては、図１０のように、右側カメラ２から壁面反射炎５ａを見た光軸２５と壁４ａに当たっているスポット光２３を見た光軸２４にずれがある。このため図１２（Ｂ）の右画像では、壁面反射炎画像２８Ｒの左側のずれた位置にスポット光３１Ｒの輝点が映っている。
【００５２】
したがって、図１１（Ａ）（Ｂ）のように、左右画面２６Ｌ，２６Ｒの同じ炎画像２７Ｌ，２７Ｒにスポット光３０Ｌ，３０Ｒが一致している場合は、監視範囲に実際に存在する炎と判別する。
【００５３】
これに対し図１２（Ａ）（Ｂ）のように、左右画面２６Ｌの炎画像２８Ｌにスポット光３０Ｌが一致し、右画面２６Ｒの同じ炎画像２８Ｒからスポット光３１Ｒがずれている場合には、ずれを起こした炎画像２８Ｒは火災による炎ではなく、壁面反射による炎であることが判別でき、処理対象から除外することができる。
【００５４】
ここで図１０は、図２の監視空間において平面的に見た炎３と壁面反射炎５ａについてのスポット光の照射による判別を例にとっているが、炎３と床面反射炎５ｂについても同様に、各炎位置を検出して順次スポット光を照射してその画像を見ることで、炎画像からずれた位置にスポット光が映っている場合にその炎画像を反射によるものと判別して処理対象から除外すればよい。
【００５５】
図１３は本発明による火災検出装置の第３実施形態であり、図９の第２実施形態におけるスポット光の投影の代わりにレーザ測距を行うようにしたことを特徴とする。
【００５６】
図１３において、左側カメラ１、右側カメラ２、左画像メモリ６、右画像メモリ７は左画像処理部１７及び右画像処理部１８は図９の第２実施形態と同じであるが、スポット光投影機能に代えてレーザ測距部３２、測距制御部３３、更にレーザ測距に対応した火災判定部３４を設けている。
【００５７】
左画像処理部１７と右画像処理部１８は、図１の画像処理部８と同様、三角測量の原理に基づいて火源位置を検出する。即ち、図６及び図７のステップＳ１〜ＳＳ１０と同じ処理により、火源位置が検出される。 レーザ測距部３２はレーザ光を対象物に照射し、その伝播時間により距離を直接検出する。このレーザ測距部３２は、例えば図１４の監視空間に対する平面図のように、カメラ光軸が平行配置された左側カメラ１，右側カメラ２の設置側に三次元的に旋回自在に配置されている。
【００５８】
このレーザ測距部３２を使用した炎とその反射炎を区別する判別処理を説明すると次のようになる。まず左画像処理部１７及び右画像処理部１８は、左側カメラ１及び右側カメラ２で撮影して左画像メモリ６及び右画像メモリ７に記憶した左画像及び右画像を取り込み、図８のような二値化及びラベリングを行った後に重心位置を求め、図２の三角測量の原理に基づきそれぞれの火源位置を検出している。
【００５９】
このようにして、各画像について火源位置が得られたならば、この火源位置情報は測距制御部３３に与えられ、火源位置ごとにレーザ測距部３２を指向制御し、火源までの距離を測定する。即ち図１４にあっては、まずレーザ測距部３２を光軸３５のように炎３に指向させ、炎３までの距離Ｌ１を測定する。続いて壁面反射炎５ａに光軸３６を指向させて距離を測距する。この場合、レーザ測距部３２の光軸３６は壁４ａに点３８で当たり、測定距離Ｌ２は壁面反射炎５ａまでの距離Ｌ２より短くなる。
【００６０】
レーザ測距部３２による炎３及び壁面反射炎５ａの測定距離Ｌ１，Ｌ２は火災判定部３４に与えられ、左画像処理部１７及び右画像処理部１８で三角測量の原理で求めた火源位置から算出されるレーザ測距部３２との算出距離と比較される。この場合、レーザ測距距離と三角測量による検出位置で求めた距離は、炎３については一致するが壁面反射炎５ａについては不一致となり、不一致となった火源を火災によらず反射によるものとして処理対象から除外する。
【００６１】
このレーザ測距は図２の監視空間における平面的に見た炎３と壁面反射炎５ａの判別を例にとっているが、炎３と床面反射炎５ｂについても同様に処理することで、床面反射炎５ｂを処理対象から除外することができる。
【００６２】
図１５は本発明の第４の実施形態である。この第４実施形態は、３台以上のカメラ１ａ，１ｂ，１ｃ，１ｄ，・・・を採用したことを特徴とする。例えば図１〜図８の第１実施形態について、３台のカメラ１ａ，１ｂ，１ｃを設置した場合には、カメラ１ａとカメラ１ｂで取り込んだ画像を画像処理部８で処理し、その結果にカメラ１ｃの画像を画像処理部８に取り込んで火災判定部９で火災位置を判定する。この場合の判定は、図１〜図８の第１実施形態で述べたような方法で可能である。
【００６３】
この場合、カメラ１ａ，１ｂ，１ｃを３次元に配置すれば立体像が得られる。また、炎３を囲むように配置することにより隈無く画像を見ることが出来る。
