JP2004257952A - Distance measuring method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enable highly precise distance measurement even if an edge noise (a rapid change of a signal) occurs at the end of only a part that is regarded as the same at one side by a parallax in the case of a subject having a high contrast when determining displacements between two signals of a subject image to be photographed for detecting the distance to the subject. <P>SOLUTION: In the distance measuring method, the presence of the edge noise is judged by comparing the difference between inclinations of the ends or between image signals of a division for the image signal regarded as the same. In the case where it is judged as the presence of the edge noise, which of the signals of the subjects is to be corrected is judged, comparing with the absolute values of the inclinations of the ends as needed, and the image signal of the ends is corrected based on the other image signal. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

【０００６】
【数３】

もし、フォトセンサアレイ５３，５４上の２つの像が相対的にｉ_０フォトセンサピッチずれているとすれば、ｆ（ｉ_０）＝０となる。ただし、通常は前記Ｘが完全にフォトセンサピッチの整数倍になることや、２つのフォトセンサからの被写体像信号の形が（相対的にずれていることを除き）完全に同じになることは少ないので、ｆ（ｉ_０）＞０となる。これ考え方に基づき、距離測定装置はｆ（ｉ）の最小値を与えるｉ＝ｉ_０を求めて、ｉ_０もしくはこれにフォトセンサピッチｐを乗じたものをＸとする。また、被写体像信号Ｌ（ｊ）およびＲ（ｊ）のうち、ｆ（ｉ）の計算に使われる部分をｆ（ｉ）の演算領域と定義する。なお、より高精度な検出を行うために通常はｆ（ｉ_０），ｆ（ｉ_０−１），ｆ（ｉ_０−２），ｆ（ｉ_０＋１），ｆ（ｉ_０＋２）を用いて補間計算を実施するが、その詳細については説明を省略する。
フォトセンサアレイ５３，５４の出力にノイズがのると前記相対変位Ｘすなわち距離検出の精度に悪影響を与えるため、例えば特許文献１のように種々の対策が考案されてきた。
特許文献１は被写体のバックに図７の７０のような照明や太陽といった光源があり、その光源からの光により被写体の像信号が変化してしまう、いわゆるフレアがある場合、フォトセンサアレイ５３とレンズ５１で構成される光学系とフォトセンサアレイ５４とレンズ５２で構成される光学系の視差により一方の光学系に強いフレアが生じた場合の対処に関するものである。すなわち、図７の構成を図９のように描きなおすと、領域ＬＲが２つのフォトセンサ５１および５３に共通に投影される領域、領域ＲＲがフォトセンサ５４にのみ投影される領域、領域ＬＬがフォトセンサ５３にのみ投影される領域となる。領域ＬＬもしくはＲＲに上述の光源があると、片方のフォトセンサのみその影響を受けることになる。例えば、図７の光源７０のように領域ＲＲにフレアの発生源があると、図１０（ａ）に示す正常時の被写体像信号が、図１０（ｂ）に示すように信号Ｒの左側の端部がフレアのため特に大きく変形したものになってしまう。特許文献１はこの場合のフレアの影響を除去して、正確な測距を行おうとするものである。なお、図９においてＬ１，Ｒ１はそれぞれ２つのフォトセンサアレイ５３，５４の最端部のセンサのセンサＮＯである。該センサＮＯは図１０の横軸となっていて、図１０の横軸の最も左側がＬ１，Ｒ１となっている。また、図１０のＬ，Ｒはそれぞれフォトセンサアレイ５３，５４から出力される被写体像信号である。
すなわち、フレアがある状況では、被写体から光に光源７０からの光が加算されて被写体像信号全体がバイアスされる形でかさ上げされ、さらに光源７０の影響が２つのフォトセンサアレイ５３，５４で異なるため、それぞれの被写体像信号がバイアス分を差し引いても一致しない部分ができてしまう。２つの被写体像信号の差がバイアス分だけであれば、そのバイアスを相殺するように２つの被写体像信号の平均値差を一方の被写体像信号に対し加算もしくは減算する補正を行えば良いが、図１０（ｂ）に示すように被写体像信号の一部変形が起こっている場合に対しては有効ではない。このような被写体に対し特許文献１では２つの被写体像信号に関し、視差の影響の出ない部分に制限してバイアス補正する方法や一方の被写体像信号に少しずつバイアスを加えていき、その都度２つの被写体像信号の一致度を判断してその一致度が所定値より低くなったときに補正完了とする方法が示されている。
【０００７】
【特許文献１】
特開２００２−２５０８５７号公報
【０００８】
【発明が解決しようとする課題】
特許文献１などのようにフレアに対し２つのフォトセンサアレイから得られる２つの被写体像信号の一方にバイアスを加える方法は、エッジノイズがあるものには適さないものである。エッジノイズとは、被写体の高いコントラストにより、上述の演算領域の一方だけに、端部の被写体像信号の急激な変化が現れるものである。これは、例えば、領域ＬＲと領域ＲＲの境界近傍に高いコントラスト変化部があり、フォトセンサアレイ５４にのみその変化部が投影されると、図１１（ａ）に示すように被写体像信号Ｒだけにエッジでの急激な変化Ｅ_Ｒが生じるからである。また、領域ＬＲと領域ＬＬの境界近傍に高いコントラスト変化部があると、同様に図１１（ｂ）に示すように被写体像信号Ｌだけにエッジでの急激な変化Ｅ_Ｌが生じることもある。さらに、Ｅ_Ｒ，Ｅ_Ｌは、領域ＬＲと領域ＲＲの境界近傍もしくは領域ＬＲと領域ＬＬの境界近傍に高いコントラスト変化部がある場合だけではなく、領域ＬＲにある被写体に対する被写体像信号の一部を切り取って上述の演算領域とするときに、切り取る部分の近傍付近に高いコントラスト変化があるときも生じることがある。Ｅ_Ｒ，Ｅ_Ｌは被写体のコントラストによるもので、正確に言えばノイズではないが、演算領域において２つの被写体像信号の片方だけに現れ、相対変位Ｘを求める際の誤差要因となるので、エッジノイズと呼ばれる。エッジノイズは、演算領域において、最端部の被写体像信号に大きな変化を与えて演算領域の相似性を崩すが、他の被写体像信号の相似性は保たれるため、主に補間計算の精度に影響を与えて高精度な距離測定を妨げる。エッジノイズは上述のように被写体自体のコントラストによるものであり、これにより他の部分の被写体像信号にバイアスを乗せるということはないため、フレア対策のように一方の被写体像信号にバイアスを与えて補正するという手法では対処できない。
