JPH09252467A - Mobile object detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect a moving object based on data digitally coded and picked up by a camera. SOLUTION: An image signal from a monitor television camera 11 is converted into digital image compression coded data by a coding section 12. The coding section 12 encodes picture element information being a difference of an image between a preceding pattern and this pattern or between a succeeding pattern and this pattern and information of a motion vector representing the magnitude and the direction of the motion between the two patterns in the unit of plural blocks. The coded data are decoded by a decoding section 14 of a monitor center or the like. A motion vector as to a moving object in the case of decoding is extracted by a mobile object detection discrimination section 16. The discrimination section 16 extracts the motion vector and extracts an object block based on a motion index to obtain a static quantity and to discriminate the object moving object through the comparison with a reference value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving object detecting apparatus used as a monitoring apparatus using a television or as an automatic photographing apparatus, and more particularly to image data encoded and transmitted as digital data.

[0002]

2. Description of the Related Art As a security system, it is common practice to install a television camera so as to capture an image of the inside of a surveillance area and monitor the image captured by this television camera on a monitor screen in a predetermined surveillance room. There is. When performing such image monitoring, sensors such as an infrared sensor, a sound sensor, and an optical sensor are used as automatic detection means for tracking a moving object whose movement cannot be predicted or detecting an intruder in the monitoring area. Kind is used. Then, all the moving body information detected by such sensors is finally judged by human eyes.

However, in a system using such sensors, it is naturally necessary to install a large number of sensors such as an infrared sensor, and such sensors detect all moving objects. Therefore, all moving objects other than the purpose for monitoring are detected. Also, many external parts such as sensors are required, and the configuration scale is inevitably large.

Also, a television camera is used as a monitoring device for a predetermined area indoors or outdoors. In this case, a large number of surveillance cameras are installed in the surveillance area, and images taken by the surveillance cameras are displayed on a monitor screen installed in the central control room. In such a system for centralized management, a method is adopted in which image information captured by each surveillance camera is compressed and transmitted as digital data to efficiently transmit a high-definition image. However, in reality, since the images of a large number of monitors provided in the control room are each monitored by human eyes, there is a possibility that important information may be overlooked. 1 shows a configuration of an encoding processing device for transmitting screen information to a centralized control room or the like in the system, in which an input screen imaged by a television camera is divided into predetermined blocks by a block division unit 51. For example, the input screen is divided into small block units of “8 pixels × 8 lines”, the information converted into the block units is orthogonally transformed by the orthogonal transformation unit 52, and further quantized by the quantization unit 53. The variable length coding unit 54 performs coding. Then, the encoded information is output as pixel data and sent to a central control room or the like.

The coded information is inversely processed by the inverse quantizer 55 and the orthogonal inverse transformer 56 and stored in the frame memory 57. The screen information accumulated in the frame memory 57 is sent to the motion detecting unit 58 together with the current screen information when a temporally different image is input to the block dividing unit 51. Then, by comparing screens that differ in time with each other in units of several blocks (called macroblocks, which corresponds to, for example, 4 blocks of 2 blocks × 2 blocks), the magnitude and direction of the motion in macroblock units can be calculated. As estimated. By applying this motion vector to the screen stored in the frame memory 57, the lower screen input from the block division unit 51 is predicted to create a predicted image, and the encoding process is performed by the difference between the original image and this predicted image. By doing so, further compression is enabled. By such processing, the information on the encoded difference pixel, the parameter information thereof, and the motion vector in macroblock units are transmitted.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the above points, and when a screen imaged by a surveillance camera is digitally encoded and transmitted, a target is selected from the imaged screens. An object of the present invention is to provide a moving object detection device that can be used as a monitoring device or an automatic image capturing device that recognizes the presence of a moving object and automatically identifies a target object that moves within a predetermined imaging area. is there.

[0007]

A moving object detecting apparatus according to the present invention extracts a motion vector from the transmitted information when digitally compressing and transmitting a screen imaged by a video camera. Then, a continuous block is extracted from the motion index based on the extracted motion vector.
Then, the motion index based on the motion vector contained in this continuous block is accumulated and calculated, the statistic amount is obtained, and the size of the extracted continuous block area and the statistic amount of the motion index are respectively compared with the reference data to move. Determine your body.

