JP3637241B2 - Image monitoring method and image monitoring apparatus - Google Patents

Description

【００２５】
存在検出の他の方法として、ヒストグラムが複数の正規分布の和で構成されていると仮定して、分布数が特徴量メモリ７に記憶されている特徴量に近い方を存在検出の結果としても良い。
【００２６】
次に、背景更新部４で背景画像を更新する。背景更新部４では存在検出部３の検出結果を利用し、対象物が存在する状態から存在しない状態に変化した場合と、対象物が存在しない状態から存在する状態に変化した場合とで、背景更新処理を切り替える。背景更新の一例を説明する。
【００２７】
背景更新が過去フレームにおける画素値の加重平均で算出されるものとする。このときに対象物が無い状態から有る状態に変化した場合には、図３に示す様に時間的に緩やかに背景更新を行い、逆に対象物が有る状態から無い状態に変化した場合には、図４に示す様に素早く背景更新を行う。緩やかに背景更新をするには、多くの過去フレームまで利用して加重平均し、素早く背景更新をするには、最近のフレームのみを利用して加重平均すれば良い。
【００２８】
このように、対象物の有無の状態変化の方向により背景更新の時間幅を変えることにより、対象物が無い状態から有る状態に変化した場合には長い時間幅にわたって変化部分が検出され、有る状態から無い状態に変化した場合には短い時間幅で変化部分が検出されることになる。つまり、移動物体が画像中にある場合などは図５のように、移動物体に対応する画素値は頻繁に変動しており、その画素は常に変化部分として検出される。静止物体が背景上に加わった場合には、図６のように画素値の大きな変動は１度だけであるので、背景更新が行われるまでの時間幅で変化部分として検出される。逆に、背景にあった物体が取り除かれた場合には、画素値は図７のように変動し、短い時間幅で変化部分として検出される。
【００２９】
つまり、背景上に新たに加わった物体は変化部分として長く検出され、検出物の認識等の後処理に必要な情報を十分提供できる。逆に背景からとり除かれた物体は、それまでの過去画像や背景画像に情報があるため、長い時間幅での変化検出は必要なく、素早い背景更新によりその後の変化検出に備えることができる。
【００３０】
背景更新の他の例として、対象物の有無の変化の方向により、背景更新までの時間的遅れを変化させる方法を説明する。図８に示すように、対象物が無い状態から有る状態に変化した場合には、長い時間的遅れを持って背景更新処理をする。逆に対象物が有る状態から無い状態に変化した場合には、短い時間的遅れを持って背景更新処理をする。このようにすることで、１つ目の例と同様、背景上に新たに加わった物体は変化部分として長く検出され、検出物の認識等の後処理に必要な情報を十分提供でき、背景からとり除かれた物体は、素早い背景更新によりその後の変化検出に備えることができる。
【００３１】
次に、出力制御部８の動作について説明する。処理の流れとしては、背景更新部４と出力制御部８に前後関係はなく、フレーム毎に並列に処理が進む。本発明の出力制御部８は、変化部分の検出結果を表示する他に、背景上に加わった変化部分と、背景から取り除かれた変化部分を区別して表示装置９に表示する。
【００３２】
例えば、図９のように、表示画面中には変化部分が重ね書きされた監視画像が表示される領域（画像表示領域）Ｇ１の他に、背景上に新たに加わった検出対象物の画像が表示される領域Ｇ２と、背景から取り除かれた検出対象物の画像が表示される領域Ｇ３とがある。
【００３３】
存在検出部３は、無しから有りと検出された対象物については、画像表示領域Ｇ１に当該対象物の存在する画像領域の位置を例えば当該対象物の輪郭や外接矩形を強調・特殊表示する等して他と区別して表示すると共に、表示領域Ｇ２に監視画像から切り出された当該対象物の部分画像を表示する。また、有りから無しと検出された対象物については、画像表示領域Ｇ１に当該対象物の存在していた画像領域の位置を例えば当該対象物の輪郭や外接矩形を強調・特殊表示する等して他と区別して表示すると共に、当該対象物の存在していたときの過去の監視画像または背景画像中から対応する部分（対象物）の画像を切り出し、表示領域Ｇ３に表示する。
【００３４】
このようにすることで、ユーザは、出力装置９の表示画面を一見すれば、監視している画像中に何が加わり、何が取り除かれたかを容易に理解することができる。特に取り除かれたものについては、従来の監視画像からだけではわからないため、本発明による表示法を用いることで、ユーザに理解容易な情報を提供することができる。
【００３５】
以上説明したように、上記実施形態によれば、検出対象物が存在する場合と存在しない場合を区別し適切な背景更新を行うことで、背景上に新たに加わった物体に対しては、検出物の認識等の後処理に必要な情報を十分提供でき、背景からとり除かれた部分に対しては、素早い背景更新によりその後の変化検出に備えることができる。
【００３６】
また、画像中に加わった部分と取り除かれた部分を区別して表示することで、監視者に理解容易な情報を提供することができ、本発明の効果は多大である。
【００３７】
以上、種々の例を説明したが、本発明はこれらの例に限定されるものではなく、種々変形して応用可能である。
【００３８】
また、発明の実施の形態に記載した本発明の手法は、コンピュータに実行させることのできるプログラムとして、磁気ディスク（フロッピーディスク、ハードディスクなど）、光ディスク（ＣＤ−ＲＯＭ、ＤＶＤなど）、半導体メモリなどの記録媒体に格納して頒布することもできる。
【００３９】
【発明の効果】
以上説明したように、本発明によれば、監視領域中の対象物の変化（有無）を検知した際に、その検知した変化に応じた適切な呈示が行える。
【図面の簡単な説明】
【図１】本発明の実施形態にかかる画像監視装置の構成例を示した図。
【図２】対象物が有る場合と無い場合の画像特徴量としての画素値のヒストグラムの違いを説明するための図。
【図３】対象物が無い状態から有る状態へ変化した場合の背景画像更新例を説明するための図。
【図４】対象物が有る状態から無い状態へ変化した場合の背景画像更新例を説明するための図。
【図５】移動物体を検出する場合の背景画像の更新について説明するための図。
【図６】静止物体を検出する場合の背景画像の更新について説明するための図。
【図７】監視画像中の対象物が有りから無しと検出された場合の背景画像の更新について説明するための図。
【図８】背景更新の他の例について説明するための図。
【図９】検出結果の表示例を示した図。
【符号の説明】
１…画像入力部
２…変化検出部
３…存在検出部
４…背景更新部
５…画像メモリ
６…背景画像メモリ
７…特徴量メモリ
８…出力制御部
９…出力装置
１１…画像監視装置[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image monitoring apparatus that monitors an object using an image.
[0002]
[Prior art]
Japanese Patent Application Laid-Open No. 7-320063 discloses an image monitoring apparatus that performs background updating by smoothing each pixel with a hysteresis characteristic in the time direction. Although this apparatus employs a background update method that is not easily affected by video noise or the like, the process is not switched depending on whether or not an object exists.
