JP3696440B2 - Distance detector - Google Patents

Description

【００１５】
ここで、ＩＲ（ｍ，ｎ）は、図６に示す右対象物像Ｒ１内の座標（ｍ，ｎ）の位置の輝度値であり、ＩＬ（ａ＋ｍ−Ｍ，ｂ＋ｎ−Ｎ）は、探索領域内の座標（ａ，ｂ）を基点とした、右対象物像Ｒ１と同一形状の局所領域Ｒ３内の座標（ｍ，ｎ）の位置の輝度値である。基点の座標（ａ，ｂ）を変化させて輝度差分総和値Ｃ（ａ，ｂ）が最小となる位置を求めることにより、左対象物像の位置が特定される。
【００１６】
ステップＳ１５の処理により、図７に示すように右対象物像Ｒ１と、この対象物像（すなわち同一の対象物）に対応する左対象物像Ｒ４とが抽出されるので、ステップＳ１６では、右対象物像Ｒ１の重心位置と、画像中心線ＬＣＴＲとの距離ｄＲ（画素数）及び左対象物像Ｒ４の重心位置と画像中心線ＬＣＴＲとの距離ｄＬ（画素数）を求め、右画像を基準画像としたＲＬ視差Δｄ１＝ｄＲ＋ｄＬを算出する。
【００１７】
次にステップＳ１７〜Ｓ１９では、左画像を基準画像とする視差（以下「ＬＲ視差」という）を求める。すなわち、左画像から左対象物像Ｒ４を抽出し（ステップＳ１７）、左対象物像Ｒ４に対応する右対象物像を探索する右画像探索領域を右画像中に設定し、前記式（１）による相関演算を実行して左対象物像Ｒ４に対応する右対象物像を抽出する（ステップＳ１８）。そして、ステップＳ１６と同様にしてＬＲ視差Δｄ２を算出する（ステップＳ１９）。
【００１８】
このようにして左対象物像Ｒ４に対応する像を右画像中で探索した結果、右対象物像Ｒ１が得られれば、ＲＬ視差Δｄ１とＬＲ視差Δｄ２とはほぼ等しくなるので、正確な視差が得られていることが検証される。そこで、ステップＳ２０では、ＲＬ視差Δｄ１と、ＬＲ視差Δｄ２との偏差の絶対値｜Δｄ１−Δｄ２｜が所定閾値ΔｄＴＨ以下か否かを判別し、｜Δｄ１−Δｄ２｜≦ΔｄＴＨであるときは、ＲＬ視差及びＬＲ視差のいずれか一方の視差または両者の平均値をΔｄとして下記式（２）に適用して、自車両１０と、対象物との距離ｚを算出する（ステップＳ２２）。
【数２】

【００１９】
ここで、Ｂは基線長、すなわち図８に示すようにカメラ１Ｒの撮像素子１１Ｒの中心位置と、カメラ１Ｌの撮像素子１１Ｌの中心位置との水平方向の距離（両カメラの光軸の間隔）、Ｆはレンズ１２Ｒ、１２Ｌの焦点距離、ｐは、撮像素子１１Ｒ、１１Ｌ内の画素間隔である。
【００２０】
一方ステップＳ２０で｜Δｄ１−Δｄ２｜＞ΔｄＴＨであるときは、視差を算出する際に実行している相関演算（ステップＳ１５，Ｓ１８）の結果から、ＲＬ視差Δｄ１を算出したときの第１の相関特性と、ＬＲ視差Δｄ２を算出したときの第２の相関特性とを比較して、より相関性の高い対象物像から得られた視差を選択可能かどうかを判別し（ステップＳ２１）、選択可能であるときは前記ステップＳ２２に進んで、より相関性の高い対象物像から得られた視差を用いて距離ｚの算出を行う。また、相関特性を比較しても選択できない、すなわち第１及び第２の相関特性がほぼ同一であるとき、または両相関特性とも相関性が低いことを示すものであるときは、距離算出を実行せずに処理を終了する。
【００２１】
図９は、上記相関特性を説明するための図あり、２つ対象物像の相関性パラメータである輝度差分総和値Ｃ（ａ，ｂ）を変数ｂを一定として（最小の総和値が得られたときのｂ値に設定して）、変数ａの関数として示したものである。この図でラインＬ１及びＬ２は、それぞれ第１の相関特性及び第２の相関特性に対応し、この相関特性を代表する相関特性パラメータとしては、例えば総和値Ｃの最小値ＣＭＩＮ１，ＣＭＩＮ２、及び／またはその最小値をとる変数値ａ１，ａ２の近傍におけるラインＬ１，Ｌ２の傾き（例えば平均の傾きを示す直線Ｌ１ａ，Ｌ１ｂ，Ｌ２ａ，Ｌ２ｂの傾き）の絶対値ＤＣを用いる。最小値ＣＭＩＮはその値が小さいほど、相関性が高いことを示し、また傾きの絶対値ＤＣはその値が大きいほど相関性が高いことを示す。図９に示す例では、ラインＬ２で示される第２の相関特性がＮＧ（相関性が低い）ので、ＬＲ視差Δｄ２の信頼性が低いと判定して、ＲＬ視差Δｄ１を用いて距離ｚの算出を行う。相関特性は、例えば最小値ＣＭＩＮが所定閾値ＣＭＩＮＴＨより小さくかつ傾きの絶対値ＣＤが所定傾き値ＣＤＴＨより大きいとき、ＯＫと判定し、それ以外のときＮＧと判定する。第１及び第２の相関特性がともにＯＫであるときは、最小値ＣＭＩＮのより小さい方を採用する。ただし、その差ΔＣＭＩＮが小さいとき（所定値ΔＣＭＩＮＴＨ以下であるとき）は、後述するように距離の算出は行わないようにする。