【０００９】
そこで、本発明は上記課題を解決するためになされたものであり、その目的は、少なくともｆ（ｉ_０）の演算領域の被写体像信号にエッジノイズがあるかどうかを判定し、エッジノイズがあると判定した場合は被写体像信号を適切に補正することにより、高精度な距離測定を行うことのできる距離測定方法を提供することにある。
【００１０】
【課題を解決するための手段】
そこで、上記課題を解決するために、請求項１に係る発明は、被写体から放射または反射される光束のうちの互いに異なる部分の光束を２つのフォトセンサアレイ上にそれぞれ投影し、該各フォトセンサアレイ上に形成される被写体像の相対変位から前記被写体までの距離を求めるために、
前記２つのフォトセンサアレイから出力される２つの被写体像信号より前記フォトセンサアレイのセンサピッチのｉ（整数）倍の変位量に対する前記２つの被写体像の不一致度を示す尺度ｆ（ｉ）とその最小値を与えるｉの値ｉ_０を求め、該ｉ_０またはこれにフォトセンサピッチを乗じた値、もしくは該ｉ_０およびその前後のｉ_０−１，ｉ_０−２，ｉ_０＋１，ｉ_０＋２に対応する関数値ｆ（ｉ_０），ｆ（ｉ_０−１），ｆ（ｉ_０−２），ｆ（ｉ_０＋１），ｆ（ｉ_０＋２）をそれぞれ求めてこれらの値の一部もしくは全部を用いて推定したｆ（ｉ）の最小値ｆ（ｘ）を与えるｘまたはこれにフォトセンサピッチを乗じた値を、前記相対変位とする距離測定方法において、
前記２つの被写体像信号において前記関数値ｆ（ｉ_０），ｆ（ｉ_０−１），ｆ（ｉ_０−２），ｆ（ｉ_０＋１），ｆ（ｉ_０＋２）の計算に使われる演算領域のうち、少なくとも前記ｆ（ｉ_０）の演算領域における前記２つの被写体像信号を比較し、一方の端部に他方の端部にはないノイズがあるかを判定し、ノイズがあると判定された場合は前記演算領域における端部の被写体像信号の補正を行ってから前記尺度ｆ（ｉ）を再計算して相対変位を求めるようにしたことを特徴。
【００１１】
請求項２に係る発明は、請求項１に係る発明において、前記距離測定方法が、前記一方の演算領域の端部の前記ノイズが、当該箇所には投影されるが前記他方の演算領域の端部には投影されない前記被写体の部分の状態に起因するものについて判定することを特徴。
請求項３に係る発明は、請求項１または２に係る発明において、前記演算領域をそれぞれの画素数が均等になるように分割領域に分割し、前記２つのフォトセンサアレイからそれぞれ出力される被写体像信号について前記分割領域ごとに不一致度を評価し、端部を含む前記分割領域の不一致度が端部を含まない分割領域の不一致度より大きいと判断されるときに、ノイズがあると判定することを特徴とする。
【００１２】
請求項４に係る発明は、請求項３に係る発明において、前記分割領域の不一致度を、前記分割領域における前記２つのフォトセンサアレイの対応する画素よりそれぞれ出力される被写体像信号の差の絶対値の総和を計算することにより求め、端部を含む領域の不一致度がそれ以外の領域の不一致度の平均値の１．２倍以上であるとき、ノイズがあると判定することを特徴とする。
請求項５に係る発明は、請求項１または２に係る発明において、前記演算領域の最端部の画素と該最端部の画素よりひとつ内側の画素より出力される被写体像信号の差を、前記二つのフォトセンサアレイ間で比較してノイズがあるかを判定することを特徴とする。
【００１３】
請求項６に係る発明は、請求項５に係る発明において、Ａ／Ｄ変換器により前記該２つのフォトセンサアレイからそれぞれ出力される被写体像信号をデジタル化し、前記前記演算領域の最端部の画素と該最端部の画素よりひとつ内側の画素より出力される被写体像信号の差が前記Ａ／Ｄ変換機の最小分解能のｎ倍以上であるときノイズがあると判定することを特徴とする。
請求項７に係る発明は、請求項６に係る発明において、前記ｎが２であることを特徴とする。
請求項８に係る発明は、請求項５に係る発明において、前記演算領域の最端部の画素と該最端部の画素よりひとつ内側の画素より出力される被写体像信号の差を、右上がり，平坦，右下がりの三つに分類し、どちらか一方の前記演算領域の端部が右上がりで、他方の前記演算領域の端部が右下がりである場合にノイズがある、と判定することを特徴とする。
【００１４】
請求項９に係る発明は、請求項１ないし８のいずれかに係る発明において、ノイズがあると判定された場合、前記演算領域の最端部の画素と該最端部の画素よりひとつ内側の画素より出力される被写体像信号の差の絶対値を前記２つの被写体像信号間で比較し、前記被写体像信号の差の絶対値が大きい方の前記被写体像信号にノイズがあると判断することを特徴とする。
請求項１０に係る発明は、請求項９に係る発明において、ノイズがあると判定された場合、ノイズがあると判断された前記被写体像信号について、前記演算領域の最端部の画素による出力される被写体像信号を、該最端部の画素よりひとつ内側の画素より出力される被写体像信号に置き換えることにより前記端部の被写体像信号の補正を行うことを特徴とする。
【００１５】
請求項１１に係る発明は、請求項１ないし８のいずれかに係る発明において、ノイズがあると判定された場合、一方の前記演算領域の最端部の画素および該最端部の画素よりひとつ内側の画素より出力される被写体像信号を、他方の前記被写体像信号の演算領域の最端部の画素および該最端部の画素よりひとつ内側の画素より出力される被写体像信号に一致するようにして前記端部の前記被写体像信号の補正を行うことを特徴とする。
請求項１２に係る発明は、請求項９に係る発明において、ノイズがあると判定された場合、ノイズがあると判断された前記被写体像信号の前記演算領域の最端部の画素および該最端部の画素よりひとつ内側の画素より出力される被写体像信号をそれぞれＳＥ１，ＳＥ２、他方の演算領域の最端部の画素および該最端部の画素よりひとつ内側の画素より出力される被写体像信号をそれぞれＳＴ１，ＳＴ２とし、ＳＥ１をＳＥ２＋（ＳＴ１−ＳＴ２）で置き換えることにより前記被写体像信号の補正を行うことを特徴とする。
【００１６】
請求項１３に係る発明は、請求項９に係る発明において、ノイズがあると判定された場合、ノイズがあると判断された前記被写体像信号について、前記演算領域の最端部の画素より出力される被写体像信号を、該最端部の画素より内側の画素より出力される被写体像信号より外挿したものに置き換えることにより前記端部の前記被写体像信号の補正を行うことを特徴とするに。
請求項１４に係る発明は、請求項１３に係る発明において、前記ノイズがあると判断された前記被写体像信号の前記演算領域の最端部の画素より出力される被写体像信号をＳＥ１、該最端部の画素より内側の２つの画素より出力される被写体像信号を前記最端部の画素に近い画素順にそれぞれＳＥ２，ＳＥ３とすると、前記外挿したものが（２×ＳＥ２−ＳＥ３）であることを特徴とする。
【００１７】
請求項１５に係る発明は、請求項１３に係る発明において、前記ノイズがあると判断された前記被写体像信号の前記演算領域の最端部の画素より出力される被写体像信号をＳＥ１、該最端部の画素より内側の３つの画素より出力される被写体像信号を前記最端部の画素に近い画素順にそれぞれＳＥ２，ＳＥ３，ＳＥ４とすると、前記外挿したものが（２．５×ＳＥ２＋０．５×ＳＥ４−２×ＳＥ３）であることを特徴とする。
【００１８】
【発明の実施の形態】
