JP3750436B2 - Object display method and apparatus - Google Patents

Description

【００６６】
各方向のうち、最小値となる方向を移動方向と定義する。ただし、各記号は（数１）、（数２）と同様であり、abs()は、()内の値の絶対値をとる関数である。
【００６７】
【数１６】

【００６８】
ただし、各記号は（数１）、（数２）と同様である。
【００６９】
各方向のうち、最小値となる方向を移動方向と定義する。
以上移動方向検出の方法の具体例をa)〜g)まで示した。これらの方法のどれを用いることもできるし、その他の方法を用いることもできる。また、これらの方法を複数使い、
以上の操作を各点について行うが、画像の端の点は、同様の所定のサンプリング点を取れないため、特殊処理を行う必要がある。特殊処理として、画像の両端がつながっているとして中心部と同様の所定のサンプリング点を取る方法や、端の点は処理をせずに原画や０を出力する方法、端のサンプリング点を少なくして処理する方法、原画のまま出力する方法など、端の部分の特殊処理方法は各種ある。
【００７０】
ステップ６０３：非線形平滑化処理
ステップ６０２により求められた方向に非線形平滑化処理を行なう。ただし、ステップ６０２と同じサンプリング点と平滑化方法を用いる場合には、ステップ６０２で得られた平滑化値をそのまま用いることができる。非線形平滑化処理の方法は各種ありステップ６０２の平滑化方法と同一である。
【００７１】
ステップ６０４：データ出力処理
平滑化して得られた探知画像を出力する。
【００７２】
以上により、着目した探知画像から時間的に−方向に船の移動量を考慮した被写体の移動方向を求めその方向に推定されたデータを参照して平滑化処理することができるようになる。
【００７３】
（２） 課題を解決する本発明第２の実施例
第２の実施例は、障害物が移動している場合に有効な例を説明する。
流木などの障害物が潮流に流され移動している場合がある。この場合は、同一の方向に信号が検出されず、時系列的に障害物の方向は変化する。この移動する障害物に対応するため図７に示すように障害物がどの方向を推移したかを推定するためX軸サーチ範囲７０１を設ける。X軸サーチ範囲は、潮流や検出する障害物などにより常識的な範囲があり自由に設定できる。この場合、探知画像t=0の時の着目点G(i,j,0)がX軸サーチ範囲７０１とY軸サーチ範囲１０６に囲まれた矩形内の点を結んだ各方向より移動方向を求め、平滑化すべき信号がどの方向に有るかを推定することになる。
【００７４】
図８は、図７を拡大したものである。X軸サーチ範囲７０１とY軸サーチ範囲１０６に囲まれた矩形内の各点と探知画像t=0の時の着目点AのG(i,j,0)を結んだ方向を定義する。白丸８０１で示した点が画像上の画素のサンプリング点で、黒丸８０２で示した点が画素上になく補間を必要とするサンプリング点である。補間方法は上記移動方向検出手段で述べたように各種ある。それぞれの方向にサンプリングされた点を用いて移動方向を検出し、最小となる方向に１次元非線形平滑化処理を行ない平滑化する。
【００７５】
また、時刻t=0より過去のデータを参照し平滑化する方法だけでなく、図９に示すように図１の探知画像t=0より後に取得される探知画像t=1からt=10を用い、前後の時刻の探知画像を用い平滑化することも可能である。任探知画像t=0の着目点AのG(i,j,0)１０１とした時、時間軸−方向はいままでと同様の構成を持ち、時間軸＋方向に時間軸＋方向の平滑化時間９０３、時間軸＋方向の船の移動量を考慮したオフセット９０１、時間軸＋方向のY軸サーチ範囲９０２を設ける。時間軸＋方向の平滑化時間は、例のように11とするだけでなく自由に設定できる。また、時間軸＋方向の船の移動量を考慮したオフセット、時間軸＋方向のＹ軸サーチ範囲も同様に自由に設定できる。探知画像t=0の着目点AのG(i,j,0)に対し時間軸＋方向、時間軸−方向に移動方向を検出し、求められたそれぞれの移動方向から１次元データを選択し１次元の非線形平滑化処理を行なうこともできる。これにより着目した探知画像から時間的に＋方向及び−方向に船の移動量を考慮した被写体の移動方向を求めその方向に推定されたデータを参照して平滑化処理することができるようになる。
【００７６】
図１０は本発明の課題を解決する方法の処理手順を示した図であり、第１の実施例と異なるステップのみ図１０の番号をステップ番号として以下に説明する。
【００７７】
ステップ１００１：データ圧縮処理１
入力された探知画像tに対し、S/N比を向上するためにY軸方向にデータを1/pに圧縮する（数１７）に従った処理を行なう。
【００７８】
【数１７】
G'(i,j,t)=press(i,j*p,t)…（数１７）
ただし、G'(i,j,t)は1/pに圧縮した出力値であり、pressはG(i,q,t)からG(i,q,t)までの最大値をとる関数であり、qはj*pからj*p+p-1までの値をとる。pは正の整数値である。以上の操作を出力画像G'(i,j,t)の各点について元画像より圧縮した値を求めるが、画像の端の点は、同様の所定のサンプリング点を取れないため、特殊処理を行う必要がある。画像の端の部分の特殊処理については、以下のステップ１００２のデータボケ化処理を行う。
【００７９】
ステップ１００２：データボケ化処理
探知画像tに対し、信号検出を容易にするために着目点G(i,j,t)とY軸方向の前後の点G(i,j-n,t)からG(i,j+n,t)の内、最大値を選択する（数１８）の処理を行なう。
【００８０】
【数１８】
G'(i,j,t)=max(i,j,t,n)…（数１８）
ただし、G'(i,j,t)はデータボケ化処理後の値であり、maxはG(i,j,t)を中心にしてY軸方向に前後n点の内最大値を返す関数である。以上の操作を各点について行うが、画像のY軸方向の端の点は、同様の所定のサンプリング点を取れないため、特殊処理を行う必要がある。
【００８１】
ステップ１００３：データ圧縮処理２
平滑化後の探知画像に対し、S/N比を向上するためにY軸方向にデータを圧縮する処理をデータ圧縮処理１と同様に行なう。
【００８２】
ステップ１００４：濃度調節処理
平滑化後の探知画像に対し、濃度を調節し障害物の信号をより検出し易くする（数１９）の処理を行なう。
【００８３】
【数１９】
G'(i,j,t)=G(i,j,t)/ave(i,j,t,n);…（数１９）
ただし、 G'(i,j,t)は濃度変換処理後の値であり、aveはG(i,j,t)を中心にしてY軸方向に前後n点の平均を返す関数である。
【００８４】
以上の操作を各点について行うが、画像のY軸方向の端の点は、同様の所定のサンプリング点を取れないため、特殊処理を行う必要がある。
【００８５】
（３） 課題を解決する本発明第３の実施例