[0003]
Japanese Patent Application Laid-Open No. 7-79429 discloses an image monitoring apparatus that updates a background when it is detected that a moving object is no longer extracted. In this apparatus, when a stationary object is added on the background, the background update is performed immediately, the change detection period is short, and there is a case where sufficient information cannot be provided for subsequent recognition processing, for example. Further, when the object is removed from the background, it is impossible to know what has been removed from only the currently monitored image.
[0004]
[Problems to be solved by the invention]
As described above, when the background image is updated when the change (presence / absence) of the object in the monitoring area is detected, the case where the detection object is changed from the non-existing state to the existing state is the background Although the time required for the image processing is different from the case where the object is present and the object disappears, the background image cannot be updated in accordance with the time width corresponding thereto, and appropriate state change detection processing is performed. Etc. cannot be performed continuously. In addition, for example, since the image processing result is not presented in synchronization with the movement of the screen display, there is a problem that the user is somewhat uncomfortable and difficult to use.
[0005]
In addition, in an image monitoring apparatus that detects a changed part due to a difference from a background image, when the detection target object changes from a non-existing state, the target object is detected by looking at the changed part. It is possible to determine what is, but when the object originally exists in the background and the object disappears, it is not possible to determine what was there by looking at the changed part There was a point.
[0006]
Therefore, the present invention uses an image monitoring method and an image monitoring method capable of appropriately presenting according to the detected change when detecting a change in the object in the monitoring area (presence / absence of the object in the monitoring area). An object is to provide an image monitoring apparatus.
[0007]
More specifically, when a background image is updated when a change in an object in the monitoring area (presence or absence of an object in the monitoring area) is detected, the detection object is changed from a non-existing state to an existing state. The background is updated according to the contents of the image processing when the object changes and when the object changes from the state originally existing in the background to the non-existing state.
[0008]
In addition, the fact that the object in the monitoring area has been newly added and the fact that it has been removed is distinguished and presented.
[0009]
[Means for Solving the Problems]
The image monitoring method of the present invention updates a background image when detecting a change in the presence / absence of an object from time-series image information, and the length of time required for updating the background image according to the change in state. By detecting the change in the object in the monitoring area (presence / absence of the object in the monitoring area), appropriate presentation according to the detected change can be performed. The processing can be continued, and it is convenient for the user.