In the moving object detecting device having such a structure, for example, a screen imaged by a television camera installed in the monitoring area is digitally encoded when being transmitted to a central control room or the like. Then, the information of the motion vector included in the encoded image information is used. That is,
Based on the statistic of the motion index based on this motion vector, the moving object in the imaging screen can be identified,
The size of the continuous block can be determined. Therefore, for example, when a person intrudes into the surveillance room, the movement of the person determines the intrusion, and a warning is issued, so that highly reliable security monitoring that does not depend on the attention of the person can be performed. I can do it.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below with reference to the drawings. FIG. 1A shows the configuration of a surveillance system that detects a moving body. The signal of the image captured by the surveillance television camera 11 is converted into digital image compression encoded data by the encoding unit 12. It The encoding unit 12 encodes and outputs pixel information of an image difference between the previous screen and the rear screen and motion vector information indicating the magnitude and direction of motion on the two screens in units of a plurality of blocks. .

The data encoded by the encoding unit 12 is transmitted to, for example, a monitoring center via a transmission line 13, and the monitoring center has a decoding unit 14 for decoding the transmitted encoded data. It is provided and reproduced as an image on the monitor 15. The decoding unit 14 naturally decodes the transmitted data. At this time, the moving object detection / determination unit 16 extracts a motion vector and determines a moving portion. The result determined by the moving body detection / determination unit 16 is supplied to the warning unit 18 and the VTR 17, and the operation thereof is controlled. That is, the VTR accumulates only the image of the portion including the moving body from the detection result.

FIG. 1B shows the structure of the moving body detecting / determining unit 16 in detail, and includes a motion vector extracting unit 161 to which the decoded data from the decoding unit 14 is input. The motion vector extraction unit 161 extracts a motion vector from the decoded data, and the motion index calculation / continuous block extraction unit 16
At 2, find the number of consecutive blocks that have movement within a specified threshold.

As shown in FIG. 2, a block existing in the drawing, which is to be contained in an object 30 corresponding to a human, naturally has a motion vector whose size and direction are almost the same as the motion of this object 30. There is.

For example, in MPEG processing, a motion vector has a horizontal component and a vertical component in a transmission sequence as shown in FIG. 3, and each component is further subjected to "Motion".
It is described separately in two parts, "Code" and "Motion r." On the decoding side, the motion vector is calculated from these two components.

Therefore, for example, as shown in FIG. 4A, from the two-dimensional data of the vector in the x and y directions, the magnitude of the vector is "r" and the direction is the rotational coordinate "θ".
(0 to 180 °) ”and the motion index is“ r × cos
If it is calculated and taken out as (θ) ”, those having the same size and direction to some extent are included within a certain threshold value of“ r × cos (θ) ”. The size of the object 30 can be estimated and the area of the moving object can be extracted by examining the number of consecutive blocks having the motion vector. Further, if the motion vectors contained in the entire area are averaged, the object 30 The whole movement can be discriminated, and this is accumulated for several screens in the statistic calculator 163 and the statistic is taken.

FIG. 4B shows the flow of the processing. First, in step 101, "r × θ" is calculated for the motion vector of each block, and in step 102, "r × θ" is calculated vertically or horizontally. It is checked whether the difference between the values of r × θ ″ is continuous blocks within a predetermined threshold value. And step 10
In step 3, the motion of the entire object, the size of the vector, and the size of the object are calculated. In step 104, the statistics for several screens are calculated and accumulated.

Here, the horizontal axis indicates time, and the vertical axis indicates the product "r" of the magnitude r and the cos θ which is a mapping of the rotation angle θ in the direction.
If xcos (θ) ”is taken, a vector that keeps a constant motion as shown in (A) of FIG. 5 is one in which a vector of approximately the same direction and the same magnitude continues in time. As shown in (B) of the figure, in the case of a small and fast-moving object such as a tree leaf, the direction and size of the movement frequently change.

The value "r × cos" thus obtained
(Θ) ”is stored as an index of motion for several screens. For example, the number of positive and negative exchanges of vector quantities, variance, etc. are obtained together with these data, and used as statistics on the motion of the block group.