探知画像のエッジ画像を用いて移動方向の検出を行ない、この検出された移動方向を用いて探知画像tの平滑化を行なう。図１１は本発明の課題を解決する方法の処理手順を示した図であり、第２の実施例と異なるステップのみ図１１の番号をステップ番号として以下に説明する。
【００８６】
ステップ１１０１：エッジ処理
探知画像tに対しエッジ画像を作成する。エッジ画像については上述の文献２に記載されている。
【００８７】
ステップ１１０２：移動方向検出処理
ステップ１１０１で作成された探知画像tのエッジ画像を用いて移動方向を検出する。移動方向の検出処理は、実施例１のステップ６０２と同様である。
【００８８】
ステップ１１０３：非線形平滑化処理
ステップ１１０２で求められた各点の移動方向に平滑化処理を行なう。非線形平滑化処理方法は実施例１のステップ６０３と同様である。
【００８９】
（４） 課題を解決する本発明第４の実施例
船の移動距離と移動方向検出手段により検出された障害物の移動方向より障害物の深度を測定することができる。図１２は、船の移動方向と障害物の移動方向が一致する場合の模擬図である。時刻t=0の時の障害物までの距離と深度および他のパラメータとの相関関係を表した。この場合の船の障害物までの水平距離yと障害物の深度dの求め方を以下に示す。
【００９０】
時刻t=0における船の位置１２０１、時刻t=-10における船の位置１２０２とした時、時刻t=-10から時刻t=0までに移動する船の距離１２０５をl、時刻t=0における障害物までの距離１２０３を 、時刻t=-10における障害物までの距離１２０４を とすると、時系列に重ねた探知画像の時間軸−距離断面図においては、移動方向検出手段により検出された方向が方向１２０８の時、時刻t=-10から時刻t=0までに移動する船の距離１２０５は、１０５にあたり、時刻t=0における障害物までの距離１２０３は１２０９にあたり、時刻t=-10における障害物までの距離１２０４は、１２１０にそれぞれあたる。
【００９１】
時刻t=0の時の障害物までの水平距離１２０６をy、障害物の深度１２０７をdとすると以下の（数２０）、（数２１）を用いて求めることができる。
【００９２】
【数２０】

【００９３】
【数２１】

【００９４】
なお、上記の実施例では、船舶について述べているが、航空機、潜水艦などの他のものでも適用可能である。また、ノイズ除去に関しては、アクティブソナーで検出されたもののみでなく、画像処理における表示全般について適用可能である。
【００９５】
【発明の効果】
本発明によれば、時系列的に得られる探知画像からノイズを除去し、船舶航行の障害物を探知し船舶の航行が安全になるという効果がある。
【図面の簡単な説明】
【図１】本発明による時系列に重ねた探知画像を最小変化方向に１次元非線形平滑化する処理の概念図である。
【図２】図１の時系列に重ねた探知画像の時間軸−距離断面図を示す図である。
【図３】船からのアクティブソナー観測領域を示す図である。
【図４】船からのビームフォーミング測定領域と測定信号を示す図である。
【図５】時系列に並べた探知画像を示す図である。
【図６】本発明第１の実施例の手順を示す図である。
【図７】障害物の移動を考慮して最小変化方向に１次元非線形平滑化する処理の概念図である。
【図８】図７を拡大した図である。
【図９】時間軸＋方向及び−方向に最小変化方向に１次元非線形平滑化する処理の概念図である。
【図１０】本発明第２の実施例の手順を示す図である。
【図１１】本発明第３の実施例の手順を示す図である。
【図１２】船の移動方向と障害物の移動方向が一致する場合の模擬図である。
【符号の説明】
１０１：探知画像t=0の時の着目点G(i,j,0)、１０２：推定最大移動点、１０３：推定最小移動点、１０４：探知画像t=-10の時の着目点(i,j,-10)、１０５：船の移動量を考慮したオフセット、１０６：ｙ軸サーチ範囲、１０７：平滑化時間、２０１：画素上のサンプリング点、２０２：補間を要すサンプリング点、３０１：t=0の時の船、３０２：t=-10の時の船、３０３：船が座礁する岩、３０４：t=0の時の観測域、３０５：t=-10の時の観測域、４０１：方向A、４０２：方向B、４０３：方向C、４０４：方向Aにビームフォーミングした測定領域、４０５：方向Bにビームフォーミングした測定領域、４０６：方向Cにビームフォーミングした測定領域、４１０：船の観測域、４１１：方向Aにビームフォーミングして測定された信号、４１２：方向Bにビームフォーミングして測定された信号、４１３：方向Cにビームフォーミングして測定された信号、５０１：t=0の時の方向Aから方向Cまでの各方向の信号を濃度変換して得られた探知画像、５０２：t=0の時に測定された障害物、５０３：t=-1の時に測定された障害物、５０４：t=-2の時に測定された障害物、５０５：t=-10の時に測定された障害物、６０１：データ入力処理、６０２：移動方向検出処理、６０３：非線形平滑化処理、６０４：データ出力処理、７０１：ｘ軸サーチ範囲、８０１：画素上のサンプリング点、８０２：補間を要すサンプリング点、９０１：時間ｔ軸＋方向の船の移動量を考慮したオフセット、９０２：時間ｔ軸＋方向のｙ軸サーチ範囲、９０３：時間ｔ軸＋方向の平滑化時間、１００１：データ圧縮処理１、１００２：データボケ化処理、１００３：データ圧縮処理２、１００４：濃度調節処理、１１０１：エッジ処理、１２０１：時刻t=0における船、１２０２：時刻t=-10における船、１２０３：時刻t=0における障害物までの距離、１２０４：時刻t=-10における障害物までの距離、１２０５：時刻t=-10から時刻t=0までに移動する船の距離、１２０６：時刻t=0の時の障害物までの水平距離、１２０７：障害物の深度、１２０８：移動方向検出手段により検出された方向、１２０９：探知画像の時間軸−距離断面図における時刻t=0における障害物までの距離、１２１０：探知画像の時間軸−距離断面図における時刻t=-10における障害物までの距離。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to object detection for detecting an object. Among them, the present invention relates to detecting an object existing in or on water and displaying the detected object. Moreover, it is related with the system which navigates a ship safely regarding ship obstacle detection. It also relates to fish school detection that detects underwater fish.