【００２２】
図１０は、ＲＬ視差Δｄ１と、ＬＲ視差Δｄ２と偏差が大きくなる場合を説明するための図であり、２人の人に対応する対象物像Ｒ１１，Ｒ１２が右画像中に存在するが、２人の人と自車両１０との間に破線ＯＢＪＸで示す他の車両が存在するため、左画像中では右対象物像Ｒ１２に対応する左対象物像が存在しない（他の車両の陰に隠れている）ような場合を示している。なおこの図はグレースケール画像であるが、分かり易くするために対象物像と関連する像のみ示し、また相関演算を行う範囲を示す外接四角形は図示を省略している。また同図に示す対象物像Ｒ１３，Ｒ１４は、他の車両のテールランプが高輝度領域として撮像されたものである。
【００２３】
この図に示す場合には、ステップＳ１４〜Ｓ１６で右画像から右対象物像Ｒ１１を抽出して左画像から対応する左対象物像Ｒ２１を抽出し、ステップＳ１７〜Ｓ１９で左対象物像Ｒ２１に対応する右対象物像として誤ってＲ１２を抽出した場合には、ＬＲ視差Δｄ２がＲＬ視差Δｄ１より大きくなるため、ステップＳ２０の答が否定（ＮＯ）となり、ステップＳ２１に進んで相関特性の比較を行う。この例では、ＲＬ視差Δｄ１を算出したときの第１の相関特性の方が、ＬＲ視差Δｄ２を算出したときの第２の相関特性より相関性が高いので、ＲＬ視差Δｄ１を正しい視差として選択して、ステップＳ２２に進む。
【００２４】
またステップＳ１４〜Ｓ１６で右画像から右対象物Ｒ１２を抽出したときは、左画像から似通ったものである左対象物像Ｒ２１が抽出されるが、ステップＳ１７〜Ｓ１９で左対象物像Ｒ２１に対応する右対象物像としてＲ１１が抽出されるので、ステップＳ２０の答が否定（ＮＯ）となり、ステップＳ２１に進む。ステップＳ２１で相関特性を比較すると今度は、ＬＲ視差Δｄ２を算出したときの第２の相関特性の方が、ＲＬ視差Δｄ１を算出したときの第１の相関特性より相関性が高いことを示すので、ＬＲ視差Δｄ２を正しい視差として、ステップＳ２２に進む。
【００２５】
また右対象物像Ｒ１１とＲ１２とが非常に似通っているときは、いずれか一方の視差を選択できないので、ステップＳ２１の答が否定（ＮＯ）となって、距離ｚを算出することなく処理を終了する。
以上のように本実施形態では、右画像を基準とするＲＬ視差Δｄ１と、左画像を基準とするＬＲ視差Δｄ２とを算出し、両者がほぼ一致したときに正しい視差が得られたと判定し、そのようにして得られた視差に基づいて距離ｚを算出するようにしたので、左右の画像中の対象物像の対応関係を誤って把握して視差が不正確なものとなることを防止し、対象物までの距離をより正確に検出することができる。
【００２６】
またＲＬ視差Δｄ１とＬＲ視差Δｄ２との偏差が大きいときは、両者を算出したときの第１の相関特性と第２の相関特性とを比較して、相関性が高い方の視差を選択するようにしたので、例えば図９に示すような場合であっても誤りなく、距離を算出することができる。また、相関特性の比較を行っても選択できないときは、そのデータによる距離の算出を行わないようにしたので、誤ったデータの取得を防止することができる。
【００２７】
画像処理ユニット２は、図３の処理により算出された対象物までの距離と、対象物の他の位置情報（自車両１０の進行方向に対して垂直な水平方向及び高さ方向の座標）とに基づいて、その対象物と自車両１０との衝突の可能性を判定し、衝突の可能性が高いときは、スピーカ３及びＨＵＤ４を介して運転者への警報を発する。
【００２８】