【００６４】
次にカメラが４台、５台と増しても上記３台の場合と同じ方法を採用して火災位置を判定することが可能である。
【００６５】
また、カメラを偶数台配置した場合、例えばカメラ１ａ〜１ｄの４台を例にとれば、カメラ１ａとカメラ１ｂで画像処理し、カメラ１ｃとカメラ１ｄとで画像処理し、各々の画像処理した結果で更に画像処理することで火災判定してもよい。
【００６６】
次に図９〜図１２の第２実施形態で図１５のようにカメラを３台以上設置した場合は、スポット光投影部１５を付けたカメラを基準に画像処理することになる。例えばカメラ１ａ〜１ｃの３台を設置して、スポット光投影部１５をカメラ１ｂに付けた場合は、まずカメラ１ａとカメラ１ｂで画像処理を実施し、次にカメラ１ｂとカメラ１ｃとで画像処理をして火災判定を実施すれば良い。
【００６７】
更にカメラ１ａ〜１ｄの４台が設置されている場合は、同じくスポット光投影部１５をカメラ１ｂに取り付けておいて、カメラ１ａ〜１ｃの３台の画像処理結果に加えて、カメラ１ｂとカメラ１ｄの画像処理をした結果により火災判定を実施する。以下カメラが５台以上でも同じように処理する。
【００６８】
他の方法として、カメラの設置台数が偶数の場合は、一対のカメラごとにスポット光投影部１５を取り付け、図９〜図１２の第２実施形態のように一対のカメラごとに画像処理を実施し、その各々の結果を総合して火災判定を実施しても良い。
【００６９】
図１３，図１４の第３実施形態について図１５のようにカメラを３台以上設置した場合も、上記の第２実施形態でカメラを３台以上設置した場合について、スポット光投影部１５に代えてレーザ測距部３２を採用することで、同様な方法で火災位置を判定することができる。
【００７０】
尚、本発明の火災検出装置は、監視範囲に固定設置してもよいし、移動可能に設置してもよい。
【００７１】
【発明の効果】
以上説明してきたように本発明によれば、光軸が実質的に平行配置された２台以上のカメラの監視画像から三角測量の原理に基づき火源位置を求め、予め定めた監視範囲との比較、スポット光の照射、あるいはレーザ測距等により、監視範囲内のものだけを火源と判断するようにしたことで、壁や床等で反射した炎を火源と誤って判断することを確実に防止でき、火源位置の検出に基づく消火等の火災検出に伴う対応処理を適切且つ確実に行うことができる。
【図面の簡単な説明】
【図１】本発明の第１実施形態のブロック図
【図２】図１の監視範囲に対するカメラ配置の説明図
【図３】図１の画像処理部による炎画像の位置検出の説明図
【図４】図１の火災判定部による壁の炎反射に対する判定処理の説明図
【図５】図１の火災判定部による床の炎反射に対する判定処理の説明図
【図６】図１の火災検出処理のフローチャート
【図７】図６に続く図１の火災検出処理のフローチャート
【図８】図１の画像処理部による元画像、二値化画像及びラベリング画像の説明図
【図９】スポット光照射を用いた本発明の第２実施形態のブロック図
【図１０】スポット光の照射による炎と炎の反射を区別する処理の説明図
【図１１】図１０で炎にスポット光を照射した時の監視画像の説明図
【図１２】図１１で炎の反射にスポット光を照射した時の監視画像の説明図
【図１３】レーザ測距を用いた本発明の第３実施形態のブロック図
【図１４】図１３で炎及び炎の反射をレーザ測距した時の説明図
【図１５】本発明の第４実施形態のブロック図
【符号の説明】
１：左側カメラ
１ａ〜１ｄ：カメラ
２：右側カメラ
３：炎
４ａ：壁
４ｂ：床
５ａ：壁面反射炎
５ｂ：床反射炎
６：左画像メモリ
７：右画像メモリ
８：画像処理部
９，１９，３４：火災判定部
１０Ｌ，１０Ｒ：監視範囲
１１Ｌ：左画像
１１Ｒ：右画像
１２ａ，２７Ｌ，２７Ｒ：炎画像
１２ｂ，２８Ｌ，２８Ｒ：壁面反射炎画像
１２ｃ：床面反射炎画像
１３ａ，１３ｂ：二値化画像
１４ａ〜１４ｂ：ラベリング画像
１５：スポット光投影部
１６：投影制御部
１７：左画像処理部
１８：右画像処理部
２０，２２：光ビーム
２１，２３：スポット光
２４，２５：カメラ光軸
３０Ｌ，３０Ｒ，３１Ｌ，３１Ｒ：スポット光画像
３２：レーザ測距部
３３：測距制御部
３５，３６：レーザビーム
３７，３８：測距位置[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a fire detection apparatus that extracts a flame image from a monitoring image of a camera and detects a fire source due to a fire.