[0002]
[Prior art]
The following documents are related to the present invention.
[0003]
(1) Literature 1: Japan Society of Ocean Acoustics Journal of Ocean Acoustic Society Vol. 25 No. 1 (Vol. 89) P25-P29, this document relates to active sonar.
[0004]
(2) Japanese Unexamined Patent Publication No. 7-181255 Kunihiko Mano Fumio Nakamura: Ship Collision Stranding Prevention System Above (1) emits a search wave in the search direction by a generally known method as an active sonar, and a reflected wave By measuring, the distance to the obstacle is measured. Thereby, a detection image is obtained in time series.
[0005]
In (2) above, an unmanned helicopter mounted on a large transport ship is caused to fly by remote control using a helicopter controller, and the water surface and depth to the bottom of the water based on the reflected light of the laser beam from the unmanned helicopter and the current flight position are determined. Receive on. The depth from the unmanned helicopter and the current flight position are received and displayed on the CRT, and when it is determined that it is a reef from that depth, the navigation body is run near the reef, Is used to obtain detailed information on the seabed or reef, and based on this detailed information, the seabed shape is displayed three-dimensionally.
[0006]
[Problems to be solved by the invention]
The prior art (1) is used to detect obstacles that hinder the navigation of ships such as dark reefs and driftwood. However, the reflected wave due to the active sonar includes noise due to sensor performance, tidal current, etc. in addition to the reflected wave of the obstacle.
[0007]
The conventional technique (2) describes a method of performing operation based on a detected image photographed using the conventional technique (1), but does not describe a noise removal method of the detected image.
[0008]
The problem to be solved by the present invention is to detect the exact position of an object, in particular the depth or height direction. Another object is to remove the noise from the signal obtained in time series, distinguish the object from the displayed signal, and extract the object. Further, it is intended to make the navigation of the ship safe by removing noise and making it easy to detect obstacles in the navigation of the ship.
[0009]
[Means for Solving the Problems]
In order to solve the above problems, the following configuration is adopted.
[0010]
(1) Moving direction detection means
FIG. 1 is a representation in which detection images t = −10 to detection image t = 0 are sequentially stacked. The time from t = 0 to t = −10 used for smoothing is referred to as smoothing time 107. In the example of FIG. The smoothing time 107 is not only 11 as in the example of FIG. 1, but may be shorter or longer than 11.
[0011]
If the ship is not moving, G (i, j, 0) 101 of the point of interest A when the detection image t = 0 is G (i, j, 0) 101 of the same point of interest D when the detection image t = -10. , -10) should be the same signal value as 104. However, when the ship moves, the relative position of the obstacle from the ship changes. In order to cope with this change, from the smoothing time 107 and the navigation speed of the ship, an offset considering the amount of movement of the ship from G (i, j, -10) of the point of interest D when the detection image t = -10. 105 is provided. Further, there are errors due to smoothing time 107, tidal current, underwater temperature and other factors, and an estimated maximum movement point B102 and an estimated minimum movement point C103 are provided from an offset considering the amount of movement of the ship. A range up to the estimated minimum movement point C103 is referred to as a Y-axis search range 106.
[0012]
The maximum estimated movement point 102, the minimum estimated movement point 103, and the offset 105 considering the movement amount of the ship are specified by using predetermined calculation values in accordance with the ship speed, the smoothing time 107, and other error factors. be able to.
[0013]
The maximum estimated moving point 102, the minimum estimated moving point 103, and the offset 105 considering the amount of movement of the ship can be freely set according to the speed of the ship, the smoothing time 107, and other error factors.
[0014]
The moving direction should be obtained from each direction connecting G (i, j, 0) 101 at the point of interest A at the detection image t = 0 and a point on the Y-axis search range of the detection image t = -10, and smoothed. Estimate in which direction the signal is.