本実施形態では、赤外線カメラ１Ｒ、１Ｌが、第１及び第２の撮像手段に相当し、画像処理ユニット２が距離検出装置を構成する。具体的には、図３のステップＳ１４〜Ｓ１６が第１の視差算出手段に相当し、同図のステップＳ１７〜Ｓ１９が第２の視差算出手段に相当し、同図のステップＳ２０が視差同一性判断手段に相当し、ステップＳ２１及びＳ２２が距離算出手段に相当する。そして、右対象物像及び左対象物像がそれぞれ第１対象物像及び第２対象物像に相当し、ＲＬ視差及びＬＲ視差が、第１及び第２の視差に相当する。
【００２９】
なお本発明は上述した実施形態に限るものではなく、種々の変形が可能である。例えば、上述した実施形態では、ＲＬ視差Δｄ１とＬＲ視差Δｄ２との同一性は、両者の偏差の絶対値に基づいて判定したが、両者の比Δｄ１／Δｄ２が１±ＲＴＨの範囲内にあるか否かにより判定してもよい。ＲＴＨは、視差の算出誤差程度の違いは同一性の範囲内と判定するための閾値である。
また撮像手段として赤外線カメラを用いたが、通常の可視光線のみ検出可能なテレビカメラなどを使用してもよい。また、２つの撮像手段は、離間して配置されていればよく、図２に示した配置に限定されるものではない。
【００３０】
【発明の効果】
以上詳述したように本発明によれば、第１の撮像手段により得られる画像に含まれる第１対象物像を基準として、第２の撮像手段により得られる画像に含まれる、前記第１対象物像に対応する第２対象物像を探索することにより第１の視差が求られ、第２の撮像手段により得られる画像に含まれる前記第２対象物像を基準として、第１の撮像手段により得られる画像に含まれる、前記第２対象物像に対応する第１対象物像を探索することにより第２の視差が求められ、第１及び第２の視差がほぼ同一であるとき、第１及び第２の視差の少なくとも一方を用いて前記対象物までの距離が算出されるので、第１及び第２の撮像手段により得られる２つ画像に含まれる対象物像の対応関係を正しく把握して正確な視差を得ることができ、対象物までの距離をより正確に検出すること可能となる。また第１及び第２の視差がほぼ同一でないと判断されたときは、第１対象物像を基準として第２対象物像を探索したときの第１の相関特性と、第２対象物像を基準として第１対象物像を探索したときの第２の相関特性とを比較して、第１及び第２の視差のいずれかを選択可能かが判断され、選択可能であると判断された場合のみ、より相関性が高い対象物像から得られた視差を用いて対象物までの距離が算出されるので、第１及び第２の視差の偏差が大きい場合でも、正しい視差を選択して正確な距離を検出することができる。
【図面の簡単な説明】
【図１】本発明の一実施形態にかかる距離検出装置を含む、車両の周辺監視装置の構成を示す図である。
【図２】赤外線カメラの配置を説明するための図である。
【図３】距離算出手順を示すフローチャートである。
【図４】赤外線カメラにより得られるグレースケール画像を説明するための図である。
【図５】右対象物像と、対応する左対象物像とを示す図である。
【図６】右対象物像から対応する左対象物像を探索する手法を説明するための図である。
【図７】右対象物像と左対象物像から視差を算出する手法を説明するための図である。
【図８】視差から対象物までの距離を算出する手法を説明するための図である。
【図９】相関特性の比較を説明するための図である。
【図１０】視差を誤って算出するおそれのある場合を説明するための図である。
【符号の説明】
１Ｒ、１Ｌ 赤外線カメラ（第１及び第２の撮像手段）
２ 画像処理ユニット（第１の視差算出手段、第２の視差算出手段、距離算出手段）[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a distance detection device that calculates a parallax of an object from two images obtained by two imaging devices and detects a distance to the object based on the parallax.