[0002]
[Prior art]
Conventionally, a fire detection device having a two-dimensional imaging device such as a CCD camera or a two-dimensional array is known.
[0003]
In such a fire detection device, for example, a monitoring image taken by a CCD camera is stored in a memory, a predetermined luminance in the stored monitoring image is extracted by determining a pixel area as a flame area, and the extracted flame area Is considered a fire source.
[0004]
[Problems to be solved by the invention]
However, in such a conventional fire detection device, there is a problem that when the reflected light of the flame is reflected on a surface such as a wall or a floor close to the fire source, it is difficult to distinguish whether it is a fire source or a reflection.
[0005]
In order to distinguish the flame from other light sources, there is a method of using a band-pass filter that passes only a wavelength around 4.3 μm generated by the resonance radiation of CO 2, which is said to be a feature of the flame. In the case of reflection of light, the light reflected from the object is divided into a specular reflection component in which the radiated light is directly reflected and a diffuse reflection component to be radiated after being once absorbed on the object surface. Since the diffuse reflection component shows a spectral distribution representing the material of the object, it can be distinguished by a band pass filter.
[0006]
However, since the specular reflection component has the same spectral distribution as the light source light, the specular reflection component increases as the angle of the reflection surface becomes shallower, and the reflection of flame and flame cannot be distinguished by the bandpass filter.
[0007]
In addition, using the property that a part of the specular reflection component becomes polarized light, a linear polarizing filter is attached to the front surface of the lens, and if the polarization of the reflected light and the angle of the polarizing filter match, the polarized portion of the specular reflection component can be removed. it can. However, the direction of polarization differs between the reflection of the floor surface and the wall surface, and cannot be removed at the same time.
[0008]
The present invention has been made in view of such problems, and an object of the present invention is to provide a fire detection apparatus that reliably detects a fire source by distinguishing between a flame and a reflection of the flame included in a monitoring image.
[0009]
[Means for Solving the Problems]
In order to achieve this object, the present invention is configured as follows.
[0012]
Of the present invention Fire detection device Is arranged at two or more different points, two or more cameras for capturing a monitoring image of the fire source in the monitoring range, a spot light projection unit that irradiates the spot light toward the fire source position of the monitoring range, An image processing unit that detects a fire source position based on the principle of triangulation from two or more fire source images included in each of monitoring images taken by two or more cameras, and a fire source position detected by the image processing unit. When the spotlight projection unit irradiates the spotlight and the position of the spotlight image and the flame image of the monitoring image taken with two or more cameras match, it is determined as a fire source due to fire, and the position is shifted. And a fire determination unit characterized in that it is determined as a fire source due to reflection and excluded from the processing target.
[0013]
In such a fire detection device of the present invention, By judging only those within the monitoring range as the fire source, it is possible to reliably prevent the flame reflected from the floor or wall from being mistakenly judged as the fire source.
[0014]
Furthermore, in another form of the present invention, two or more cameras that are arranged at two or more different points and take a monitoring image of a fire source in the monitoring range, and two or more cameras are arranged, Based on the principle of triangulation based on the principle of triangulation from a distance measuring unit that measures the distance to an object existing in the monitoring range and two or more fire source images included in each of the monitoring images taken by two or more cameras Measure the distance by pointing the distance measuring device to the image processing unit to be detected and the fire source position detected by the image processing unit, and if the measured distance of the image processing unit and the measured distance of the distance measuring device match, If it is determined as a fire source and does not match, the fire is determined as a fire source by reflection and excluded from the processing target.
[0015]
Also in this case, by determining the fire source based on the distance measurement, it is possible to reliably prevent the flame reflected from the floor or wall from being erroneously determined as the fire source.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows a first embodiment of a fire detection apparatus according to the present invention, which detects the position of a flame image included in a monitoring image based on the principle of triangulation, and determines whether it is a flame caused by a fire or a flame caused by reflection. It is characterized by that.
[0017]
In FIG. 1, the fire detection apparatus according to the first embodiment includes a left camera 1, a right camera 2, a left image memory 6, a right image memory 7, an image processing unit 8, and a fire determination unit 9. The left camera 1 and the right camera 2 have monitoring ranges 10L and 10R, respectively, and monitor the flame 3 that is a detection target in the monitoring space.
[0018]
As shown in FIG. 2, the left camera 1 and the right camera 2 are arranged so that the camera optical axes 1a and 2a are parallel or intersect with the surveillance section, and the relative relationship between the camera optical axes 1a and 2a is maintained. The left camera 1 and the right camera 2 are directed toward the monitoring space, and images of the monitoring ranges 10L and 10R are taken.