[0015]
FIG. 2 is a projection of the example of FIG. 1 on a t-axis-y-axis cross section. Each point on the straight line of the Y-axis search range 106 from the estimated maximum movement amount B102 to the estimated minimum movement amount C103 is connected to G (i, j, 0) 101 of the point of interest A when the detection image t = 0, and in each direction Is defined. Points indicated by white circles 201 are sampling points of pixels on the image, and points indicated by black circles 202 are sampling points that are not on the pixels and require interpolation. There are various interpolation methods. A method of averaging two nearby points on the detection image, a method of performing cubic convolution from four or 16 points in the vicinity, and a detection image before and after the 64 points in the vicinity are used. There are a method of three-dimensional cubic convolution from a point, a method of setting a specific pixel value among the closest pixel points, and the like.
[0016]
A smoothing process is performed for each sampled direction, and a direction that is the maximum value smoothed in each direction is defined as a moving direction.
[0017]
The moving direction selected by the moving direction detecting means described above is obtained by estimating the positional deviation of the subject image due to the movement of the ship for each point.
[0018]
(2) Non-linear smoothing means
For each point of the image, (1) one-dimensional nonlinear smoothing is performed in a predetermined moving direction obtained by the moving direction detecting means. Here, the sampling points to be smoothed can be selected from the same sampling points as the sampling points used in the moving direction detecting means, or different points can be selected to increase or decrease the number of sampling points. However, (1) when using the same sampling point and smoothing method as the moving direction detecting means, (1) the smoothing value of the moving direction calculated by the moving direction detecting means can be used as it is.
[0019]
The above (1) moving direction detecting means and (2) nonlinear smoothing means may be used together, or one of them may be used independently.
[0020]
The configuration of the present invention is as follows.
[0021]
Input detection images from the active sonar device in time series, select multiple detection images from the detection images input in time series, and associate and associate locations indicating the same object for each of the selected detection images In the object display method, display is performed using a value of a predetermined rank among the values of display data included in each of the given locations.
[0022]
In the object display method, the predetermined order is a value that is a fourth of the number of detected images selected from the maximum value.
[0023]
Further, in this object display method, when the fourth is not an integer, the object display method is characterized in that the display is performed using an average of two rank values closest to the quarter.
[0024]
Further, the present invention provides an obstacle detection device for detecting an obstacle in a moving body, and means for inputting a detection image obtained in a time series from an active sonar device that monitors the traveling direction of the moving body. Means for creating a plurality of groups of possible locations corresponding to obstacles as predetermined locations in the detected images among the detected images using the input detected images, and display data of each created group And an obstacle detection device comprising: means for obtaining a sum of the values of the plurality of groups; and means for selecting a group having the largest sum of the values of the display data among the plurality of groups as a group representing the obstacle. is there.
[0025]
Further, in this obstacle detection device, the obstacles indicated by the selected group are displayed by using the values having the predetermined magnitudes among the display data values of the respective groups. It is also an obstacle detection device characterized by having a display means.
[0026]
Furthermore, this invention is also the following structures.
[0027]
In a system for navigating ship obstacles, in a system for safely navigating a ship, an active sonar device for monitoring the front of the ship, and a device for processing a detection image obtained in time series from the active sonar device, each time The detection images obtained in step 3 are arranged in time series to form three-dimensional data in which a time axis is newly added to the two-dimensional image, paying attention to the detection image at a predetermined time, and in the + direction in time from the target detection image For the partial three-dimensional data limited to a predetermined time in the and-directions, the movement direction of the subject in the partial three-dimensional data taking into account the amount of movement of the ship by the partial three-dimensional data from the target detection image is determined. A ship obstacle detection method characterized in that smoothing processing is performed with reference to data obtained and estimated in the direction.
[0028]
Further, in this ship obstacle detection method, the ship obstacle detection method is characterized by smoothing only in the time axis direction.
[0029]
Further, in this ship obstacle detection method, as a process for obtaining the moving direction in the partial three-dimensional data of the subject in consideration of the movement amount of the ship in the partial three-dimensional data from the focused detection image, Smoothing is performed for each point in a predetermined direction including the time axis, and one-dimensional data in the direction having the maximum value is selected from the smoothed values in the predetermined direction, and the one-dimensional data is smoothed. It is also a ship obstacle detection method characterized by this.
[0030]
Furthermore, in this ship obstacle detection method, as the smoothing method, when the selected one-dimensional data is arranged in ascending order, data values at or about 3/4 points of the number of data are output. It is also a ship obstacle detection method.
[0031]
Further, in this ship obstacle detection method, the movement distance obtained from the movement amount of the ship and the distance change from the movement direction in the three-dimensional data to the subject are obtained, and the movement distance of the ship and the distance change to the subject are obtained. It is also a ship obstacle detection method characterized by measuring the horizontal distance from the ship to the subject and the depth of the subject.
[0032]
DETAILED DESCRIPTION OF THE INVENTION
Each embodiment described below is an invention of a ship obstacle detection method, and can be realized by creating a program or hardware for performing the following method. An image of a reflected wave signal obtained by firing an active sonar at a certain time T is called a detected image t = T. The x-axis on the image represents the direction of the reflected wave signal, and the y-axis represents a ship. Represents the distance from. The amplitude P of the signal at an arbitrary point (i, j) on the detected image t = T is expressed as P = G (i, j, T).
[0033]
As shown in FIG. 3, the ship at each time can fire an active sonar so that the ship 302 at time t = -10 can obtain a reflected wave signal of the observation region 305 at time t = -10. The ship 301 at time t = 0 can obtain a reflected wave signal from the observation region 304 at time t = 0. There is a rock 303 on which the ship is stranded on the route of the ship, and the stranded rock 303 is detected using a reflected wave signal from time t = −10 to time t = 0.
[0034]
FIG. 4 is a model of the observation area of the ship when the reflected wave of the obstacle is measured at a certain time T. The ship 407 can observe the reflected wave in the observation region 400 by beam forming. A measurement region 404 beamformed in the direction A401 can be obtained as a signal 411 measured by beamforming in the direction A401. A measurement region 405 beamformed in the direction B402 can be obtained as a signal 412 measured by beamforming in the direction B402. Similarly, the measurement region 406 beamformed in the direction C403 can be obtained as a signal 413 measured by beamforming in the direction C403. The two-dimensional array 410 is an array of signals in each direction from direction A to direction C. A detected image is obtained by density-converting the two-dimensional array 401 with an appropriate value.