[0002]
[Prior art]
As a method for detecting the distance to the object included in the image based on the two images obtained by the two imaging devices, the two images are divided into, for example, blocks of 4 pixels × 4 pixels, 2. Description of the Related Art Conventionally, it is known to find a matching block in the other image based on a block of an image, calculate a parallax, and detect a distance to an object based on the parallax (Japanese Patent Laid-Open No. 6-266828). Issue gazette).
[0003]
[Problems to be solved by the invention]
However, in the conventional method, in order to find a matching block, it is necessary to perform a correlation operation on the entire image, and there is a problem that the amount of calculation becomes enormous. Further, when there are a plurality of object images having similar shapes (similar as images) close to each other in the image, the correspondence between the two images may be erroneously recognized. If the parallax is calculated based on the corresponding relationship and the distance is calculated from the parallax, there is a problem that the accuracy of detecting the distance is lowered.
[0004]
The present invention has been made paying attention to this point, and correctly grasps the correspondence between the object images included in the images obtained by the two imaging devices, calculates an accurate parallax, and improves the distance detection accuracy. It is an object of the present invention to provide a distance detection device that can perform the above-described process.
[0005]
[Means for Solving the Problems]
In order to achieve the above object, the invention described in claim 1 calculates the parallax of an object from two images obtained by the first and second imaging means, and calculates the distance to the object based on the parallax. In the distance detection device to detect, the first object image included in the image obtained by the second image pickup unit with reference to the first object image included in the image obtained by the first image pickup unit. The first parallax calculating means for obtaining the first parallax by searching for the corresponding second object image, and the second object image included in the image obtained by the second imaging means as a reference, Second parallax calculating means for obtaining a second parallax by searching for a first object image corresponding to the second object image included in an image obtained by the first imaging means; Whether the second parallax is almost the same When the parallax identity determining means for determining whether or not the parallax identity determining means determines that the parallax is substantially the same, at least one of the first and second parallaxes is used to reach the object. A first correlation characteristic (when the second object image is searched with respect to the first object image when the distance is calculated and the disparity identity determining unit determines that the disparity is not substantially the same). L1) is compared with the second correlation characteristic (L2) when the first object image is searched with reference to the second object image, and one of the first and second parallaxes is compared. And a distance calculation unit that calculates the distance using a parallax obtained from an object image having a higher correlation only when it is determined that selection is possible and only when it is determined that selection is possible. .
[0006]
Here, “the parallax is substantially the same” is determined based on, for example, whether or not the absolute value of the deviation between the first and second parallaxes is equal to or less than a predetermined threshold value. In this case, the predetermined threshold is a threshold for determining the identity of the first and second parallaxes, and is set based on an error when the parallax is calculated a plurality of times by changing the reference image for one target object. . That is, if the deviation between the first and second parallaxes is within the error range, the values are determined to be the same. Also, the identity of the parallax may be determined based on whether or not the ratio of the first and second parallax is within a predetermined range centered at 1 (1 ± ΔRTH).
[0007]
According to this configuration, the first object image included in the image obtained by the second imaging unit is used as a reference, and the first object image included in the image obtained by the first imaging unit is used as a reference. The first parallax is obtained by searching for the two object images, and is included in the image obtained by the first imaging means with the second object image contained in the image obtained by the second imaging means as a reference Second parallax is obtained by searching for a first object image corresponding to the second object image, and the first and second parallaxes are substantially the same. Since the distance to the object is calculated using at least one of the above, the correspondence between the object images included in the two images obtained by the first and second imaging means is correctly grasped and accurate parallax is obtained. And get more accurate distance to the object It becomes possible to detect. When it is determined that the first and second parallaxes are not substantially the same, the first correlation characteristic when the second object image is searched with reference to the first object image and the second object image are obtained. When the first correlation image is compared with the second correlation characteristic when the first object image is searched for as a reference, and it is determined whether one of the first and second parallaxes can be selected. Since the distance to the object is calculated using the parallax obtained from the object image having higher correlation, the correct parallax can be selected accurately even when the deviation between the first and second parallaxes is large. A simple distance can be detected.