[0019]
The monitoring space in FIG. 2 is partitioned by a wall 4a and a floor 4b. A fire is generated in this monitoring space and a flame 3 is present. A wall reflection flame 5a is reflected on the wall 4a facing the flame 3 due to fire. A floor reflection flame 5b caused by the flame 4 is also reflected on the floor 4b. For this reason, when the monitoring space is imaged by the left camera 1 and the right camera 2, the three monitoring images include the flame 3, the wall reflection flame 5a, and the floor reflection flame 5b.
[0020]
The monitoring images of the monitoring ranges 10L and 10R simultaneously captured by the left camera 1 and the right camera 2 in FIG. 1 are written into the left image memory 6 and the right image memory 7, respectively, and then read out by the image processing unit 8. The fire source position is detected from a pair of fire source images included in each monitoring image based on the principle of triangulation.
[0021]
In the case of FIG. 2, the fire source positions detected by the image processing unit 8 are detected as the three fire source positions of the flame 3, the wall reflection flame 5a, and the floor reflection flame 5b. The positions of one or more fire source images detected by the image processing unit 8 are given to the fire determination unit 9. In the fire determination unit 9, coordinate information of the boundary position of the wall 4a, the floor 4b, etc. in the monitoring space of FIG. 2 is set in advance, and the fire source position detected by the image processing unit 8 is set in advance. If it is inside the set boundary position of 4b, it is determined as a fire source due to a fire. On the other hand, when the detected fire source position exceeds the set boundary range, it is excluded from the processing target as a fire source by reflection.
[0022]
FIG. 3 is an explanatory diagram of a fire source position detection process using the principle of triangulation by the image processing unit 8 of FIG. In FIG. 3, if the camera position corresponding to the imaging surface of the left camera 1 in FIG. 1 is F1, and the camera position corresponding to the imaging surface of the right camera 2 is F2, the distance between the camera positions F1 and F2 is, for example, a distance d. It is determined in advance.
[0023]
In the left image 11L and the right image 11R photographed by the camera optical axes 1a and 2a arranged from the camera positions F1 and F2, the fire source images P1 and P2 of the fire source position P of the flame due to fire are at the illustrated positions. Suppose it was reflected. In this case, the fire source position P in the actual monitoring space includes a straight line extending from the camera position F1 and the fire source position P1 on the left image 11L, and a fire source position P2 on the right image 11R. It is given as an intersection on the extended line of connected lines. For this reason, the three-dimensional coordinates (X, Y, Z) of the fire source position P can be obtained by the principle of triangulation.
[0024]
To simplify the description, if the camera optical axes 1a and 2a are completely parallel and the vertical coordinate positions z of the camera positions F1 and F2 are the same, the fire source position P1 of the left image 11L The coordinates are (x1, y1), and the coordinates of the fire source position P2 in the right image 11R are (x2, y2). The focal length of the camera positions F1 and F2 with respect to the left image 11L and the right image 11R is assumed to be f. In this case, the value of the three-dimensional coordinates (X, Y, Z) of the fire source position P in the monitoring space is given by the following equation.
[0025]
[Expression 1]

FIG. 4 is an explanatory diagram of a determination process for distinguishing the flame 3 of FIG. 2 from the wall reflection flame 5a by the fire determination unit 9 of FIG. That is, FIG. 4 is a plan view of FIG. 2. From the detection result of the fire source position P (X, Y, Z) based on the principle of triangulation in FIG. The distance between the flame 3 and the wall reflection flame 5a in the direction is Z.
[0026]
If the position of the wall 4a in the monitoring space is set as a distance in the X-axis direction with respect to the camera optical axis 1a of the left camera 1, for example, the set boundary distance from the camera optical axis 1a to the wall surface 4a is Xth. It has become. Therefore, the distance X0 in the X-axis direction from the camera optical axis 1a is obtained based on the fire source position coordinates based on the triangulation of the flame 3, and the fire source in which the wall reflection flame 5a is detected based on the triangulation principle. Similarly, the distance Xr in the X-axis direction with respect to the camera optical axis 1a is obtained from the position.
[0027]
The distance X0 of the flame 3 and the distance Xr of the wall reflection flame 5a thus obtained are compared with a preset boundary distance Xth of the wall 4a. Since the distance X0 of the flame 3 is smaller than the set boundary distance Xth of the wall 4a, it can be seen that the flame 3 is in the monitoring space. On the other hand, it can be seen that the distance Xr of the wall reflection flame 5a exceeds the set boundary distance Xth of the wall 4a and exists outside the monitoring range.
[0028]
Since there is no flame outside the monitoring range in this way, the flame 5a having the distance Xr is excluded from the processing target as a flame due to reflection, and only the flame 3 is determined to be a flame due to a fire and processed. Do.
[0029]
FIG. 5 is an explanatory diagram of determination processing by the fire determination unit 9 of FIG. 1 for distinguishing between the flame 3 and the floor reflection flame 5b in the monitoring space of FIG. Also in FIG. 5, based on the principle of triangulation in FIG. 3, the fire source position is detected for the floor reflection flame 5b of the floor 4b in addition to the fire source 3.