[0010]
Preferably, the time required to complete the update of the background image when it is detected that there is no change in the state of the object is from the time it is completed until the update of the background image is detected when it is detected that there is no change. Longer than required.
[0011]
Further, in the image monitoring method of the present invention, when a change in the state of presence / absence of an object is detected from time-series image information, the object in which the change in state is detected as being present is detected, and the presence / absence of being detected as being absent. By distinguishing and presenting each object, when a change in the object in the monitoring area (presence / absence of the object in the monitoring area) is detected, an appropriate presentation according to the detected change can be performed, and the user Easy to use.
[0012]
Preferably, at least one of the image of the object detected as having at least the state change and the position of the image of the object is presented.
[0013]
The image monitoring apparatus of the present invention detects a state change of the presence / absence of an object from time-series image information, and updates the background image when the detection means detects a state change of the object, A change in an object in the monitoring area (presence or absence of an object in the monitoring area) is detected by providing an updating unit that changes the length of time required to update the background image according to the state change. At this time, appropriate presentation according to the detected change can be performed, processing for appropriate change detection and the like can be continued, and user convenience is good.
[0014]
Preferably, the time required to complete the update of the background image when it is detected that the state change of the object is absent from the absence by the detection means is the update of the background image when the presence or absence is detected. It takes longer to complete.
[0015]
Further, the image monitoring apparatus of the present invention includes a detection unit that detects a change in the presence / absence of an object from time-series image information, an object that is detected as having a change in state by the detection unit, and an absence of the object. When a change in the object in the monitoring area (presence / absence of the object in the monitoring area) is detected by providing a presentation means for distinguishing and presenting the detected object from the presence, the detection is performed. Appropriate presentation according to the change can be performed, which is convenient for the user.
[0016]
Preferably, at least one of the image of the object detected as having at least the state change and the position of the image of the object is presented.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
[0018]
FIG. 1 is a diagram showing a configuration example of an image monitoring apparatus according to an embodiment of the present invention.
[0019]
In the present embodiment, an image input unit 1 that inputs image data input from an imaging unit (not shown), an image memory 5 that stores image data digitized by the image input unit 1, and a background that stores a background image An image memory 6, a change detection unit 2 that detects a changed portion from the difference between the input image and the background image, a feature amount memory 7 that stores image feature amounts when the target is present and absent, and calculation from the input image The presence determination unit 3 that determines the presence of the object, the background update unit 4 that updates the background, and the detection result are displayed by comparing the image feature amount to be stored with the feature amount stored in the feature amount memory. And an output control unit 8 that performs the operation.
[0020]
A description will be given below along the flow of processing. First, an image feature amount in each case is calculated in advance from an image when the detection target exists and an image when the detection target does not exist, and is stored in the feature amount memory 7. As the image feature amount, for example, a histogram of pixel values can be used. FIG. 2 shows the difference in the histogram of pixel values when the object does not exist (FIG. 2A) and when it exists (FIG. 2B). As the feature amount, for example, a histogram value is normalized and stored.
[0021]
In the monitoring process, first, an image signal input from the outside is digitized by the image input unit 1 and stored in the image memory 5. The image memory 5 can store data for a plurality of frames, for example, image data for the past n frames (n is a positive integer).
[0022]
Next, the change detection unit 2 calculates the difference between the background image stored in the background image memory 6 in advance and the input image for each pixel, and changes if this difference is equal to or greater than a predetermined threshold value. Detect that it is a part.
[0023]
Next, the presence detection unit 3 calculates the image feature amount of the input image and compares it with the feature amount stored in the feature amount memory 7 to determine whether or not the detection target exists. Presence detection may be performed on the entire image, or may be performed for each area obtained by dividing the image. When the pixel value histogram is used as the image feature amount, the histogram of each pixel value in the changed portion is normalized, and the histogram when the target exists is compared with the histogram when the target does not exist. The normalized frequency of the pixel value i of the histogram having pixel values “0” to “m” stored in the feature amount memory 7 is D (i), and the normalization of the histogram of one change portion of the input image When the converted frequency is expressed as G (i), a value v is calculated for each of the cases where the detection target exists and does not exist, using the following formula, and the smaller value of v is calculated for the presence detection. As a result.
[0024]
[Expression 1]

[0025]
As another method of presence detection, assuming that the histogram is composed of the sum of a plurality of normal distributions, the one whose distribution number is closer to the feature quantity stored in the feature quantity memory 7 can be used as the result of the presence detection. good.