[0008]
The first and second parallax calculation means are correlations indicating a correlation between a reference object image included in one image and a local region having the same shape as the object image included in the other image. sexual parameter (C (a, b)) is calculated, Rukoto to search probe corresponding object image is desired on the basis of the correlation parameter.
[0009]
Also the distance calculating means, when the first and second parallax is not substantially the same, can not select one of the first and the second parallax even by referring to the first and second correlation characteristics In some cases, it is desirable not to calculate the distance to the object.
Thereby, it is possible to prevent erroneous distance data from being acquired when there are a plurality of object images that are remarkably similar in the image.
[0010]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a diagram showing a configuration of a vehicle periphery monitoring device including a distance detection device according to an embodiment of the present invention. This device includes two infrared cameras 1R and 1L capable of detecting far infrared rays, Obtained by a yaw rate sensor 5 that detects the yaw rate of the vehicle, a vehicle speed sensor 6 that detects the travel speed (vehicle speed) VCAR of the vehicle, a brake sensor 7 that detects the amount of brake operation, and these cameras 1R and 1L. An image processing unit 2 that detects an object such as an animal in front of the vehicle based on the image data and issues a warning when there is a high possibility of a collision; a speaker 3 that issues a warning by sound; and a camera 1R or 1L The head-up display (hereinafter referred to as “HUD”) for displaying the image obtained by the above and for allowing the driver to recognize an object having a high possibility of collision. And a 4.
[0011]
As shown in FIG. 2, the cameras 1R and 1L are arranged in front of the vehicle 10 at almost target positions with respect to the central axis in the lateral direction of the vehicle 10. The optical axes of the two cameras 1R and 1L are They are parallel to each other and fixed so that their height from the road surface is equal. The infrared cameras 1R and 1L have a characteristic that the output signal level increases (the luminance increases) as the temperature of the object increases.
[0012]
The image processing unit 2 includes an A / D conversion circuit that converts an input analog signal into a digital signal, an image memory that stores a digitized image signal, a CPU (Central Processing Unit) that performs various arithmetic processes, and data that is being calculated by the CPU Output circuit for outputting RAM (Random Access Memory) used for storing CPU, ROM (Read Only Memory) for storing programs and tables executed by CPU, map, etc., driving signal for speaker 3, display signal for HUD4, etc. The output signals of the cameras 1R and 1L and the sensors 5 to 7 are converted into digital signals and input to the CPU.
As shown in FIG. 2, the HUD 4 is provided so that the screen 4 a is displayed at the front position of the driver on the front window of the vehicle 10.
[0013]
FIG. 3 is a flowchart showing a processing procedure in the image processing unit 2. First, the output signals of the cameras 1R and 1L are A / D converted and stored in the image memory (steps S11, S12, and S13). The image stored in the image memory is a grayscale image including luminance information. FIGS. 5A and 5B are diagrams for explaining gray scale images obtained by the cameras 1R and 1L (a right image is obtained by the camera 1R and a left image is obtained by the camera 1L), respectively. The hatched area is an intermediate gradation (gray) area, and the area surrounded by a thick solid line has a high luminance level (at a high temperature) and is an area of an object displayed as white on the screen (hereinafter, “ It is referred to as a “high luminance region”. In the right image and the left image, the horizontal positions of the same target object on the screen are shifted and displayed, and the distance to the target object can be calculated from this shift (parallax).
[0014]
In subsequent steps S14 to S16, a parallax (hereinafter referred to as “RL parallax”) using the right image as a reference image is obtained. First, in step S14, by selecting one of the object images included in the right image, as shown in FIG. 5A, the right object image R1 (here, the entire region surrounded by the circumscribed rectangle is displayed). Extract right object image). In the subsequent step S15, a left image search area for searching for a left object image corresponding to the right object image R1 is set in the left image, and a correlation operation is executed to extract a left object image. Specifically, as shown in FIG. 5B, the left image search region R2 is set in the left image according to each vertex coordinate of the right object image R1, and the right object image R1 is set in the search region R2. The luminance difference total value C (a, b) indicating the height of the correlation with is calculated by the following equation (1). The sum value C (a, b) as the correlation parameter indicates that the smaller the value is, the higher the correlation is. Therefore, an area where the sum value C (a, b) is minimum is set as the left object image. Extract. When there is past position data for the same object, an area R2a (indicated by a broken line in FIG. 5B) narrower than the search area R2 is set as the left image search area based on the position data.