本発明は、まず２つのセンサアレイから出力される被写体像信号Ｌ（ｊ）およびＲ（ｊ）から、２つの像の相対変位に対する像の不一致度を示す評価関数ｆ（ｉ）を計算してその最小値ｆ（ｉ_０）を求め、ｆ（ｉ_０）のみの演算領域、もしくはｆ（ｉ_０），ｆ（ｉ_０−１），ｆ（ｉ_０−２），ｆ（ｉ_０＋１），ｆ（ｉ_０＋２）のうち補間演算に使われるものの演算領域において、その部分の被写体像信号にエッジノイズがあるかどうかを判定して、エッジノイズがあると判定した場合は被写体像信号を適切に補正することにより、高精度な距離測定方法を提供するものである。
【００１９】
以下、図に沿って本発明の実施形態を説明する。
【００２０】
【実施例１】
図１は、エッジノイズの判定を行うこの発明の第１の実施例について説明するための図である。図１に示すＬ，Ｒは例えば図１０（ｂ）のＬ，Ｒの右側部分を演算領域として抽出し、横軸の位置を揃えて示したものでＥ_Ｌがある場合の例である。この演算領域のセンサを５つの分割領域Ａ１，Ａ２，Ａ３，Ａ４，Ａ５にそれぞれが同じ数のセンサを含むように等分割する。そして、分割領域Ａｉごとに当該分割領域内の被写体像信号Ｌ_Ａｉ（ｊ），Ｒ_Ａｉ（ｊ）の不一致度ｇ（Ａｉ）を次式により求める。
【００２１】
【数４】

（３）式で求められる不一致度は、図１において実線で示されるＬと破線で示されるＲに挟まれた領域の面積に相当し、Ｅ_Ｌのある分割領域Ａ５を除き、大差のないものになる。分割領域Ａ５のみは、Ｅ_Ｌのため曲線Ｌと曲線Ｒの相似性が著しくくずれ、不一致度ｇ（Ａ５）が他の分割領域より大きくなる。図１はＥ_Ｌを想定したものであるが、Ｅ_Ｒを想定するとＥ_Ｒは分割領域Ａ１に現れるため不一致度ｇ（Ａ１）が他の分割領域より大きくなる。また、Ａ２〜Ａ４はエッジノイズの影響を受けることがないため、この部分の不一致度を基準値として分割領域Ａ１およびＡ５の不一致度と比較し、分割領域Ａ１およびＡ５の不一致度の方が大きいと判断されたら、端部の分割領域Ａ１もしくはＡ５にエッジノイズがあると判断できる。すなわち、ｇ（Ａ２）〜ｇ（Ａ４）の平均値ｇ_ＡＶＥでｇ（Ａ１），ｇ（Ａ５）を除した比をそれぞれｈ_Ａ１，ｈ_Ａ５とすると、ｈ_Ａ１，ｈ_Ａ５が基準値Ｋ１以上であるときにエッジノイズがあると判断する。実験ではＫ１＝１．２程度で良好な結果を得ることができた。
【００２２】
【実施例２】
図２は、エッジノイズの判定を行うこの発明の第２の実施例について説明するための図である。図２は図１の分割領域Ａ５を拡大表示し、さらに個々のセンサ出力もプロットしたものである。図中に丸で示したＬ（ｎ−４），・・・，Ｌ（ｎ）およびＲ（ｎ−４），・・・，Ｒ（ｎ）はそれぞれ分割領域Ａ５に含まれるフォトセンサアレイ５３（Ｌ），５４（Ｒ）の（ｎ−４），・・・，ｎ番目のセンサ出力を示す点である。本実施例は、エッジノイズがあればＬとＲで端部の傾きは異なる、という考え方に基づき判定するものである。すなわち、フォトセンサアレイの出力Ｌ，Ｒの端部の傾きをそれぞれ（Ｌ（ｎ）−Ｌ（ｎ−１））および（Ｒ（ｎ）−Ｒ（ｎ−１））により求めて比較し、２つの傾きの差の絶対値が基準値Ｋ２以上であるときに、端部の被写体像信号に差がある、すなわちエッジノイズありと判定する。図８に示すように、通常はフォトセンサアレイより出力される被写体像信号はまずＡ／Ｄ変換器によりデジタル化されるため、基準値Ｋ２についてはＡ／Ｄ変換器の最小分解能を単位としてそのｎ倍とするのがよい。Ａ／Ｄ変換器の量子化誤差等を考えると、ｎとしては２もしくはそれ以上の値とするのが適当である。
なお、エッジノイズがあると判定された場合にＬとＲのどちらにエッジノイズがあるかを判断するためには、（Ｌ（ｎ）−Ｌ（ｎ−１））および（Ｒ（ｎ）−Ｒ（ｎ−１））の絶対値を比較し、大きい方の被写体像信号にエッジノイズがあるとすればよい。この判断方法は、実施例１においても適用できる。
【００２３】
【実施例３】
図３は、エッジノイズの判定を行うこの発明の第３の実施例について説明するための図である。図３は図２の右端の２つセンサ出力（被写体像信号）Ｌ（ｎ−１），Ｌ（ｎ）およびＲ（ｎ−１），Ｒ（ｎ）に注目した図である。本実施例では傾き（Ｌ（ｎ）−Ｌ（ｎ−１））および（Ｒ（ｎ）−Ｒ（ｎ−１））をその値により右上がり，平坦，右下がりに分類し、その結果によりＬ，Ｒ２つの被写体像信号の比較を行う。分類については、実施例２と同様にＡ／Ｄ変換器の最小分解能を単位として、（Ｌ（ｎ）−Ｌ（ｎ−１））もしくは（Ｒ（ｎ）−Ｒ（ｎ−１））がｎ１（＞０）倍以上であれば右上がり、−ｎ１倍より小さければ（絶対値は大きい）右下がり、それ以外なら平坦とすればよい。これは、図２に示すようにＬ（ｎ−１），Ｒ（ｎ−１）を基準にしてそれに対しＡ／Ｄ変換器の最小分解能の−ｎ１〜＋ｎ１倍の範囲にある領域Ｃ２にＬ（ｎ），Ｒ（ｎ）があれば平坦、Ｃ２より上の領域Ｃ１にあれば上向き、領域Ｃ３にあれば下向きと判断していることに相当する。ＬとＲが同じもしくは隣接した（右上がりと平坦、もしくは平坦と右下がり）分類になる場合は両者に大きな差はなく、エッジノイズはないと判定する。隣接した分類でも大きな差はないとするのは、例えば上述のＣ１とＣ２の境界近傍にＬ（ｎ）とＲ（ｎ）がある場合を考えているためである。一方が右上がりで他方が右下がりのときに、両者は差があり、すなわちエッジノイズがあると判定する。図３はＥ_Ｒに関する説明を行うための図であるが、Ｅ_Ｌについても領域Ａ１における左端の２センサの被写体像信号出力を用いて同様に判定することができる。
また、エッジノイズがあると判断された場合にＬとＲのどちらにエッジノイズがあるかを判断する方法は、実施例２と同じである。
【００２４】
【実施例４】
図４は、エッジノイズがあると判定された場合、その対策として被写体像信号出力の補正を行うこの発明の第４の実施例について説明するための図である。図４は図２の被写体像信号を補正する例である。上述の判断方法により、補正すべき信号はＬではなくＲであることが既に判断されていて、その結果を受けてＲ（ｎ）を補正するものである。本実施例においては、相対変位Ｘを求める際にはＲ（ｎ）は誤差要因でしかないのでこのデータを廃棄し、隣接した画素からの像出力Ｒ（ｎ−１）に置き換えて補正を行う。図４はＥ_Ｒの信号Ｒに対する処理を示すものであるが、補正すべき信号がＬである場合やＥ_Ｌに対しても同様に処理を行うことができる。この方法は、後述する外挿方式に比べ演算時間が短いという特徴がある。
【００２５】
【実施例５】
図５は、エッジノイズがあると判定された場合、その対策として被写体像信号出力の補正を行うこの発明の第５の実施例について説明するための図である。本実施例は端部の波形をＬとＲで同じものにするという考えに基づき、Ｒ（ｎ−１）Ｒ（ｎ）間の傾きをＬ（ｎ−１）Ｌ（ｎ）間の傾きに揃える補正を行っている。すなわち、Ｒ（ｎ）について次式による置き換えを行う。
【００２６】
【数５】
Ｒ（ｎ）＝Ｒ（ｎ−１）＋（Ｌ（ｎ）−Ｌ（ｎ−１）） ・・・（４）
これにより、図５に示すように、Ｒ（ｎ−１）とＲ（ｎ）を結ぶ直線とＬ（ｎ−１）とＬ（ｎ）を結ぶ直線が平行となり、同じ波形とすることができる。図５はＥ_Ｒに対する例であるが、Ｅ_Ｌに対しても同様に処理を行うことができる。なお、本補正方法においては補正を適用する被写体像信号をエッジノイズがあると判断された被写体像信号に限るものではない。これは、どちらの被写体像信号を補正しても補正後は同じ評価関数ｆ（ｉ）を与えるからである。
【００２７】
【実施例６】
図６は、エッジノイズがあると判定された場合、その対策として被写体像信号出力の補正を行うこの発明の第６の実施例および後述する第７の実施例について説明するための図である。図６も実施例４と同様にエッジノイズがあると判断されたＲ（ｎ）を補正するものである。第６および第７に実施例は関数のテイラー展開の考え方に基づき補正を行う。関数Ｆ（ｘ）に関するテイラー展開は次式で与えられる。