[0035]
FIG. 5 shows detected images obtained from time t = −10 to time t = 0. On the detection image 501 at the time t = 0, there is a stranded rock 502 measured at the time t = 0. The detected images include a grounded rock 503 measured at time t = -1, a grounded rock 504 measured at time t = -2, and a grounded rock measured at time t = -10. 505 is measured. From these detection images from time t = −10 to time t = 0, noise is removed, and stranded rocks that become obstacles for ship navigation are detected.
[0036]
The outline of the present invention is shown below.
The detected image may include burst noise that suddenly becomes a large signal, and the reflected wave of the obstacle may become small and may not be recognized. For example, when detection images can be obtained every second, ten detection images can be obtained in 10 seconds. A process of stably acquiring the reflected wave of the obstacle from the ten images is performed. If there are obstacles such as rocks in the ten detected images, the obstacles appear to move as the ship moves (FIG. 5). The position where the detection image appears is also moved by the movement of the ship, but if there is an obstacle such as a rock on the sea surface or there is an obstacle such as a rock of a certain depth in the sea, the place where the detection image appears on the detection image The moving distance that changes is different. Therefore, the moving direction corresponding to the depth is estimated for each pixel shown in each detected image from the 10 images. Therefore, for the moving direction from depth 0 to the maximum depth observed with sonar, the moving direction according to the depth is defined as shown in FIG. 2, and the object is stabilized from the data value on the line of each moving direction. Judge whether it is reflected.
[0037]
The ten pieces of data in each moving direction include burst noise or data with a small reflected wave signal. Therefore, 10 pieces of data are smoothed. For example, when 10 pieces of data are arranged in ascending order, the largest value or the second largest value may be a burst noise, and the first smallest value or the second smallest value may be an unclear image. There is sex. Therefore, the number of data × 3 / 4th data value when it is arranged in ascending order of data as a value that is reflected as strongly as possible and does not take burst noise is obtained. The number of data when arranged in ascending order of data in each moving direction x 3 / 4th data value is obtained, and the moving direction that takes the maximum value among each moving direction is estimated as the moving direction that seems to actually have a reflector To do. The estimated number of data in the moving direction × 3 / 4th data value is set as the output value. By performing this process on each point of the detection image, noise such as burst noise can be removed, and a stable detection image can be obtained.
[0038]
The first embodiment of the present invention will be described in section (1), the second embodiment in section (2), the third embodiment in section (3), and the fourth embodiment in section (4).
[0039]
(1) First embodiment of the present invention for solving the problems
FIG. 6 is a diagram showing a processing procedure of a method for solving the problems of the present invention, and will be described below using the numbers in FIG. 6 as step numbers.
[0040]
Step 601: Data input processing
The detection images from time t = -10 to time t = 0 are input.
[0041]
Step 602: Movement direction detection processing
The one-dimensional data sampled in each movement direction is smoothed, and the direction having the maximum value among the smoothed values in each movement direction is defined as the movement direction.
[0042]
There are various ways of calculating the movement amount in each movement direction, and specific examples are shown in the following a) to g).
[0043]
a) 3/4 median method
3/4 median value (Expression 1) for a predetermined sampling point in each moving direction.
[0044]
[Expression 1]
S (i, j, f, d) = median3_4 (I (i, j, f, d, m)) ... (Equation 1)
However, i, j is the position of the point of interest of the image, f is the detected image t = f, d is the point of interest G (i, j, f) of the detected image t = f from each point in the Y-axis search range 106 In each direction, m is the order in which the points sampled in each direction from the point of interest G (i, j, f) are ordered, and when m = 0, the point of interest of the detection image t = 0, and m ≧ 1 The time indicates a point on the detection image t = m acquired after the detection image t = 0, and when m ≦ −1, the point on the detection image t = −m acquired before t = 0. (i, j, f, d, m) represents the numerical value of a point that has advanced m sampling points in the d direction from the point of interest G (i, j, f), and median3_4 (i, j, f, d, m) is , A function that returns the (n-1) * 3 / 4th value in ascending order from m sampling points in the d direction, and the closest value if there is no (n-1) * 3 / 4th array , Taking the average of two close array elements, or taking a weighted average of close array elements to give a 3/4 median value. S is the amount of change in the direction d at the point of interest (i, j, f). There are various definitions other than the above (Equation 1) for calculating the amount of change, and examples are shown in the following b) to f).
[0045]
b) Average method
[0046]
[Expression 2]
S (i, j, f, d) = total / m ... (Equation 2)
[0047]
[Equation 3]
total = Σ (I (i, j, f, d, m)) ... (Equation 3)
In this method, a value obtained by averaging the predetermined sampling points in the moving direction is used as a smoothing value (Equation 2).
[0048]
However, each symbol is the same as in (Equation 1), total is the sum of the predetermined sampling points in each moving direction, H is the average of total by the number of samplings, Σ is all the predetermined points for m It is a sum. Note that Σ may not be all the predetermined points. For example, the sum of the values excluding the highest point may be used.
[0049]
c) Median method
[0050]
[Expression 4]
S (i, j, f, d) = median (i, j, f, d, m) ... (Equation 4)
In this method, the median value at a predetermined sampling point in the moving direction is set to a smoothing value (Equation 4). The median value takes an intermediate value of the permutation when the sampling point values are arranged in ascending or descending order. In the case of an odd number, the center value is taken, and in the case of an even number, one of the center two values or the average of the two values is taken. However, each symbol is the same as in (Formula 1), and median (i, j, f, d, m) is a function that returns a median value from m sampling points in the d direction.