[0030]
In the fire determination unit 9 in FIG. 1, for example, the distance to the floor 4b is registered in advance as the set boundary distance Yth with reference to the optical axes 1a and 2a of the left camera 1 and the right camera 2 viewed from the side. . If the fire source positions of the flame 3 and the floor reflection flame 5b can be detected in this state, the distance Y0 to the flame 3 and the distance Yr to the floor reflection flame 5b are detected based on the camera optical axes 1a and 2a. Then, each is compared with a preset boundary distance Yth to the floor 4b set in advance.
[0031]
In this case, since the distance Y0 to the flame 3 is smaller than the set boundary distance Yth, it can be determined that the flame 3 is a flame due to a fire in the monitoring range. On the other hand, the distance Yr of the flame 5b exceeds the set boundary distance Yth and exists under the floor 4b. Since the flame 5b at a position exceeding the set boundary distance Yth is not a flame due to a fire but a floor reflection flame 5b due to the floor 4b, it is excluded from the processing target.
[0032]
Next, the fire detection process of the first embodiment of FIG. 1 will be described with reference to the flowcharts of FIGS. Here, Step S1 in FIG. 6 to Step S10 in FIG. 7 are detection processing of the fire source position by the image processing unit 8 in FIG. 1, and Steps S11 to S14 in FIG. 7 are fire determinations by the fire determination unit 9 in FIG. It is processing.
[0033]
First, in step S1 of FIG. 6, for example, the right image captured by the right camera 2 and stored in the right image memory 7 is taken into the image processing unit 8, and in step S2, a pixel that may be a fire source is set to 1 (white pixel). ), And pixels having no possibility of fire source are set to 0 (black pixels). Subsequently, in step S3, a pixel region set to 1 as the possibility of a fire source is set as a pixel object, and labeling is performed by assigning a unique number to each object.
[0034]
FIG. 8 is a processed image of steps S1 to S3 in FIG. First, FIG. 8A is an original image obtained by reading the right image, for example, and shows three images of a flame image 12a, a wall reflection flame image 12b, and a floor reflection flame image 12c. In the original image of FIG. 8A, 1 (white pixel) is set for each pixel unit when there is a possibility of fire, and 0 (black pixel) is set when there is no possibility of fire. Thus, the binarized image of FIG.
[0035]
For example, when an infrared camera is used as the camera, the possibility of a fire source when generating the binarized image is determined to be a possibility of fire for pixels having a predetermined temperature or more. That's fine. In the case of a normal camera, the luminance is high due to the radiated light from the flame, so that a pixel having a predetermined luminance or higher may be determined as a possibility of fire.
[0036]
In this case, it is possible to improve the determination accuracy by determining the possibility of a fire by capturing fluctuations, presence / absence of movement, movement of the center of gravity, and the like, which are characteristics peculiar to flames. Furthermore, in the case of a color camera, the color of the flame at the time of fire may be determined in addition to the luminance. That is, since the color of the flame at the time of fire is usually red (diffusion flame), for example, when the color image is RGB data, the pixel of the RGB component that becomes red is determined to be a fire possibility. 1 (white pixel) may be set.
[0037]
By binarization as shown in FIG. 8B, the binarized images 13a and 13b filled with 1 can be recognized as fire source images. Subsequently, as shown in FIG. 8C, regarding the binarized images 13a and 13b set to 1 (white pixels) as the possibility of a fire source for the binarized images 13a and 13b in FIG. 8B, Labeled images 14a, 14b, and 14c labeled as objects 1, 2, and 3 are generated.
[0038]
Such binarization and labeling for the captured image are similarly performed for the left image in steps S4 to S6 in FIG. Subsequently, focusing on one object on the right screen and the left screen after labeling, the right center of gravity of the object focused on the right image is measured in step S8, and the left center of gravity of the object focused on the left image is measured in step S9. Measure.
[0039]
Through these steps S8 and S9, the fire source position of the object of interest is specified. That is, the coordinate positions of the fire source positions P1, P2 on the images in the left image 11L and the right image 11R in FIG. 3 are obtained. In step S3, the three-dimensional position of the fire source position P in the monitoring range is calculated based on the principle of triangulation shown in FIG.
[0040]
Next, the process proceeds to step S11, and as shown in FIGS. 4 and 5, for example, it is determined whether or not it is within the monitoring range by comparing with the set boundary distances Xth and Yth with respect to the preset wall 4a and floor 4b. If it is within the monitoring range, the process proceeds to step S12, and the object currently processed is determined to be a fire source due to fire.
[0041]
If it is not in the monitoring range in step S11, the process proceeds to step S13, where it is checked whether or not all objects have been determined. If not, the process returns to step S7, and the center of gravity is measured for the next object to obtain the three-dimensional position. In step S11, it is determined whether or not it is within the monitoring range.