【００２８】
【数６】
Ｆ（ｘ_０＋ｘ）＝Ｆ（ｘ_０）＋ｘＦ^（１）（ｘ_０）＋（ｘ^２／２）Ｆ^（２）（ｘ_０）＋・・・ ・・・（５）
ここで、Ｆ^（１）（ｘ）およびＦ^（２）（ｘ）はそれぞれ関数Ｆ（ｘ）の一次微分関数、二次微分関数である。実施例６では、（５）式の一次微分の項まで使い、一次微分値を差分で近似することにより、Ｒ（ｎ）について次式による置き換えを行う。
【００２９】
【数７】

これはＥ_Ｒの信号Ｒに対する処理を示すものであるが、補正すべき信号がＬである場合やＥ_Ｌに対しても同様に処理を行うことができる。
【００３０】
【実施例７】
実施例７は（５）式のニ次微分の項まで使って補正を行うものである。二次微分は差分の差をとることで近似することにより、Ｒ（ｎ）について次式による置き換えを行う。
【００３１】
【数８】

これはＥ_Ｒの信号Ｒに対する処理を示すものであるが、補正すべき信号がＬである場合やＥ_Ｌに対しても同様に処理を行うことができる。
また、本発明はテイラー展開の二次微分の項までを利用しているが、より高次の微分項を使ったものも同様に考えることができる。
【００３２】
【発明の効果】
以上説明したように、本発明によれば、２つのフォトセンサアレイから出力される被写体像信号にエッジノイズがあるかを判定し、エッジノイズがあると判定された場合は必要に応じてどちらの被写体像信号にエッジノイズがあるかを判断してそのエッジノイズがある部分の被写体像信号を適切に補正することにより、高精度な距離測定を行うことのできる距離測定方法を提供することができる。
【図面の簡単な説明】
【図１】第１の実施例を説明するための図である。
【図２】第２の実施例を説明するための図である。
【図３】第３の実施例を説明するための図である。
【図４】第４の実施例を説明するための図である。
【図５】第５の実施例を説明するための図である。
【図６】第５および第６の実施例を説明するための図である。
【図７】三角測距の原理に基づく距離測定原理図である。
【図８】距離検出装置の構成を示す図である。
【図９】距離測定における視差を説明するための図である。
【図１０】フォトセンサアレイより出力される被写体像信号に関しそれぞれ、（ａ）は正常な被写体像信号の例、（ｂ）はフレアのある被写体像信号の例を示す図である。
【図１１】エッジノイズのある被写体像信号の例を示す図である。
【符号の説明】
５１，５２ レンズ
５３，５４ フォトセンサアレイ
５５ 被写体
５６，５７ 被写体像
５８，５９ 光軸
６２，６３ Ａ／Ｄ変換器
６４ 演算処理装置（マイコン）
７０ 光源
Ｌ フォトセンサアレイ５３より出力される被写体像信号
Ｒ フォトセンサアレイ５４より出力される被写体像信号
ＬＬ，ＬＲ，ＲＲ フォトセンサアレイの視野を区分した領域
Ａ１〜Ａ５ 演算領域を等分割した領域[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention provides, for example, a pair of optical sensor arrays on which an image of an object to be measured is formed via an optical path through spatially different lenses, in a distance measuring method used for an automatic focusing mechanism of a camera. The present invention relates to an improvement in a distance measuring method for measuring a distance to the target by using a video output signal corresponding to a received light intensity of each optical sensor of the optical sensor array.
[0002]
[Prior art]
FIG. 7 is a principle diagram of distance measurement based on the principle of triangulation. Light from the subject 55 is formed as subject images 56 and 57 on the photosensor arrays 53 and 54 through the lenses 51 and 52. The points P ₆ and P ₇ are the intersections of the light beams (optical axes 58 and 59) passing through the center points P ₂ and P ₃ of the lenses 51 and 52 from the front at infinity and the photosensor arrays 53 and 54. B the distance between the points _{P 6} and the point _{P 7,} the distance between the photosensor array 53 and the lens 51 and f (substantially equal to the focal length of the lens 51, 52). Also, let X1 and X2 be the deviations of the subject images 56 and 57 from the optical axes 58 and 59, and let X be the sum of X1 and X2. Incidentally, reference numeral 61 denotes a measurement system.
[0003]
Here, since the triangles P ₁ P ₂ P ₄ and P ₂ P ₅ P ₆ and the triangles P ₁ P ₄ P ₃ and P ₃ P ₇ P ₈ are similar to each other, the distance d to the subject 55 is expressed by the following equation. Is required.
[0004]
(Equation 1)
d = B · f / (X1 + X2) = B · f / X
X is a relative displacement between two images when the subject 55 is at infinity, that is, when two subject images 56 and 57 are at the intersection of the optical axes 58 and 59 of the lenses 51 and 52 and the photosensor array. . Since B and f are constants, the distance d can be obtained by detecting X.