[0051]
d) Average value-Weighted average method of original picture
[0052]
[Equation 5]
S (i, j, f, d) = a * H1 + (1-a) * G (i, j, f) (Expression 5)
[0053]
[Formula 6]
a = ρ0 ** 2 / (ρ1 ** 2 + (αDρ) ** 2) (Equation 6)
In this method, the smoothing value (Equation 5) is set so as to be between the average value and the original value. However, each symbol is the same as in (Equation 1), H1 is a value obtained by averaging the predetermined sampling points in the moving direction, and ρ1 is a value obtained by calculating a deviation for the predetermined sampling points in the moving direction. ρ is a value that determines the degree of noise from the image. For example, it is an average value of points that are 5% or less from the maximum value from the detected image t = 0 to t = −10, and α is a predetermined constant that determines the degree of smoothing. It is.
[0054]
e) Median value-Weighted average method of original picture
[0055]
[Expression 7]
S (i, j, f, d) = a * M1 + (1-a) * G (i, j, f) (Expression 7)
This is a method in which a value between the median value and the original image value is used as a smoothing value (Equation 7). However, each symbol is the same as in (Equation 1), and M1 is a median value for a predetermined sampling point in the moving direction.
[0056]
f) Weighting method according to luminance difference
[0057]
[Equation 8]
S (i, j, f, d) = Σ (W (i, j, f, d, m) * I (i, j, f, d, m)) (Expression 8)
[0058]
[Equation 9]
W (i, j, f, d, m) = w (i, j, f, d, m) / wt (Equation 9)
[0059]
[Expression 10]
When m ≧ 1, w (i, j, f, d, m) = w (i, j, f, d, m-1) / (1 + ((I (i, j, f, d, m ) -G (i, j, f)) / (αDρ)) ** k) (Equation 10)
[0060]
## EQU11 ##
When m = 0, w (i, j, f, d, 0) = 1 (Equation 11)
[0061]
[Expression 12]
When m ≦ −1, w (i, j, f, d, m) = w (i, j, f, d, m + 1) / (1 + ((I (i, j, f, d, m) -G (i, j, f)) / (αDρ)) ** k) (Equation 12)
However, each symbol is the same as (Equation 1), (Equation 2), and (Equation 6), and W (i, j, f, d, m) is smoothed by I (i, j, f, d , m) indicates the weight when adding, w (i, j, f, d, m) and wt are intermediate terms when calculating the weight W, k is a predetermined constant, and ** k is k Means power.
[0062]
There are various methods other than (Equation 8) to (Equation 12). For example, (Equation 10) can be changed to (Equation 13) and (Equation 12) can be changed to (Equation 14).
[0063]
[Formula 13]
When m ≧ 1, w (i, j, f, d, m) = w (i, j, q, m-1) / (1 + exp ((I (i, j, f, d, m) -G (i, j, f)) / (αDρ))) (Equation 13)
[0064]
[Expression 14]
When m ≦ −1, w (i, j, f, d, m) = w (i, j, q, m + 1) / (1 + exp ((I (i, j, f, d, m ) -G (i, j, f)) / (αDρ))) (Expression 14)
g) Minimum change method
Further, the direction can also be detected by using the following (Equation 15) to (Equation 16) without using the above smoothing method. In this case, the direction taking the minimum value is defined as the moving direction.
[0065]
[Expression 15]

[0066]
The direction having the minimum value among the directions is defined as the moving direction. However, each symbol is the same as in (Equation 1) and (Equation 2), and abs () is a function that takes the absolute value of the value in ().
[0067]
[Expression 16]

[0068]
However, each symbol is the same as (Equation 1) and (Equation 2).
[0069]
The direction having the minimum value among the directions is defined as the moving direction.
Specific examples of the method of detecting the moving direction have been described above from a) to g). Any of these methods can be used, and other methods can be used. You can also use multiple of these methods,
The above operation is performed for each point. However, since the same predetermined sampling point cannot be taken at the end point of the image, it is necessary to perform special processing. As special processing, both end points of the image are connected, the same sampling point as the central part is taken, the end point is not processed, the original image or 0 is output, and the end sampling points are reduced. There are various special processing methods for the edge part, such as a method for processing the image and a method for outputting the original image as it is.
[0070]
Step 603: Nonlinear smoothing process
Nonlinear smoothing processing is performed in the direction obtained in step 602. However, when the same sampling point and smoothing method as in step 602 are used, the smoothed value obtained in step 602 can be used as it is. There are various non-linear smoothing methods, and the same smoothing method as in step 602 is used.
[0071]
Step 604: Data output processing
The detection image obtained by smoothing is output.
[0072]
As described above, it is possible to obtain the moving direction of the subject in consideration of the moving amount of the ship in the negative direction from the focused detection image and perform the smoothing process with reference to the data estimated in the direction.
[0073]
(2) Second embodiment of the present invention for solving the problems
The second embodiment describes an example effective when an obstacle is moving.
Obstacles such as driftwood may be moved by the tide. In this case, no signal is detected in the same direction, and the direction of the obstacle changes in time series. In order to cope with this moving obstacle, as shown in FIG. 7, an X-axis search range 701 is provided to estimate which direction the obstacle has changed. The X-axis search range can be freely set because there is a common-sense range depending on the current and obstacles to be detected. In this case, the point of interest G (i, j, 0) at the time of the detection image t = 0 changes the moving direction from each direction connecting the points in the rectangle surrounded by the X-axis search range 701 and the Y-axis search range 106. Thus, the direction of the signal to be smoothed is estimated.