[0026]
Next, the background update unit 4 updates the background image. The background update unit 4 uses the detection result of the presence detection unit 3 to change the background between a state where the target object is present and a state where the target object is not present and a case where the target object is changed from a state where the target is not present. Switch the update process. An example of background update will be described.
[0027]
It is assumed that the background update is calculated by a weighted average of pixel values in the past frame. At this time, when the state is changed from the state where there is no object, the background is gradually updated as shown in FIG. 3, and when the state is changed from the state where the object is present to the state where there is no object. The background is updated quickly as shown in FIG. In order to update the background slowly, a weighted average is used using many past frames, and in order to update the background quickly, a weighted average may be used using only recent frames.
[0028]
In this way, by changing the time width of the background update according to the direction of the state change of the presence / absence of the target object, when changing from the state without the target object to the state with the target object, the changed part is detected over a long time width, In the case where the state is changed to a state where there is no change, the changed portion is detected in a short time width. That is, when the moving object is in the image, as shown in FIG. 5, the pixel value corresponding to the moving object frequently fluctuates, and the pixel is always detected as a changed portion. When a stationary object is added on the background, since the pixel value has a large change only once as shown in FIG. 6, it is detected as a changed portion in the time width until the background update is performed. On the contrary, when the object in the background is removed, the pixel value fluctuates as shown in FIG. 7, and is detected as a changed portion with a short time width.
[0029]
That is, an object newly added on the background is detected as a changed part for a long time, and information necessary for post-processing such as recognition of the detected object can be sufficiently provided. On the other hand, since the object removed from the background has information in the past image and background image so far, it is not necessary to detect change in a long time width, and it is possible to prepare for subsequent change detection by quick background update.
[0030]
As another example of background update, a method of changing the time delay until background update according to the direction of change in the presence or absence of an object will be described. As shown in FIG. 8, when the state is changed from the state where there is no object, the background update process is performed with a long time delay. On the other hand, when the object is changed from being present to not being present, the background update process is performed with a short time delay. In this way, as in the first example, the newly added object is detected as a changed part for a long time, and can provide sufficient information necessary for post-processing such as recognition of the detected object. The removed object can be prepared for subsequent change detection by quick background update.
[0031]
Next, the operation of the output control unit 8 will be described. As for the flow of processing, the background update unit 4 and the output control unit 8 have no context, and the processing proceeds in parallel for each frame. The output control unit 8 of the present invention displays the change part detection result on the display device 9 separately from the change part added on the background and the change part removed from the background.
[0032]
For example, as shown in FIG. 9, in addition to an area (image display area) G1 in which a monitoring image in which a changed portion is overwritten is displayed on the display screen, an image of a detection object newly added on the background is displayed. There is a region G2 to be displayed and a region G3 in which an image of the detection target removed from the background is displayed.
[0033]
The presence detection unit 3 highlights and specially displays the position of the image area where the target object exists in the image display area G1, for example, the outline or circumscribed rectangle of the target object, for the target object detected as being absent. In addition to displaying them separately from others, a partial image of the target object cut out from the monitoring image is displayed in the display area G2. Further, for an object detected as being present or absent, the position of the image area where the object is present is emphasized / specially displayed in the image display area G1, for example, the outline or circumscribed rectangle of the object. While distinguishing it from others, an image of a corresponding portion (object) is cut out from the past monitoring image or background image when the object exists, and displayed in the display area G3.
[0034]
By doing so, the user can easily understand what has been added to and removed from the image being monitored at a glance at the display screen of the output device 9. Since the information that has been removed in particular cannot be understood from the conventional monitoring image alone, information that is easy to understand can be provided to the user by using the display method according to the present invention.
[0035]
As described above, according to the above-described embodiment, it is possible to detect an object newly added on the background by distinguishing between the case where the detection target exists and the case where the detection target does not exist and performing appropriate background updating. Information necessary for post-processing such as recognition of an object can be sufficiently provided, and a portion removed from the background can be prepared for subsequent change detection by quick background update.
[0036]
Further, by distinguishing and displaying the added part and the removed part in the image, it is possible to provide easy-to-understand information to the supervisor, and the effect of the present invention is great.
[0037]
Although various examples have been described above, the present invention is not limited to these examples, and various modifications can be applied.
[0038]
The method of the present invention described in the embodiment of the present invention is a program that can be executed by a computer, such as a magnetic disk (floppy disk, hard disk, etc.), an optical disk (CD-ROM, DVD, etc.), a semiconductor memory, etc. It can also be stored in a recording medium and distributed.