FIG. 8 shows a configuration of a distance detection device configured based on the principle of the triangulation. 8, 53 and 54 are the same photo sensor arrays as those shown in FIG. 7, and 62 and 63 are A / D converters. Analog outputs from the sensors in the photosensor arrays 53 and 54 are converted into digital signals by A / D converters 62 and 63. Reference numeral 64 denotes an arithmetic processing device (hereinafter, referred to as a microcomputer) such as a microcomputer, which obtains a relative displacement X between the two images, and controls the imaging lens using the X. As described above, such a distance detection device generally forms a system using a microcomputer. The operation of obtaining the relative variation X in the distance measuring device is roughly as follows.
Now, the output values of the photosensor arrays 53 and 54 after the A / D conversion are represented by L (1), L (2),..., L (M1), R (1), R (2),. , R (M2), and an evaluation function f (i) indicating the degree of inconsistency of the images with respect to the relative displacement of the two images of X = ip (p is the photosensor pitch) is obtained. The evaluation function f (i) is given by, for example, Expression (1) or Expression (2).
[0005]
(Equation 2)

[0006]
[Equation 3]

If, if two of the image on the photosensor array 53 is shifted relatively _{i 0} photosensor pitch, the f _(i 0) = 0. However, usually, the X does not completely become an integral multiple of the photosensor pitch, and the shapes of the subject image signals from the two photosensors become completely the same (except that they are relatively displaced). Since the number is small, f (i ₀ )> 0. Based on this idea, the distance measuring apparatus obtains i = i ₀ that gives the minimum value of f (i), and sets i ₀ or a value obtained by multiplying i ₀ or the photo sensor pitch p as X. Further, of the subject image signals L (j) and R (j), a part used for calculating f (i) is defined as a calculation area of f (i). In order to perform more accurate detection, f (i ₀ ), f (i ₀ -1), f (i ₀ -2), f (i ₀ +1) and f (i ₀ +2) are usually used. The interpolation calculation is performed by using the above, but the details are omitted.
Since noises on the outputs of the photosensor arrays 53 and 54 adversely affect the relative displacement X, that is, the accuracy of distance detection, various countermeasures have been devised, for example, as in Patent Document 1.
Japanese Patent Application Laid-Open No. H11-163873 discloses a case in which a light source such as illumination or the sun as shown in FIG. 7 is provided on the back of a subject, and when there is a so-called flare in which the image signal of the subject changes due to the light from the light source, the photo sensor array 53 This relates to a case where a strong flare occurs in one of the optical systems due to parallax between the optical system including the lens 51 and the optical system including the photosensor array 54 and the lens 52. That is, when the configuration of FIG. 7 is redrawn as shown in FIG. 9, the region where the region LR is commonly projected to the two photosensors 51 and 53, the region where the region RR is projected only to the photosensor 54, and the region LL are The area is projected only on the photosensor 53. If the above-mentioned light source exists in the region LL or RR, only one of the photo sensors is affected by the light source. For example, when there is a flare source in the region RR as in the light source 70 in FIG. 7, the subject image signal in the normal state shown in FIG. 10A is changed to the left side of the signal R as shown in FIG. The ends are particularly greatly deformed due to flares. Patent Literature 1 attempts to perform accurate distance measurement by removing the influence of flare in this case. In FIG. 9, L1 and R1 are the sensor numbers of the sensors at the end portions of the two photosensor arrays 53 and 54, respectively. The sensor NO is on the horizontal axis in FIG. 10, and the leftmost of the horizontal axis in FIG. 10 is L1, R1. L and R in FIG. 10 are subject image signals output from the photosensor arrays 53 and 54, respectively.
That is, in a situation where there is a flare, the light from the light source 70 is added to the light from the object to raise the object image signal in a biased manner, and the influence of the light source 70 differs between the two photosensor arrays 53 and 54. Therefore, there is a portion where the respective subject image signals do not match even if the bias is subtracted. If the difference between the two subject image signals is only the amount of the bias, correction may be made to add or subtract the average difference between the two subject image signals to or from one of the subject image signals so as to cancel the bias. This is not effective when the subject image signal is partially deformed as shown in FIG. In Japanese Patent Application Laid-Open No. H10-163, with respect to such a subject, a bias correction method is applied to two subject image signals by restricting them to a portion where the influence of parallax does not occur, or a bias is gradually added to one subject image signal. A method is shown in which the coincidence between two subject image signals is determined, and when the coincidence is lower than a predetermined value, the correction is completed.
[0007]
[Patent Document 1]
JP-A-2002-250857
[Problems to be solved by the invention]
A method of applying a bias to one of two subject image signals obtained from two photosensor arrays with respect to flare as in Patent Literature 1 is not suitable for a device having edge noise. The edge noise means a sharp change in the image signal of the object at the end in only one of the above-described calculation regions due to the high contrast of the object. This is because, for example, when there is a high contrast change portion near the boundary between the regions LR and RR and the change portion is projected only on the photosensor array 54, only the subject image signal R as shown in FIG. This is because an abrupt change in the edge E _R occurs. Also, there when there is a high contrast change unit near the boundary of the region LR and the region LL, Similarly, only the subject image signals L as shown in that sudden changes in edge E _L occurs FIG 11 (b). Furthermore, E _R, E _L is not only when there is a high contrast change unit near the boundary of the region LR and the region RR near the boundary or region LR and the region LL, part of the subject image signal with respect to the subject in the area LR Is cut off to obtain the above-described calculation area, there may be a case where there is a high contrast change near the cut-out portion. E _R and E _L are due to the contrast of the subject, and are not noise, to be precise. However, they appear in only one of the two subject image signals in the calculation area, and cause an error when calculating the relative displacement X. Called noise. Edge noise changes the similarity of the calculation area by giving a large change to the image signal at the extreme end in the calculation area.However, since the similarity of other object image signals is maintained, the accuracy of interpolation calculation is mainly determined. To prevent accurate distance measurement. Since the edge noise is due to the contrast of the subject itself as described above, this does not cause a bias to be applied to the subject image signal of the other part. Therefore, a bias is applied to one subject image signal as a measure against flare. It cannot be dealt with by the technique of correcting.
[0009]
Therefore, the present invention has been made to solve the above-described problem, and an object of the present invention is to determine whether or not there is edge noise in a subject image signal in at least a calculation area of f (i ₀ ). It is another object of the present invention to provide a distance measurement method capable of performing a highly accurate distance measurement by appropriately correcting the subject image signal when it is determined that
[0010]
[Means for Solving the Problems]
Therefore, in order to solve the above problem, the invention according to claim 1 projects light beams of different portions of light beams emitted or reflected from a subject onto two photosensor arrays, respectively, and To determine the distance to the subject from the relative displacement of the subject image formed on the array,
A scale f (i) indicating the degree of inconsistency of the two subject images with respect to a displacement of i (integer) times the sensor pitch of the photosensor array from the two subject image signals output from the two photosensor arrays, and its scale. obtains the value _{i 0} of i giving the minimum value, the _{i 0} or this value multiplied by a photosensor pitch or the _{i 0} and before and after the _{i 0 -1,, i 0 -2} , i 0 + 1, i 0 +2, function values f (i ₀ ), f (i ₀ -1), f (i ₀ -2), f (i ₀ +1), f (i ₀ +2) are obtained, and one of these values is calculated. A distance measuring method in which x giving the minimum value f (x) of f (i) estimated by using a part or the whole or a value obtained by multiplying x by a photosensor pitch is used as the relative displacement,
The two object image signals are used for calculating the function values f (i ₀ ), f (i ₀ -1), f (i ₀ -2), f (i ₀ +1), and f (i ₀ +2). The two object image signals in at least the calculation area of f (i ₀ ) in the calculation area are compared, and it is determined whether one end has noise that is not present in the other end. If it is determined, the object image signal at the end in the calculation area is corrected, and then the scale f (i) is recalculated to obtain the relative displacement.