There is a difference in size and movement between the target object and other objects. Therefore, for example, when it is intended to detect a somewhat large object such as a human in the screen, the area of the moving object extracted from continuous blocks having substantially the same motion vector becomes large to some extent. At the same time, the magnitude of the motion vector does not change drastically.

Therefore, regarding the target object motion in advance, the size of the object size reference value based on the number of consecutive blocks is stored in the size change reference data section 164, and the movement index value changes within a certain time. The reference value of the statistical information of the motion index statistic is input to the reference statistic data unit 165 respectively, and the reference values input in the comparison units 166 and 167, the continuous block extraction unit 162, and the statistic calculation unit 163 are input. Each is compared with the output. Then, the comparison results of the comparison units 166 and 167 are supplied to the determination unit 168, and when a certain specified value is satisfied, it is determined that the target moving object (human) is detected, and the warning unit
Warning information is transmitted to 18, and at the same time, a command is given to the VTR 17 to record a high-definition image at that time.

In the embodiment shown here, “r × cos (θ)” is used as the index of the movement of the object, but other quantities such as f (r, θ) and f (x, y)
Of course, it is possible to use.

In the above embodiment, it is considered that the reference statistic data of the motion vector is first learned according to the monitoring situation. As shown in FIG. 6, if the inside of the monitoring site such as the room is within a fixed range, 1
The range included in one block is automatically determined.

Therefore, for example, when the target is to detect the presence or absence of a person invading the room, the person is put in the room at the initial stage and passed through the image pickup screen of the camera, whereby the intruder is In a certain case, the size of a moving object extracted from a continuous block having a motion vector having a size within the threshold value and its statistics are learned. Then, a reference value for comparison and determination is set based on the learned value, which facilitates obtaining a high recognition rate.

FIG. 7 shows a second embodiment for performing such a learning operation. In the motion index calculation / continuous block extraction unit 162 and the statistic amount calculation unit 163, within a predetermined monitoring range in the initial state. For example, a human being is intruded into the learning memories 169 and 17 and the values obtained at that time are learned.
Learn and memorize at 0. Then, the data stored in the learning memories 169 and 170 are supplied to the comparison units 166 and 167 as reference values, respectively.

FIG. 8 shows image information in the case of detecting an intruder at the entrance, for example. In such a case, the fixed narrow space called the entrance is the monitoring target, and the location of the intruder is determined. Has been done. Therefore, the size of the intruder can be determined fairly accurately based on the number of blocks. In addition, since the intruder gradually moves closer to the entrance, the size of the block changes so as to gradually expand, and as a whole, it proceeds in the approach direction.

FIG. 9 illustrates a third embodiment corresponding to such an example, in which the output from each of the motion index calculation / continuous block extraction unit 162 and the statistic amount calculation unit 163 is converted into a time series analysis unit. Input to 171 and 172 to detect the state of temporal change, and to obtain reference data when a person enters the front door in the initial state, and to the learning memories 169 and 170 as in the previous embodiment. Each of them is learned and stored as reference data. By comparing these reference data with the obtained statistical data, etc., the accuracy of recognition of an intruder at the entrance, etc. is surely improved, and, for example, the recognition comparison with the case where an object is collapsed at the entrance is high. It is performed with high accuracy to prevent false recognition.

FIG. 10 shows the fourth embodiment, in which the moving object detection / determination unit 16 controls the camera so that the object extracted by the determined moving block group is always in the center of the screen. It has a tracking function to perform tracking in the direction of the visual field.

That is, when the moving object detection / judgment unit 16 judges the object determined to be the target from the captured image, the position information of the moving object extracted from the continuous block group is input to the address calculation unit 19, Calculate the position of a moving object. Then, the control unit 20 controls the image pickup direction of the television camera 11 based on the calculated value of the address calculation unit 19,
The imaging direction is directed to the target object.

[0028]

As described above, according to the moving object detecting apparatus of the present invention, by detecting the block of the moving object within the specified image pickup range, special intrusion detecting means such as sensors is used. Without this, a motion vector is extracted on the decoding side of the image signal, an intruding object is determined based on the extracted motion vector, and a moving object is accurately detected.