[0042]
For this reason, if the object to be processed is within the monitoring range, the fire determination is performed immediately. If an object that may be a fire source is not actually due to a fire, the object is not in the monitoring range in step S13, and when all the objects have been determined, a non-fire is detected in step S4. Is determined.
[0043]
FIG. 9 is a block diagram of a second embodiment of the fire detection device according to the present invention. In this second embodiment, spot light is irradiated to the fire source position determined from the monitoring image, and it is determined from the state of the image irradiated with the spot light whether it is a fire flame or a flame reflection image. It is characterized by that.
[0044]
In FIG. 9, the left camera 1, the right camera 2, the left image memory 6, and the right image memory 7 are the same as those in the embodiment shown in FIG. 1, but in the second embodiment, the spot light projection unit 15, the projection A control unit 16, a left image processing unit 17, a right image processing unit 18, and a fixed unit 19 are provided. The left image processing unit 17 and the right image processing unit 18 detect the fire source position based on the principle of triangulation, like the image processing unit 8 of FIG. That is, the fire source position is detected by the same processing as steps S1 to SS10 in FIGS.
[0045]
As shown in the plan view of FIG. 10, the spot light projection unit 15 is rotatably disposed on the left camera 1 of the left camera 1 and the right camera 2 in which the camera optical axes are arranged in parallel, for example. The spot light projection unit 15 is, for example, a laser emitting device, and the spot light 20 can be irradiated to, for example, the position of the fire source 3 in the monitoring range by controlling the directing direction thereof.
[0046]
The discrimination process between the flame 3 of FIG. 10 using the spot light projection unit 15 and the wall reflection flame 5a reflected on the wall 4a, for example, is as follows. First, the left image processing unit 17 in FIG. 9 captures an image of the monitoring range 10L taken by the left camera 1 from the left image memory 6, and uses the positions of the flame 3 and the wall reflection flame 5a on the left screen as the principle of triangulation. Detect based on.
[0047]
If the positions of the flame 3 and the wall reflection flame 5a can be detected from the photographed image of the left camera 1, the optical axis of the spot light projection unit 15 is directed toward the optical axis 15a toward the flame 3, and the spot 3 is irradiated with the spot light. 20 is irradiated. When images taken by the left camera 1 and the right camera 2 in the state of irradiation of the spot light 20 with respect to the flame 3 are captured, it becomes as shown in FIGS.
[0048]
A flame image 27L and a wall reflection flame image 28L are shown in the left image 26L of FIG. 11A, and a bright spot of the spot light 30L irradiated by the spot light projection unit 15 is shown in the flame image 27L. A flame image 27R and a wall reflection flame image 28R are also shown in the right image 26R of FIG. 11B, and the bright spot of the spot light 30R emitted from the spot light projection unit 15 is shown in the flame image 27R.
[0049]
Next, in FIG. 10, the beam spot is irradiated by controlling the direction of the optical axis of the spot light projection unit 15 to the wall reflection flame 5a as in the optical axis 15b. However, the irradiation of the spot light by the optical axis 15b hits the wall 4a and becomes the spot light 23. When images taken by the left camera 1 and the right camera 2 are captured in the spot light irradiation state by the optical axis 15b, the images are as shown in FIGS.
[0050]
First, in the left image 26L of FIG. 12 (A), the wall surface reflection flame 5a and the spot light 23 that hits the wall 4a are positioned on a straight line of the optical axis 15b from the spot light projection unit 15. The bright spot of the spot light 31L is shown in the flame image 28L.
[0051]
On the other hand, in the right image 26R in FIG. 12B, as shown in FIG. 10, the optical axis 25 when the right camera 2 looks at the wall reflection flame 5a and the optical axis when the spot light 23 hitting the wall 4a is seen. There is a shift in 24. For this reason, in the right image of FIG. 12B, the bright spot of the spot light 31R is shown at a position shifted to the left of the wall reflection flame image 28R.
[0052]
Accordingly, as shown in FIGS. 11A and 11B, when the spot lights 30L and 30R coincide with the same flame images 27L and 27R on the left and right screens 26L and 26R, it is determined that the flame is actually present in the monitoring range. To do.
[0053]
On the other hand, as shown in FIGS. 12A and 12B, when the spot light 30L matches the flame image 28L on the left and right screens 26L, and the spot light 31R deviates from the same flame image 28R on the right screen 26R, It is possible to determine that the flame image 28R that has shifted is not a flame due to a fire but a flame due to wall reflection, and can be excluded from the processing target.
[0054]
Here, FIG. 10 shows an example of discrimination by irradiation of spot light for the flame 3 and the wall reflection flame 5a viewed in plan in the monitoring space of FIG. 2, but the same applies to the flame 3 and the floor reflection flame 5b. Detects each flame position, sequentially irradiates spot light and looks at the image, and if spot light is reflected at a position deviated from the flame image, the flame image is determined to be a reflection object. Can be excluded.