[0011]
According to a second aspect of the present invention, in the invention according to the first aspect, the distance measuring method is configured such that the noise at an end of the one operation area is projected to the corresponding location, but the end of the other operation area is projected. The determination is made on what is caused by the state of the part of the subject that is not projected on the part.
According to a third aspect of the present invention, in the first or second aspect of the present invention, the calculation region is divided into divided regions so that the number of pixels of each of the calculation regions is equal, and the object output from each of the two photosensor arrays is provided. The degree of mismatch is evaluated for each of the divided regions of the image signal, and when it is determined that the degree of mismatch of the divided region including the end is larger than the degree of mismatch of the divided region not including the end, it is determined that there is noise. It is characterized by the following.
[0012]
According to a fourth aspect of the present invention, in the invention according to the third aspect, the degree of inconsistency of the divided areas is determined by calculating an absolute value of a difference between subject image signals output from corresponding pixels of the two photosensor arrays in the divided areas. It is determined by calculating the sum of the values, and when the degree of inconsistency of the area including the end is 1.2 times or more the average value of the degree of inconsistency of the other areas, it is determined that there is noise. .
The invention according to claim 5 is the invention according to claim 1 or 2, wherein a difference between a subject image signal output from a pixel at the extreme end of the calculation area and a pixel one inside the pixel at the extreme end is calculated by: It is characterized in that it is determined whether there is noise by comparing the two photosensor arrays.
[0013]
According to a sixth aspect of the present invention, in the invention according to the fifth aspect, an object image signal output from each of the two photosensor arrays is digitized by an A / D converter, and When a difference between a pixel and a subject image signal output from a pixel one inner side of the pixel at the end is at least n times the minimum resolution of the A / D converter, it is determined that there is noise. .
The invention according to claim 7 is the invention according to claim 6, wherein n is 2.
According to an eighth aspect of the present invention, in the invention according to the fifth aspect, a difference between a subject image signal output from a pixel at the extreme end of the calculation area and a pixel one inside the pixel at the extreme end is increased to the right. , Flat, and lower right, and it is determined that there is noise when one end of the operation area is upward right and the other end of the operation area is lower right. It is characterized by.
[0014]
According to a ninth aspect of the present invention, in the invention according to any one of the first to eighth aspects, when it is determined that there is noise, the pixel at the extreme end of the calculation area and one pixel inside the pixel at the extreme end are calculated. Comparing the absolute value of the difference between the subject image signals output from the pixels between the two subject image signals, and determining that the subject image signal having the greater absolute value of the difference between the subject image signals has noise; It is characterized by.
According to a tenth aspect of the present invention, in the invention according to the ninth aspect, when it is determined that there is noise, the subject image signal determined to have the noise is output by a pixel at an end portion of the calculation area. The object image signal at the edge is corrected by replacing the object image signal with an object image signal output from a pixel one inside the pixel at the end.
[0015]
According to an eleventh aspect of the present invention, in the invention according to any one of the first to eighth aspects, when it is determined that there is noise, one pixel is located at the end of one of the operation regions and one of the pixels at the end. The subject image signal output from the inner pixel is made to coincide with the subject image signal output from the pixel at the extreme end of the calculation area of the other subject image signal and the pixel one inside the pixel at the extreme end. And correcting the subject image signal at the end portion.
According to a twelfth aspect of the present invention, in the invention according to the ninth aspect, when it is determined that there is noise, a pixel at an extreme end of the calculation region of the subject image signal determined to have noise and the extreme end SE1 and SE2, respectively, the subject image signals output from the pixels one inward from the pixels of the part, and the subject image signals output from the pixels at the extreme end of the other operation area and the pixels one inside the extreme end pixels Are ST1 and ST2, respectively, and the subject image signal is corrected by replacing SE1 with SE2 + (ST1-ST2).
[0016]
According to a thirteenth aspect of the present invention, in the invention according to the ninth aspect, when it is determined that there is noise, the subject image signal determined to have the noise is output from a pixel at an end portion of the calculation area. And correcting the subject image signal at the end by replacing the subject image signal with a signal extrapolated from a subject image signal output from a pixel inside the pixel at the extreme end. .
According to a fourteenth aspect of the present invention, in the thirteenth aspect, the subject image signal output from the pixel at the end of the operation area of the subject image signal determined to have the noise is SE1, Assuming that the subject image signals output from the two pixels inside the end pixels are SE2 and SE3 in the order of pixels closest to the end pixel, the extrapolated one is (2 × SE2−SE3). It is characterized by the following.
[0017]
According to a fifteenth aspect of the present invention, in the thirteenth aspect, the subject image signal output from the pixel at the end of the operation area of the subject image signal determined to have the noise is SE1, Assuming that the subject image signals output from the three pixels inside the end pixel are SE2, SE3, and SE4 in the order of pixels closest to the end pixel, the extrapolated one is (2.5 × SE2 + 0. 5 × SE4-2 × SE3).
[0018]
BEST MODE FOR CARRYING OUT THE INVENTION
According to the present invention, first, an evaluation function f (i) indicating the degree of inconsistency of the images with respect to the relative displacement of the two images is calculated from the subject image signals L (j) and R (j) output from the two sensor arrays. the determined minimum value _{f (i 0), f (} i 0) calculation region only or _{f (i 0),, f} (i 0 -1), f (i 0 -2), f (i 0 +1) , F (i ₀ +2), in the calculation area of the one used for the interpolation calculation, it is determined whether or not there is edge noise in the subject image signal in that part. If it is determined that there is edge noise, the subject image signal is determined. By providing an appropriate correction, a highly accurate distance measurement method is provided.
[0019]
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0020]
Embodiment 1
FIG. 1 is a diagram for explaining a first embodiment of the present invention for determining edge noise. L shown in FIG. 1, R extracts Figure 10 (b) L, the right portion of the R a calculation region for example, an example in which in shows align the position of the horizontal axis is E _L. The sensors in this calculation area are equally divided into five divided areas A1, A2, A3, A4, and A5 so that each includes the same number of sensors. Then, the degree of mismatch g (Ai) between the subject image signals L _Ai (j) and R _Ai (j) in each of the divided areas Ai is determined by the following equation.