[0074]
FIG. 8 is an enlarged view of FIG. A direction connecting each point in the rectangle surrounded by the X-axis search range 701 and the Y-axis search range 106 and G (i, j, 0) of the point of interest A when the detection image t = 0 is defined. A point indicated by a white circle 801 is a sampling point of a pixel on the image, and a point indicated by a black circle 802 is a sampling point that is not on the pixel and requires interpolation. There are various interpolation methods as described in the moving direction detecting means. The moving direction is detected using the points sampled in each direction, and smoothing is performed by performing a one-dimensional nonlinear smoothing process in the minimum direction.
[0075]
Further, not only a method of smoothing with reference to past data from time t = 0, but also detection images t = 1 to t = 10 acquired after the detection image t = 0 in FIG. 1 as shown in FIG. It is also possible to use and smooth the detected images at the previous and subsequent times. When G (i, j, 0) 101 at the point of interest A of the target detection image t = 0, the time axis-direction has the same configuration as before, and the time axis + direction is smoothed in the time axis + direction. A time 903, an offset 901 in consideration of the amount of movement of the ship in the time axis + direction, and a Y axis search range 902 in the time axis + direction are provided. The smoothing time in the time axis + direction can be set freely as well as 11 as in the example. Similarly, the offset considering the amount of movement of the ship in the time axis + direction and the Y axis search range in the time axis + direction can be set freely. For G (i, j, 0) of point of interest A of detection image t = 0, the movement direction is detected in the time axis + direction and the time axis-direction, and one-dimensional data is selected from each obtained movement direction. One-dimensional nonlinear smoothing processing can also be performed. As a result, it is possible to obtain the moving direction of the subject in consideration of the movement amount of the ship in the + direction and the − direction in time from the focused detection image, and to perform the smoothing process with reference to the data estimated in that direction. .
[0076]
FIG. 10 is a diagram showing a processing procedure of a method for solving the problems of the present invention, and only steps different from those in the first embodiment will be described below with the numbers in FIG. 10 as step numbers.
[0077]
Step 1001: Data compression process 1
In order to improve the S / N ratio, the input detected image t is subjected to processing in accordance with (Equation 17) in which data is compressed to 1 / p in the Y-axis direction.
[0078]
[Expression 17]
G ′ (i, j, t) = press (i, j * p, t) (Equation 17)
Where G '(i, j, t) is the output value compressed to 1 / p, and press is a function that takes the maximum value from G (i, q, t) to G (i, q, t). Yes, q takes a value from j * p to j * p + p-1. p is a positive integer value. The value obtained by compressing the above operation from the original image for each point of the output image G ′ (i, j, t) is obtained. However, since the end point of the image cannot take the same predetermined sampling point, special processing is performed. There is a need to do. As for the special processing of the edge portion of the image, the data blurring processing in the following step 1002 is performed.
[0079]
Step 1002: Data blur processing
For the detected image t, in order to facilitate signal detection, the point of interest G (i, j, t) and the point G (i, jn, t) before and after the Y-axis direction to G (i, j + n, t ), The process of selecting the maximum value (Equation 18) is performed.
[0080]
[Formula 18]
G ′ (i, j, t) = max (i, j, t, n) (Equation 18)
However, G '(i, j, t) is the value after data blurring processing, and max is a function that returns the maximum value of n points before and after in the Y-axis direction centering on G (i, j, t) It is. The above operation is performed for each point. However, since the same predetermined sampling point cannot be taken at the end point in the Y-axis direction of the image, it is necessary to perform special processing.
[0081]
Step 1003: Data compression processing 2
A process for compressing data in the Y-axis direction is performed in the same manner as the data compression process 1 on the smoothed detection image in order to improve the S / N ratio.
[0082]
Step 1004: Density adjustment processing
The smoothed detection image is subjected to the process of adjusting the density to make it easier to detect an obstacle signal (Equation 19).
[0083]
[Equation 19]
G ′ (i, j, t) = G (i, j, t) / ave (i, j, t, n);
However, G ′ (i, j, t) is a value after density conversion processing, and ave is a function that returns the average of n points before and after the G (i, j, t) in the Y-axis direction.
[0084]
The above operation is performed for each point. However, since the same predetermined sampling point cannot be taken at the end point in the Y-axis direction of the image, it is necessary to perform special processing.
[0085]
(3) Third embodiment of the present invention for solving the problems
The movement direction is detected using the edge image of the detection image, and the detection image t is smoothed using the detected movement direction. FIG. 11 is a diagram showing a processing procedure of a method for solving the problems of the present invention, and only steps different from those of the second embodiment will be described below using the numbers in FIG. 11 as step numbers.
[0086]
Step 1101: Edge processing
An edge image is created for the detected image t. The edge image is described in Document 2 described above.
[0087]
Step 1102: Movement direction detection processing
The moving direction is detected using the edge image of the detected image t created in step 1101. The moving direction detection process is the same as that in step 602 of the first embodiment.
[0088]
Step 1103: Nonlinear smoothing process
Smoothing is performed in the moving direction of each point obtained in step 1102. The nonlinear smoothing processing method is the same as that in step 603 of the first embodiment.
[0089]
(4) Fourth embodiment of the present invention for solving the problems
The depth of the obstacle can be measured from the moving distance of the ship and the moving direction of the obstacle detected by the moving direction detecting means. FIG. 12 is a simulation diagram when the moving direction of the ship and the moving direction of the obstacle coincide. The correlation between the distance to the obstacle and the depth and other parameters at time t = 0 was expressed. In this case, how to obtain the horizontal distance y to the obstacle of the ship and the depth d of the obstacle will be described below.
[0090]
When the ship position 1201 at time t = 0 and the ship position 1202 at time t = −10, the distance 1205 of the ship moving from time t = −10 to time t = 0 is l, and at time t = 0. Assuming that the distance 1203 to the obstacle is the distance 1204 to the obstacle at time t = −10, in the time axis-distance cross-sectional view of the detection images superimposed in time series, the direction detected by the moving direction detection means Is in the direction 1208, the distance 1205 of the ship moving from time t = -10 to time t = 0 corresponds to 105, the distance 1203 to the obstacle at time t = 0 corresponds to 1209, and at time t = -10 The distance 1204 to the obstacle corresponds to 1210, respectively.
