JP2006303897A - Device and method for extracting facial image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a facial image extracting device for surely acquiring a facial image, which is required for monitoring, with small amount of arithmetic, and to provide a facial image extracting method. <P>SOLUTION: Inputted videos are accumulated in a memory, and a frame rate converting means thins the inputted video in the temporal direction, so as to convert the frame rate; the face of a person is extracted, based on the input image from the memory or the frame rate converting means, and then, the face image is outputted; the face image is evaluated by a prescribed method, so as to calculate an evaluation value. The calculated evaluation values are accumulated; a range is determined, from which the highest evaluation value is believed to be capable of being obtained from among the accumulated evaluation values; the reproduction of the videos within the applicable range is indicated to the memory; and then, the highest evaluation value is selected among the calculated evaluation values, so as to output the corresponding face image. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a face image extraction apparatus and a face image extraction method for extracting a face image necessary for monitoring from a captured video.

  In recent years, security video surveillance systems using surveillance cameras have been spreading. From small-scale systems targeting financial institutions and private shops, to a wide range of airports and stations using dozens of surveillance cameras. Various systems are used up to large-scale systems to be monitored.

  In such a monitoring system, not only the monitoring image directly displayed on the monitor is monitored, but also the recorded image is recorded on a recording device such as a time lapse recorder or a hard disk recorder, and the recorded image is later recorded as needed. In many cases, a confirmation method is used. In the method of always recording video, the capacity of the recording device is wasted, and it is necessary to check all the useless parts where no people are shown at the time of confirmation. For this reason, there is known a recording apparatus that can record only an image determined to be necessary for monitoring, for example, a portion where the monitoring target has moved, using an infrared sensor or an image processing apparatus connected to the recording apparatus. Yes.

  For example, Patent Document 1 proposes a face image monitoring system that can search for a specific person by extracting a person's face image from a captured video and registering it as a database, thereby improving the efficiency of monitoring work. ing.

Japanese Patent Laid-Open No. 2002-191044

  However, in the conventional face image extraction device used in the face image monitoring system disclosed in Patent Document 1, it is necessary to perform image processing to determine whether or not to record a video, so that the processing is realized. Therefore, there is a problem that a high-speed computing device is required, and the cost of the face image extracting device increases. In addition, there is a method in which processing is performed by thinning out the input video in order to reduce the amount of calculation, but this method may miss a face image.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a face image extraction device and a face image extraction method that can reliably acquire a face image necessary for monitoring with a small amount of calculation. .

The above object can be achieved by the following constitution and method.
(1) A face image extraction device, a memory for storing input video, a frame rate conversion means for thinning out the video in the time direction and converting a frame rate, an output of the memory, and the frame rate conversion means Switching means for switching the output of the face, face detection means for extracting a human face from the output of the switching means and outputting a face image, and calculating the evaluation value of the face image output from the face detection means by a predetermined method Evaluation value calculation means, output switching means for switching the output destination of the calculated evaluation value, evaluation value storage means for storing the evaluation value from the output switching means, and stored in the evaluation value storage means A detailed analysis range determining means for determining a range in which the highest evaluation value can be obtained from the evaluation value, and instructing reproduction of a video in a range corresponding to the memory; and the output switching means And a maximum evaluation value selection means for selecting the highest evaluation value from the evaluation values and outputting a corresponding face image.

  With this configuration, the highest evaluation value is selected by the maximum evaluation value selection means from the evaluation values from the output switching means, and the corresponding face image is output, so the face image necessary for monitoring can be reliably acquired with a small amount of computation. can do.

(2) The face image extraction device according to (1) above includes an image deletion unit that narrows down a detailed analysis range each time the evaluation value storage unit is updated and deletes an unnecessary image from the memory.

  With this configuration, every time the evaluation value storage unit is updated, the detailed analysis range is narrowed down by the image deletion unit, and unnecessary images are deleted from the memory, so that the amount of images that must be stored in the memory is reduced. be able to.

(3) A monitoring system, comprising the face image extraction device according to (1) or (2).

  With this configuration, it is possible to provide a monitoring system having the same effect as the effect obtained by the face image extracting device described in (1) or (2) above.

(4) A face image extraction method, wherein input video is stored in a memory, and the input video is thinned out in a time direction by a frame rate conversion means to convert a frame rate, and from the memory or the frame rate conversion means The face of the person is extracted from the input image and the face image is output, the face image is evaluated by a predetermined method to calculate an evaluation value, the calculated evaluation value is accumulated, and the accumulated evaluation value is A range in which the highest evaluation value is considered to be obtained is determined, an instruction to reproduce video in the range corresponding to the memory is given, the highest evaluation value is selected from the calculated evaluation values, and a corresponding face image is selected. Output.

  With this configuration, since the highest evaluation value is selected from the calculated evaluation values and the corresponding face image is output, the face image necessary for monitoring can be reliably acquired with a relatively small amount of calculation.