[0021]
(Equation 4)

(3) the inconsistency sought expression is equivalent to the area of the region between the R represented by L and the broken line indicated by the solid line in FIG. 1, except for the divided area A5 with E _L, having no significant difference become. Only the divided region A5 is collapsed significantly similarity of the curve L and the curve R for E _L, inconsistency g (A5) is greater than the other divided area. Although FIG. 1 is obtained by assuming the E _L, E _R assuming E _R is the inconsistency g to appear in the divided area A1 (A1) is larger than the other divided area. Since A2 to A4 are not affected by the edge noise, the degree of mismatch between the divided areas A1 and A5 is larger than the degree of mismatch between the divided areas A1 and A5 using the degree of mismatch at this portion as a reference value. , It can be determined that there is edge noise in the end divided region A1 or A5. That is, assuming that the ratios obtained by dividing g (A1) and g (A5) by the average value g _AVE of g (A2) to g (A4) are h _A1 and h _A5 , respectively, h _A1 and h _A5 are equal to or more than the reference value K1. Is determined to have edge noise. In the experiment, good results were obtained with K1 = about 1.2.
[0022]
Embodiment 2
FIG. 2 is a diagram for explaining a second embodiment of the present invention for determining edge noise. FIG. 2 is an enlarged view of the divided area A5 in FIG. 1 and plots individual sensor outputs. L (n−4),..., L (n) and R (n−4),..., R (n) indicated by circles in FIG. (L), (n-4) of 54 (R),..., The point indicating the n-th sensor output. In the present embodiment, the determination is made based on the idea that if there is edge noise, the inclination of the end portions differs between L and R. That is, the inclinations of the ends of the outputs L and R of the photosensor array are obtained by (L (n) -L (n-1)) and (R (n) -R (n-1)) and compared. When the absolute value of the difference between the two inclinations is equal to or greater than the reference value K2, it is determined that there is a difference in the subject image signal at the end, that is, there is edge noise. As shown in FIG. 8, the subject image signal normally output from the photosensor array is first digitized by the A / D converter, so that the reference value K2 is set in units of the minimum resolution of the A / D converter. It is good to make it n times. Considering the quantization error of the A / D converter and the like, it is appropriate to set n to a value of 2 or more.
When it is determined that there is edge noise, to determine which of L and R has the edge noise, it is necessary to determine (L (n) −L (n−1)) and (R (n) − R (n-1)) may be compared to determine that the larger subject image signal has edge noise. This determination method can be applied to the first embodiment.
[0023]
Embodiment 3
FIG. 3 is a diagram for explaining a third embodiment of the present invention for determining edge noise. FIG. 3 is a diagram focusing on two sensor outputs (subject image signals) L (n-1), L (n) and R (n-1), R (n) at the right end of FIG. In this embodiment, the slopes (L (n) -L (n-1)) and (R (n) -R (n-1)) are classified into right-up, flat, and right-down according to their values. L and R two object image signals are compared. Regarding the classification, (L (n) -L (n-1)) or (R (n) -R (n-1)) is used in units of the minimum resolution of the A / D converter as in the second embodiment. If it is at least n1 (> 0) times, it will rise to the right, if it is less than -n1 times (it will have a large absolute value), it will fall to the right, otherwise it will be flat. This is because, as shown in FIG. 2, L (n-1) and R (n-1) are set as reference values, and the area C2 is in the range of -n1 to + n1 times the minimum resolution of the A / D converter. If (n) and R (n) are present, it is determined that the area is flat, if the area C1 is above C2, it is upward, and if it is in the area C3, it is downward. If L and R are the same or adjacent (upward and flat, or flat and downward), it is determined that there is no large difference between them and that there is no edge noise. The reason why there is no large difference between adjacent classifications is that L (n) and R (n) are considered near the boundary between C1 and C2, for example. When one is rising to the right and the other is falling to the right, it is determined that there is a difference between them, that is, that there is edge noise. Figure 3 is a diagram for an explanation about E _R, it can be determined similarly using the subject image signal output of the leftmost 2 sensor in the well region A1 for E _L.
When it is determined that there is edge noise, the method of determining which of L and R has edge noise is the same as in the second embodiment.
[0024]
Embodiment 4
FIG. 4 is a diagram for explaining a fourth embodiment of the present invention in which, when it is determined that there is edge noise, subject image signal output is corrected as a countermeasure. FIG. 4 shows an example of correcting the subject image signal of FIG. According to the above-described determination method, it is already determined that the signal to be corrected is R instead of L, and R (n) is corrected based on the result. In the present embodiment, when calculating the relative displacement X, R (n) is only an error factor, so this data is discarded, and correction is performed by replacing the data with the image output R (n-1) from an adjacent pixel. . Figure 4 is illustrates a process for signal R E _R, can be the signal to be corrected performs processing similarly for the case and E _L is L. This method is characterized in that the calculation time is shorter than the extrapolation method described later.
[0025]
Embodiment 5
FIG. 5 is a diagram for explaining a fifth embodiment of the present invention in which, when it is determined that there is edge noise, a subject image signal output is corrected as a countermeasure. In this embodiment, the slope between R (n-1) R (n) is changed to the slope between L (n-1) L (n) based on the idea that the waveform at the end is the same between L and R. Corrections are made to align. That is, R (n) is replaced by the following equation.
[0026]
(Equation 5)
R (n) = R (n-1) + (L (n) -L (n-1)) (4)
Thereby, as shown in FIG. 5, the straight line connecting R (n-1) and R (n) and the straight line connecting L (n-1) and L (n) become parallel, and can have the same waveform. . Figure 5 is an example for E _R, it can be processed similarly for E _L. In the present correction method, the subject image signal to which the correction is applied is not limited to the subject image signal determined to have edge noise. This is because the same evaluation function f (i) is given after the correction of either object image signal.
[0027]
Embodiment 6
FIG. 6 is a diagram for explaining a sixth embodiment of the present invention and a seventh embodiment which will be described later, in which the subject image signal output is corrected as a countermeasure when it is determined that there is edge noise. FIG. 6 also corrects R (n) determined to have edge noise as in the fourth embodiment. Sixth and seventh embodiments perform correction based on the concept of Taylor expansion of a function. The Taylor expansion for the function F (x) is given by the following equation.
[0028]
(Equation 6)
^{_{F (x 0 + x) =}} F (x 0) + xF (1) (x 0) + (x 2/2) F (2) (x 0) + ··· ··· (5)
Here, F ⁽¹⁾ (x) and F ⁽²⁾ (x) are a primary differential function and a secondary differential function of the function F (x), respectively. In the sixth embodiment, R (n) is replaced by the following equation by approximating the primary differential value with the difference by using the first derivative term of the equation (5).
[0029]
(Equation 7)

This is intended to indicate a process for the signal R E _R, can be the signal to be corrected performs processing similarly for the case and E _L is L.
[0030]
Embodiment 7
In the seventh embodiment, the correction is performed using the second derivative term of the equation (5). R (n) is replaced by the following equation by approximating the second derivative by taking the difference between the differences.
[0031]
(Equation 8)

This is intended to indicate a process for the signal R E _R, can be the signal to be corrected performs processing similarly for the case and E _L is L.
Further, although the present invention uses the second derivative term of the Taylor expansion, a term using a higher derivative term can be similarly considered.
[0032]
【The invention's effect】
As described above, according to the present invention, it is determined whether or not there is edge noise in the subject image signals output from the two photosensor arrays. It is possible to provide a distance measuring method capable of performing high-accuracy distance measurement by determining whether or not the subject image signal has edge noise and appropriately correcting the subject image signal in a portion where the edge noise is present. .