[0091]
If the horizontal distance 1206 to the obstacle at time t = 0 is y and the obstacle depth 1207 is d, the following equations (20) and (21) can be used.
[0092]
[Expression 20]

[0093]
[Expression 21]

[0094]
In the above-described embodiment, the ship is described, but other types such as an aircraft and a submarine can be applied. In addition, noise removal is applicable not only to those detected by active sonar but also to general display in image processing.
[0095]
【The invention's effect】
According to the present invention, there is an effect that noise is removed from a detection image obtained in time series, an obstacle for navigation of a ship is detected, and navigation of the ship becomes safe.
[Brief description of the drawings]
FIG. 1 is a conceptual diagram of processing for performing one-dimensional non-linear smoothing of detection images superimposed in time series in the minimum change direction according to the present invention.
FIG. 2 is a diagram illustrating a time axis-distance cross-sectional view of detection images superimposed in time series in FIG. 1;
FIG. 3 is a diagram showing an active sonar observation region from a ship.
FIG. 4 is a diagram showing a beamforming measurement region and a measurement signal from a ship.
FIG. 5 is a diagram showing detection images arranged in time series.
FIG. 6 is a diagram showing the procedure of the first embodiment of the present invention.
FIG. 7 is a conceptual diagram of processing for one-dimensional nonlinear smoothing in the minimum change direction in consideration of the movement of an obstacle.
FIG. 8 is an enlarged view of FIG. 7;
FIG. 9 is a conceptual diagram of processing for one-dimensional nonlinear smoothing in the minimum change direction in the time axis + direction and − direction.
FIG. 10 is a diagram showing a procedure of a second embodiment of the present invention.
FIG. 11 is a diagram showing the procedure of a third embodiment of the present invention.
FIG. 12 is a simulation diagram in a case where the moving direction of the ship matches the moving direction of the obstacle.
[Explanation of symbols]
101: Point of interest G (i, j, 0) when detection image t = 0, 102: Estimated maximum movement point, 103: Estimation minimum movement point, 104: Point of interest when detection image t = -10 (i , j, -10), 105: offset considering the amount of movement of the ship, 106: y-axis search range, 107: smoothing time, 201: sampling point on the pixel, 202: sampling point requiring interpolation, 301: ship at t = 0, 302: ship at t = -10, 303: rock on which the ship is stranded, 304: observation area at t = 0, 305: observation area at t = -10, 401: Direction A, 402: Direction B, 403: Direction C, 404: Measurement region beamformed in direction A, 405: Measurement region beamformed in direction B, 406: Measurement region beamformed in direction C, 410: Ship observation area, 411: Signal measured by beamforming in direction A, 412: Beam in direction B Signal measured by forming, 413: signal measured by beam forming in direction C, 501: detection obtained by density conversion of signals in each direction from direction A to direction C when t = 0 Image, 502: Obstacle measured at t = 0, 503: Obstacle measured at t = -1, 504: Obstacle measured at t = -2, 505: t = -10 Occasionally measured obstacles, 601: data input processing, 602: movement direction detection processing, 603: nonlinear smoothing processing, 604: data output processing, 701: x-axis search range, 801: sampling point on pixel, 802: Sampling points that require interpolation, 901: offset taking into account the amount of movement of the ship in the time t-axis + direction, 902: time t-axis + y-axis search range in the direction, 903: smoothing time in the time t-axis + direction, 1001 : Data compression processing 1, 1002: Data Tablization processing, 1003: Data compression processing 2, 1004: Density adjustment processing, 1101: Edge processing, 1201: Ship at time t = 0, 1202: Ship at time t = -10, 1203: Obstacle at time t = 0 1204: Distance to the obstacle at time t = -10 1205: Distance of the ship moving from time t = -10 to time t = 0, 1206: To obstacle at time t = 0 1207: Depth of obstacle 1208: Direction detected by moving direction detection means 1209: Distance to obstacle at time t = 0 in time axis-distance cross-sectional view of detection image 1210: Detection image The distance to the obstacle at time t = -10 in the time axis-distance cross-sectional view of FIG.

Claims (4)

Input detection images from active sonar devices in time series,
Select multiple detection images from detection images input in time series,
For each sensed image that has been selected, sampled between a predetermined maximum estimated moving point and the minimum estimated travel point where the moving object may be estimated to be the same object, connecting the sampling points obtained by the target point and the sampling direction Smoothing every time, the direction of taking the maximum value from the smoothed value as the moving direction of the moving body,
An object display method comprising displaying the maximum value as a value of each detection image . The object display method according to claim 1,
When smoothing is performed for each direction, the smoothed value is obtained from the (data number × 3/4) th data value when the data to be smoothed for each direction are arranged in ascending order. An object display method characterized by the above. Means for inputting detected images in time series from an active sonar device;
  Means for selecting a plurality of detection images from detection images input in time series;
  For each selected detection image, sample between the predetermined maximum estimated moving point and the minimum estimated moving point where the moving object can be estimated as the same object, and connect the point of interest and the sampling point obtained by sampling Smoothing every time, a means for setting the direction of taking the maximum value from the smoothed value as the moving direction of the moving body,
  An object display device comprising: means for displaying the maximum value as a value of each detection image. The object display device according to claim 3,
  The means for setting the moving direction of the moving body is the (number of data × 3/4) th when data to be smoothed for each direction is arranged in ascending order when performing smoothing for each direction. An object display device characterized in that the smoothed value is obtained from a data value.

Images