[Expression 1]

[0015]
Here, IR (m, n) is a luminance value at the position of the coordinate (m, n) in the right object image R1 shown in FIG. 6, and IL (a + m−M, b + n−N) is a search region. The luminance value at the position of the coordinate (m, n) in the local region R3 having the same shape as the right object image R1, with the coordinate (a, b) in the base as the base point. The position of the left object image is specified by changing the base point coordinates (a, b) to obtain the position where the luminance difference total value C (a, b) is minimized.
[0016]
As shown in FIG. 7, the right object image R1 and the left object image R4 corresponding to this object image (that is, the same object) are extracted by the process of step S15. A distance dR (number of pixels) between the center of gravity of the object image R1 and the image center line LCTR and a distance dL (number of pixels) between the center of gravity of the left object image R4 and the image center line LCTR are obtained, and the right image is used as a reference. RL parallax Δd1 = dR + dL for an image is calculated.
[0017]
Next, in steps S17 to S19, a parallax (hereinafter referred to as “LR parallax”) using the left image as a reference image is obtained. That is, the left object image R4 is extracted from the left image (step S17), the right image search area for searching for the right object image corresponding to the left object image R4 is set in the right image, and the equation (1) The correlation calculation is performed to extract the right object image corresponding to the left object image R4 (step S18). Then, the LR parallax Δd2 is calculated in the same manner as in step S16 (step S19).
[0018]
Thus, if the right object image R1 is obtained as a result of searching the right image for the image corresponding to the left object image R4, the RL parallax Δd1 and the LR parallax Δd2 are substantially equal. It is verified that it is obtained. Therefore, in step S20, it is determined whether or not the absolute value | Δd1−Δd2 | of the deviation between the RL parallax Δd1 and the LR parallax Δd2 is equal to or smaller than a predetermined threshold value ΔdTH. The distance z between the host vehicle 10 and the object is calculated by applying the parallax of either one of the parallax and the LR parallax or the average value of both to Δd as shown in the following equation (2) (step S22).
[Expression 2]

[0019]
Here, B is the base length, that is, the horizontal distance between the center position of the image sensor 11R of the camera 1R and the center position of the image sensor 11L of the camera 1L as shown in FIG. 8 (interval between the optical axes of both cameras). , F are focal lengths of the lenses 12R and 12L, and p is a pixel interval in the image sensors 11R and 11L.
[0020]
On the other hand, if | Δd1−Δd2 |> ΔdTH in step S20, the first correlation when the RL parallax Δd1 is calculated from the result of the correlation calculation (steps S15 and S18) executed when calculating the parallax. The characteristic and the second correlation characteristic when the LR parallax Δd2 is calculated are compared, and it is determined whether or not the parallax obtained from the object image with higher correlation can be selected (step S21). If YES, the process proceeds to step S22, and the distance z is calculated using the parallax obtained from the object image with higher correlation. If the correlation characteristics cannot be selected, that is, if the first and second correlation characteristics are almost the same, or if both correlation characteristics indicate low correlation, the distance calculation is executed. The process is terminated without doing so.
[0021]
FIG. 9 is a diagram for explaining the above-mentioned correlation characteristics. The luminance difference total value C (a, b), which is a correlation parameter between two object images, is set to a constant variable b (the minimum total value is obtained). It is shown as a function of the variable a). In this figure, lines L1 and L2 correspond to the first correlation characteristic and the second correlation characteristic, respectively. As correlation characteristic parameters representing this correlation characteristic, for example, minimum values CMIN1, CMIN2, and / or Alternatively, the absolute value DC of the slopes of the lines L1 and L2 (for example, slopes of straight lines L1a, L1b, L2a, and L2b indicating average slopes) in the vicinity of the variable values a1 and a2 that take the minimum values is used. The minimum value CMIN indicates that the smaller the value is, the higher the correlation is, and the absolute value DC of the slope indicates that the larger the value is, the higher the correlation is. In the example shown in FIG. 9, since the second correlation characteristic indicated by the line L2 is NG (low correlation), it is determined that the reliability of the LR parallax Δd2 is low, and the distance z is calculated using the RL parallax Δd1. I do. For example, when the minimum value CMIN is smaller than the predetermined threshold value CMINTH and the absolute value CD of the inclination is larger than the predetermined inclination value CDTH, the correlation characteristic is determined to be OK, and otherwise determined to be NG. When both the first and second correlation characteristics are OK, the smaller one of the minimum values CMIN is adopted. However, when the difference ΔCMIN is small (when it is equal to or smaller than the predetermined value ΔCMINTH), the distance is not calculated as will be described later.