[Brief description of the drawings]
FIG. 1 is a diagram for explaining a first embodiment.
FIG. 2 is a diagram for explaining a second embodiment.
FIG. 3 is a diagram for explaining a third embodiment.
FIG. 4 is a diagram for explaining a fourth embodiment.
FIG. 5 is a diagram for explaining a fifth embodiment;
FIG. 6 is a diagram for explaining fifth and sixth embodiments.
FIG. 7 is a principle diagram of distance measurement based on the principle of triangulation.
FIG. 8 is a diagram illustrating a configuration of a distance detection device.
FIG. 9 is a diagram for explaining parallax in distance measurement.
10A is a diagram illustrating an example of a normal subject image signal, and FIG. 10B is a diagram illustrating an example of a subject image signal having flare, with respect to a subject image signal output from a photosensor array.
FIG. 11 is a diagram illustrating an example of a subject image signal having edge noise.
[Explanation of symbols]
51, 52 Lens 53, 54 Photosensor array 55 Subject 56, 57 Subject image 58, 59 Optical axis 62, 63 A / D converter 64 Arithmetic processing unit (microcomputer)
70 Light Source L Subject Image Signal R Output from Photo Sensor Array 53 Subject Image Signals LL, LR, RR Output from Photo Sensor Array 54 Areas A1 to A5 Dividing Field of View of Photo Sensor Array

Claims (15)

The luminous fluxes of different portions of the luminous flux radiated or reflected from the subject are respectively projected onto the two photosensor arrays, and the distance to the subject is calculated from the relative displacement of the subject image formed on each of the photosensor arrays. To ask
A scale f (i) indicating the degree of inconsistency of the two subject images with respect to a displacement of i (integer) times the sensor pitch of the photosensor array from the two subject image signals output from the two photosensor arrays, and its scale. obtains the value _{i 0} of i giving the minimum value, the _{i 0} or this value multiplied by a photosensor pitch or the _{i 0} and before and after the _{i 0 -1,, i 0 -2} , i 0 + 1, i 0 +2, function values f (i ₀ ), f (i ₀ -1), f (i ₀ -2), f (i ₀ +1), f (i ₀ +2) are obtained, and one of these values is calculated. A distance measuring method in which x giving the minimum value f (x) of f (i) estimated by using a part or the whole or a value obtained by multiplying x by a photosensor pitch is used as the relative displacement,
The two object image signals are used for calculating the function values f (i ₀ ), f (i ₀ -1), f (i ₀ -2), f (i ₀ +1), and f (i ₀ +2). The two object image signals in at least the calculation area of f (i ₀ ) in the calculation area are compared, and it is determined whether one end has noise that is not present in the other end. If the determination is made, a subject image signal at an end in the calculation area is corrected, and then the scale f (i) is recalculated to obtain a relative displacement. The method for measuring the distance may be based on a state of the part of the subject, in which the noise at the end of the one calculation area is projected to the corresponding place but not to the end of the other calculation area. The distance measuring method according to claim 1, wherein the determination is performed. The operation area is divided into divided areas such that the number of pixels is equal, and the degree of inconsistency is evaluated for each of the divided areas with respect to the subject image signal output from each of the two photosensor arrays. 3. The distance measuring method according to claim 1, wherein when it is determined that the degree of mismatch of the divided areas is greater than the degree of mismatch of the divided areas not including the end, it is determined that there is noise. The degree of inconsistency of the divided area is obtained by calculating the sum of absolute values of the differences between subject image signals output from the corresponding pixels of the two photosensor arrays in the divided area. 4. The distance measuring method according to claim 3, wherein when the degree of mismatch is equal to or more than 1.2 times the average value of the degree of mismatch in other areas, it is determined that there is noise. The difference between the pixel at the extreme end of the operation area and the subject image signal output from the pixel one inside the pixel at the extreme end is compared between the two photosensor arrays to determine whether there is noise. The distance measuring method according to claim 1 or 2, wherein: The subject image signal output from each of the two photosensor arrays is digitized by an A / D converter, and is output from a pixel at the extreme end of the calculation area and a pixel one inside the pixel at the extreme end. 6. The distance measuring method according to claim 5, wherein it is determined that there is noise when the difference between the subject image signals is at least n times the minimum resolution of the A / D converter. 7. The distance measuring method according to claim 6, wherein n is 2. The difference between the pixel at the extreme end of the calculation region and the subject image signal output from the pixel one inside the pixel at the extreme end is classified into three types of rising right, flat, and falling right. The distance measuring method according to claim 5, wherein it is determined that there is noise when an end of the operation area is rising to the right and another end of the operation area is to the right. If it is determined that there is noise, the absolute value of the difference between the pixel at the extreme end of the calculation area and the subject image signal output from a pixel one inside the pixel at the extreme end is calculated between the two subject image signals. 9. The distance measuring method according to claim 1, wherein a comparison is made to determine that the subject image signal having a larger absolute value of the difference between the subject image signals has noise. When it is determined that there is noise, for the subject image signal that is determined to have noise, the subject image signal output by the pixel at the extreme end of the calculation area is one inner side of the pixel at the extreme end. 10. The distance measuring method according to claim 9, wherein the object image signal at the end is corrected by replacing the object image signal output from a pixel. If it is determined that there is noise, the object image signal output from the pixel at the extreme end of one of the operation regions and the pixel one inside the pixel at the extreme end is converted to the operation region of the other object image signal. And correcting the subject image signal at the end portion so as to match the subject image signal output from the pixel at the extreme end and the pixel one pixel inside the pixel at the extreme end. 9. The distance measuring method according to any one of 1 to 8. When it is determined that there is noise, the subject image signal output from the pixel at the extreme end of the computation region of the subject image signal determined to have noise and a pixel one inside the pixel at the extreme end is SE1 and SE2, the subject image signals output from the pixels at the extreme end of the other operation area and the pixel one inward from the extreme end pixels are ST1 and ST2, respectively, and SE1 is SE2 + (ST1-ST2). The distance measuring method according to claim 9, wherein the correction of the subject image signal is performed by replacement. When it is determined that there is noise, for the subject image signal that is determined to have noise, the subject image signal output from the pixel at the extreme end of the calculation area is a pixel inside the pixel at the extreme end. 10. The distance measuring method according to claim 9, wherein the subject image signal at the end is corrected by replacing the subject image signal with an extrapolated subject image signal. SE1 is a subject image signal output from the pixel at the extreme end of the calculation area of the subject image signal determined to have noise, and a subject image output from two pixels inside the pixel at the extreme end. 14. The distance measurement method according to claim 13, wherein the extrapolated signal is (2 * SE2-SE3), where the signals are SE2 and SE3, respectively, in the order of pixels closest to the pixel at the extreme end. SE1 is a subject image signal output from a pixel at the extreme end of the computation region of the subject image signal determined to have the noise, and a subject image output from three pixels inside the pixel at the extreme end. If the signals are SE2, SE3, and SE4 in the order of pixels closest to the pixel at the end, the extrapolated signal is (2.5 × SE2 + 0.5 × SE4-2 × SE3). Item 14. The distance measuring method according to Item 13.

Images