[0055]
FIG. 13 shows a third embodiment of the fire detection apparatus according to the present invention, which is characterized in that laser ranging is performed instead of spot light projection in the second embodiment of FIG.
[0056]
In FIG. 13, the left camera 1, the right camera 2, the left image memory 6, and the right image memory 7 are the same as the second embodiment of FIG. 9, but the left image processing unit 17 and the right image processing unit 18 are the same as those in the second embodiment of FIG. In place of the function, a laser distance measurement unit 32, a distance measurement control unit 33, and a fire determination unit 34 corresponding to laser distance measurement are provided.
[0057]
The left image processing unit 17 and the right image processing unit 18 detect the fire source position based on the principle of triangulation, like the image processing unit 8 of FIG. That is, the fire source position is detected by the same processing as steps S1 to SS10 in FIGS. The laser distance measuring unit 32 irradiates the object with laser light and directly detects the distance based on the propagation time. For example, as shown in the plan view of the monitoring space in FIG. 14, the laser distance measuring unit 32 is arranged in a three-dimensionally rotatable manner on the installation side of the left camera 1 and the right camera 2 in which the camera optical axes are arranged in parallel. Yes.
[0058]
A discrimination process for distinguishing between a flame using the laser distance measuring unit 32 and its reflection flame will be described as follows. First, the left image processing unit 17 and the right image processing unit 18 take in the left image and the right image which are taken by the left camera 1 and the right camera 2 and stored in the left image memory 6 and the right image memory 7, as shown in FIG. After performing binarization and labeling, the position of the center of gravity is obtained, and each fire source position is detected based on the principle of triangulation in FIG.
[0059]
In this way, if the fire source position is obtained for each image, this fire source position information is given to the distance measurement control unit 33, and the laser distance measurement unit 32 is directed and controlled for each fire source position. Measure the distance to. That is, in FIG. 14, first, the laser distance measuring unit 32 is directed toward the flame 3 like the optical axis 35, and the distance L <b> 1 to the flame 3 is measured. Subsequently, the distance is measured by directing the optical axis 36 toward the wall reflection flame 5a. In this case, the optical axis 36 of the laser distance measuring unit 32 hits the wall 4a at a point 38, and the measurement distance L2 is shorter than the distance L2 to the wall reflection flame 5a.
[0060]
The measurement distances L1 and L2 of the flame 3 and the wall reflection flame 5a by the laser ranging unit 32 are given to the fire determining unit 34, and the fire source position obtained by the left image processing unit 17 and the right image processing unit 18 on the principle of triangulation Is compared with the calculated distance from the laser distance measuring unit 32. In this case, the distance obtained by the laser ranging distance and the detection position by triangulation is the same for the flame 3 but not for the wall reflection flame 5a, and the mismatched fire source is assumed to be due to reflection regardless of the fire. Exclude from processing.
[0061]
The laser distance measurement uses an example of discrimination between the flame 3 and the wall reflection flame 5a as viewed in plan in the monitoring space of FIG. 2, but the flame 3 and the floor reflection flame 5b are processed in the same manner to obtain the floor surface. The reflection flame 5b can be excluded from the processing target.
[0062]
FIG. 15 shows a fourth embodiment of the present invention. The fourth embodiment is characterized in that three or more cameras 1a, 1b, 1c, 1d,. For example, in the first embodiment shown in FIGS. 1 to 8, when three cameras 1a, 1b, and 1c are installed, images captured by the cameras 1a and 1b are processed by the image processing unit 8, and the result is obtained. The image of the camera 1c is taken into the image processing unit 8, and the fire determination unit 9 determines the fire position. The determination in this case can be made by the method described in the first embodiment of FIGS.
[0063]
In this case, if the cameras 1a, 1b, and 1c are arranged three-dimensionally, a stereoscopic image can be obtained. In addition, the image can be seen without hesitation by arranging it so as to surround the flame 3.
[0064]
Next, even if the number of cameras is increased to four or five, it is possible to determine the fire position by adopting the same method as in the case of the above three.
[0065]
Further, when an even number of cameras are arranged, for example, when four cameras 1a to 1d are taken as an example, image processing is performed by the camera 1a and the camera 1b, image processing is performed by the camera 1c and the camera 1d, and each image processing is performed. The fire may be determined by further image processing based on the result.
[0066]
Next, in the second embodiment of FIGS. 9 to 12, when three or more cameras are installed as shown in FIG. 15, image processing is performed based on the camera with the spot light projection unit 15. For example, when three cameras 1a to 1c are installed and the spot light projection unit 15 is attached to the camera 1b, image processing is first performed by the camera 1a and the camera 1b, and then an image is captured by the camera 1b and the camera 1c. It is only necessary to perform a fire judgment after processing.
[0067]
Further, when four cameras 1a to 1d are installed, the spot light projection unit 15 is similarly attached to the camera 1b, and in addition to the image processing results of the three cameras 1a to 1c, the camera 1b and the camera 1b Fire determination is performed based on the result of the 1d image processing. The same processing is performed for five or more cameras.