[0039]
【The invention's effect】
As described above, according to the present invention, when a change (presence / absence) of an object in a monitoring area is detected, an appropriate presentation according to the detected change can be performed.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration example of an image monitoring apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram for explaining a difference between histograms of pixel values as image feature amounts when there is an object and when there is no object.
FIG. 3 is a diagram for explaining an example of a background image update when a state is changed from a state where there is no object to a state where there is an object;
FIG. 4 is a diagram for explaining an example of a background image update when the object is changed from being present to not being present.
FIG. 5 is a diagram for explaining updating of a background image when a moving object is detected.
FIG. 6 is a diagram for explaining update of a background image when a stationary object is detected.
FIG. 7 is a diagram for explaining update of a background image when an object in a monitoring image is detected as being present or absent.
FIG. 8 is a diagram for explaining another example of background update.
FIG. 9 is a diagram showing a display example of detection results.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Image input part 2 ... Change detection part 3 ... Presence detection part 4 ... Background update part 5 ... Image memory 6 ... Background image memory 7 ... Feature-value memory 8 ... Output control part 9 ... Output device 11 ... Image monitoring apparatus

Claims (8)

A first step of inputting a time-series frame image;
By comparing the background image stored in the storage means for detecting the change of each object from the input frame image and the input first frame image, the object is obtained from the first frame image. A second step of determining whether or not an object exists from the first frame image when a change is detected,
A third step of updating the background image stored in the storage unit with a new background image generated by weighted averaging pixel values of a plurality of past frame images input;
Have
The third step includes
When it is determined that the target object exists in the frame image input before the first frame image and the target object does not exist in the first frame image, thereafter, from the first frame image, Less than the case where it is determined that the target object does not exist in the previously input frame image and the target object exists in the first frame image, and the new background image is used by using the latest frame image. Produces
When it is determined that no object exists in the frame image input before the first frame image and the object exists in the first frame image, the first frame image is thereafter determined. The new background image is used by using a larger number of past frame images than when it is determined that the target object exists in the frame image input before and the target object does not exist in the first frame image. The image monitoring method characterized by producing | generating.   The frame image input before the first frame image does not exist in the frame image input before the first frame image, and the object existing in the first frame image and the frame image input before the first frame image The image monitoring method according to claim 1, further comprising: a fourth step of distinguishing and presenting an object that exists and is not present in the first frame image.   The fourth step is not present in the frame image input before the first frame image, but is based on the image of the object existing in the first frame image and the first frame image. 3. The image monitoring method according to claim 2, wherein at least one of images of an object existing in a previously input frame image and not present in the first frame image is presented.   In the fourth step, an image of an object that is present in a frame image input before the first frame image and does not exist in the first frame image is present in the object. The image monitoring method according to claim 2, wherein the image is cut out from a past frame image or background image and presented. Image input means for inputting time-series frame images;
Storage means for storing a background image for detecting a change in each object from the frame image input by the image input means;
Detecting means for comparing the background image stored in the storage means with the first frame image input by the image input means and detecting a change in the object from the first frame image;
Determination means for determining whether or not the object exists from the first frame image in which the change of the object is detected;
Updating means for updating the background image stored in the storage means with a new background image generated by weighted averaging pixel values of a plurality of past frame images;
With
The updating means includes
When it is determined that the target object exists in the frame image input before the first frame image and the target object does not exist in the first frame image, the first frame image is subsequently processed. The object is not present in the frame image input before the frame image, and is smaller than the case where it is determined that the object is present in the first frame image, and the new frame image is used by using the latest frame image. A simple background image,
When it is determined that no object exists in the frame image input before the first frame image and the object exists in the first frame image, the first frame image is thereafter determined. The new background image is used by using a larger number of past frame images than when it is determined that the target object exists in the frame image input before and the target object does not exist in the first frame image. Generating an image monitoring apparatus.   The frame image input before the first frame image does not exist in the frame image input before the first frame image, and the object existing in the first frame image and the frame image input before the first frame image The image monitoring apparatus according to claim 5, further comprising a presenting unit that presents each object that is present and is not present in the first frame image.   The presenting means is not present in the frame image input before the first frame image, and the object image existing in the first frame image and the first frame image are present before the first frame image. The image monitoring apparatus according to claim 6, wherein the image monitoring apparatus presents at least one of the images of the object that is present in the frame image input to and not present in the first frame image.   The presenting means presents an image of an object that is present in a frame image input before the first frame image and does not exist in the first frame image when the object is present. The image monitoring apparatus according to claim 6, wherein the image monitoring apparatus is cut out from a past frame image or background image and presented.

Images