[Brief description of drawings]

FIG. 1A is a configuration diagram illustrating an image monitoring system that configures a moving body detection device according to an embodiment of the present invention, and FIG. 1B is a diagram illustrating a moving body / detection determination unit in this system. Diagram.

FIG. 2 is a diagram for explaining the size and motion vector of an object on an imaging screen.

FIG. 3 is a diagram illustrating motion vector data in an MPEG sequence of an image signal.

FIG. 4A is a diagram illustrating an example for obtaining a motion index, and FIG. 4B is a diagram illustrating a flow of a statistical amount calculation therefor.

FIG. 5A and FIG. 5B are views for explaining a difference in temporal change of a motion vector depending on an object.

FIG. 6 is a diagram for explaining an example of initial learning according to the second embodiment.

FIG. 7 is a configuration diagram illustrating the above embodiment.

FIG. 8 is a diagram illustrating a state of an intruder for explaining an intruder from the entrance according to the third embodiment.

FIG. 9 is a configuration diagram illustrating the above embodiment.

FIG. 10 is a configuration diagram illustrating a fourth embodiment.

FIG. 11 is a configuration diagram for explaining a conventionally known encoding process of a digital moving image.

[Explanation of symbols]

 11 ... TV camera 12 ... Encoding unit 13 ... Transmission line 14 ... Decoding unit 15 ... Monitor 16 ... Moving body detection / determination unit 17 ... VTR 18 ... Warning unit 19 ... Address calculation unit 20 ... Control unit 161 ... Motion vector extraction unit 162 ... Motion index calculation / continuous block extraction unit 163 ... Statistics calculation unit 164 ... Size change reference data unit 165 ... Motion index statistics reference measurement data unit 166, 167 ... Comparison unit 168 ... Judgment unit 169, 170 ... Learning memory 171, 172 ... Time series analysis section

Claims (7)

[Claims] 1. A motion vector extracting means for extracting a motion vector representing motions of mutually different screens in units of a plurality of blocks; a motion vector extracting means for extracting the transmitted encoded motion vector; A motion block extraction means for extracting a continuous block from the motion index based on the extracted motion vector; and a statistical amount calculation means for accumulating and calculating a motion index based on the motion vector included in the continuous block, and obtaining a statistic thereof. The size of the region of the extracted continuous block and the statistic obtained by the statistic calculation means are respectively compared with reference data, and a determination means for determining a moving body is provided, and the image is captured by the video camera. A moving object detection device characterized in that an object moving on the screen is recognized. 2. A motion vector extraction means for extracting a motion vector representing motions of mutually different screens in units of a plurality of blocks, and a motion block extraction for extracting a continuous block from a motion index based on the extracted motion vector. Means, a motion vector extraction means for extracting the transmitted encoded motion vector, a continuous block extraction means for extracting a continuous block of a motion index based on the extracted motion vector, and a continuous block included in the continuous block A motion index based on a motion vector that is accumulated and calculated, and a statistic calculating unit that obtains the statistic, a size of the continuous block region extracted from the continuous block, and a statistic obtained by the statistic calculating unit. The determination means for comparing the reference data with the determination of the moving body, and the movement of the moving body determined by this determination means A moving body detecting apparatus, wherein the moving body is positioned within a screen imaged by the video camera. . 3. The statistical amount computing means includes means for accumulating temporal changes of a motion index based on the motion vector for a plurality of frames, and moves from comparison with a preset statistical property of the motion index. The moving object detection apparatus according to claim 1, wherein the target object is fixed and detected. 4. A means for learning and storing the initial state of the statistical property of the motion index obtained by the statistical amount calculating means, and the initial state of the statistical property learned and stored in the storage means is stored. The moving body detection device according to claim 1, wherein the moving body detection device is used as reference statistical data. 5. The continuous block extracting means sets a threshold value of a motion index, and extracts a region of a target moving object from successive blocks having a motion index within the threshold value. The moving object detection apparatus according to claim 1, wherein the size of the moving object is predicted. 6. The motion index uses the magnitude and vector direction of the extracted motion vector, or is constituted by a function thereof.
The moving object detection device according to any one of 1. 7. The moving body detection device according to claim 1, wherein the motion index is composed of a vertical component and a horizontal component of the extracted motion vector and a function thereof.