[0022]
FIG. 10 is a diagram for explaining a case where the deviation between the RL parallax Δd1 and the LR parallax Δd2 becomes large. The object images R11 and R12 corresponding to two people are present in the right image. Since there is another vehicle indicated by the broken line OBJX between the person and the host vehicle 10, there is no left object image corresponding to the right object image R12 in the left image (hidden behind other vehicles). Shows the case. Although this figure is a grayscale image, only the image related to the object image is shown for easy understanding, and the circumscribed rectangle indicating the range where the correlation calculation is performed is omitted. Further, the object images R13 and R14 shown in the figure are obtained by capturing the tail lamps of other vehicles as a high luminance area.
[0023]
In the case shown in this figure, the right object image R11 is extracted from the right image in steps S14 to S16, the corresponding left object image R21 is extracted from the left image, and the left object image R21 is extracted in steps S17 to S19. If R12 is mistakenly extracted as the corresponding right object image, the LR parallax Δd2 is larger than the RL parallax Δd1, so the answer to step S20 is negative (NO), and the flow proceeds to step S21 to compare the correlation characteristics. Do. In this example, since the first correlation characteristic when the RL parallax Δd1 is calculated has a higher correlation than the second correlation characteristic when the LR parallax Δd2 is calculated, the RL parallax Δd1 is selected as the correct parallax. Then, the process proceeds to step S22.
[0024]
Further, when the right object R12 is extracted from the right image in steps S14 to S16, a similar left object image R21 is extracted from the left image. However, in steps S17 to S19, the left object image R21 corresponds to the left object image R21. Since R11 is extracted as the right object image to be performed, the answer to step S20 is negative (NO), and the process proceeds to step S21. When the correlation characteristics are compared in step S21, this time, the second correlation characteristic when the LR parallax Δd2 is calculated indicates that the correlation is higher than the first correlation characteristic when the RL parallax Δd1 is calculated. The LR parallax Δd2 is set as the correct parallax, and the process proceeds to step S22.
[0025]
Further, when the right object images R11 and R12 are very similar, since either one of the parallaxes cannot be selected, the answer to step S21 is negative (NO), and the process is performed without calculating the distance z. finish.
As described above, in the present embodiment, the RL parallax Δd1 based on the right image and the LR parallax Δd2 based on the left image are calculated, and it is determined that the correct parallax is obtained when the two substantially coincide with each other. Since the distance z is calculated based on the parallax obtained in this way, it is possible to prevent the parallax from becoming inaccurate by erroneously grasping the correspondence between the object images in the left and right images. The distance to the object can be detected more accurately.
[0026]
When the deviation between the RL parallax Δd1 and the LR parallax Δd2 is large, the first correlation characteristic and the second correlation characteristic obtained by calculating both are compared, and the parallax having the higher correlation is selected. Thus, even in the case shown in FIG. 9, for example, the distance can be calculated without error. In addition, when the correlation characteristics cannot be selected even if they are compared, the calculation of the distance based on the data is not performed, so that erroneous data acquisition can be prevented.
[0027]
The image processing unit 2 calculates the distance to the object calculated by the process of FIG. 3 and other position information of the object (the horizontal and vertical coordinates perpendicular to the traveling direction of the host vehicle 10). Based on the above, the possibility of a collision between the object and the host vehicle 10 is determined. When the possibility of the collision is high, an alarm is issued to the driver via the speaker 3 and the HUD 4.