(5) In the face image extraction method according to (4), the detailed analysis range is narrowed down each time the accumulated evaluation value is updated, and unnecessary images are deleted from the memory.

  With this configuration, every time the accumulated evaluation value is updated, the detailed analysis range is narrowed down and unnecessary images are deleted from the memory, so that the amount of images that the memory must accumulate can be reduced. .

(6) A recording medium on which a program for executing the face image extraction method described in (4) or (5) is recorded.

  With this configuration, it is possible to provide a recording medium having the same effect as that obtained by the face image extraction method described in (4) or (5).

  Since the present invention selects the highest evaluation value from the evaluation values from the output switching means by the maximum evaluation value selection means and outputs the corresponding face image, the face image necessary for monitoring is reliably acquired with a small amount of computation. Therefore, it is possible to provide a face image extraction device having an effect that it can be performed.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration of a face image extraction apparatus according to Embodiment 1 of the present invention. In this figure, a face image extraction apparatus 100 according to the present embodiment includes a memory 101 that stores input video (input video), and a frame rate conversion unit 102 that converts the frame rate by thinning out the input video in the time direction. An input switching unit 103 that is connected to the memory 101 and the frame rate conversion unit 102 and switches input to the subsequent stage; a face detection unit 104 that extracts a human face from the output of the input switching unit 103 and outputs a face image; An evaluation value calculation unit 105 that evaluates the face image output from the detection unit 104 by a predetermined method to calculate an evaluation value, an output switching unit 108 that switches an output destination of the calculated evaluation value, and an output switching unit 108 An evaluation value accumulating unit 107 that accumulates the evaluation value, and a range in which the highest evaluation value can be obtained from the evaluation value accumulated in the evaluation value accumulating unit 107 is determined , A detailed analysis range selection unit 106 for instructing reproduction of a video in a range (detailed analysis range) corresponding to the memory 101, and a video stored in the memory 101 corresponding to the range where the highest evaluation value is considered to be obtained. A maximum evaluation value selection unit 109 that selects the highest evaluation value from the range based on the output from the output switching unit 108 and outputs a corresponding face image;

  The operation of the face image extraction apparatus 100 configured as described above will be described with reference to the flowcharts of FIGS. First, the input switching unit 103 is set so that the frame rate conversion unit 102 is the input source, and the output switching unit 108 is the evaluation value storage unit 107 as the output destination. The captured video (input video) is subjected to thinning processing at regular intervals by the frame rate conversion unit 102 (step S201), and is input to the face detection unit 104 via the input switching unit 103. The face detection unit 104 extracts a face image from the input image (step S202), and the evaluation value calculation unit 105 calculates an evaluation value for the extracted face image by a predetermined method, and uses the calculated evaluation value. The evaluation value is stored in the evaluation value storage unit 107 via the output switching unit 108 (step S203).

  For example, when the evaluation value is equal to or lower than the threshold value (step S204), the detailed analysis range selection unit 106 checks the evaluation value storage unit 107 and selects the highest evaluation value among the evaluation values accumulated so far. Based on the evaluation values before and after that, the range of the video to be analyzed in detail is determined (step S205). Then, the input source of the input switching unit 103 is set in the memory 101, the output destination of the output switching unit 108 is set in the maximum evaluation value selection unit 109 (step S206), and the memory 101 is instructed to reproduce the video in the range to be analyzed in detail. (Step S207).

  The face detection unit 104 and the evaluation value calculation unit 105 calculate the evaluation value for each frame of the image reproduced by the memory 101 by the same method as described above, and notify the maximum evaluation value selection unit 109 (steps S208 and S209). ). When all the evaluation values in the corresponding range have been calculated, the maximum evaluation value selection unit 109 selects the frame with the highest evaluation value and outputs it as the final output (step S210). After the output, the maximum evaluation value selection unit 109 sets the input source of the input switching unit 103 to the frame rate conversion unit 102 and the output destination of the output switching unit 108 to the evaluation value storage unit 107 (step S211).

  Various methods have been proposed for the face detection unit 104. For example, the difference between the input image and the background image is taken and binarized with a threshold value, and pattern matching is performed on each area of the binary image, which is human. Alternatively, a method of extracting the upper part of the area determined as a face image may be used.

  The evaluation value calculation unit 105 may be anything as long as it can evaluate the quality of the output image. In general, face size, image contrast, matching with a standard face template, or the like can be used.

  Note that the detailed analysis range is selected before and after the frame in which the highest evaluation value is obtained. For example, the data stored in the evaluation value storage unit 107 is approximated by a function to obtain the highest evaluation value. It is also possible to predict a frame that is considered to be capable of being selected, and select the frame before and after that frame as the detailed analysis range.

  In addition, although the maximum evaluation value selection unit 109 outputs a frame having the maximum evaluation value, for example, it may be set to output a plurality of higher evaluation values.