[0068]
As another method, when the number of installed cameras is an even number, a spot light projection unit 15 is attached to each pair of cameras, and image processing is performed for each pair of cameras as in the second embodiment of FIGS. However, the fire determination may be performed by combining the results.
[0069]
In the third embodiment shown in FIGS. 13 and 14, even when three or more cameras are installed as shown in FIG. 15, the case where three or more cameras are installed in the second embodiment is replaced with the spot light projector 15. By adopting the laser distance measuring unit 32, the fire position can be determined by the same method.
[0070]
The fire detection device of the present invention may be fixedly installed in the monitoring range or may be installed movably.
[0071]
【The invention's effect】
As described above, according to the present invention, the fire source position is obtained based on the principle of triangulation from the monitoring images of two or more cameras having optical axes arranged substantially in parallel, and the predetermined monitoring range is obtained. By comparing, spot light irradiation, or laser ranging, etc., only fires within the monitoring range are judged as fire sources, so that flames reflected on walls and floors are mistakenly judged as fire sources. It can prevent reliably and can perform appropriately and reliably the response | compatibility process accompanying fire detection, such as fire extinguishing based on the detection of a fire source position.
[Brief description of the drawings]
FIG. 1 is a block diagram of a first embodiment of the present invention.
FIG. 2 is an explanatory diagram of camera arrangement with respect to the monitoring range of FIG.
FIG. 3 is an explanatory diagram of flame image position detection by the image processing unit of FIG. 1;
FIG. 4 is an explanatory diagram of determination processing for flame reflection on a wall by the fire determination unit in FIG.
5 is an explanatory diagram of a determination process for floor flame reflection by the fire determination unit of FIG.
FIG. 6 is a flowchart of the fire detection process of FIG.
7 is a flowchart of the fire detection process of FIG. 1 following FIG.
8 is an explanatory diagram of an original image, a binarized image, and a labeling image by the image processing unit of FIG.
FIG. 9 is a block diagram of a second embodiment of the present invention using spot light irradiation.
FIG. 10 is an explanatory diagram of a process for distinguishing between flames reflected by spot light and flame reflections
11 is an explanatory diagram of a monitoring image when spot light is irradiated on the flame in FIG. 10;
12 is an explanatory diagram of a monitoring image when spot light is irradiated on the reflection of the flame in FIG. 11;
FIG. 13 is a block diagram of a third embodiment of the present invention using laser ranging.
FIG. 14 is an explanatory diagram when the laser and the flame reflection of the flame and the reflection of the flame are measured in FIG.
FIG. 15 is a block diagram of a fourth embodiment of the present invention.
[Explanation of symbols]
1: Left camera
1a to 1d: Camera
2: Right camera
3: Flame
4a: Wall
4b: Floor
5a: Wall reflection flame
5b: Floor reflection flame
6: Left image memory
7: Right image memory
8: Image processing unit
9, 19, 34: Fire judgment part
10L, 10R: Monitoring range
11L: Left image
11R: Right image
12a, 27L, 27R: Flame images
12b, 28L, 28R: Wall reflection flame image
12c: Floor reflection flame image
13a, 13b: binarized image
14a-14b: Labeling images
15: Spot light projection unit
16: Projection control unit
17: Left image processing unit
18: Right image processing unit
20, 22: Light beam
21, 23: Spot light
24, 25: Camera optical axis
30L, 30R, 31L, 31R: Spot light image
32: Laser ranging unit
33: Ranging control unit
35, 36: Laser beam
37, 38: Ranging position

Claims (2)

Two or more cameras that are arranged at two or more different points and take a monitoring image of a fire source in the monitoring range;
A spot light projection unit that emits spot light toward the fire source position of the monitoring range;
An image processing unit for detecting a fire source position based on the principle of triangulation from two or more fire source images included in each of the monitoring images taken by the two or more cameras;
The positions of the flame image and the spot light image of the monitoring image taken by the two or more cameras coincide with the spot position projected from the spot light projection unit on the fire source position detected by the image processing unit. A fire determination unit characterized in that it is determined as a fire source due to fire, and if it is out of position, it is determined as a fire source due to reflection and excluded from the processing target;
A fire detection device characterized by comprising: Two or more cameras that are arranged at two or more different points and take a monitoring image of a fire source in the monitoring range;
A distance measuring unit that is disposed on the side of the two or more cameras and measures a distance to an object existing in a range;
An image processing unit for detecting a fire source position based on the principle of triangulation from two or more fire source images included in each of the monitoring images taken by the two or more cameras;
A distance is measured by directing the distance measuring device to the fire source position detected by the image processing unit, and when the measurement distance matches the measurement distance of the image processing unit, it is determined as a fire source due to a fire and does not match A fire detection device comprising: a fire determination unit that is excluded from a processing target as a fire source due to reflection.

Images