[0028]
In the present embodiment, the infrared cameras 1R and 1L correspond to first and second imaging means, and the image processing unit 2 constitutes a distance detection device. Specifically, steps S14 to S16 in FIG. 3 correspond to first parallax calculation means, steps S17 to S19 in FIG. 3 correspond to second parallax calculation means, and step S20 in FIG. It corresponds to the determination means, and steps S21 and S22 correspond to the distance calculation means. The right object image and the left object image correspond to the first object image and the second object image, respectively, and the RL parallax and the LR parallax correspond to the first and second parallaxes, respectively.
[0029]
The present invention is not limited to the embodiment described above, and various modifications can be made. For example, in the above-described embodiment, the identity between the RL parallax Δd1 and the LR parallax Δd2 is determined based on the absolute value of the deviation between the two, but is the ratio Δd1 / Δd2 between the two in the range of 1 ± RTH? It may be determined depending on whether or not. RTH is a threshold value for determining that the difference in the parallax calculation error is within the same range.
Further, although an infrared camera is used as the imaging means, a television camera that can detect only normal visible light may be used. Further, the two image pickup means need only be arranged apart from each other, and is not limited to the arrangement shown in FIG.
[0030]
【The invention's effect】
As described above in detail, according to the present invention, the first object included in the image obtained by the second imaging means on the basis of the first object image contained in the image obtained by the first imaging means. By searching for a second object image corresponding to the object image, the first parallax is obtained, and the first imaging unit is based on the second object image included in the image obtained by the second imaging unit. When the second parallax is obtained by searching for the first object image corresponding to the second object image included in the image obtained by the above, and the first and second parallaxes are substantially the same, Since the distance to the object is calculated using at least one of the first and second parallaxes, the correspondence between the object images included in the two images obtained by the first and second imaging means is correctly grasped. Can get accurate parallax, Distance becomes possible to more accurately detect. When it is determined that the first and second parallaxes are not substantially the same, the first correlation characteristic when the second object image is searched with reference to the first object image and the second object image are obtained. When the first correlation image is compared with the second correlation characteristic when searching for the first object image as a reference, it is determined whether one of the first and second parallaxes can be selected, and it is determined that the selection is possible Since the distance to the object is calculated using the parallax obtained from the object image with higher correlation, the correct parallax can be selected and accurately selected even when the deviation between the first and second parallaxes is large. A simple distance can be detected.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration of a vehicle periphery monitoring device including a distance detection device according to an embodiment of the present invention.
FIG. 2 is a diagram for explaining the arrangement of an infrared camera.
FIG. 3 is a flowchart showing a distance calculation procedure.
FIG. 4 is a diagram for explaining a grayscale image obtained by an infrared camera.
FIG. 5 is a diagram illustrating a right object image and a corresponding left object image.
FIG. 6 is a diagram for explaining a method of searching for a corresponding left object image from a right object image.
FIG. 7 is a diagram for explaining a method of calculating parallax from a right object image and a left object image.
FIG. 8 is a diagram for explaining a method of calculating a distance from a parallax to an object.
FIG. 9 is a diagram for explaining a comparison of correlation characteristics.
FIG. 10 is a diagram for explaining a case where there is a possibility that parallax may be erroneously calculated.
[Explanation of symbols]
1R, 1L infrared camera (first and second imaging means)
2 Image processing unit (first parallax calculating means, second parallax calculating means, distance calculating means)

Claims (1)

In a distance detection apparatus that calculates a parallax of an object from two images obtained by the first and second imaging means, and detects a distance to the object based on the parallax.
The second object image corresponding to the first object image included in the image obtained by the second image pickup means with reference to the first object image included in the image obtained by the first image pickup means. First parallax calculating means for obtaining the first parallax by searching for;
The first object corresponding to the second object image included in the image obtained by the first image pickup means with reference to the second object image included in the image obtained by the second image pickup means. Second parallax calculating means for obtaining a second parallax by searching for an image;
Parallax identity determining means for determining whether or not the first and second parallaxes are substantially the same;
When the parallax identity determining means determines that the parallax is substantially the same, the distance to the object is calculated using at least one of the first and second parallaxes, and the parallax identity determining means When the parallax is determined to be substantially the same, the first correlation characteristic when the second object image is searched with reference to the first object image, and the second object image with reference to the second object image. Only when the first correlation image is compared with the second correlation characteristic when the first object image is searched to determine whether one of the first parallax and the second parallax can be selected. And a distance calculating unit that calculates the distance using a parallax obtained from an object image having higher correlation .

Images