  With the above configuration and operation, it is possible to reduce the amount of calculation without causing face image omission. Here, the amount of calculation in the present embodiment will be described with reference to FIG. 3A shows the number of frames to be processed in the conventional example, and FIG. 3B shows the number of frames to be processed in the present embodiment. The vertical axis represents the evaluation value, the horizontal axis represents the time axis, and the Δ mark below the horizontal axis represents the frame subjected to face detection and evaluation value calculation processing.

  In FIG. 3 (a), evaluation values for all 31 frames are calculated, but in FIG. 3 (b), first, frames are thinned out to 1/6 to calculate evaluation values for 6 frames indicated by large circles. From the result, further evaluation values are calculated for five frames in which high evaluation values are expected. As a result, the evaluation value is calculated for 31 frames in the conventional example, whereas the evaluation value is calculated for 11 frames in the present embodiment, so that the calculation amount is reduced by 65% without reducing accuracy. Reduced.

(Embodiment 2)
FIG. 4 is a block diagram showing the configuration of the face image extraction apparatus according to Embodiment 2 of the present invention. In this figure, the same components as those in FIG. 1 described above are denoted by the same reference numerals and description thereof is omitted.

  In FIG. 4, the face image extraction apparatus 200 according to the present embodiment adds an image deletion unit 201 that deletes unnecessary images from the memory 101 in addition to the same configuration as the face image extraction apparatus 100 according to the first embodiment described above. It has the structure.

  The operation of the face image extraction apparatus 200 configured as described above will be described with reference to the flowcharts of FIGS. In FIG. 5, the steps common to FIG. 2 are denoted by the same reference numerals as those in FIG. In FIG. 5, every time the evaluation value storage unit 107 is updated, the image deletion unit 201 narrows down the detailed analysis range (step S <b> 301), and images in a range that cannot be the detailed analysis range at that time (unnecessary images) ) Is deleted from the memory 101 (step S302).

  With such a configuration and operation, the amount of images that must be stored in the memory 101 can be reduced.

  The present invention has an effect that a face image necessary for monitoring can be reliably acquired with a small amount of calculation, and is useful for a video monitoring system using a monitoring camera.

1 is a block diagram showing the configuration of a face image extraction device according to Embodiment 1 of the present invention. The flowchart for demonstrating operation | movement of the face image extraction apparatus which concerns on Embodiment 1 of this invention. The figure for demonstrating the amount of calculations of the face image extraction apparatus which concerns on Embodiment 1 of this invention. The block diagram which shows the structure of the face image extraction apparatus which concerns on Embodiment 2 of this invention. The flowchart for demonstrating operation | movement of the face image extraction apparatus which concerns on Embodiment 2 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100, 200 Face image extraction apparatus 101 Memory 102 Frame rate conversion part 103 Input switching part 104 Face detection part 105 Evaluation value calculation part 106 Detailed analysis range selection part 107 Evaluation value storage part 108 Output switching part 109 Maximum evaluation value selection part 201 Image Delete part

Claims (6)

A memory for storing the input video,
Frame rate conversion means for converting the frame rate by thinning out the video in the time direction;
Switching means for switching between the output of the memory and the output of the frame rate conversion means;
Face detection means for extracting a human face from the output of the switching means and outputting a face image;
Evaluation value calculating means for calculating an evaluation value of the face image output from the face detecting means by a predetermined method;
Output switching means for switching the output destination of the calculated evaluation value;
Evaluation value storage means for storing the evaluation value from the output switching means;
A detailed analysis range determining unit that determines a range in which the highest evaluation value can be obtained from the evaluation value stored in the evaluation value storage unit, and that instructs the reproduction of the video corresponding to the memory;
Maximum evaluation value selection means for selecting the highest evaluation value from the evaluation values from the output switching means and outputting a corresponding face image;
A face image extraction apparatus comprising:   The face image extraction apparatus according to claim 1, further comprising an image deletion unit that narrows down a detailed analysis range and deletes an unnecessary image from the memory each time the evaluation value storage unit is updated.   A monitoring system comprising the face image extraction device according to claim 1.   The input video is stored in a memory, the frame rate is converted by thinning the input video in the time direction by a frame rate conversion unit, and a human face is extracted from the input image from the memory or the frame rate conversion unit. An image is output, the face image is evaluated by a predetermined method to calculate an evaluation value, the calculated evaluation value is accumulated, and a range in which the highest evaluation value can be obtained from the accumulated evaluation value , Instructing the reproduction of the video in the range corresponding to the memory, selecting the highest evaluation value from the calculated evaluation values, and outputting a corresponding face image.   The face image extraction method according to claim 4, wherein a detailed analysis range is narrowed every time the accumulated evaluation value is updated, and unnecessary images are deleted from the memory.   A recording medium on which a program for executing the face image extraction method according to claim 4 is recorded.

Images